Waqed H. Hassan| Zahraa Mohammad Fadhe*| Rifqa F. Thiab| Karrar Mahdi Civil Engineering Department, Faculty of Engineering, University of Warith Al-Anbiyaa, Kerbala 56001, Iraq Civil Engineering Department, Faculty of Engineering, University of Kerbala, Kerbala 56001, Iraq Corresponding Author Email: Waqed.hammed@uowa.edu.iq
OPEN ACCESS
Abstract:
This work investigates numerically a local scour moves in irregular waves around tripods. It is constructed and proven to use the numerical model of the seabed-tripod-fluid with an RNG k turbulence model. The present numerical model then examines the flow velocity distribution and scour characteristics. After that, the suggested computational model Flow-3D is a useful tool for analyzing and forecasting the maximum scour development and the flow field in random waves around tripods. The scour values affecting the foundations of the tripod must be studied and calculated, as this phenomenon directly and negatively affects the structure of the structure and its design life. The lower diagonal braces and the main column act as blockages, increasing the flow accelerations underneath them. This increases the number of particles that are moved, which in turn creates strong scouring in the area. The numerical model has a good agreement with the experimental model, with a maximum percentage of error of 10% between the experimental and numerical models. In addition, Based on dimensional analysis parameters, an empirical equation has been devised to forecast scour depth with flow depth, median size ratio, Keulegan-Carpenter (Kc), Froud number flow, and wave velocity that the results obtained in this research at various flow velocities and flow depths demonstrated that the maximum scour depth rate depended on wave height with rising velocities and decreasing particle sizes (d50) and the scour depth attains its steady-current value for Vw < 0.75. As the Froude number rises, the maximum scour depth will be large.
Keywords:
local scour, tripod foundation, Flow-3D, waves
1. Introduction
New energy sources have been used by mankind since they become industrialized. The main energy sources have traditionally been timber, coal, oil, and gas, but advances in the science of new energies, such as nuclear energy, have emerged [1, 2]. Clean and renewable energy such as offshore wind has grown significantly during the past few decades. There are numerous different types of foundations regarding offshore wind turbines (OWTs), comprising the tripod, jacket, gravity foundation, suction anchor (or bucket), and monopile [3, 4]. When the water depth is less than 30 meters, Offshore wind farms usually employ the monopile type [4]. Engineers must deal with the wind’s scouring phenomenon turbine foundations when planning and designing wind turbines for an offshore environment [5]. Waves and currents generate scour, this is the erosion of soil near a submerged foundation and at its location [6]. To predict the regional scour depth at a bridge pier, Jalal et al. [7-10] developed an original gene expression algorithm using artificial neural networks. Three monopiles, one main column, and several diagonal braces connecting the monopiles to the main column make up the tripod foundation, which has more complicated shapes than a single pile. The design of the foundation may have an impact on scour depth and scour development since the foundation’s form affects the flow field [11, 12]. Stahlmann [4] conducted several field investigations. He discovered that the main column is where the greatest scour depth occurred. Under the main column is where the maximum scour depth occurs in all experiments. The estimated findings show that higher wave heights correspond to higher flow velocities, indicating that a deeper scour depth is correlated with finer silt granularity [13] recommends as the design value for a single pile. These findings support the assertion that a tripod may cause the seabed to scour more severely than a single pile. The geography of the scour is significantly more influenced by the KC value (Keulegan–Carpenter number)
The capability of computer hardware and software has made computational fluid dynamics (CFD) quite popular to predict the behavior of fluid flow in industrial and environmental applications has increased significantly in recent years [14].
Finding an acceptable piece of land for the turbine’s construction and designing the turbine pile precisely for the local conditions are the biggest challenges. Another concern related to working in a marine environment is the effect of sea waves and currents on turbine piles and foundations. The earth surrounding the turbine’s pile is scoured by the waves, which also render the pile unstable.
In this research, the main objective is to investigate numerically a local scour around tripods in random waves. It is constructed and proven to use the tripod numerical model. The present numerical model is then used to examine the flow velocity distribution and scour characteristics.
2. Numerical Model
To simulate the scouring process around the tripod foundation, the CFD code Flow-3D was employed. By using the fractional area/volume method, it may highlight the intricate boundaries of the solution domain (FAVOR).
This model was tested and validated utilizing data derived experimentally from Schendel et al. [15] and Sumer and Fredsøe [6]. 200 runs were performed at different values of parameters.
2.1 Momentum equations
The incompressible viscous fluid motion is described by the three RANS equations listed below [16]:
where, respectively, u, v, and w represent the x, y, and z flow velocity components; volume fraction (VF), area fraction (Ai; I=x, y, z), water density (f), viscous force (fi), and body force (Gi) are all used in the formula.
2.2 Model of turbulence
Several turbulence models would be combined to solve the momentum equations. A two-equation model of turbulence is the RNG k-model, which has a high efficiency and accuracy in computing the near-wall flow field. Therefore, the flow field surrounding tripods was captured using the RNG k-model.
2.3 Model of sediment scour
2.3.1 Induction and deposition
Eq. (4) can be used to determine the particle entrainment lift velocity [17].
α𝛼i is the Induction parameter, ns the normal vector is parallel to the seafloor, and for the present numerical model, ns=(0,0,1), θ𝜃cr is the essential Shields variable, g is the accelerated by gravity, di is the size of the particles, ρi is species density in beds, and d∗ The diameter of particles without dimensions; these values can be obtained in Eq. (5).
fbis the essential particle packing percentage, qb, i is the bed load transportation rate, and cb, I the percentage of sand by volume i. These variables can be found in Eq. (9), Eq. (10), fb, δ𝛿i the bed load thickness.
In this paper, after the calibration of numerous trials, the selection of parameters for sediment scour is crucial. Maximum packing fraction is 0.64 with a shields number of 0.05, entrainment coefficient of 0.018, the mass density of 2650, bed load coefficient of 12, and entrainment coefficient of 0.01.
3. Model Setup
To investigate the scour characteristics near tripods in random waves, the seabed-tripod-fluid numerical model was created as shown in Figure 1. The tripod basis, a seabed, and fluid and porous medium were all components of the model. The seabed was 240 meters long, 40 meters wide, and three meters high. It had a median diameter of d50 and was composed of uniformly fine sand. The 2.5-meter main column diameter D. The base of the main column was three dimensions above the original seabed. The center of the seafloor was where the tripod was, 130 meters from the offshore and 110 meters from the onshore. To prevent wave reflection, the porous media were positioned above the seabed on the onshore side.
Figure 1. An illustration of the numerical model for the seabed-tripod-fluid
3.1 Generation of meshes
Figure 2 displays the model’s mesh for the Flow-3D software grid. The current model made use of two different mesh types: global mesh grid and nested mesh grid. A mesh grid with the following measurements was created by the global hexahedra mesh grid: 240m length, 40m width, and 32m height. Around the tripod, a finer nested mesh grid was made, with dimensions of 0 to 32m on the z-axis, 10 to 30 m on the x-axis, and 25 to 15 m on the y-axis. This improved the calculation’s precision and mesh quality.
To increase calculation efficiency, the top side, The model’s two x-z plane sides, as well as the symmetry boundaries, were all specified. For u, v, w=0, the bottom boundary wall was picked. The offshore end of the wave boundary was put upstream. For the wave border, random waves were generated using the wave spectrum from the Joint North Sea Wave Project (JONSWAP). Boundary conditions are shown in Figure 3.
Figure 3. Boundary conditions of the typical problem
The wave spectrum peak enhancement factor (=3.3 for this work) and can be used to express the unidirectional JONSWAP frequency spectrum.
3.3 Mesh sensitivity
Before doing additional research into scour traits and scour depth forecasting, mesh sensitivity analysis is essential. Three different mesh grid sizes were selected for this section: Mesh 1 has a 0.45 by 0.45 nested fine mesh and a 0.6 by 0.6 global mesh size. Mesh 2 has a 0.4 global mesh size and a 0.35 nested fine mesh size, while Mesh 3 has a 0.25 global mesh size and a nested fine mesh size of 0.15. Comparing the relative fine mesh size (such as Mesh 2 or Mesh 3) to the relatively coarse mesh size (such as Mesh 1), a larger scour depth was seen; this shows that a finer mesh size can more precisely represent the scouring and flow field action around a tripod. Significantly, a lower mesh size necessitates a time commitment and a more difficult computer configuration. Depending on the sensitivity of the mesh guideline utilized by Pang et al., when Mesh 2 is applied, the findings converge and the mesh size is independent [20]. In the next sections, scouring the area surrounding the tripod was calculated using Mesh 2 to ensure accuracy and reduce computation time. The working segment generates a total of 14, 800,324 cells.
3.4 Model validation
Comparisons between the predicted outcomes from the current model and to confirm that the current numerical model is accurate and suitably modified, experimental data from Sumer and Fredsøe [6] and Schendel et al. [15] were used. For the experimental results of Run 05, Run 15, and Run 22 from Sumer and Fredsøe [6], the experimental A9, A13, A17, A25, A26, and A27 results from Schendel et al. [15], and the numerical results from the current model are shown in Figure 4. The present model had d50=0.051cm, the height of the water wave(h)=10m, and wave velocity=0.854 m.s-1.
Figure 5. Comparison of the present study’s maximum scour depth with that authored by Sumer and Fredsøe [6] and Schendel et al. [15]
According to Figure 5, the highest discrepancy between the numerical results and experimental data is about 10%, showing that overall, there is good agreement between them. The ability of the current numerical model to accurately depict the scour process and forecast the maximum scour depth (S) near foundations is demonstrated by this. Errors in the simulation were reduced by using the calibrated values of the parameter. Considering these results, a suggested simulated scouring utilizing a Flow-3D numerical model is confirmed as a superior way for precisely forecasting the maximum scour depth near a tripod foundation in random waves.
3.5 Dimensional analysis
The variables found in this study as having the greatest impacts, variables related to flow, fluid, bed sediment, flume shape, and duration all had an impact on local scouring depth (t). Hence, scour depth (S) can be seen as a function of these factors, shown as:
With the aid of dimensional analysis, the 14-dimensional parameters in Eq. (11) were reduced to 6 dimensionless variables using Buckingham’s -theorem. D, V, and were therefore set as repetition parameters and others as constants, allowing for the ignoring of their influence. Eq. (12) thus illustrates the relationship between the effect of the non-dimensional components on the depth of scour surrounding a tripod base.
(12)
\frac{S}{D}=f\left(\frac{h}{D}, \frac{d 50}{D}, \frac{V}{V W}, F r, K c\right)
where, SD𝑆𝐷 are scoured depth ratio, VVw𝑉𝑉𝑤 is flow wave velocity, d50D𝑑50𝐷 median size ratio, $Fr representstheFroudnumber,and𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑠𝑡ℎ𝑒𝐹𝑟𝑜𝑢𝑑𝑛𝑢𝑚𝑏𝑒𝑟,𝑎𝑛𝑑Kc$is the Keulegan-Carpenter.
4. Result and Discussion
4.1 Development of scour
Similar to how the physical model was used, this numerical model was also used. The numerical model’s boundary conditions and other crucial variables that directly influence the outcomes were applied (flow depth, median particle size (d50), and wave velocity). After the initial 0-300 s, the scour rate reduced as the scour holes grew quickly. The scour depths steadied for about 1800 seconds before reaching an asymptotic value. The findings of scour depth with time are displayed in Figure 6.
4.2 Features of scour
Early on (t=400s), the scour hole began to appear beneath the main column and then began to extend along the diagonal bracing connecting to the wall-facing pile. Gradually, the geography of the scour; of these results is similar to the experimental observations of Stahlmann [4] and Aminoroayaie Yamini et al. [1]. As the waves reached the tripod, there was an enhanced flow acceleration underneath the main column and the lower diagonal braces as a result of the obstructing effects of the structural elements. More particles are mobilized and transported due to the enhanced near-bed flow velocity, it also increases bed shear stress, turbulence, and scour at the site. In comparison to a single pile, the main column and structural components of the tripod have a significant impact on the flow velocity distribution and, consequently, the scour process and morphology. The main column and seabed are separated by a gap, therefore the flow across the gap may aid in scouring. The scour hole first emerged beneath the main column and subsequently expanded along the lower structural components, both Aminoroayaie Yamini et al. [1] and Stahlmann [4] made this claim. Around the tripod, there are several different scour morphologies and the flow velocity distribution as shown in Figures 7 and 8.
Figure 8. Random waves of flow velocity distribution around a tripod
4.3 Wave velocity’s (Vw) impact on scour depth
In this study’s section, we looked at how variations in wave current velocity affected the scouring depth. Bed scour pattern modification could result from an increase or decrease in waves. As a result, the backflow area produced within the pile would become stronger, which would increase the depth of the sediment scour. The quantity of current turbulence is the primary cause of the relationship between wave height and bed scour value. The current velocity has increased the extent to which the turbulence energy has changed and increased in strength now present. It should be mentioned that in this instance, the Jon swap spectrum random waves are chosen. The scour depth attains its steady-current value for Vw<0.75, Figure 9 (a) shows that effect. When (V) represents the mean velocity=0.5 m.s-1.
Figure 9. Main effects on maximum scour depth (Smax) as a function of column diameter (D)
4.4 Impact of a median particle (d50) on scour depth
In this section of the study, we looked into how variations in particle size affected how the bed profile changed. The values of various particle diameters are defined in the numerical model for each run numerical modeling, and the conditions under which changes in particle diameter have an impact on the bed scour profile are derived. Based on Figure 9 (b), the findings of the numerical modeling show that as particle diameter increases the maximum scour depth caused by wave contact decreases. When (d50) is the diameter of Sediment (d50). The Shatt Al-Arab soil near Basra, Iraq, was used to produce a variety of varied diameters.
4.5 Impact of wave height and flow depth (h) on scour depth
One of the main elements affecting the scour profile brought on by the interaction of the wave and current with the piles of the wind turbines is the height of the wave surrounding the turbine pile causing more turbulence to develop there. The velocity towards the bottom and the bed both vary as the turbulence around the pile is increased, modifying the scour profile close to the pile. According to the results of the numerical modeling, the depth of scour will increase as water depth and wave height in random waves increase as shown in Figure 9 (c).
4.6 Froude number’s (Fr) impact on scour depth
No matter what the spacing ratio, the Figure 9 shows that the Froude number rises, and the maximum scour depth often rises as well increases in Figure 9 (d). Additionally, it is crucial to keep in mind that only a small portion of the findings regarding the spacing ratios with the smallest values. Due to the velocity acceleration in the presence of a larger Froude number, the range of edge scour downstream is greater than that of upstream. Moreover, the scouring phenomena occur in the region farthest from the tripod, perhaps as a result of the turbulence brought on by the collision of the tripod’s pile. Generally, as the Froude number rises, so does the deposition height and scour depth.
4.7 Keulegan-Carpenter (KC) number
The geography of the scour is significantly more influenced by the KC value. Greater KC causes a deeper equilibrium scour because an increase in KC lengthens the horseshoe vortex’s duration and intensifies it as shown in Figure 10.
The result can be attributed to the fact that wave superposition reduced the crucial KC for the initiation of the scour, particularly under small KC conditions. The primary variable in the equation used to calculate This is the depth of the scouring hole at the bed. The following expression is used to calculate the Keulegan-Carpenter number:
Kc=Vw∗TpD𝐾𝑐=𝑉𝑤∗𝑇𝑝𝐷 (13)
where, the wave period is Tp and the wave velocity is shown by Vw.
Figure 10. Relationship between the relative maximum scour depth and KC
5. Conclusion
(1) The existing seabed-tripod-fluid numerical model is capable of faithfully reproducing the scour process and the flow field around tripods, suggesting that it may be used to predict the scour around tripods in random waves.
(2) Their results obtained in this research at various flow velocities and flow depths demonstrated that the maximum scour depth rate depended on wave height with rising velocities and decreasing particle sizes (d50).
(3) A diagonal brace and the main column act as blockages, increasing the flow accelerations underneath them. This raises the magnitude of the disturbance and the shear stress on the seafloor, which in turn causes a greater number of particles to be mobilized and conveyed, as a result, causes more severe scour at the location.
(4) The Froude number and the scouring process are closely related. In general, as the Froude number rises, so does the maximum scour depth and scour range. The highest maximum scour depth always coincides with the bigger Froude number with the shortest spacing ratio.
Since the issue is that there aren’t many experiments or studies that are relevant to this subject, therefore we had to rely on the monopile criteria. Therefore, to gain a deeper knowledge of the scouring effect surrounding the tripod in random waves, further numerical research exploring numerous soil, foundation, and construction elements as well as upcoming physical model tests will be beneficial.
Nomenclature
CFD
Computational fluid dynamics
FAVOR
Fractional Area/Volume Obstacle Representation
VOF
Volume of Fluid
RNG
Renormalized Group
OWTs
Offshore wind turbines
Greek Symbols
ε, ω
Dissipation rate of the turbulent kinetic energy, m2s-3
Subscripts
d50
Median particle size
Vf
Volume fraction
GT
Turbulent energy of buoyancy
KT
Turbulent velocity
PT
Kinetic energy of the turbulence
Αi
Induction parameter
ns
Induction parameter
ΘΘcr
The essential Shields variable
Di
Diameter of sediment
d∗
The diameter of particles without dimensions
µf
Dynamic viscosity of the fluid
qb,i
The bed load transportation rate
Cs,i
Sand particle’s concentration of mass
D
Diameter of pile
Df
Diffusivity
D
Diameter of main column
Fr
Froud number
Kc
Keulegan–Carpenter number
G
Acceleration of gravity g
H
Flow depth
Vw
Wave Velocity
V
Mean Velocity
Tp
Wave Period
S
Scour depth
References
[1] Aminoroayaie Yamini, O., Mousavi, S.H., Kavianpour, M.R., Movahedi, A. (2018). Numerical modeling of sediment scouring phenomenon around the offshore wind turbine pile in marine environment. Environmental Earth Sciences, 77: 1-15. https://doi.org/10.1007/s12665-018-7967-4
[2] Hassan, W.H., Hashim, F.S. (2020). The effect of climate change on the maximum temperature in Southwest Iraq using HadCM3 and CanESM2 modelling. SN Applied Sciences, 2(9): 1494. https://doi.org/10.1007/s42452-020-03302-z
[3] Fazeres-Ferradosa, T., Rosa-Santos, P., Taveira-Pinto, F., Pavlou, D., Gao, F.P., Carvalho, H., Oliveira-Pinto, S. (2020). Preface: Advanced research on offshore structures and foundation design part 2. In Proceedings of the Institution of Civil Engineers-Maritime Engineering. Thomas Telford Ltd, 173(4): 96-99. https://doi.org/10.1680/jmaen.2020.173.4.96
[4] Stahlmann, A. (2013). Numerical and experimental modeling of scour at foundation structures for offshore wind turbines. In ISOPE International Ocean and Polar Engineering Conference. ISOPE, pp. ISOPE-I.
[5] Petersen, T.U., Sumer, B.M., Fredsøe, J. (2014). Edge scour at scour protections around offshore wind turbine foundations. In 7th International Conference on Scour and Erosion. CRC Press, pp. 587-592.
[6] Sumer, B.M., Fredsøe, J. (2001). Scour around pile in combined waves and current. Journal of Hydraulic Engineering, 127(5): 403-411. https://doi.org/10.1061/(ASCE)0733-9429(2001)127:5(403)
[7] Jalal, H.K., Hassan, W.H. (2020). Effect of bridge pier shape on depth of scour. In IOP Conference Series: Materials Science and Engineering. IOP Publishing, 671(1): 012001. https://doi.org/10.1088/1757-899X/671/1/012001
[8] Hassan, W.H., Jalal, H.K. (2021). Prediction of the depth of local scouring at a bridge pier using a gene expression programming method. SN Applied Sciences, 3(2): 159. https://doi.org/10.1007/s42452-020-04124-9
[9] Jalal, H.K., Hassan, W.H. (2020). Three-dimensional numerical simulation of local scour around circular bridge pier using Flow-3D software. In IOP Conference Series: Materials Science and Engineering. IOP Publishing, 745(1): 012150. https://doi.org/10.1088/1757-899X/745/1/012150
[10] Hassan, W.H., Attea, Z.H., Mohammed, S.S. (2020). Optimum layout design of sewer networks by hybrid genetic algorithm. Journal of Applied Water Engineering and Research, 8(2): 108-124. https://doi.org/10.1080/23249676.2020.1761897
[11] Hassan, W.H., Hussein, H.H., Alshammari, M.H., Jalal, H.K., Rasheed, S.E. (2022). Evaluation of gene expression programming and artificial neural networks in PyTorch for the prediction of local scour depth around a bridge pier. Results in Engineering, 13: 100353. https://doi.org/10.1016/j.rineng.2022.100353
[12] Hassan, W.H., Hh, H., Mohammed, S.S., Jalal, H.K., Nile, B.K. (2021). Evaluation of gene expression programming to predict the local scour depth around a bridge pier. Journal of Engineering Science and Technology, 16(2): 1232-1243. https://doi.org/10.1016/j.rineng.2022.100353
[13] Nerland, C. (2010). Offshore wind energy: Balancing risk and reward. In Proceedings of the Canadian Wind Energy Association’s 2010 Annual Conference and Exhibition, Canada, p. 2000.
[14] Hassan, W.H., Nile, B.K., Mahdi, K., Wesseling, J., Ritsema, C. (2021). A feasibility assessment of potential artificial recharge for increasing agricultural areas in the kerbala desert in Iraq using numerical groundwater modeling. Water, 13(22): 3167. https://doi.org/10.3390/w13223167
[15] Schendel, A., Welzel, M., Schlurmann, T., Hsu, T.W. (2020). Scour around a monopile induced by directionally spread irregular waves in combination with oblique currents. Coastal Engineering, 161: 103751. https://doi.org/10.1016/j.coastaleng.2020.103751
[16] Yakhot, V., Orszag, S.A. (1986). Renormalization group analysis of turbulence. I. Basic theory. Journal of Scientific Computing, 1(1): 3-51. https://doi.org/10.1007/BF01061452
[17] Mastbergen, D.R., Van Den Berg, J.H. (2003). Breaching in fine sands and the generation of sustained turbidity currents in submarine canyons. Sedimentology, 50(4): 625-637. https://doi.org/10.1046/j.1365-3091.2003.00554.x
[18] Soulsby, R. (1997). Dynamics of marine sands. https://doi.org/10.1680/doms.25844
[19] Van Rijn, L.C. (1984). Sediment transport, part I: Bed load transport. Journal of Hydraulic Engineering, 110(10): 1431-1456. https://doi.org/10.1061/(ASCE)0733-9429(1984)110:10(1431)
[20] Pang, A.L.J., Skote, M., Lim, S.Y., Gullman-Strand, J., Morgan, N. (2016). A numerical approach for determining equilibrium scour depth around a mono-pile due to steady currents. Applied Ocean Research, 57: 114-124. https://doi.org/10.1016/j.apor.2016.02.010
This work investigates numerically a local scour moves in irregular waves around tripods. It is constructed and proven to use the numerical model of the seabed-tripodfluid with an RNG k turbulence model. The present numerical model then examines the flow velocity distribution and scour characteristics. After that, the suggested computational model Flow-3D is a useful tool for analyzing and forecasting the maximum scour development and the flow field in random waves around tripods. The scour values affecting the foundations of the tripod must be studied and calculated, as this phenomenon directly and negatively affects the structure of the structure and its design life. The lower diagonal braces and the main column act as blockages, increasing the flow accelerations underneath them. This increases the number of particles that are moved, which in turn creates strong scouring in the area. The numerical model has a good agreement with the experimental model, with a maximum percentage of error of 10% between the experimental and numerical models. In addition, Based on dimensional analysis parameters, an empirical equation has been devised to forecast scour depth with flow depth, median size ratio, Keulegan-Carpenter (Kc), Froud number flow, and wave velocity that the results obtained in this research at various flow velocities and flow depths demonstrated that the maximum scour depth rate depended on wave height with rising velocities and decreasing particle sizes (d50) and the scour depth attains its steady-current value for Vw < 0.75. As the Froude number rises, the maximum scour depth will be large.
금속 적층 제조 중 고체 상 변형 예측: Inconel-738의 전자빔 분말층 융합에 대한 사례 연구
Nana Kwabena Adomako a, Nima Haghdadi a, James F.L. Dingle bc, Ernst Kozeschnik d, Xiaozhou Liao bc, Simon P. Ringer bc, Sophie Primig a
Abstract
Metal additive manufacturing (AM) has now become the perhaps most desirable technique for producing complex shaped engineering parts. However, to truly take advantage of its capabilities, advanced control of AM microstructures and properties is required, and this is often enabled via modeling. The current work presents a computational modeling approach to studying the solid-state phase transformation kinetics and the microstructural evolution during AM. Our approach combines thermal and thermo-kinetic modelling. A semi-analytical heat transfer model is employed to simulate the thermal history throughout AM builds. Thermal profiles of individual layers are then used as input for the MatCalc thermo-kinetic software. The microstructural evolution (e.g., fractions, morphology, and composition of individual phases) for any region of interest throughout the build is predicted by MatCalc. The simulation is applied to an IN738 part produced by electron beam powder bed fusion to provide insights into how γ′ precipitates evolve during thermal cycling. Our simulations show qualitative agreement with our experimental results in predicting the size distribution of γ′ along the build height, its multimodal size character, as well as the volume fraction of MC carbides. Our findings indicate that our method is suitable for a range of AM processes and alloys, to predict and engineer their microstructures and properties.
Additive manufacturing (AM) is an advanced manufacturing method that enables engineering parts with intricate shapes to be fabricated with high efficiency and minimal materials waste. AM involves building up 3D components layer-by-layer from feedstocks such as powder [1]. Various alloys, including steel, Ti, Al, and Ni-based superalloys, have been produced using different AM techniques. These techniques include directed energy deposition (DED), electron- and laser powder bed fusion (E-PBF and L-PBF), and have found applications in a variety of industries such as aerospace and power generation[2], [3], [4]. Despite the growing interest, certain challenges limit broader applications of AM fabricated components in these industries and others. One of such limitations is obtaining a suitable and reproducible microstructure that offers the desired mechanical properties consistently. In fact, the AM as-built microstructure is highly complex and considerably distinctive from its conventionally processed counterparts owing to the complicated thermal cycles arising from the deposition of several layers upon each other [5], [6].
Several studies have reported that the solid-state phases and solidification microstructure of AM processed alloys such as CMSX-4, CoCr [7], [8], Ti-6Al-4V [9], [10], [11], IN738[6], 304L stainless steel[12], and IN718 [13], [14] exhibit considerable variations along the build direction. For instance, references [9], [10] have reported that there is a variation in the distribution of α and β phases along the build direction in Ti-alloys. Similarly, the microstructure of an L-PBF fabricated martensitic steel exhibits variations in the fraction of martensite [15]. Furthermore, some of the present authors and others [6], [16], [17], [18], [19], [20] have recently reviewed and reported that there is a difference in the morphology and fraction of nanoscale precipitates as a function of build height in Ni-based superalloys. These non-uniformities in the as-built microstructure result in an undesired heterogeneity in mechanical and other important properties such as corrosion and oxidation[19], [21], [22], [23]. To obtain the desired microstructure and properties, additional processing treatments are utilized, but this incurs extra costs and may lead to precipitation of detrimental phases and grain coarsening. Therefore, a through-process understanding of the microstructure evolution under repeated heating and cooling is now needed to further advance 3D printed microstructure and property control.
It is now commonly understood that the microstructure evolution during printing is complex, and most AM studies concentrate on the microstructure and mechanical properties of the final build only. Post-printing studies of microstructure characteristics at room temperature miss crucial information on how they evolve. In-situ measurements and modelling approaches are required to better understand the complex microstructural evolution under repeated heating and cooling. Most in-situ measurements in AM focus on monitoring the microstructural changes, such as phase transformations and melt pool dynamics during fabrication using X-ray scattering and high-speed X-ray imaging [24], [25], [26], [27]. For example, Zhao et al. [25] measured the rate of solidification and described the α/β phase transformation during L-PBF of Ti-6Al-4V in-situ. Also, Wahlmann et al. [21] recently used an L-PBF machine coupled with X-ray scattering to investigate the changes in CMSX-4 phase during successive melting processes. Although these techniques provide significant understanding of the basic principles of AM, they are not widely accessible. This is due to the great cost of the instrument, competitive application process, and complexities in terms of the experimental set-up, data collection, and analysis [26], [28].
Computational modeling techniques are promising and more widely accessible tools that enable advanced understanding, prediction, and engineering of microstructures and properties during AM. So far, the majority of computational studies have concentrated on physics based process models for metal AM, with the goal of predicting the temperature profile, heat transfer, powder dynamics, and defect formation (e.g., porosity) [29], [30]. In recent times, there have been efforts in modeling of the AM microstructure evolution using approaches such as phase-field [31], Monte Carlo (MC) [32], and cellular automata (CA) [33], coupled with finite element simulations for temperature profiles. However, these techniques are often restricted to simulating the evolution of solidification microstructures (e.g., grain and dendrite structure) and defects (e.g., porosity). For example, Zinovieva et al. [33] predicted the grain structure of L-PBF Ti-6Al-4V using finite difference and cellular automata methods. However, studies on the computational modelling of the solid-state phase transformations, which largely determine the resulting properties, remain limited. This can be attributed to the multi-component and multi-phase nature of most engineering alloys in AM, along with the complex transformation kinetics during thermal cycling. This kind of research involves predictions of the thermal cycle in AM builds, and connecting it to essential thermodynamic and kinetic data as inputs for the model. Based on the information provided, the thermokinetic model predicts the history of solid-state phase microstructure evolution during deposition as output. For example, a multi-phase, multi-component mean-field model has been developed to simulate the intermetallic precipitation kinetics in IN718 [34] and IN625 [35] during AM. Also, Basoalto et al. [36] employed a computational framework to examine the contrasting distributions of process-induced microvoids and precipitates in two Ni-based superalloys, namely IN718 and CM247LC. Furthermore, McNamara et al. [37] established a computational model based on the Johnson-Mehl-Avrami model for non-isothermal conditions to predict solid-state phase transformation kinetics in L-PBF IN718 and DED Ti-6Al-4V. These models successfully predicted the size and volume fraction of individual phases and captured the repeated nucleation and dissolution of precipitates that occur during AM.
In the current study, we propose a modeling approach with appreciably short computational time to investigate the detailed microstructural evolution during metal AM. This may include obtaining more detailed information on the morphologies of phases, such as size distribution, phase fraction, dissolution and nucleation kinetics, as well as chemistry during thermal cycling and final cooling to room temperature. We utilize the combination of the MatCalc thermo-kinetic simulator and a semi-analytical heat conduction model. MatCalc is a software suite for simulation of phase transformations, microstructure evolution and certain mechanical properties in engineering alloys. It has successfully been employed to simulate solid-state phase transformations in Ni-based superalloys [38], [39], steels [40], and Al alloys[41] during complex thermo-mechanical processes. MatCalc uses the classical nucleation theory as well as the so-called Svoboda-Fischer-Fratzl-Kozeschnik (SFFK) growth model as the basis for simulating precipitation kinetics [42]. Although MatCalc was originally developed for conventional thermo-mechanical processes, we will show that it is also applicable for AM if the detailed time-temperature profile of the AM build is known. The semi-analytical heat transfer code developed by Stump and Plotkowski [43] is used to simulate these profile throughout the AM build.
1.1. Application to IN738
Inconel-738 (IN738) is a precipitation hardening Ni-based superalloy mainly employed in high-temperature components, e.g. in gas turbines and aero-engines owing to its exceptional mechanical properties at temperatures up to 980 °C, coupled with high resistance to oxidation and corrosion [44]. Its superior high-temperature strength (∼1090 MPa tensile strength) is provided by the L12 ordered Ni3(Al,Ti) γ′ phase that precipitates in a face-centered cubic (FCC) γ matrix [45], [46]. Despite offering great properties, IN738, like most superalloys with high γ′ fractions, is challenging to process owing to its propensity to hot cracking [47], [48]. Further, machining of such alloys is challenging because of their high strength and work-hardening rates. It is therefore difficult to fabricate complex INC738 parts using traditional manufacturing techniques like casting, welding, and forging.
The emergence of AM has now made it possible to fabricate such parts from IN738 and other superalloys. Some of the current authors’ recent research successfully applied E-PBF to fabricate defect-free IN738 containing γ′ throughout the build [16], [17]. The precipitated γ′ were heterogeneously distributed. In particular, Haghdadi et al. [16] studied the origin of the multimodal size distribution of γ′, while Lim et al. [17] investigated the gradient in γ′ character with build height and its correlation to mechanical properties. Based on these results, the present study aims to extend the understanding of the complex and site-specific microstructural evolution in E-PBF IN738 by using a computational modelling approach. New experimental evidence (e.g., micrographs not published previously) is presented here to support the computational results.
2. Materials and Methods
2.1. Materials preparation
IN738 Ni-based superalloy (59.61Ni-8.48Co-7.00Al-17.47Cr-3.96Ti-1.01Mo-0.81W-0.56Ta-0.49Nb-0.47C-0.09Zr-0.05B, at%) gas-atomized powder was used as feedstock. The powders, with average size of 60 ± 7 µm, were manufactured by Praxair and distributed by Astro Alloys Inc. An Arcam Q10 machine by GE Additive with an acceleration voltage of 60 kV was used to fabricate a 15 × 15 × 25 mm3 block (XYZ, Z: build direction) on a 316 stainless steel substrate. The block was 3D-printed using a ‘random’ spot melt pattern. The random spot melt pattern involves randomly selecting points in any given layer, with an equal chance of each point being melted. Each spot melt experienced a dwell time of 0.3 ms, and the layer thickness was 50 µm. Some of the current authors have previously characterized the microstructure of the very same and similar builds in more detail [16], [17]. A preheat temperature of ∼1000 °C was set and kept during printing to reduce temperature gradients and, in turn, thermal stresses [49], [50], [51]. Following printing, the build was separated from the substrate through electrical discharge machining. It should be noted that this sample was simultaneously printed with the one used in [17] during the same build process and on the same build plate, under identical conditions.
2.2. Microstructural characterization
The printed sample was longitudinally cut in the direction of the build using a Struers Accutom-50, ground, and then polished to 0.25 µm suspension via standard techniques. The polished x-z surface was electropolished and etched using Struers A2 solution (perchloric acid in ethanol). Specimens for image analysis were polished using a 0.06 µm colloidal silica. Microstructure analyses were carried out across the height of the build using optical microscopy (OM) and scanning electron microscopy (SEM) with focus on the microstructure evolution (γ′ precipitates) in individual layers. The position of each layer being analyzed was determined by multiplying the layer number by the layer thickness (50 µm). It should be noted that the position of the first layer starts where the thermal profile is tracked (in this case, 2 mm from the bottom). SEM images were acquired using a JEOL 7001 field emission microscope. The brightness and contrast settings, acceleration voltage of 15 kV, working distance of 10 mm, and other SEM imaging parameters were all held constant for analysis of the entire build. The ImageJ software was used for automated image analysis to determine the phase fraction and size of γ′ precipitates and carbides. A 2-pixel radius Gaussian blur, following a greyscale thresholding and watershed segmentation was used [52]. Primary γ′ sizes (>50 nm), were measured using equivalent spherical diameters. The phase fractions were considered equal to the measured area fraction. Secondary γ′ particles (<50 nm) were not considered here. The γ′ size in the following refers to the diameter of a precipitate.
2.3. Hardness testing
A Struers DuraScan tester was utilized for Vickers hardness mapping on a polished x-z surface, from top to bottom under a maximum load of 100 mN and 10 s dwell time. 30 micro-indentations were performed per row. According to the ASTM standard [53], the indentations were sufficiently distant (∼500 µm) to assure that strain-hardened areas did not interfere with one another.
2.4. Computational simulation of E-PBF IN738 build
2.4.1. Thermal profile modeling
The thermal history was generated using the semi-analytical heat transfer code (also known as the 3DThesis code) developed by Stump and Plotkowski [43]. This code is an open-source C++ program which provides a way to quickly simulate the conductive heat transfer found in welding and AM. The key use case for the code is the simulation of larger domains than is practicable with Computational Fluid Dynamics/Finite Element Analysis programs like FLOW-3D AM. Although simulating conductive heat transfer will not be an appropriate simplification for some investigations (for example the modelling of keyholding or pore formation), the 3DThesis code does provide fast estimates of temperature, thermal gradient, and solidification rate which can be useful for elucidating microstructure formation across entire layers of an AM build. The mathematics involved in the code is as follows:
In transient thermal conduction during welding and AM, with uniform and constant thermophysical properties and without considering fluid convection and latent heat effects, energy conservation can be expressed as:(1)��∂�∂�=�∇2�+�̇where � is density, � specific heat, � temperature, � time, � thermal conductivity, and �̇ a volumetric heat source. By assuming a semi-infinite domain, Eq. 1 can be analytically solved. The solution for temperature at a given time (t) using a volumetric Gaussian heat source is presented as:(2)��,�,�,�−�0=33�����32∫0�1������exp−3�′�′2��+�′�′2��+�′�′2����′(3)and��=12��−�′+��2for�=�,�,�(4)and�′�′=�−���′Where � is the vector �,�,� and �� is the location of the heat source.
The numerical integration scheme used is an adaptive Gaussian quadrature method based on the following nondimensionalization:(5)�=��xy2�,�′=��xy2�′,�=��xy,�=��xy,�=��xy,�=���xy
A more detailed explanation of the mathematics can be found in reference [43].
The main source of the thermal cycling present within a powder-bed fusion process is the fusion of subsequent layers. Therefore, regions near the top of a build are expected to undergo fewer thermal cycles than those closer to the bottom. For this purpose, data from the single scan’s thermal influence on multiple layers was spliced to represent the thermal cycles experienced at a single location caused by multiple subsequent layers being fused.
The cross-sectional area simulated by this model was kept constant at 1 × 1 mm2, and the depth was dependent on the build location modelled with MatCalc. For a build location 2 mm from the bottom, the maximum number of layers to simulate is 460. Fig. 1a shows a stitched overview OM image of the entire build indicating the region where this thermal cycle is simulated and tracked. To increase similarity with the conditions of the physical build, each thermal history was constructed from the results of two simulations generated with different versions of a random scan path. The parameters used for these thermal simulations can be found in Table 1. It should be noted that the main purpose of the thermal profile modelling was to demonstrate how the conditions at different locations of the build change relative to each other. Accurately predicting the absolute temperature during the build would require validation via a temperature sensor measurement during the build process which is beyond the scope of the study. Nonetheless, to establish the viability of the heat source as a suitable approximation for this study, an additional sensitivity analysis was conducted. This analysis focused on the influence of energy input on γ′ precipitation behavior, the central aim of this paper. This was achieved by employing varying beam absorption energies (0.76, 0.82 – the values utilized in the simulation, and 0.9). The direct impact of beam absorption efficiency on energy input into the material was investigated. Specifically, the initial 20 layers of the build were simulated and subsequently compared to experimental data derived from SEM. While phase fractions were found to be consistent across all conditions, disparities emerged in the mean size of γ′ precipitates. An absorption efficiency of 0.76 yielded a mean size of approximately 70 nm. Conversely, absorption efficiencies of 0.82 and 0.9 exhibited remarkably similar mean sizes of around 130 nm, aligning closely with the outcomes of the experiments.
The numerical analyses of the evolution of precipitates was performed using MatCalc version 6.04 (rel 0.011). The thermodynamic (‘mc_ni.tdb’, version 2.034) and diffusion (‘mc_ni.ddb’, version 2.007) databases were used. MatCalc’s basic principles are elaborated as follows:
The nucleation kinetics of precipitates are computed using a computational technique based on a classical nucleation theory[54] that has been modified for systems with multiple components [42], [55]. Accordingly, the transient nucleation rate (�), which expresses the rate at which nuclei are formed per unit volume and time, is calculated as:(6)�=�0��*∙�xp−�*�∙�∙exp−��where �0 denotes the number of active nucleation sites, �* the rate of atomic attachment, � the Boltzmann constant, � the temperature, �* the critical energy for nucleus formation, τ the incubation time, and t the time. � (Zeldovich factor) takes into consideration that thermal excitation destabilizes the nucleus as opposed to its inactive state [54]. Z is defined as follows:(7)�=−12�kT∂2∆�∂�2�*12where ∆� is the overall change in free energy due to the formation of a nucleus and n is the nucleus’ number of atoms. ∆�’s derivative is evaluated at n* (critical nucleus size). �* accounts for the long-range diffusion of atoms required for nucleation, provided that the matrix’ and precipitates’ composition differ. Svoboda et al. [42] developed an appropriate multi-component equation for �*, which is given by:(8)�*=4��*2�4�∑�=1��ki−�0�2�0��0�−1where �* denotes the critical radius for nucleation, � represents atomic distance, and � is the molar volume. �ki and �0� represent the concentration of elements in the precipitate and matrix, respectively. The parameter �0� denotes the rate of diffusion of the ith element within the matrix. The expression for the incubation time � is expressed as [54]:(9)�=12�*�2
and �*, which represents the critical energy for nucleation:(10)�*=16�3�3∆�vol2where � is the interfacial energy, and ∆Gvol the change in the volume free energy. The critical nucleus’ composition is similar to the γ′ phase’s equilibrium composition at the same temperature. � is computed based on the precipitate and matrix compositions, using a generalized nearest neighbor broken bond model, with the assumption of interfaces being planar, sharp, and coherent [56], [57], [58].
In Eq. 7, it is worth noting that �* represents the fundamental variable in the nucleation theory. It contains �3/∆�vol2 and is in the exponent of the nucleation rate. Therefore, even small variations in γ and/or ∆�vol can result in notable changes in �, especially if �* is in the order of �∙�. This is demonstrated in [38] for UDIMET 720 Li during continuous cooling, where these quantities change steadily during precipitation due to their dependence on matrix’ and precipitate’s temperature and composition. In the current work, these changes will be even more significant as the system is exposed to multiple cycles of rapid cooling and heating.
Once nucleated, the growth of a precipitate is assessed using the radius and composition evolution equations developed by Svoboda et al. [42] with a mean-field method that employs the thermodynamic extremal principle. The expression for the total Gibbs free energy of a thermodynamic system G, which consists of n components and m precipitates, is given as follows:(11)�=∑���0��0�+∑�=1�4���33��+∑�=1��ki�ki+∑�=1�4���2��.
The chemical potential of component � in the matrix is denoted as �0�(�=1,…,�), while the chemical potential of component � in the precipitate is represented by �ki(�=1,…,�,�=1,…,�). These chemical potentials are defined as functions of the concentrations �ki(�=1,…,�,�=1,…,�). The interface energy density is denoted as �, and �� incorporates the effects of elastic energy and plastic work resulting from the volume change of each precipitate.
Eq. (12) establishes that the total free energy of the system in its current state relies on the independent state variables: the sizes (radii) of the precipitates �� and the concentrations of each component �ki. The remaining variables can be determined by applying the law of mass conservation to each component �. This can be represented by the equation:(12)��=�0�+∑�=1�4���33�ki,
Furthermore, the global mass conservation can be expressed by equation:(13)�=∑�=1���When a thermodynamic system transitions to a more stable state, the energy difference between the initial and final stages is dissipated. This model considers three distinct forms of dissipation effects [42]. These include dissipations caused by the movement of interfaces, diffusion within the precipitate and diffusion within the matrix.
Consequently, �̇� (growth rate) and �̇ki (chemical composition’s rate of change) of the precipitate with index � are derived from the linear system of equation system:(14)�ij��=��where �� symbolizes the rates �̇� and �̇ki [42]. Index i contains variables for precipitate radius, chemical composition, and stoichiometric boundary conditions suggested by the precipitate’s crystal structure. Eq. (10) is computed separately for every precipitate �. For a more detailed description of the formulae for the coefficients �ij and �� employed in this work please refer to [59].
The MatCalc software was used to perform the numerical time integration of �̇� and �̇ki of precipitates based on the classical numerical method by Kampmann and Wagner [60]. Detailed information on this method can be found in [61]. Using this computational method, calculations for E-PBF thermal cycles (cyclic heating and cooling) were computed and compared to experimental data. The simulation took approximately 2–4 hrs to complete on a standard laptop.
3. Results
3.1. Microstructure
Fig. 1 displays a stitched overview image and selected SEM micrographs of various γ′ morphologies and carbides after observations of the X-Z surface of the build from the top to 2 mm above the bottom. Fig. 2 depicts a graph that charts the average size and phase fraction of the primary γ′, as it changes with distance from the top to the bottom of the build. The SEM micrographs show widespread primary γ′ precipitation throughout the entire build, with the size increasing in the top to bottom direction. Particularly, at the topmost height, representing the 460th layer (Z = 22.95 mm), as seen in Fig. 1b, the average size of γ′ is 110 ± 4 nm, exhibiting spherical shapes. This is representative of the microstructure after it solidifies and cools to room temperature, without experiencing additional thermal cycles. The γ′ size slightly increases to 147 ± 6 nm below this layer and remains constant until 0.4 mm (∼453rd layer) from the top. At this position, the microstructure still closely resembles that of the 460th layer. After the 453rd layer, the γ′ size grows rapidly to ∼503 ± 19 nm until reaching the 437th layer (1.2 mm from top). The γ′ particles here have a cuboidal shape, and a small fraction is coarser than 600 nm. γ′ continue to grow steadily from this position to the bottom (23 mm from the top). A small fraction of γ′ is > 800 nm.
Besides primary γ′, secondary γ′ with sizes ranging from 5 to 50 nm were also found. These secondary γ′ precipitates, as seen in Fig. 1f, were present only in the bottom and middle regions. A detailed analysis of the multimodal size distribution of γ′ can be found in [16]. There is no significant variation in the phase fraction of the γ′ along the build. The phase fraction is ∼ 52%, as displayed in Fig. 2. It is worth mentioning that the total phase fraction of γ′ was estimated based on the primary γ′ phase fraction because of the small size of secondary γ′. Spherical MC carbides with sizes ranging from 50 to 400 nm and a phase fraction of 0.8% were also observed throughout the build. The carbides are the light grey precipitates in Fig. 1g. The light grey shade of carbides in the SEM images is due to their composition and crystal structure [52]. These carbides are not visible in Fig. 1b-e because they were dissolved during electro-etching carried out after electropolishing. In Fig. 1g, however, the sample was examined directly after electropolishing, without electro-etching.
Table 2 shows the nominal and measured composition of γ′ precipitates throughout the build by atom probe microscopy as determined in our previous study [17]. No build height-dependent composition difference was observed in either of the γ′ precipitate populations. However, there was a slight disparity between the composition of primary and secondary γ′. Among the main γ′ forming elements, the primary γ′ has a high Ti concentration while secondary γ′ has a high Al concentration. A detailed description of the atom distribution maps and the proxigrams of the constituent elements of γ′ throughout the build can be found in [17].
Table 2. Bulk IN738 composition determined using inductively coupled plasma atomic emission spectroscopy (ICP-AES). Compositions of γ, primary γ′, and secondary γ′ at various locations in the build measured by APT. This information is reproduced from data in Ref. [17] with permission.
at%
Ni
Cr
Co
Al
Mo
W
Ti
Nb
C
B
Zr
Ta
Others
Bulk
59.12
17.47
8.48
7.00
1.01
0.81
3.96
0.49
0.47
0.05
0.09
0.56
0.46
γ matrix
Top
50.48
32.91
11.59
1.94
1.39
0.82
0.44
0.8
0.03
0.03
0.02
–
0.24
Mid
50.37
32.61
11.93
1.79
1.54
0.89
0.44
0.1
0.03
0.02
0.02
0.01
0.23
Bot
48.10
34.57
12.08
2.14
1.43
0.88
0.48
0.08
0.04
0.03
0.01
–
0.12
Primary γ′
Top
72.17
2.51
3.44
12.71
0.25
0.39
7.78
0.56
–
0.03
0.02
0.05
0.08
Mid
71.60
2.57
3.28
13.55
0.42
0.68
7.04
0.73
–
0.01
0.03
0.04
0.04
Bot
72.34
2.47
3.86
12.50
0.26
0.44
7.46
0.50
0.05
0.02
0.02
0.03
0.04
Secondary γ′
Mid
70.42
4.20
3.23
14.19
0.63
1.03
5.34
0.79
0.03
–
0.04
0.04
0.05
Bot
69.91
4.06
3.68
14.32
0.81
1.04
5.22
0.65
0.05
–
0.10
0.02
0.11
3.2. Hardness
Fig. 3a shows the Vickers hardness mapping performed along the entire X-Z surface, while Fig. 3b shows the plot of average hardness at different build heights. This hardness distribution is consistent with the γ′ precipitate size gradient across the build direction in Fig. 1, Fig. 2. The maximum hardness of ∼530 HV1 is found at ∼0.5 mm away from the top surface (Z = 22.5), where γ′ particles exhibit the smallest observed size in Fig. 2b. Further down the build (∼ 2 mm from the top), the hardness drops to the 440–490 HV1 range. This represents the region where γ′ begins to coarsen. The hardness drops further to 380–430 HV1 at the bottom of the build.
3.3. Modeling of the microstructural evolution during E-PBF
3.3.1. Thermal profile modeling
Fig. 4 shows the simulated thermal profile of the E-PBF build at a location of 23 mm from the top of the build, using a semi-analytical heat conduction model. This profile consists of the time taken to deposit 460 layers until final cooling, as shown in Fig. 4a. Fig. 4b-d show the magnified regions of Fig. 4a and reveal the first 20 layers from the top, a single layer (first layer from the top), and the time taken for the build to cool after the last layer deposition, respectively.
The peak temperatures experienced by previous layers decrease progressively as the number of layers increases but never fall below the build preheat temperature (1000 °C). Our simulated thermal cycle may not completely capture the complexity of the actual thermal cycle utilized in the E-PBF build. For instance, the top layer (Fig. 4c), also representing the first deposit’s thermal profile without additional cycles (from powder heating, melting, to solidification), recorded the highest peak temperature of 1390 °C. Although this temperature is above the melting range of the alloy (1230–1360 °C) [62], we believe a much higher temperature was produced by the electron beam to melt the powder. Nevertheless, the solidification temperature and dynamics are outside the scope of this study as our focus is on the solid-state phase transformations during deposition. It takes ∼25 s for each layer to be deposited and cooled to the build temperature. The interlayer dwell time is 125 s. The time taken for the build to cool to room temperature (RT) after final layer deposition is ∼4.7 hrs (17,000 s).
3.3.2. MatCalc simulation
During the MatCalc simulation, the matrix phase is defined as γ. γ′, and MC carbide are included as possible precipitates. The domain of these precipitates is set to be the matrix (γ), and nucleation is assumed to be homogenous. In homogeneous nucleation, all atoms of the unit volume are assumed to be potential nucleation sites. Table 3 shows the computational parameters used in the simulation. All other parameters were set at default values as recommended in the version 6.04.0011 of MatCalc. The values for the interfacial energies are automatically calculated according to the generalized nearest neighbor broken bond model and is one of the most outstanding features in MatCalc [56], [57], [58]. It should be noted that the elastic misfit strain was not included in the calculation. The output of MatCalc includes phase fraction, size, nucleation rate, and composition of the precipitates. The phase fraction in MatCalc is the volume fraction. Although the experimental phase fraction is the measured area fraction, it is relatively similar to the volume fraction. This is because of the generally larger precipitate size and similar morphology at the various locations along the build [63]. A reliable phase fraction comparison between experiment and simulation can therefore be made.
Table 3. Computational parameters used in the simulation.
γ′ = 0.080–0.140 J/m2 and MC carbide = 0.410–0.430 J/m2
3.3.2.1. Precipitate phase fraction
Fig. 5a shows the simulated phase fraction of γ′ and MC carbide during thermal cycling. Fig. 5b is a magnified view of 5a showing the simulated phase fraction at the center points of the top 70 layers, whereas Fig. 5c corresponds to the first two layers from the top. As mentioned earlier, the top layer (460th layer) represents the microstructure after solidification. The microstructure of the layers below is determined by the number of thermal cycles, which increases with distance to the top. For example, layers 459, 458, 457, up to layer 1 (region of interest) experience 1, 2, 3 and 459 thermal cycles, respectively. In the top layer in Fig. 5c, the volume fraction of γ′ and carbides increases with temperature. For γ′, it decreases to zero when the temperature is above the solvus temperature after a few seconds. Carbides, however, remain constant in their volume fraction reaching equilibrium (phase fraction ∼ 0.9%) in a short time. The topmost layer can be compared to the first deposit, and the peak in temperature symbolizes the stage where the electron beam heats the powder until melting. This means γ′ and carbide precipitation might have started in the powder particles during heating from the build temperature and electron beam until the onset of melting, where γ′ dissolves, but carbides remain stable [28].
During cooling after deposition, γ′ reprecipitates at a temperature of 1085 °C, which is below its solvus temperature. As cooling progresses, the phase fraction increases steadily to ∼27% and remains constant at 1000 °C (elevated build temperature). The calculated equilibrium fraction of phases by MatCalc is used to show the complex precipitation characteristics in this alloy. Fig. 6 shows that MC carbides form during solidification at 1320 °C, followed by γ′, which precipitate when the solidified layer cools to 1140 °C. This indicates that all deposited layers might contain a negligible amount of these precipitates before subsequent layer deposition, while being at the 1000 °C build temperature or during cooling to RT. The phase diagram also shows that the equilibrium fraction of the γ′ increases as temperature decreases. For instance, at 1000, 900, and 800 °C, the phase fractions are ∼30%, 38%, and 42%, respectively.
Deposition of subsequent layers causes previous layers to undergo phase transformations as they are exposed to several thermal cycles with different peak temperatures. In Fig. 5c, as the subsequent layer is being deposited, γ′ in the previous layer (459th layer) begins to dissolve as the temperature crosses the solvus temperature. This is witnessed by the reduction of the γ′ phase fraction. This graph also shows how this phase dissolves during heating. However, the phase fraction of MC carbide remains stable at high temperatures and no dissolution is seen during thermal cycling. Upon cooling, the γ′ that was dissolved during heating reprecipitates with a surge in the phase fraction until 1000 °C, after which it remains constant. This microstructure is similar to the solidification microstructure (layer 460), with a similar γ′ phase fraction (∼27%).
The complete dissolution and reprecipitation of γ′ continue for several cycles until the 50th layer from the top (layer 411), where the phase fraction does not reach zero during heating to the peak temperature (see Fig. 5d). This indicates the ‘partial’ dissolution of γ′, which continues progressively with additional layers. It should be noted that the peak temperatures for layers that underwent complete dissolution were much higher (1170–1300 °C) than the γ′ solvus.
The dissolution and reprecipitation of γ′ during thermal cycling are further confirmed in Fig. 7, which summarizes the nucleation rate, phase fraction, and concentration of major elements that form γ′ in the matrix. Fig. 7b magnifies a single layer (3rd layer from top) within the full dissolution region in Fig. 7a to help identify the nucleation and growth mechanisms. From Fig. 7b, γ′ nucleation begins during cooling whereby the nucleation rate increases to reach a maximum value of approximately 1 × 1020 m−3s−1. This fast kinetics implies that some rearrangement of atoms is required for γ′ precipitates to form in the matrix [65], [66]. The matrix at this stage is in a non-equilibrium condition. Its composition is similar to the nominal composition and remains unchanged. The phase fraction remains insignificant at this stage although nucleation has started. The nucleation rate starts declining upon reaching the peak value. Simultaneously, diffusion-controlled growth of existing nuclei occurs, depleting the matrix of γ′ forming elements (Al and Ti). Thus, from (7), (11), ∆�vol continuously decreases until nucleation ceases. The growth of nuclei is witnessed by the increase in phase fraction until a constant level is reached at 27% upon cooling to and holding at build temperature. This nucleation event is repeated several times.
At the onset of partial dissolution, the nucleation rate jumps to 1 × 1021 m−3s−1, and then reduces sharply at the middle stage of partial dissolution. The nucleation rate reaches 0 at a later stage. Supplementary Fig. S1 shows a magnified view of the nucleation rate, phase fraction, and thermal profile, underpinning this trend. The jump in nucleation rate at the onset is followed by a progressive reduction in the solute content of the matrix. The peak temperatures (∼1130–1160 °C) are lower than those in complete dissolution regions but still above or close to the γ′ solvus. The maximum phase fraction (∼27%) is similar to that of the complete dissolution regions. At the middle stage, the reduction in nucleation rate is accompanied by a sharp drop in the matrix composition. The γ′ fraction drops to ∼24%, where the peak temperatures of the layers are just below or at γ′ solvus. The phase fraction then increases progressively through the later stage of partial dissolution to ∼30% towards the end of thermal cycling. The matrix solute content continues to drop although no nucleation event is seen. The peak temperatures are then far below the γ′ solvus. It should be noted that the matrix concentration after complete dissolution remains constant. Upon cooling to RT after final layer deposition, the nucleation rate increases again, indicating new nucleation events. The phase fraction reaches ∼40%, with a further depletion of the matrix in major γ′ forming elements.
3.3.2.2. γ′ size distribution
Fig. 8 shows histograms of the γ′ precipitate size distributions (PSD) along the build height during deposition. These PSDs are predicted at the end of each layer of interest just before final cooling to room temperature, to separate the role of thermal cycles from final cooling on the evolution of γ′. The PSD for the top layer (layer 460) is shown in Fig. 8a (last solidified region with solidification microstructure). The γ′ size ranges from 120 to 230 nm and is similar to the 44 layers below (2.2 mm from the top).
Further down the build, γ′ begins to coarsen after layer 417 (44th layer from top). Fig. 8c shows the PSD after the 44th layer, where the γ′ size exhibits two peaks at ∼120–230 and ∼300 nm, with most of the population being in the former range. This is the onset of partial dissolution where simultaneously with the reprecipitation and growth of fresh γ′, the undissolved γ′ grows rapidly through diffusive transport of atoms to the precipitates. This is shown in Fig. 8c, where the precipitate class sizes between 250 and 350 represent the growth of undissolved γ′. Although this continues in the 416th layer, the phase fractions plot indicates that the onset of partial dissolution begins after the 411th layer. This implies that partial dissolution started early, but the fraction of undissolved γ′ was too low to impact the phase fraction. The reprecipitated γ′ are mostly in the 100–220 nm class range and similar to those observed during full dissolution.
As the number of layers increases, coarsening intensifies with continued growth of more undissolved γ′, and reprecipitation and growth of partially dissolved ones. Fig. 8d, e, and f show this sequence. Further down the build, coarsening progresses rapidly, as shown in Figs. 8d, 8e, and 8f. The γ′ size ranges from 120 to 1100 nm, with the peaks at 160, 180, and 220 nm in Figs. 8d, 8e, and 8f, respectively. Coarsening continues until nucleation ends during dissolution, where only the already formed γ′ precipitates continue to grow during further thermal cycling. The γ′ size at this point is much larger, as observed in layers 361 and 261, and continues to increase steadily towards the bottom (layer 1). Two populations in the ranges of ∼380–700 and ∼750–1100 nm, respectively, can be seen. The steady growth of γ′ towards the bottom is confirmed by the gradual decrease in the concentration of solute elements in the matrix (Fig. 7a). It should be noted that for each layer, the γ′ class with the largest size originates from continuous growth of the earliest set of the undissolved precipitates.
Fig. 9, Fig. 10 and supplementary Figs. S2 and S3 show the γ′ size evolution during heating and cooling of a single layer in the full dissolution region, and early, middle stages, and later stages of partial dissolution, respectively. In all, the size of γ′ reduces during layer heating. Depending on the peak temperature of the layer which varies with build height, γ′ are either fully or partially dissolved as mentioned earlier. Upon cooling, the dissolved γ′ reprecipitate.
In Fig. 9, those layers that underwent complete dissolution (top layers) were held above γ′ solvus temperature for longer. In Fig. 10, layers at the early stage of partial dissolution spend less time in the γ′ solvus temperature region during heating, leading to incomplete dissolution. In such conditions, smaller precipitates are fully dissolved while larger ones shrink [67]. Layers in the middle stages of partial dissolution have peak temperatures just below or at γ′ solvus, not sufficient to achieve significant γ′ dissolution. As seen in supplementary Fig. S2, only a few smaller γ′ are dissolved back into the matrix during heating, i.e., growth of precipitates is more significant than dissolution. This explains the sharp decrease in concentration of Al and Ti in the matrix in this layer.
The previous sections indicate various phenomena such as an increase in phase fraction, further depletion of matrix composition, and new nucleation bursts during cooling. Analysis of the PSD after the final cooling of the build to room temperature allows a direct comparison to post-printing microstructural characterization. Fig. 11 shows the γ′ size distribution of layer 1 (460th layer from the top) after final cooling to room temperature. Precipitation of secondary γ′ is observed, leading to the multimodal size distribution of secondary and primary γ′. The secondary γ′ size falls within the 10–80 nm range. As expected, a further growth of the existing primary γ′ is also observed during cooling.
3.3.2.3. γ′ chemistry after deposition
Fig. 12 shows the concentration of the major elements that form γ′ (Al, Ti, and Ni) in the primary and secondary γ′ at the bottom of the build, as calculated by MatCalc. The secondary γ′ has a higher Al content (13.5–14.5 at% Al), compared to 13 at% Al in the primary γ′. Additionally, within the secondary γ′, the smallest particles (∼10 nm) have higher Al contents than larger ones (∼70 nm). In contrast, for the primary γ′, there is no significant variation in the Al content as a function of their size. The Ni concentration in secondary γ′ (71.1–72 at%) is also higher in comparison to the primary γ′ (70 at%). The smallest secondary γ′ (∼10 nm) have higher Ni contents than larger ones (∼70 nm), whereas there is no substantial change in the Ni content of primary γ′, based on their size. As expected, Ti shows an opposite size-dependent variation. It ranges from ∼ 7.7–8.7 at% Ti in secondary γ′ to ∼9.2 at% in primary γ′. Similarly, within the secondary γ′, the smallest (∼10 nm) have lower Al contents than the larger ones (∼70 nm). No significant variation is observed for Ti content in primary γ′.
4. Discussion
A combined modelling method is utilized to study the microstructural evolution during E-PBF of IN738. The presented results are discussed by examining the precipitation and dissolution mechanism of γ′ during thermal cycling. This is followed by a discussion on the phase fraction and size evolution of γ′ during thermal cycling and after final cooling. A brief discussion on carbide morphology is also made. Finally, a comparison is made between the simulation and experimental results to assess their agreement.
4.1. γ′ morphology as a function of build height
4.1.1. Nucleation of γ′
The fast precipitation kinetics of the γ′ phase enables formation of γ′ upon quenching from higher temperatures (above solvus) during thermal cycling [66]. In Fig. 7b, for a single layer in the full dissolution region, during cooling, the initial increase in nucleation rate signifies the first formation of nuclei. The slight increase in nucleation rate during partial dissolution, despite a decrease in the concentration of γ′ forming elements, may be explained by the nucleation kinetics. During partial dissolution and as the precipitates shrink, it is assumed that the regions at the vicinity of partially dissolved precipitates are enriched in γ′ forming elements [68], [69]. This differs from the full dissolution region, in which case the chemical composition is evenly distributed in the matrix. Several authors have attributed the solute supersaturation of the matrix around primary γ′ to partial dissolution during isothermal ageing [69], [70], [71], [72]. The enhanced supersaturation in the regions close to the precipitates results in a much higher driving force for nucleation, leading to a higher nucleation rate upon cooling. This phenomenon can be closely related to the several nucleation bursts upon continuous cooling of Ni-based superalloys, where second nucleation bursts exhibit higher nucleation rates [38], [68], [73], [74].
At middle stages of partial dissolution, the reduction in the nucleation rate indicates that the existing composition and low supersaturation did not trigger nucleation as the matrix was closer to the equilibrium state. The end of a nucleation burst means that the supersaturation of Al and Ti has reached a low level, incapable of providing sufficient driving force during cooling to or holding at 1000 °C for further nucleation [73]. Earlier studies on Ni-based superalloys have reported the same phenomenon during ageing or continuous cooling from the solvus temperature to RT [38], [73], [74].
4.1.2. Dissolution of γ′ during thermal cycling
γ′ dissolution kinetics during heating are fast when compared to nucleation due to exponential increase in phase transformation and diffusion activities with temperature [65]. As shown in Fig. 9, Fig. 10, and supplementary Figs. S2 and S3, the reduction in γ′ phase fraction and size during heating indicates γ′ dissolution. This is also revealed in Fig. 5 where phase fraction decreases upon heating. The extent of γ′ dissolution mostly depends on the temperature, time spent above γ′ solvus, and precipitate size[75], [76], [77]. Smaller γ′ precipitates are first to be dissolved [67], [77], [78]. This is mainly because more solute elements need to be transported away from large γ′ precipitates than from smaller ones [79]. Also, a high temperature above γ′ solvus temperature leads to a faster dissolution rate[80]. The equilibrium solvus temperature of γ′ in IN738 in our MatCalc simulation (Fig. 6) and as reported by Ojo et al. [47] is 1140 °C and 1130–1180 °C, respectively. This means the peak temperature experienced by previous layers decreases progressively from γ′ supersolvus to subsolvus, near-solvus, and far from solvus as the number of subsequent layers increases. Based on the above, it can be inferred that the degree of dissolution of γ′ contributes to the gradient in precipitate distribution.
Although the peak temperatures during later stages of partial dissolution are much lower than the equilibrium γ′ solvus, γ′ dissolution still occurs but at a significantly lower rate (supplementary Fig. S3). Wahlmann et al. [28] also reported a similar case where they observed the rapid dissolution of γ′ in CMSX-4 during fast heating and cooling cycles at temperatures below the γ′ solvus. They attributed this to the γ′ phase transformation process taking place in conditions far from the equilibrium. While the same reasoning may be valid for our study, we further believe that the greater surface area to volume ratio of the small γ′ precipitates contributed to this. This ratio means a larger area is available for solute atoms to diffuse into the matrix even at temperatures much below the solvus [81].
4.2. γ′ phase fraction and size evolution
4.2.1. During thermal cycling
In the first layer, the steep increase in γ′ phase fraction during heating (Fig. 5), which also represents γ′ precipitation in the powder before melting, has qualitatively been validated in [28]. The maximum phase fraction of 27% during the first few layers of thermal cycling indicates that IN738 theoretically could reach the equilibrium state (∼30%), but the short interlayer time at the build temperature counteracts this. The drop in phase fraction at middle stages of partial dissolution is due to the low number of γ′ nucleation sites [73]. It has been reported that a reduction of γ′ nucleation sites leads to a delay in obtaining the final volume fraction as more time is required for γ′ precipitates to grow and reach equilibrium [82]. This explains why even upon holding for 150 s before subsequent layer deposition, the phase fraction does not increase to those values that were observed in the previous full γ′ dissolution regions. Towards the end of deposition, the increase in phase fraction to the equilibrium value of 30% is as a result of the longer holding at build temperature or close to it [83].
During thermal cycling, γ′ particles begin to grow immediately after they first precipitate upon cooling. This is reflected in the rapid increase in phase fraction and size during cooling in Fig. 5 and supplementary Fig. S2, respectively. The rapid growth is due to the fast diffusion of solute elements at high temperatures [84]. The similar size of γ′ for the first 44 layers from the top can be attributed to the fact that all layers underwent complete dissolution and hence, experienced the same nucleation event and growth during deposition. This corresponds with the findings by Balikci et al. [85], who reported that the degree of γ′ precipitation in IN738LC does not change when a solution heat treatment is conducted above a certain critical temperature.
The increase in coarsening rate (Fig. 8) during thermal cycling can first be ascribed to the high peak temperature of the layers [86]. The coarsening rate of γ′ is known to increase rapidly with temperature due to the exponential growth of diffusion activity. Also, the simultaneous dissolution with coarsening could be another reason for the high coarsening rate, as γ′ coarsening is a diffusion-driven process where large particles grow by consuming smaller ones [78], [84], [86], [87]. The steady growth of γ′ towards the bottom of the build is due to the much lower layer peak temperature, which is almost close to the build temperature, and reduced dissolution activity, as is seen in the much lower solute concentration in γ′ compared to those in the full and partial dissolution regions.
4.2.2. During cooling
The much higher phase fraction of ∼40% upon cooling signifies the tendency of γ′ to reach equilibrium at lower temperatures (Fig. 4). This is due to the precipitation of secondary γ′ and a further increase in the size of existing primary γ′, which leads to a multimodal size distribution of γ′ after cooling [38], [73], [88], [89], [90]. The reason for secondary γ′ formation during cooling is as follows: As cooling progresses, it becomes increasingly challenging to redistribute solute elements in the matrix owing to their lower mobility [38], [73]. A higher supersaturation level in regions away from or free of the existing γ′ precipitates is achieved, making them suitable sites for additional nucleation bursts. More cooling leads to the growth of these secondary γ′ precipitates, but as the temperature and in turn, the solute diffusivity is low, growth remains slow.
4.3. Carbides
MC carbides in IN738 are known to have a significant impact on the high-temperature strength. They can also act as effective hardening particles and improve the creep resistance [91]. Precipitation of MC carbides in IN738 and several other superalloys is known to occur during solidification or thermal treatments (e.g., hot isostatic pressing) [92]. In our case, this means that the MC carbides within the E-PBF build formed because of the thermal exposure from the E-PBF thermal cycle in addition to initial solidification. Our simulation confirms this as MC carbides appear during layer heating (Fig. 5). The constant and stable phase fraction of MC carbides during thermal cycling can be attributed to their high melting point (∼1360 °C) and the short holding time at peak temperatures [75], [93], [94]. The solvus temperature for most MC carbides exceeds most of the peak temperatures observed in our simulation, and carbide dissolution kinetics at temperatures above the solvus are known to be comparably slow [95]. The stable phase fraction and random distribution of MC carbides signifies the slight influence on the gradient in hardness.
4.4. Comparison of simulations and experiments
4.4.1. Precipitate phase fraction and morphology as a function of build height
A qualitative agreement is observed for the phase fraction of carbides, i.e. ∼0.8% in the experiment and ∼0.9% in the simulation. The phase fraction of γ′ differs, with the experiment reporting a value of ∼51% and the simulation, 40%. Despite this, the size distribution of primary γ′ along the build shows remarkable consistency between experimental and computational analyses. It is worth noting that the primary γ′ morphology in the experimental analysis is observed in the as-fabricated state, whereas the simulation (Fig. 8) captures it during deposition process. The primary γ′ size in the experiment is expected to experience additional growth during the cooling phase. Regardless, both show similar trends in primary γ′ size increments from the top to the bottom of the build. The larger primary γ’ size in the simulation versus the experiment can be attributed to the fact that experimental and simulation results are based on 2D and 3D data, respectively. The absence of stereological considerations [96] in our analysis could have led to an underestimation of the precipitate sizes from SEM measurements. The early starts of coarsening (8th layer) in the experiment compared to the simulation (45th layer) can be attributed to a higher actual γ′ solvus temperature than considered in our simulation [47]. The solvus temperature of γ′ in a Ni-based superalloy is mainly determined by the detailed composition. A high amount of Cr and Co are known to reduce the solvus temperature, whereas Ta and Mo will increase it [97], [98], [99]. The elemental composition from our experimental work was used for the simulation except for Ta. It should be noted that Ta is not included in the thermodynamic database in MatCalc used, and this may have reduced the solvus temperature. This could also explain the relatively higher γ′ phase fraction in the experiment than in simulation, as a higher γ′ solvus temperature will cause more γ′ to precipitate and grow early during cooling [99], [100].
Another possible cause of this deviation can be attributed to the extent of γ′ dissolution, which is mainly determined by the peak temperature. It can be speculated that individual peak temperatures at different layers in the simulation may have been over-predicted. However, one needs to consider that the true thermal profile is likely more complicated in the actual E-PBF process [101]. For example, the current model assumes that the thermophysical properties of the material are temperature-independent, which is not realistic. Many materials, including IN738, exhibit temperature-dependent properties such as thermal conductivity, specific heat capacity, and density [102]. This means that heat transfer simulations may underestimate or overestimate the temperature gradients and cooling rates within the powder bed and the solidified part. Additionally, the model does not account for the reduced thermal diffusivity through unmelted powder, where gas separating the powder acts as insulation, impeding the heat flow [1]. In E-PBF, the unmelted powder regions with trapped gas have lower thermal diffusivity compared to the fully melted regions, leading to localized temperature variations, and altered solidification behavior. These limitations can impact the predictions, particularly in relation to the carbide dissolution, as the peak temperatures may be underestimated.
While acknowledging these limitations, it is worth emphasizing that achieving a detailed and accurate representation of each layer’s heat source would impose tough computational challenges. Given the substantial layer count in E-PBF, our decision to employ a semi-analytical approximation strikes a balance between computational feasibility and the capture of essential trends in thermal profiles across diverse build layers. In future work, a dual-calibration strategy is proposed to further reduce simulation-experiment disparities. By refining temperature-independent thermophysical property approximations and absorptivity in the heat source model, and by optimizing interfacial energy descriptions in the kinetic model, the predictive precision could be enhanced. Further refining the simulation controls, such as adjusting the precipitate class size may enhance quantitative comparisons between modeling outcomes and experimental data in future work.
4.4.2. Multimodal size distribution of γ′ and concentration
Another interesting feature that sees qualitative agreement between the simulation and the experiment is the multimodal size distribution of γ′. The formation of secondary γ′ particles in the experiment and most E-PBF Ni-based superalloys is suggested to occur at low temperatures, during final cooling to RT [16], [73], [90]. However, so far, this conclusion has been based on findings from various continuous cooling experiments, as the study of the evolution during AM would require an in-situ approach. Our simulation unambiguously confirms this in an AM context by providing evidence for secondary γ′ precipitation during slow cooling to RT. Additionally, it is possible to speculate that the chemical segregation occurring during solidification, due to the preferential partitioning of certain elements between the solid and liquid phases, can contribute to the multimodal size distribution during deposition [51]. This is because chemical segregation can result in variations in the local composition of superalloys, which subsequently affects the nucleation and growth of γ′. Regions with higher concentrations of alloying elements will encourage the formation of larger γ′ particles, while regions with lower concentrations may favor the nucleation of smaller precipitates. However, it is important to acknowledge that the elevated temperature during the E-PBF process will largely homogenize these compositional differences [103], [104].
A good correlation is also shown in the composition of major γ′ forming elements (Al and Ti) in primary and secondary γ′. Both experiment and simulation show an increasing trend for Al content and a decreasing trend for Ti content from primary to secondary γ′. The slight composition differences between primary and secondary γ′ particles are due to the different diffusivity of γ′ stabilizers at different thermal conditions [105], [106]. As the formation of multimodal γ′ particles with different sizes occurs over a broad temperature range, the phase chemistry of γ′ will be highly size dependent. The changes in the chemistry of various γ′ (primary, secondary, and tertiary) have received significant attention since they have a direct influence on the performance [68], [105], [107], [108], [109]. Chen et al. [108], [109], reported a high Al content in the smallest γ′ precipitates compared to the largest, while Ti showed an opposite trend during continuous cooling in a RR1000 Ni-based superalloy. This was attributed to the temperature and cooling rate at which the γ′ precipitates were formed. The smallest precipitates formed last, at the lowest temperature and cooling rate. A comparable observation is evident in the present investigation, where the secondary γ′ forms at a low temperature and cooling rate in comparison to the primary. The temperature dependence of γ′ chemical composition is further evidenced in supplementary Fig. S4, which shows the equilibrium chemical composition of γ′ as a function of temperature.
5. Conclusions
A correlative modelling approach capable of predicting solid-state phase transformations kinetics in metal AM was developed. This approach involves computational simulations with a semi-analytical heat transfer model and the MatCalc thermo-kinetic software. The method was used to predict the phase transformation kinetics and detailed morphology and chemistry of γ′ and MC during E-PBF of IN738 Ni-based superalloy. The main conclusions are:
1.The computational simulations are in qualitative agreement with the experimental observations. This is particularly true for the γ′ size distribution along the build height, the multimodal size distribution of particles, and the phase fraction of MC carbides.
2.The deviations between simulation and experiment in terms of γ′ phase fraction and location in the build are most likely attributed to a higher γ′ solvus temperature during the experiment than in the simulation, which is argued to be related to the absence of Ta in the MatCalc database.
3.The dissolution and precipitation of γ′ occur fast and under non-equilibrium conditions. The level of γ′ dissolution determines the gradient in γ′ size distribution along the build. After thermal cycling, the final cooling to room temperature has further significant impacts on the final γ′ size, morphology, and distribution.
4.A negligible amount of γ′ forms in the first deposited layer before subsequent layer deposition, and a small amount of γ′ may also form in the powder induced by the 1000 °C elevated build temperature before melting.
Our findings confirm the suitability of MatCalc to predict the microstructural evolution at various positions throughout a build in a Ni-based superalloy during E-PBF. It also showcases the suitability of a tool which was originally developed for traditional thermo-mechanical processing of alloys to the new additive manufacturing context. Our simulation capabilities are likely extendable to other alloy systems that undergo solid-state phase transformations implemented in MatCalc (various steels, Ni-based superalloys, and Al-alloys amongst others) as well as other AM processes such as L-DED and L-PBF which have different thermal cycle characteristics. New tools to predict the microstructural evolution and properties during metal AM are important as they provide new insights into the complexities of AM. This will enable control and design of AM microstructures towards advanced materials properties and performances.
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This research was sponsored by the Department of Industry, Innovation, and Science under the auspices of the AUSMURI program – which is a part of the Commonwealth’s Next Generation Technologies Fund. The authors acknowledge the facilities and the scientific and technical assistance at the Electron Microscope Unit (EMU) within the Mark Wainwright Analytical Centre (MWAC) at UNSW Sydney and Microscopy Australia. Nana Adomako is supported by a UNSW Scientia PhD scholarship. Michael Haines’ (UNSW Sydney) contribution to the revised version of the original manuscript is thankfully acknowledged.
[1]T. Debroy, H.L. Wei, J.S. Zuback, T. Mukherjee, J.W. Elmer, J.O. Milewski, A.M. Beese, A. Wilson-heid, A. De, W. ZhangAdditive manufacturing of metallic components – process, structure and propertiesProg. Mater. Sci., 92 (2018), pp. 112-224, 10.1016/j.pmatsci.2017.10.001View PDFView articleView in ScopusGoogle Scholar
[4]N.K. Adomako, J.J. Lewandowski, B.M. Arkhurst, H. Choi, H.J. Chang, J.H. KimMicrostructures and mechanical properties of multi-layered materials composed of Ti-6Al-4V, vanadium, and 17–4PH stainless steel produced by directed energy depositionAddit. Manuf., 59 (2022), Article 103174, 10.1016/j.addma.2022.103174View PDFView articleView in ScopusGoogle Scholar
[5]H. Wang, Z.G. Zhu, H. Chen, A.G. Wang, J.Q. Liu, H.W. Liu, R.K. Zheng, S.M.L. Nai, S. Primig, S.S. Babu, S.P. Ringer, X.Z. LiaoEffect of cyclic rapid thermal loadings on the microstructural evolution of a CrMnFeCoNi high-entropy alloy manufactured by selective laser meltingActa Mater., 196 (2020), pp. 609-625, 10.1016/J.ACTAMAT.2020.07.006View PDFView articleView in ScopusGoogle Scholar
[10]S.S. Al-Bermani, M.L. Blackmore, W. Zhang, I. ToddThe origin of microstructural diversity, texture, and mechanical properties in electron beam melted Ti-6Al-4VMetall. Mater. Trans. A Phys. Metall. Mater. Sci., 41 (2010), pp. 3422-3434, 10.1007/s11661-010-0397-xView article View in ScopusGoogle Scholar
[13]H. Helmer, A. Bauereiß, R.F. Singer, C. KörnerErratum to: ‘Grain structure evolution in Inconel 718 during selective electron beam melting’ (Materials Science & Engineering A (2016) 668 (180–187 (S0921509316305536) (10.1016/j.msea.2016.05.046))Mater. Sci. Eng. A., 676 (2016), p. 546, 10.1016/j.msea.2016.09.016View PDFView articleView in ScopusGoogle Scholar
[16]N. Haghdadi, E. Whitelock, B. Lim, H. Chen, X. Liao, S.S. Babu, S.P. Ringer, S. PrimigMultimodal γ′ precipitation in Inconel-738 Ni-based superalloy during electron-beam powder bed fusion additive manufacturingJ. Mater. Sci., 55 (2020), pp. 13342-13350, 10.1007/s10853-020-04915-wView article View in ScopusGoogle Scholar
[17]B. Lim, H. Chen, Z. Chen, N. Haghdadi, X. Liao, S. Primig, S.S. Babu, A. Breen, S.P. RingerMicrostructure–property gradients in Ni-based superalloy (Inconel 738) additively manufactured via electron beam powder bed fusionAddit. Manuf. (2021), Article 102121, 10.1016/j.addma.2021.102121View PDFView articleView in ScopusGoogle Scholar
[18]P. Karimi, E. Sadeghi, P. Åkerfeldt, J. Ålgårdh, J. AnderssonInfluence of successive thermal cycling on microstructure evolution of EBM-manufactured alloy 718 in track-by-track and layer-by-layer designMater. Des., 160 (2018), pp. 427-441, 10.1016/j.matdes.2018.09.038View PDFView articleView in ScopusGoogle Scholar
[19]E. Chauvet, P. Kontis, E.A. Jägle, B. Gault, D. Raabe, C. Tassin, J.J. Blandin, R. Dendievel, B. Vayre, S. Abed, G. MartinHot cracking mechanism affecting a non-weldable Ni-based superalloy produced by selective electron Beam MeltingActa Mater., 142 (2018), pp. 82-94, 10.1016/j.actamat.2017.09.047View PDFView articleView in ScopusGoogle Scholar
[20]M. Ramsperger, R.F. Singer, C. KörnerMicrostructure of the nickel-base superalloy CMSX-4 fabricated by selective electron beam meltingMetall. Mater. Trans. A Phys. Metall. Mater. Sci., 47 (2016), pp. 1469-1480, 10.1007/s11661-015-3300-y View PDF This article is free to access.View in ScopusGoogle Scholar
[21]B. Zhang, P. Wang, Y. Chew, Y. Wen, M. Zhang, P. Wang, G. Bi, J. WeiMechanical properties and microstructure evolution of selective laser melting Inconel 718 along building direction and sectional dimensionMater. Sci. Eng. A, 794 (2020), Article 139941, 10.1016/j.msea.2020.139941View PDFView articleView in ScopusGoogle Scholar
[22]C. Körner, M. Ramsperger, C. Meid, D. Bürger, P. Wollgramm, M. Bartsch, G. EggelerMicrostructure and mechanical properties of CMSX-4 single crystals prepared by additive manufacturingMetall. Mater. Trans. A Phys. Metall. Mater. Sci., 49 (2018), pp. 3781-3792, 10.1007/s11661-018-4762-5 View PDF This article is free to access.View in ScopusGoogle Scholar
[23]B. Lim, H. Chen, K. Nomoto, Z. Chen, A.I. Saville, S. Vogel, A.J. Clarke, A. Paradowska, M. Reid, S. Primig, X. Liao, S.S. Babu, A.J. Breen, S.P. RingerAdditively manufactured Haynes-282 monoliths containing thin wall struts of varying thicknessesAddit. Manuf., 59 (2022), Article 103120, 10.1016/j.addma.2022.103120View PDFView articleView in ScopusGoogle Scholar
[24]C.L.A. Leung, S. Marussi, R.C. Atwood, M. Towrie, P.J. Withers, P.D. LeeIn situ X-ray imaging of defect and molten pool dynamics in laser additive manufacturingNat. Commun., 9 (2018), pp. 1-9, 10.1038/s41467-018-03734-7 View PDF This article is free to access.View in ScopusGoogle Scholar
[25]C. Zhao, K. Fezzaa, R.W. Cunningham, H. Wen, F. De Carlo, L. Chen, A.D. Rollett, T. SunReal-time monitoring of laser powder bed fusion process using high-speed X-ray imaging and diffractionSci. Rep., 7 (2017), pp. 1-11, 10.1038/s41598-017-03761-2 View PDF This article is free to access.View in ScopusGoogle Scholar
[26]C. Kenel, D. Grolimund, X. Li, E. Panepucci, V.A. Samson, D.F. Sanchez, F. Marone, C. LeinenbachIn situ investigation of phase transformations in Ti-6Al-4V under additive manufacturing conditions combining laser melting and high-speed micro-X-ray diffractionSci. Rep., 7 (2017), pp. 1-10, 10.1038/s41598-017-16760-0 View PDF This article is free to access.Google Scholar
[27]W.L. Bevilaqua, J. Epp, H. Meyer, J. Dong, H. Roelofs, A. da, S. Rocha, A. RegulyRevealing the dynamic transformation of austenite to bainite during uniaxial warm compression through in-situ synchrotron X-ray diffractionMetals, 11 (2021), pp. 1-14, 10.3390/met11030467View article View in ScopusGoogle Scholar
[28]B. Wahlmann, E. Krohmer, C. Breuning, N. Schell, P. Staron, E. Uhlmann, C. KörnerIn situ observation of γ′ phase transformation dynamics during selective laser melting of CMSX-4Adv. Eng. Mater., 23 (2021), 10.1002/adem.202100112 View PDF This article is free to access.Google Scholar
[35]M.J. Anderson, J. Benson, J.W. Brooks, B. Saunders, H.C. BasoaltoPredicting precipitation kinetics during the annealing of additive manufactured Inconel 625 componentsIntegr. Mater. Manuf. Innov., 8 (2019), pp. 154-166, 10.1007/S40192-019-00134-7/FIGURES/11 View PDFThis article is free to access.View in ScopusGoogle Scholar
[36]H.C. Basoalto, C. Panwisawas, Y. Sovani, M.J. Anderson, R.P. Turner, B. Saunders, J.W. BrooksA computational study on the three-dimensional printability of precipitate-strengthened nickel-based superalloysProc. R. Soc. A, 474 (2018), 10.1098/RSPA.2018.0295View article Google Scholar
[37]K. McNamara, Y. Ji, F. Lia, P. Promoppatum, S.C. Yao, H. Zhou, Y. Wang, L.Q. Chen, R.P. MartukanitzPredicting phase transformation kinetics during metal additive manufacturing using non-isothermal Johnson-Mehl-Avrami models: application to Inconel 718 and Ti-6Al-4VAddit. Manuf., 49 (2022), Article 102478, 10.1016/J.ADDMA.2021.102478View PDFView articleView in ScopusGoogle Scholar
[39]A. Drexler, B. Oberwinkler, S. Primig, C. Turk, E. Povoden-karadeniz, A. Heinemann, W. Ecker, M. StockingerMaterials Science & Engineering A Experimental and numerical investigations of the γ ″ and γ ′ precipitation kinetics in Alloy 718Mater. Sci. Eng. A., 723 (2018), pp. 314-323, 10.1016/j.msea.2018.03.013View PDFView articleView in ScopusGoogle Scholar
[44]A.V. Sotov, A.V. Agapovichev, V.G. Smelov, V.V. Kokareva, M.O. Dmitrieva, A.A. Melnikov, S.P. Golanov, Y.M. AnurovInvestigation of the IN-738 superalloy microstructure and mechanical properties for the manufacturing of gas turbine engine nozzle guide vane by selective laser meltingInt. J. Adv. Manuf. Technol., 107 (2020), pp. 2525-2535, 10.1007/s00170-020-05197-xView article View in ScopusGoogle Scholar
[49]S. Sanchez, P. Smith, Z. Xu, G. Gaspard, C.J. Hyde, W.W. Wits, I.A. Ashcroft, H. Chen, A.T. ClarePowder Bed Fusion of nickel-based superalloys: a reviewInt. J. Mach. Tools Manuf., 165 (2021), 10.1016/j.ijmachtools.2021.103729 View PDF This article is free to access.Google Scholar
[50]C.L.A. Leung, R. Tosi, E. Muzangaza, S. Nonni, P.J. Withers, P.D. LeeEffect of preheating on the thermal, microstructural and mechanical properties of selective electron beam melted Ti-6Al-4V componentsMater. Des., 174 (2019), Article 107792, 10.1016/j.matdes.2019.107792View PDFView articleView in ScopusGoogle Scholar
[51]S. Griffiths, H. Ghasemi Tabasi, T. Ivas, X. Maeder, A. De Luca, K. Zweiacker, R. Wróbel, J. Jhabvala, R.E. Logé, C. LeinenbachCombining alloy and process modification for micro-crack mitigation in an additively manufactured Ni-base superalloyAddit. Manuf., 36 (2020), 10.1016/j.addma.2020.101443View article Google Scholar
[52]P. Soille, L. Vincent Pierre Soille, L.M. Vincent, Determining watersheds in digital pictures via flooding simulations, Https://Doi.Org/10.1117/12.24211. 1360 (1990) 240–250. https://doi.org/10.1117/12.24211.Google Scholar
[53]ASTM Standard Test Method for Microindentation Hardness of MaterialsKnoop and Vickers Hardness of Materials 1, Annu. B ASTM Stand. i 2010 1 42.Google Scholar
[56]B. Sonderegger, E. KozeschnikGeneralized nearest-neighbor broken-bond analysis of randomly oriented coherent interfaces in multicomponent Fcc and Bcc structuresMetall. Mater. Trans. A Phys. Metall. Mater. Sci., 40 (2009), pp. 499-510, 10.1007/S11661-008-9752-6/FIGURES/8View articleView in ScopusGoogle Scholar
[58]B. Sonderegger, E. KozeschnikInterfacial energy of diffuse phase boundaries in the generalized broken-bond approachMetall. Mater. Trans. A Phys. Metall. Mater. Sci., 41 (2010), pp. 3262-3269, 10.1007/S11661-010-0370-8/FIGURES/4View articleView in ScopusGoogle Scholar
[67]P. Strunz, M. Petrenec, J. Polák, U. Gasser, G. FarkasFormation and dissolution of’ precipitates in IN792 superalloy at elevated temperaturesMetals, 6 (2016), 10.3390/met6020037View article Google Scholar
[68]A.R.P. Singh, S. Nag, J.Y. Hwang, G.B. Viswanathan, J. Tiley, R. Srinivasan, H.L. Fraser, R. BanerjeeInfluence of cooling rate on the development of multiple generations of γ′ precipitates in a commercial nickel base superalloyMater. Charact., 62 (2011), pp. 878-886, 10.1016/j.matchar.2011.06.002View PDFView articleView in ScopusGoogle Scholar
[69]E. Balikci, A. Raman, R. MirshamsMicrostructure evolution in polycrystalline IN738LC in the range 1120 to 1250C, Zeitschrift FuerMet, 90 (1999), pp. 132-140View in ScopusGoogle Scholar
[71]Ł. Rakoczy, M. Grudzień-Rakoczy, F. Hanning, G. Cempura, R. Cygan, J. Andersson, A. Zielińska-LipiecInvestigation of the γ′ precipitates dissolution in a Ni-based superalloy during stress-free short-term annealing at high homologous temperaturesMetall. Mater. Trans. A Phys. Metall. Mater. Sci., 52 (2021), pp. 4767-4784, 10.1007/s11661-021-06420-4 View PDF This article is free to access.View in ScopusGoogle Scholar
[73]F. Masoumi, D. Shahriari, M. Jahazi, J. Cormier, A. DevauxKinetics and Mechanisms of γ′ Reprecipitation in a Ni-based SuperalloySci. Rep., 6 (2016), pp. 1-16, 10.1038/srep28650View articleGoogle Scholar
[74]A.R.P. Singh, S. Nag, S. Chattopadhyay, Y. Ren, J. Tiley, G.B. Viswanathan, H.L. Fraser, R. BanerjeeMechanisms related to different generations of γ′ precipitation during continuous cooling of a nickel base superalloyActa Mater., 61 (2013), pp. 280-293, 10.1016/j.actamat.2012.09.058View PDFView articleView in ScopusGoogle Scholar
[76]N.D. Souza, M.C. Hardy, B. Roebuck, W.E.I. Li, G.D. West, D.M. Collins, On the Rate Dependence of Precipitate Formation and Dissolution in a Nickel-Base Superalloy, Metall. Mater. Trans. A. (n.d.). https://doi.org/10.1007/s11661–022-06680–8.Google Scholar
[79]H. Huang, G. Liu, H. Wang, A. Ullah, B. HuDissolution behavior and kinetics of γ′ phase during solution treatment in powder metallurgy nickel-based superalloyMetall. Mater. Trans. A Phys. Metall. Mater. Sci., 51 (2020), pp. 1075-1084, 10.1007/s11661-019-05581-7View article View in ScopusGoogle Scholar
[80]A.J. Goodfellow, E.I. Galindo-Nava, K.A. Christofidou, N.G. Jones, T. Martin, P.A.J. Bagot, C.D. Boyer, M.C. Hardy, H.J. StoneGamma prime precipitate evolution during aging of a model nickel-based superalloyMetall. Mater. Trans. A Phys. Metall. Mater. Sci., 49 (2018), pp. 718-728, 10.1007/s11661-017-4336-y View PDF This article is free to access.View in ScopusGoogle Scholar
[81]T.P. Gabb, D.G. Backman, D.Y. Wei, D.P. Mourer, D. Furrer, A. Garg, D.L. Ellis, #947;’ Form. a Nickel-Base Disk Superalloy 2012 405 414 doi: 10.7449/2000/superalloys_2000_405_414.Google Scholar
[82]A. PlatiModelling of γ precipitation in superalloys University of CambridgeMater. Sci. (2003), p. 73Google Scholar
[91]F. Theska, W.F. Tse, B. Schulz, R. Buerstmayr, S.R. Street, M. Lison-Pick, S. PrimigReview of microstructure–mechanical property relationships in cast and wrought ni-based superalloys with boron, carbon, and zirconium microalloying additionsAdv. Eng. Mater. (2022), p. 2201514, 10.1002/ADEM.202201514 View PDF This article is free to access.Google Scholar
[93]L. Zhang, Y. Li, Q. Zhang, S. ZhangMicrostructure evolution, phase transformation and mechanical properties of IN738 superalloy fabricated by selective laser melting under different heat treatmentsMater. Sci. Eng. A, 844 (2022), Article 142947, 10.1016/j.msea.2022.142947View PDFView articleView in ScopusGoogle Scholar
[94]J.C. Franco-Correa, E. Martínez-Franco, J.M. Alvarado-Orozco, L.A. Cáceres-Díaz, D.G. Espinosa-Arbelaez, J.A. VilladaEffect of conventional heat treatments on the microstructure and microhardness of IN718 obtained by wrought and additive manufacturingJ. Mater. Eng. Perform., 30 (2021), pp. 7035-7045, 10.1007/s11665-021-06138-9View article View in ScopusGoogle Scholar
[96]N. Li, M.J. Anderson, H.C. BasoaltoAutomated stereology and uncertainty quantification considering spherical non-penetrating dispersionsPage 464.Cryst 2023, Vol. 13 (13) (2023), p. 464, 10.3390/CRYST13030464View article Google Scholar
[97]W.T. Loomis, J.W. Freeman, D.L. SponsellerInfluence of molybdenum on the γ′- phase in experimental nickelbase superalloysMet. Trans., 3 (1972), pp. 989-1000Google Scholar
[98]A.S. Shaikh Development of a γ’ Precipitation Hardening Ni-Base Superalloy for Additive Manufacturing Thesis 2018 102.〈https://odr.chalmers.se/handle/20.500.12380/255645%0Ahttps://www.researchgate.net/profile/Abdul_Shaafi_Shaikh2/publication/326226200%0Ahttps://drive.google.com/open?id=1BIez-aJyBTjnSgazzzTvv3jlrYpFC0N-〉.Google Scholar
[102]P.N. Quested, R.F. Brooks, L. Chapman, R. Morrell, Y. Youssef, K.C. MillsMeasurement and estimation of thermophysical properties of nickel based superalloysMater. Sci. Technol., 25 (2009), pp. 154-162, 10.1179/174328408×361454View articleView in ScopusGoogle Scholar
Analysis on inundation characteristics by compound external forces in coastal areas
연안 지역의 복합 외력에 의한 침수 특성 분석
Taeuk Kanga, Dongkyun Sunb, Sangho Leec* 강 태욱a, 선 동균b, 이 상호c*
aResearch Professor, Disaster Prevention Research Institute, Pukyong National University, Busan, Korea bResearcher, Disaster Prevention Research Institute, Pukyong National University, Busan, Korea cProfessor, Department of Civil Engineering, Pukyong National University, Busan, Korea a부경대학교 방재연구소 전임연구교수 b부경대학교 방재연구소 연구원 c부경대학교 공과대학 토목공학과 교수 *Corresponding Author
ABSTRACT
연안 지역은 강우, 조위, 월파 등 여러가지 외력에 의해 침수가 발생될 수 있다. 이에 이 연구에서는 연안 지역에서 발생될 수 있는 단일 및 복합 외력에 의한 지역별 침수 특성을 분석하였다. 연구에서 고려한 외력은 강우와 폭풍 해일에 의한 조위 및 월파이고, 분석 대상지역은 남해안 및 서해안의 4개 지역이다. 유역의 강우-유출 및 2차원 지표면 침수 분석에는 XP-SWMM이 사용되었고, 폭풍 해일에 의한 외력인 조위 및 월파량 산정에는 ADCSWAN (ADCIRC와 UnSWAN) 모형과 FLOW-3D 모형이 각각 활용되었다. 단일 외력을 이용한 분석 결과, 대부분의 연안 지역에서는 강우에 의한 침수 영향보다 폭풍 해일에 의한 침수 영향이 크게 나타났다. 복합 외력에 의한 침수 분석 결과는 대체로 단일 외력에 의한 침수 모의 결과를 중첩시켜 나타낸 결과와 유사하였다. 다만, 특정 지역에서는 복합 외력을 고려함에 따라 단일 외력만을 고려한 침수모의에서 나타나지 않았던 새로운 침수 영역이 발생하기도 하였다. 이러한 지역의 침수 피해 저감을 위해서는 복합 외력을 고려한 분석이 요구되는 것으로 판단되었다.
키워드
연안 지역
침수 분석
강우
폭풍 해일
복합 외력
The various external forces can cause inundation in coastal areas. This study is to analyze regional characteristics caused by single or compound external forces that can occur in coastal areas. Storm surge (tide level and wave overtopping) and rainfall were considered as the external forces in this study. The inundation analysis were applied to four coastal areas, located on the west and south coast in Republic of Korea. XP-SWMM was used to simulate rainfall-runoff phenomena and 2D ground surface inundation for watershed. A coupled model of ADCIRC and SWAN (ADCSWAN) was used to analyze tide level by storm surge and the FLOW-3D model was used to estimate wave overtopping. As a result of using a single external force, the inundation influence due to storm surge in most of the coastal areas was greater than rainfall. The results of using compound external forces were quite similar to those combined using one external force independently. However, a case of considering compound external forces sometimes created new inundation areas that didn’t appear when considering only a single external force. The analysis considering compound external forces was required to reduce inundation damage in these areas.
우리나라는 반도에 위치하여 삼면이 바다로 둘러싸여 있는 지리적 특성을 가지고 있다. 이에 따라 해양 산업을 중심으로 부산, 인천, 울산 등 대규모의 광역도시가 발달하였을 뿐만 아니라, 창원, 포항, 군산, 목포, 여수 등의 중․소규모 도시들도 발달되어 있다. 또한, 최근에는 연안 지역이 바다를 전망으로 하는 입지 조건을 가지고 있어 개발 선호도가 높고, 이에 따라 부산시 해운대의 마린시티, 엘시티와 같은 주거 및 상업시설의 개발이 지속되고 있다(Kang et al., 2019b).
한편, 최근 기후변화에 따른 지구 온난화 현상으로 평균 해수면이 상승하고, 해수면 온도도 상승하면서 태풍 및 강우의 강도가 커지고 있어 전 세계적으로 자연 재해로 인한 피해가 증가하고 있다(Kim et al., 2016). 실제로 2020년에는 최장기간의 장마가 발생하여 부산, 울산은 물론, 전국에서 50명의 인명 피해와 3,489세대의 이재민이 발생하였다1). 특히, 연안 지역은 강우, 만조 시 해수면 상승, 폭풍 해일(storm surge)에 의한 월파(wave overtopping) 등 복합적인 외력(compound external forces)에 의해 침수될 수 있다(Lee et al., 2020). 일례로, 2016년 태풍 차바 시 부산시 해운대구의 마린시티는 강우와 폭풍 해일에 의한 월파가 발생함에 따라 대규모 침수를 유발하였다(Kang et al., 2019b). 또한, 2020년 7월 23일에 부산에서는 시간당 81.6 mm의 집중호우와 약최고고조위를 상회하는 만조가 동시에 발생하였고, 이로 인해 감조 하천인 동천의 수위가 크게 상승하여 하천이 범람하였다(KSCE, 2021).
연안 지역의 복합 외력을 고려한 침수 분석에 관한 사례로서, 우선 강우와 조위를 고려한 연구 사례는 다음과 같다. Han et al. (2014)은 XP-SWMM을 이용하여 창원시 배수 구역을 대상으로 침수 모의를 수행하였는데, 연안 도시의 침수 모의에는 조위의 영향을 반드시 고려해야 함을 제시하였다. Choi et al. (2018a)은 경남 사천시 선구동 일대에 대하여 초과 강우 및 해수면 상승 시나리오를 조합하여 침수 분석을 수행하였다. Choi et al. (2018b)은 XP-SWMM을 이용하여 여수시 연등천 및 여수시청 지역에 대하여 강우 시나리오와 해수위 상승 시나리오를 고려한 복합 원인에 의한 침수 모의를 수행하여 홍수예경보 기준표를 작성하였다. 한편, 강우, 조위, 월파를 고려한 연구 사례로서, Song et al. (2017)은 부산시 해운대구 수영만 일원에 대하여 XP-SWMM으로 월파량의 적용 유무에 따른 침수 면적을 비교하였다. Suh and Kim (2018)은 부산시 마린시티 지역을 대상으로 태풍 차바 때 EurOtop의 경험식을 ADSWAN에 적용하여 월파량을 반영하였다. Chen et al. (2017)은 TELEMAC-2D 및 SWMM을 기반으로 한 극한 강우, 월파 및 조위를 고려하여 중국 해안 원자력 발전소의 침수를 예측하고 분석하기 위한 결합 모델을 개발한 바 있다. 한편, Lee et al. (2020)은 수리‧수문학 분야와 해양공학 분야에서 사용되는 물리 모형의 기술적 연계를 통해 연안 지역의 침수 모의의 재현성을 높였다.
상기의 연구들은 공통적으로 연안 지역에 대하여 복합 외력을 고려했을 때 발생되는 침수 현상의 재현 또는 예측을 목적으로 수행되었다. 이 연구는 이와 차별하여 복합 외력을 고려하는 경우 나타날 수 있는 연안 지역의 침수 특성 분석을 목적으로 수행되었다. 이를 위해 단일 외력을 독립적으로 고려했을 때 발생되는 침수 양상과 동시에 고려하는 경우의 침수 현상을 비교, 분석하였다. 복합 외력에 의한 지역적 침수 특성 분석은 우리나라 남해안과 서해안에 위치한 4개 지역에 대하여 적용되었다.
연안 지역의 침수는 크게 세 가지의 메카니즘으로 발생될 수 있다. 우선, 연안 지역은 바다와 인접하고 있기 때문에 그 영향을 직접적으로 받는다. Kim (2018)에 의하면, 연안 지역의 침수는 폭풍 해일에 의해 상승한 조위와 월파로 인해 발생될 수 있다(Table 1). 특히, 경상남도의 창원과 통영, 인천광역시의 소래포구 어시장 등 남해안 및 서해안 지역의 일부는 백중사리, 슈퍼문(super moon) 등 만조 시 조위의 상승으로 인한 침수가 발생하는 지역이 존재한다(Kang et al., 2019a). 두 번째는 강우에 의한 내수 침수 발생이다. ME (2011)에서는 도시 지역의 우수 관거를 10 ~ 30년 빈도로 계획하도록 지정하고 있고, 펌프 시설은 30 ~ 50년 빈도의 홍수를 배수시킬 수 있도록 정하고 있다. 하지만 최근에는 기후변화의 영향으로 도시 지역 배수시설의 설계 빈도를 초과하는 강우가 빈번하게 나타나고 있다. 실제로 2016년의 태풍 차바 시 울산 기상관측소에 관측된 시간 최대 강우량은 106.0 mm로서, 이는 300년 빈도 이상의 강우량에 해당하였다(Kang et al., 2019a). 따라서 배수시설의 설계 빈도 이상의 강우는 연안 도시 지역의 침수를 유발할 수 있다. 세 번째, 하천이 인접한 연안 도시에서는 하천의 범람으로 인해 침수가 발생할 수 있다. 하천의 경우, 기본계획이 수립되기는 하지만, 설계 빈도를 상회하는 강우의 발생, 제방, 수문 등 홍수 방어시설의 기능 저하, 예산 등의 문제로 하천기본계획 이행의 지연 등에 의해 범람할 가능성이 존재한다.
Table 1.
Type of natural hazard damage in coastal areas (Kim, 2018)
Item
Risk factor
Facilities damage
∙ Breaking of coastal facilities by wave – Breakwater, revetment, lighters wharf etc. ∙ Local scouring at the toe of the structures by wave ∙ Road collapse by wave overtopping
Inundation damage
∙ Inundation damage by wave overtopping ∙ Inundation of coastal lowlands by storm surge
Erosion damage
∙ Backshore erosion due to high swell waves ∙ Shoreline changes caused by construction of coastal erosion control structure ∙ Sediment transport due to the construction of artificial structures
상기의 내용을 종합하면, 연안 지역은 조위 및 월파에 의한 침수, 강우에 의한 내수 침수, 하천 범람에 의한 침수로 구분될 수 있다. 이 연구에서는 폭풍 해일에 의한 조위 상승 및 월파와 강우를 연안 지역의 침수 유발 외력으로 고려하였다. 하천 범람의 경우, 상대적으로 사례가 희소하여 제외하였다.
2.2 복합 외력을 고려한 침수 모의 방법
이 연구에서는 조위 및 월파와 강우를 연안 지역의 침수 발생에 관한 외력 조건으로 고려하였다. 따라서 해당 외력 조건을 고려하여 침수 분석을 수행할 수 있어야 한다. 이와 관련하여 Lee et al. (2020)은 Fig. 1과 같이 수리‧수문 및 해양공학 분야에서 사용되는 물리 기반 모형의 연계를 통해 조위, 월파, 강우를 고려한 침수 분석 방법을 제시하였고, 이 연구에서는 해당 방법을 이용하였다.
Fig. 1.
Connection among the models for inundation analysis in coastal areas (Lee et al., 2020)
우선, 태풍에 의해 발생되는 폭풍 해일의 영향을 분석하기 위해서는 태풍에 의해 발생되는 기압 강하, 해상풍, 진행 속도 등을 고려하여 해수면의 변화 양상 및 조석-해일-파랑을 충분히 재현 가능해야 한다. 이 연구에서는 국내․외에서 검증 및 공인된 폭풍 해일 모형인 ADCIRC 모형과 파랑 모형인 UnSWAN이 결합된 ADCSWAN (coupled model of ADCIRC and UnSWAN)을 이용하였다. 정수압 가정의 ADCSWAN은 월파량 산정에 단순 경험식을 적용하는 단점이 있지만 넓은 영역을 모의할 수 있고, FLOW-3D는 해안선의 경계를 고해상도로 재현이 가능하다. 이에 연구에서는 먼 바다 영역에 대해서는 ADCSWAN을 이용하여 분석하였고, 연안 주변의 바다 영역과 월파량 산정에 대해서는 FLOW-3D 모형을 이용하였다. 한편, 연안 지역의 침수 모의를 위해서는 유역에서 발생하는 강우-유출 현상과 우수 관거 등의 배수 체계에 대한 분석이 가능해야 한다. 또한, 배수 체계로부터 범람한 물이 지표면을 따라 흘러가는 현상을 해석할 수 있어야 하고, 바다의 조위 및 월파량을 경계조건으로 반영할 수 있어야 한다. 이 연구에서는 이러한 현상을 모의할 수 있고, 도시 침수 모의에 활용도가 높은 XP-SWMM을 이용하였다.
2.3 침수 분석 대상지역
연구의 대상지역은 조위 및 월파에 의한 침수와 강우에 의한 내수 침수의 영향이 복합적으로 발생할 수 있는 남해안과 서해안에 위치한 4개 지역이다. Table 2는 침수 분석 대상지역을 정리하여 나타낸 표이고, Fig. 2는 각 지역의 유역 경계를 나타낸 그림이다.
Table 2.
Target region for inundation analysis
Classification
Administrative district
Target region
Area (km2)
Main cause of inundation
Pump facility
Number of major outfall
The south coast
Haundae-gu, Busan
Marine City area
0.53
Wave overtopping
–
9
Haundae-gu, Busan
Centum City area
4.76
Poor interior drainage at high tide level
1
2
The west coast
Gunsan
Jungang-dong area
0.79
Poor interior drainage at high tide level
2
3
Boryeong
Ocheon Port area
0.41
High tide level
–
5
Fig. 2.
Watershed area
남해안의 분석 대상지역 중 부산시 해운대구의 마린시티는 바다 조망을 중심으로 조성된 주거지 및 상업시설 중심의 개발지역이다. 마린시티는 2016년 태풍 차바 및 2018년 태풍 콩레이 등 태풍 내습 시 월파에 의한 해수 월류로 인해 도로 및 상가 일부가 침수를 겪은 지역이다. 부산시 해운대구의 센텀시티는 과거 수영만 매립지였던 곳에 조성된 주거지 및 상업시설 중심의 신도시 지역이다. 센텀시티 유역의 북쪽은 해발고도 El. 634 m의 장산이 위치하는 등 산지 특성도 가지고 있어 상대적으로 유역 면적이 넓고, 배수시설의 규모도 크고 복잡하다. 하지만 수영강 하구의 저지대 지역에 위치함에 따라 강우 시 내수 배제가 불량하고, 특히 만조 시 침수가 잦은 지역이다.
서해안 분석 대상지역 중 전라북도 군산시의 중앙동 일원은 군산시 내항 내측에 조성된 구도시로서, 금강 및 경포천 하구에 위치하는 저지대이다. 이에 따라 군산시 풍수해저감종합계획에서는 해당 지역을 3개의 영역으로 구분하여 내수재해 위험지구(영동지구, 중동지구, 경암지구)로 지정하였고, 이 연구에서는 해당 지역을 모두 고려하였다. 한편, 군산시 중앙동 일원은 특히, 만조 시 내수 배제가 매우 불량하여 2개의 펌프시설이 운영되고 있다. 충청남도 보령시의 오천면에 위치한 오천항은 배후의 산지를 포함한 소규모 유역에 위치한다. 서해안의 특성에 따라 조석 간만의 차가 크고, 특히 태풍 내습 시 폭풍 해일에 의한 침수가 잦은 지역이다. 산지의 강우-유출수는 복개된 2개의 수로를 통해 바다로 배제되고, 상가들이 위치한 연안 주변 지역에는 강우-유출수 배제를 위한 3개의 배수 체계가 구성되어 있다.
3. 연구 결과
3.1 침수 모의 모형 구축
XP-SWMM을 이용하여 분석 대상지역별 침수 모의 모형을 구축하였다. 적절한 침수 분석 수행을 위해 지역별 수치지형도, 도시 공간 정보 시스템(urban information system, UIS), 하수 관망도 등의 수치 자료와 현장 조사를 통해 유역의 배수 체계를 구성하였다. 그리고 2차원 침수 분석을 위해 무인 드론 및 육상 라이다(LiDAR) 측량을 수행하여 평면해상도가 1 m 이하인 고해상도 수치지형모형(digital terrain model, DTM)을 구성하였고, 침수 모의 격자를 생성하였다.
Fig. 3은 XP-SWMM의 상세 구축 사례로서 부산시 마린시티 배수 유역에 대한 소유역 및 관거 분할 등을 통해 구성한 배수 체계와 고해상도 측량 결과를 이용하여 구성한 수치표면모형(digital surface model, DSM)을 나타낸다. Fig. 4는 각 대상지역에 대해 XP-SWMM을 이용하여 구축한 침수 모의 모형을 나타낸다. 침수 분석을 위해서는 침수 모의 영역에 대한 설정이 필요한데, 다수의 사전 모의를 통해 유역 내에서 침수가 발생되는 지역을 검토하여 결정하였다.
Fig. 3.
Analysis of watershed drainage system and high-resolution survey for Marine City
Fig. 4.
Simulation model for inundation analysis by target region using XP-SWMM
복합 외력에 의한 침수 영향을 검토하기 위해서는 외력 조건에 대한 빈도와 지속기간의 설정이 필요하다. 이 연구에서는 재해 현상이 충분히 나타날 수 있도록 강우와 조위 및 월파의 빈도를 모두 100년으로 설정하였다. 이때, 조위와 월파량의 산정에는 만조(약최고고조위) 시, 100년 빈도에 해당하는 태풍 내습에 따른 폭풍 해일의 발생 조건을 고려하였다.
지역별 강우 발생 특성과 유역 특성을 고려하기 위해 MOIS (2017)의 방재성능목표 기준에 따라 임계 지속기간을 결정하여 대상지역별 강우의 지속기간으로 설정하였다. 이때, 강우의 시간 분포는 MLTM (2011)의 Huff 3분위를 이용하였다. 그리고 조위와 월파의 경우, 일반적인 폭풍 해일의 지속기간을 고려하여 5시간으로 결정하였다. 한편, 침수 모의를 위한 계산 시간 간격, 2차원 모의 격자 등의 입력자료는 분석 대상지역의 유역 규모와 침수 분석 대상 영역을 고려하여 결정하였다. 참고로 침수 분석에 사용된 수치지형모형은 1 m 급의 고해상도로 구성되었지만, 2차원 침수 모의 격자의 크기는 지역별로 3 ~ 4 m이다. 이는 연구에서 사용된 XP-SWMM의 격자 수(100,000개) 제약에 따른 설정이나, Sun (2021)은 민감도 분석을 통해 2차원 침수 분석을 위한 적정 격자 크기를 3 ~ 4.5 m로 제시한 바 있다.
Table 3은 이 연구에서 설정한 침수 모의 조건과 분석 방법을 정리하여 나타낸 표이다.
Table 3.
Simulation condition and method
Classification
Target region
Simulation condition
Simulation method
Rainfall
Storm surge
Simulation time interval
2D grid size
Return period
Duration
Temporal distribution
Return period
Duration
Watershed routing
Channel routing
2D inundation
The south coast
Marine City area
100 yr
1 hr
3rd quartile of Huff’s method
100
5 hr
5 min
10 sec
1 sec
3 m
Centum City area
1 hr
100
5 min
10 sec
1 sec
4 m
The west coast
Jungang-dong area
2 hr
100
5 min
10 sec
1 sec
3.5 m
Ocheon Port area
1 hr
100
1 min
10 sec
1 sec
3 m
3.2.2 복합 재해의 동시 고려
이 연구의 대상지역들은 모두 소규모의 해안가 도시지역이고, 이러한 지역에 대한 강우의 임계지속기간은 1시간 ~ 2시간이나, 이 연구에서 분석한 폭풍 해일의 지속기간은 5시간으로 강우의 지속기간과 폭풍 해일의 지속기간이 상이하다. 이에 이 연구에서는 서로 다른 지속기간을 가진 강우와 폭풍 해일 또는 조위를 고려하기 위해 강우의 중심과 폭풍 해일의 중심이 동일한 시간에 위치하도록 설정하였다(Fig. 5).
XP-SWMM은 폭풍 해일이 지속되는 5시간 전체를 모의하도록 설정하였고, 폭풍 해일이 가장 큰 시점에 강우의 중심이 위치하도록 강우 발생 시기를 결정하였다. 다만, 부산 마린시티의 경우, 폭풍 해일에 의한 피해가 주로 월파에 의해 발생되므로 강우의 중심과 월파의 중심을 일치시켰고(Fig. 5(a)), 상대적으로 조위의 영향이 큰 3개 지역은 강우의 중심과 조위의 중심을 맞추었다. Fig. 5(b)는 군산시 중앙동 지역의 복합 외력에 의한 침수 분석에 사용된 강우와 조위의 조합이다.
한편, 100년 빈도의 확률강우량만을 고려한 침수 분석에서는 유역 유출부의 경계조건으로 우수 관거의 설계 조건을 고려하여 약최고고조위가 일정하게 유지되도록 설정하였다.
Fig. 5.
Consideration of external force conditions with different durations
3.2.3 XP-SWMM의 월파량 고려
XP-SWMM에 ADCSWAN 및 FLOW-3D 모형에 의해 산정된 월파량을 입력하기 위해 해안가 지역에 절점을 생성하여 월파 현상을 구현하였다. XP-SWMM에서 월파량을 입력하기 위한 절점의 위치는 FLOW-3D 모형에서 월파량을 산정한 격자의 중심 위치이다.
Fig. 6(a)는 마린시티 지역에 대한 월파량 입력 지점을 나타낸 것으로서, 유역 경계 주변에 동일 간격으로 원으로 표시한 지점들이 해당된다. Fig. 6(b)는 XP-SWMM에 월파량 입력 지점들을 반영하고, 하나의 절점에 월파량 시계열을 입력한 화면을 나타낸다.
Fig. 6.
Considering wave overtopping on XP-SWMM
3.3 침수 모의 결과
3.3.1 단일 외력에 의한 침수 모의 결과
Fig. 7은 단일 외력을 고려한 지역별 침수 모의 결과이다. 즉, Fig. 7의 왼쪽 그림들은 지역별로 100년 빈도 강우에 의한 침수 모의 결과를 나타내고, Fig. 7의 오른쪽 그림들은 만조 시 100년 빈도 폭풍 해일에 의한 침수 모의 결과이다. 대체로 강우에 의한 침수 영역은 유역 중․상류 지역의 유역 전반에 걸쳐 발생하였고, 폭풍 해일에 의한 침수 영역은 해안가 전면부에 위치하는 것을 볼 수 있다. 이는 폭풍 해일에 의한 조위 상승과 월파의 영향이 상류로 갈수록 감소하기 때문이다.
한편, 4개 지역 모두에서 공통적으로 강우에 비해 폭풍 해일에 의한 침수 영향이 상대적으로 크게 분석되었다. 이러한 결과는 연안 지역의 경우, 폭풍 해일에 대비한 침수 피해 저감 노력이 보다 중요함을 의미한다.
Fig. 7.
Simulation results by single external force (left: rainfall, right: storm surge)
3.3.2 복합 외력에 의한 침수 모의 결과
Fig. 8은 복합 외력을 고려한 지역별 침수 모의 결과이다. 즉, 강우 및 폭풍 해일을 동시에 고려함에 따라 발생된 침수 영역을 나타낸다. 복합 외력을 고려하는 경우, 단일 외력만을 고려한 분석 결과(Fig. 7)보다 침수 영역은 넓어졌고, 침수심은 깊어졌다.
복합 외력에 의한 침수 분석 결과는 대체로 단일 외력에 의한 침수 모의 결과를 중첩시켜 나타낸 결과와 유사하였고, 이는 일반적으로 예상할 수 있는 결과이다. 주목할만한 결과는 군산시 중앙동의 침수 분석에서 나타났다. 즉, 군산시 중앙동의 경우, 단일 외력만을 고려한 침수 모의 결과에서 나타나지 않았던 새로운 침수 영역이 발생하였다(Fig. 8(c)). 이와 관련된 상세 내용은 3.4절의 고찰에서 기술하였다.
Fig. 8.
Simulation results by compound external forces
3.4 결과 고찰
외력 조건별 침수의 영향을 정량적으로 비교하기 위해 침수 면적을 이용하였다. 이 연구에서는 강우만에 의해 유발된 침수 면적을 기준(기준값: 1)으로 하고, 폭풍 해일(조위+월파량)에 의한 침수 면적과 복합 외력에 의한 침수 면적의 상대적 비율로 분석하였다(Table 4).
Table 4.
Impact evaluation for inundation area by external force
Condition
Marine City, Busan
Centum City, Busan
Jungang-dong area, Gunsan
Ocheon Port area, Boryeong
Inundation area (km2)
Rate
Inundation area (km2)
Rate
Inundation area (km2)
Rate
Inundation area (km2)
Rate
Single external force
Rainfall (①)
0.0164
1.0
0.0759
1.0
0.0457
1.0
0.0175
1.0
Storm surge (②)
0.0363
2.21
0.0685
0.90
0.1463
3.20
0.0412
2.35
Compound external forces
Combination (①+②)
0.0524
3.19
0.1505
1.98
0.2632
5.76
0.0473
2.70
분석 결과, 부산 센텀시티를 제외한 3개 지역은 모두 폭풍 해일에 의한 침수 면적이 강우에 의한 침수 면적에 비해 2.2 ~ 3.2배 넓은 것으로 분석되었다. 한편, 복합 외력에 의한 침수 면적은 마린시티와 센텀시티의 경우, 각각의 외력에 의한 침수 면적의 합과 유사하게 나타났다. 이는 각각의 외력에 의한 침수 영역이 상이하여 거의 중복되지 않음을 의미한다. 반면에, 오천항에서는 각각의 외력에 의한 침수 면적의 합이 복합 외력에 의한 면적보다 크게 나타났다. 이는 오천항의 경우, 유역면적이 작고 배수 체계가 비교적 단순하여 강우와 폭풍 해일에 의한 침수 영역이 중복되기 때문인 것으로 분석되었다(Fig. 7(d)).
군산시 중앙동 일대의 경우, 복합 외력에 의한 침수 면적이 각각의 독립적인 외력 조건에 의한 침수 면적의 합에 비해 37.1% 크게 나타났다. 이러한 현상의 원인을 분석하기 위해 복합 외력 조건에서만 나타난 우수 관거(Fig. 8(c)의 A 구간)에 대하여 종단을 검토하였다(Fig. 9). Fig. 9(a)는 강우만에 의해 분석된 우수 관거 내 흐름 종단을 나타내고, Fig. 9(b)는 폭풍 해일만에 의한 우수 관거의 종단이다. 그림을 통해 각각의 독립적인 외력 조건 하에서는 해당 구간에서 침수가 발생되지 않은 것을 볼 수 있다. 다만, 강우만을 고려하더라도 우수 관거는 만관이 된 상태를 확인할 수 있다(Fig. 9(a)). 반면에, 만관 상태에서 폭풍 해일이 함께 고려됨에 따라 해수 범람과 조위 상승에 의해 우수 배제가 불량하게 되었고, 이로 인해 침수가 유발된 것으로 분석되었다(Fig. 9(c)). 따라서 이러한 지역은 복합 외력에 대한 취약지구로 판단할 수 있고, 단일 외력의 고려만으로는 침수를 예상하기 어려운 지역임을 알 수 있다.
Fig. 9.
A part of drainage profiles by external force in Jungang-dong area, Gunsan
4. 결 론
이 연구에서는 외력 조건에 따른 연안 지역의 침수 특성을 분석하였다. 연구에서 고려된 외력 조건은 두 가지로서 강우와 폭풍 해일(조위와 월파)이다. 분석 대상 연안 지역으로는 남해안에 위치하는 2개 지역(부산시 해운대구의 마린시티와 센텀시티)과 서해안의 2개 지역(군산시 중앙동 일원 및 보령시 오천항)이 선정되었다.
복합 외력을 고려한 연안 지역의 침수 모의를 위해서는 유역의 강우-유출 현상과 바다의 조위 및 월파량을 경계조건으로 반영할 수 있는 침수 모의 모형이 요구되는데, 이 연구에서는 XP-SWMM을 이용하였다. 한편, 조위 및 월파량 산정에는 ADCSWAN (ADCIRC와 UnSWAN) 및 FLOW-3D 모형이 이용되었다.
연안 지역별 침수 모의는 100년 빈도의 강우와 폭풍 해일을 독립적으로 고려한 경우와 복합적으로 고려한 경우를 구분하여 수행되었다. 우선, 외력을 독립적으로 고려한 결과, 대체로 폭풍 해일만 고려한 경우가 강우만 고려한 경우에 비해 침수 영향이 크게 나타났다. 따라서 연안 지역의 경우, 폭풍 해일에 의한 침수 피해 방지 계획이 상대적으로 중요한 것으로 분석되었다. 두 번째, 복합 외력에 의한 침수 분석 결과는 대체로 단일 외력에 의한 침수 모의 결과를 중첩시켜 나타낸 결과와 유사하였다. 다만, 특정 지역에서는 복합 외력을 고려함에 따라 단일 외력만을 고려한 침수 모의에서 나타나지 않았던 새로운 침수 영역이 발생하기도 하였다. 이러한 결과는 독립적인 외력 조건에서는 우수 관거가 만관 또는 그 이하의 상태가 되지만, 두 가지의 외력이 동시에 고려됨에 따라 우수 관거의 통수능 한계를 초과하여 나타났다. 이러한 지역은 복합 외력에 대한 취약지구로 판단되었고, 해당 지역의 적절한 침수 방지 대책 수립을 위해서는 복합적인 외력 조건이 고려되어야 함을 시사하였다.
현행, 자연재해저감종합계획에서는 침수와 관련된 재해 원인 지역을 내수재해, 해안재해, 하천재해 등으로 구분하고 있다. 하지만 이 연구에서 검토된 바와 같이, 연안 지역의 침수 원인은 복합적으로 나타날 뿐만 아니라, 복합 외력을 고려함에 따라 추가적으로 나타날 수 있는 침수 위험 지역도 존재한다. 따라서 기존의 획일적인 재해 원인의 구분보다는 지역의 특성에 맞는 복합적인 재해 원인을 검토할 필요가 있음을 제안한다.
Acknowledgements
본 논문은 행정안전부 극한 재난대응 기반기술 개발사업의 일환인 “해안가 복합재난 위험지역 피해저감 기술개발(연구과제번호: 2018-MOIS31-008)”의 지원으로 수행되었습니다.
References
1
Chen, X., Ji, P., Wu, Y., and Zhao, L. (2017). “Coupling simulation of overland flooding and underground network drainage in a coastal nuclear power plant.” Nuclear Engineering and Design, Vol. 325, pp. 129-134. 10.1016/j.nucengdes.2017.09.028
2
Choi, G., Song, Y., and Lee, J. (2018a). “Analysis of flood occurrence type according to complex characteristics of coastal cities.” 2018 Conference of the Korean Society of Hazard Mitigation, KOSHAM, p. 180.
3
Choi, J., Park, K., Choi, S., and Jun, H. (2018b). “A forecasting and alarm system for reducing damage from inland inundation in coastal urban areas: A case study of Yeosu City.” Journal of Korean Society of Hazard Mitigation, Vol. 18, No. 7, pp. 475-484. 10.9798/KOSHAM.2018.18.7.475
4
Han, H., Kim, Y., Kang, N., and, Kim, H.S. (2014). “Inundation analysis of a coastal urban area considering tide level.” 2014 Conference of Korean Society of Civil Engineers, KSCE, pp. 1507-1508.
5
Kang, T., Lee, S., and Sun, D. (2019a). “A technical review for reducing inundation damage to high-rise and underground-linked complex buildings in Coastal Areas (1): Proposal for analytical method.” Journal of Korean Society of Hazard Mitigation, Vol. 19, No. 5, pp. 35-43. 10.9798/KOSHAM.2019.19.5.35
6
Kang, T., Lee, S., Choi, H., and Yoon, S. (2019b). “A technical review for reducing inundation damage to high-rise and underground-linked complex buildings in coastal areas (2): Case analysis for application.” Journal of Korean Society of Hazard Mitigation, Vol. 19, No. 5, pp. 45-53. 10.9798/KOSHAM.2019.19.5.45
7
Kim, J.O., Kim, J.Y., and Lee, W.H. (2016). “Analysis on complex disaster information contents for building disaster map of coastal cities.” Journal of the Korean Association of Geographic Information Studies, Vol. 19, No. 3, pp. 43-60. 10.11108/kagis.2016.19.3.043
8
Kim, P.J. (2018). Improvement measures on the risk area designation of coastal disaster in consideration of natural hazards. Ph.D. dissertation, Chonnam National University.
9
Korean Society of Civil Engineers (KSCE) (2021). A report on the cause analysis and countermeasures establishment for Dongcheon flooding and lowland inundation. Busan/Ulsan, Gyungnam branch.
10
Lee, S., Kang, T., Sun, D., and Park, J.J. (2020). “Enhancing an analysis method of compound flooding in coastal areas by linking flow simulation models of coasts and watershed.” Sustainability, Vol. 12, No. 16, 6572. 10.3390/su12166572
11
Ministry of Environment (ME) (2011). Standard for sewerage facilities. Korea Water and Wastewater Works Association.
12
Ministry of Land, Transport and Maritime Affairs (MLTM) (2011). Improvement and complementary research for probability rainfall.
13
Ministry of the Interior and Safety (MOIS) (2017). Criteria for establishment and operation of disaster prevention performance target by region: Considering future climate change impacts.
14
Song, Y., Joo, J., Lee, J., and Park, M. (2017). “A study on estimation of inundation area in coastal urban area applying wave overtopping.” Journal of Korean Society of Hazard Mitigation, Vol. 17, No. 2, pp. 501-510. 10.9798/KOSHAM.2017.17.2.501
15
Suh, S.W., and Kim, H.J. (2018). “Simulation of wave overtopping and inundation over a dike caused by Typhoon Chaba at Marine City, Busan, Korea.” Journal of Coastal Research, Vol. 85, pp. 711-715.
16
Sun, D. (2021). Sensitivity analysis of XP-SWMM for inundation analysis in coastal area. M.Sc. Thesis, Pukyong National University.
Fatemehsadat Mirshafiee1, Emad Shahbazi 2, Mohadeseh Safi 3, Rituraj Rituraj 4,* 1Department of Electrical and Computer Engineering, K.N. Toosi University of Technology, Tehran 1999143344 , Iran 2Department of Mechatronic, Amirkabir University of Technology, Tehran 158754413, Iran 3Department of Mechatronic, Electrical and Computer Engineering, University of Tehran, Tehran 1416634793, Iran 4 Faculty of Informatics, Obuda University, 1023, Budapest, Hungary
Correspondence: rituraj88@stud.uni-obuda.hu
ABSTRACT
본 연구는 지속가능한 에너지 변환기의 전력 및 수소 발생 모델링을 위한 데이터 기반 방법론을 제안합니다. 파고와 풍속을 달리하여 파고와 수소생산을 예측합니다.
또한 이 연구는 파도에서 수소를 추출할 수 있는 가능성을 강조하고 장려합니다. FLOW-3D 소프트웨어 시뮬레이션에서 추출한 데이터와 해양 특수 테스트의 실험 데이터를 사용하여 두 가지 데이터 기반 학습 방법의 비교 분석을 수행합니다.
결과는 수소 생산의 양은 생성된 전력의 양에 비례한다는 것을 보여줍니다. 제안된 재생 에너지 변환기의 신뢰성은 지속 가능한 스마트 그리드 애플리케이션으로 추가로 논의됩니다.
This study proposes a data-driven methodology for modeling power and hydrogen generation of a sustainable energy converter. The wave and hydrogen production at different wave heights and wind speeds are predicted. Furthermore, this research emphasizes and encourages the possibility of extracting hydrogen from ocean waves. By using the extracted data from FLOW-3D software simulation and the experimental data from the special test in the ocean, the comparison analysis of two data-driven learning methods is conducted. The results show that the amount of hydrogen production is proportional to the amount of generated electrical power. The reliability of the proposed renewable energy converter is further discussed as a sustainable smart grid application.
Key words
Cavity, Combustion efficiency, hydrogen fuel, Computational Fluent and Gambit.
REFERENCES
Kalbasi, R., Jahangiri, M., Dehshiri, S.J.H., Dehshiri, S.S.H., Ebrahimi, S., Etezadi, Z.A.S. and Karimipour, A., 2021. Finding the best station in Belgium to use residential-scale solar heating, one-year dynamic simulation with considering all system losses: economic analysis of using ETSW. Sustainable Energy Technologies and Assessments, 45, p.101097.
Megura M, Gunderson R. Better poison is the cure? Critically examining fossil fuel companies, climate change framing, and corporate sustainability reports. Energy Research & Social Science. 2022 Mar 1;85:102388.
Holechek JL, Geli HM, Sawalhah MN, Valdez R. A global assessment: can renewable energy replace fossil fuels by 2050?. Sustainability. 2022 Jan;14(8):4792.
Ahmad M, Kumar A, Ranjan R. Recent Developments of Tidal Energy as Renewable Energy: An Overview. River and Coastal Engineering. 2022:329-43.
Amini E, Mehdipour H, Faraggiana E, Golbaz D, Mozaffari S, Bracco G, Neshat M. Optimization of hydraulic power take-off system settings for point absorber wave energy converter. Renewable Energy. 2022 Jun 4.
Claywell, R., Nadai, L., Felde, I., Ardabili, S. 2020. Adaptive neuro-fuzzy inference system and a multilayer perceptron model trained with grey wolf optimizer for predicting solar diffuse fraction. Entropy, 22(11), p.1192.
McLeod I, Ringwood JV. Powering data buoys using wave energy: a review of possibilities. Journal of Ocean Engineering and Marine Energy. 2022 Jun 20:1-6.
Olsson G. Water interactions: A systemic view: Why we need to comprehend the water-climate-energy-food-economics-lifestyle connections.
Malkowska A, Malkowski A. Green Energy in the Political Debate. InGreen Energy 2023 (pp. 17-39). Springer, Cham.
Mayon R, Ning D, Ding B, Sergiienko NY. Wave energy converter systems–status and perspectives. InModelling and Optimisation of Wave Energy Converters (pp. 3-58). CRC Press.
Available online at: https://www.offshore-energy.biz/uk-ecotricity-introduces-wave-power-device-searaser/ (9/27/2022)
Mousavi SM, et al.,. Deep learning for wave energy converter modeling using long short-term memory. Mathematics. 2021 Apr 15;9(8):871.
Mega V. The Energy Race to Decarbonisation. InHuman Sustainable Cities 2022 (pp. 105-141). Springer, Cham.
Li R, Tang BJ, Yu B, Liao H, Zhang C, Wei YM. Cost-optimal operation strategy for integrating large scale of renewable energy in China’s power system: From a multi-regional perspective. Applied Energy. 2022 Nov 1;325:119780.
Ardabili S., Abdolalizadeh L., Mako C., Torok B., Systematic Review of Deep Learning and Machine Learning for Building Energy, Frontiers in Energy Research, 10, 2022.
Penalba M, Aizpurua JI, Martinez-Perurena A, Iglesias G. A data-driven long-term metocean data forecasting approach for the design of marine renewable energy systems. Renewable and Sustainable Energy Reviews. 2022 Oct 1;167:112751.
Torabi, M., Hashemi, S., Saybani, M.R., 2019. A Hybrid clustering and classification technique for forecasting short‐term energy consumption. Environmental progress & sustainable energy, 38(1), pp.66-76.
Rivera FP, Zalamea J, Espinoza JL, Gonzalez LG. Sustainable use of spilled turbinable energy in Ecuador: Three different energy storage systems. Renewable and Sustainable Energy Reviews. 2022 Mar 1;156:112005.
Raza SA, Jiang J. Mathematical foundations for balancing single-phase residential microgrids connected to a three-phase distribution system. IEEE Access. 2022 Jan 6;10:5292-303.
Takach M, Sarajlić M, Peters D, Kroener M, Schuldt F, von Maydell K. Review of Hydrogen Production Techniques from Water Using Renewable Energy Sources and Its Storage in Salt Caverns. Energies. 2022 Feb 15;15(4):1415.
Lv Z, Li W, Wei J, Ho F, Cao J, Chen X. Autonomous Chemistry Enabling Environment-Adaptive Electrochemical Energy Storage Devices. CCS Chemistry. 2022 Jul 7:1-9.
Dehghan Manshadi, Mahsa, Milad Mousavi, M. Soltani, Amir Mosavi, and Levente Kovacs. 2022. “Deep Learning for Modeling an Offshore Hybrid Wind–Wave Energy System” Energies 15, no. 24: 9484. https://doi.org/10.3390/en15249484
Ishaq H, Dincer I, Crawford C. A review on hydrogen production and utilization: Challenges and opportunities. International Journal of Hydrogen Energy. 2022 Jul 22;47(62):26238-64.
Maguire JF, Woodcock LV. On the Thermodynamics of Aluminum Cladding Oxidation: Water as the Catalyst for Spontaneous Combustion. Journal of Failure Analysis and Prevention. 2022 Sep 10:1-5.
Mohammadi, M. R., Hadavimoghaddam, F., Pourmahdi, M., Atashrouz, S., Munir, M. T., Hemmati-Sarapardeh, A., … & Mohaddespour, A. (2021). Modeling hydrogen solubility in hydrocarbons using extreme gradient boosting and equations of state. Scientific reports, 11(1).
Ma S, Qin J, Xiu X, Wang S. Design and performance evaluation of an underwater hybrid system of fuel cell and battery. Energy Conversion and Management. 2022 Jun 15;262:115672.
Ahamed R, McKee K, Howard I. A Review of the Linear Generator Type of Wave Energy Converters’ Power Take-Off Systems. Sustainability. 2022 Jan;14(16):9936.
Nejad, H.D., Nazari, M., Nazari, M., Mardan, M.M.S., 2022. Fuzzy State-Dependent Riccati Equation (FSDRE) Control of the Reverse Osmosis Desalination System With Photovoltaic Power Supply. IEEE Access, 10, pp.95585-95603.
Zou S, Zhou X, Khan I, Weaver WW, Rahman S. Optimization of the electricity generation of a wave energy converter using deep reinforcement learning. Ocean Engineering. 2022 Jan 15;244:110363.
Wu J, Qin L, Chen N, Qian C, Zheng S. Investigation on a spring-integrated mechanical power take-off system for wave energy conversion purpose. Energy. 2022 Apr 15;245:123318.
Papini G, Dores Piuma FJ, Faedo N, Ringwood JV, Mattiazzo G. Nonlinear Model Reduction by Moment-Matching for a Point Absorber Wave Energy Conversion System. Journal of Marine Science and Engineering. 2022 May;10(5):656.
Forbush DD, Bacelli G, Spencer SJ, Coe RG, Bosma B, Lomonaco P. Design and testing of a free floating dual flap wave energy converter. Energy. 2022 Feb 1;240:122485.
Rezaei, M.A., 2022. A New Hybrid Cascaded Switched-Capacitor Reduced Switch Multilevel Inverter for Renewable Sources and Domestic Loads. IEEE Access, 10, pp.14157-14183.
Lin Z, Cheng L, Huang G. Electricity consumption prediction based on LSTM with attention mechanism. IEEJ Transactions on Electrical and Electronic Engineering. 2020;15(4):556-562.
Tavoosi, J., Mohammadzadeh, A., Pahlevanzadeh, B., Kasmani, M.B., 2022. A machine learning approach for active/reactive power control of grid-connected doubly-fed induction generators. Ain Shams Engineering Journal, 13(2), p.101564.
Ghalandari, M., 2019. Flutter speed estimation using presented differential quadrature method formulation. Engineering Applications of Computational Fluid Mechanics, 13(1), pp.804-810.
Li Z, Bouscasse B, Ducrozet G, Gentaz L, Le Touzé D, Ferrant P. Spectral wave explicit navier-stokes equations for wavestructure interactions using two-phase computational fluid dynamics solvers. Ocean Engineering. 2021 Feb 1;221:108513.
Zhou Y. Ocean energy applications for coastal communities with artificial intelligencea state-of-the-art review. Energy and AI. 2022 Jul 29:100189.
Miskati S, Farin FM. Performance evaluation of wave-carpet in wave energy extraction at different coastal regions: an analytical approach (Doctoral dissertation, Department of Mechanical and Production Engineering).
Gu C, Li H. Review on Deep Learning Research and Applications in Wind and Wave Energy. Energies. 2022 Feb 17;15(4):1510.
Aazami, R., 2022. Optimal Control of an Energy-Storage System in a Microgrid for Reducing Wind-Power Fluctuations. Sustainability, 14(10), p.6183.
Kabir M, Chowdhury MS, Sultana N, Jamal MS, Techato K. Ocean renewable energy and its prospect for developing economies. InRenewable Energy and Sustainability 2022 Jan 1 (pp. 263-298). Elsevier.
Babajani A, Jafari M, Hafezisefat P, Mirhosseini M, Rezania A, Rosendahl L. Parametric study of a wave energy converter (Searaser) for Caspian Sea. Energy Procedia. 2018 Aug 1;147:334-42.
He J. Coherence and cross-spectral density matrix analysis of random wind and wave in deep water. Ocean Engineering. 2020;197:106930
Ijadi Maghsoodi, A., 2018. Renewable energy technology selection problem using integrated h-swara-multimoora approach. Sustainability, 10(12), p.4481.
Band, S.S., Ardabili, S., Sookhak, M., Theodore, A., Elnaffar, S., Moslehpour, M., Csaba, M., Torok, B., Pai, H.T., 2022. When Smart Cities Get Smarter via Machine Learning: An In-depth Literature Review. IEEE Access.
Shamshirband, S., Rabczuk, T., Nabipour, N. and Chau, K.W., 2020. Prediction of significant wave height; comparison between nested grid numerical model, and machine learning models of artificial neural networks, extreme learning and support vector machines. Engineering Applications of Computational Fluid Mechanics, 14(1), pp.805-817.
Liu, Z., Mohammadzadeh, A., Turabieh, H., Mafarja, M., 2021. A new online learned interval type-3 fuzzy control system for solar energy management systems. IEEE Access, 9, pp.10498-10508.
Bavili, R.E., Mohammadzadeh, A., Tavoosi, J., Mobayen, S., Assawinchaichote, W., Asad, J.H. 2021. A New Active Fault Tolerant Control System: Predictive Online Fault Estimation. IEEE Access, 9, pp.118461-118471.
Akbari, E., Teimouri, A.R., Saki, M., Rezaei, M.A., Hu, J., Band, S.S., Pai, H.T., 2022. A Fault-Tolerant Cascaded SwitchedCapacitor Multilevel Inverter for Domestic Applications in Smart Grids. IEEE Access.
Band, S.S., Ardabili, S., 2022. Feasibility of soft computing techniques for estimating the long-term mean monthly wind speed. Energy Reports, 8, pp.638-648.
Tavoosi, J., Mohammadzadeh, A., Pahlevanzadeh, B., Kasmani, M.B., 2022. A machine learning approach for active/reactive power control of grid-connected doubly-fed induction generators. Ain Shams Engineering Journal, 13(2), p.101564.
Ponnusamy, V. K., Kasinathan, P., Madurai Elavarasan, R., Ramanathan, V., Anandan, R. K., Subramaniam, U., … & Hossain, E. A Comprehensive Review on Sustainable Aspects of Big Data Analytics for the Smart Grid. Sustainability, 2021; 13(23), 13322.
Ahmad, T., Zhang, D., Huang, C., Zhang, H., Dai, N., Song, Y., & Chen, H. Artificial intelligence in sustainable energy industry: Status Quo, challenges and opportunities. Journal of Cleaner Production, 2021; 289, 125834.
Wang, G., Chao, Y., Cao, Y., Jiang, T., Han, W., & Chen, Z. A comprehensive review of research works based on evolutionary game theory for sustainable energy development. Energy Reports, 2022; 8, 114-136.
Iranmehr H., Modeling the Price of Emergency Power Transmission Lines in the Reserve Market Due to the Influence of Renewable Energies, Frontiers in Energy Research, 9, 2022
Farmanbar, M., Parham, K., Arild, Ø., & Rong, C. A widespread review of smart grids towards smart cities. Energies, 2019; 12(23), 4484.
Quartier, N., Crespo, A. J., Domínguez, J. M., Stratigaki, V., & Troch, P. Efficient response of an onshore Oscillating Water Column Wave Energy Converter using a one-phase SPH model coupled with a multiphysics library. Applied Ocean Research, 2021; 115, 102856.
Mahmoodi, K., Nepomuceno, E., & Razminia, A. Wave excitation force forecasting using neural networks. Energy, 2022; 247, 123322.
Wang, H., Alattas, K.A., 2022. Comprehensive review of load forecasting with emphasis on intelligent computing approaches. Energy Reports, 8, pp.13189-13198.
Clemente, D., Rosa-Santos, P., & Taveira-Pinto, F. On the potential synergies and applications of wave energy converters: A review. Renewable and Sustainable Energy Reviews, 2021; 135, 110162.
Felix, A., V. Hernández-Fontes, J., Lithgow, D., Mendoza, E., Posada, G., Ring, M., & Silva, R. Wave energy in tropical regions: deployment challenges, environmental and social perspectives. Journal of Marine Science and Engineering, 2019; 7(7), 219.
Farrok, O., Ahmed, K., Tahlil, A. D., Farah, M. M., Kiran, M. R., & Islam, M. R. Electrical power generation from the oceanic wave for sustainable advancement in renewable energy technologies. Sustainability, 2020; 12(6), 2178.
Guo, B., & Ringwood, J. V. A review of wave energy technology from a research and commercial perspective. IET Renewable Power Generation, 2021; 15(14), 3065-3090.
López-Ruiz, A., Bergillos, R. J., Lira-Loarca, A., & Ortega-Sánchez, M. A methodology for the long-term simulation and uncertainty analysis of the operational lifetime performance of wave energy converter arrays. Energy, 2018; 153, 126-135.
Safarian, S., Saryazdi, S. M. E., Unnthorsson, R., & Richter, C. Artificial neural network integrated with thermodynamic equilibrium modeling of downdraft biomass gasification-power production plant. Energy, 2020; 213, 118800.
Kushwah, S. An oscillating water column (OWC): the wave energy converter. Journal of The Institution of Engineers (India): Series C, 2021; 102(5), 1311-1317.
Pap, J., Mako, C., Illessy, M., Kis, N., 2022. Modeling Organizational Performance with Machine Learning. Journal of Open Innovation: Technology, Market, and Complexity, 8(4), p.177.
Pap, J., Mako, C., Illessy, M., Dedaj, Z., Ardabili, S., Torok, B., 2022. Correlation Analysis of Factors Affecting Firm Performance and Employees Wellbeing: Application of Advanced Machine Learning Analysis. Algorithms, 15(9), p.300.
Alanazi, A., 2022. Determining Optimal Power Flow Solutions Using New Adaptive Gaussian TLBO Method. Applied Sciences, 12(16), p.7959.
Shakibjoo, A.D., Moradzadeh, M., Din, S.U., 2021. Optimized Type-2 Fuzzy Frequency Control for Multi-Area Power Systems. IEEE access, 10, pp.6989-7002.
Zhang, G., 2021. Solar radiation estimation in different climates with meteorological variables using Bayesian model averaging and new soft computing models. Energy Reports, 7, pp.8973-8996.
Cao, Y., Raise, A., Mohammadzadeh, A., Rathinasamy, S., 2021. Deep learned recurrent type-3 fuzzy system: Application for renewable energy modeling/prediction. Energy Reports, 7, pp.8115-8127.
Tavoosi, J., Suratgar, A.A., Menhaj, M.B., 2021. Modeling renewable energy systems by a self-evolving nonlinear consequent part recurrent type-2 fuzzy system for power prediction. Sustainability, 13(6), p.3301.
Bourouis, S., Band, S.S., 2022. Meta-Heuristic Algorithm-Tuned Neural Network for Breast Cancer Diagnosis Using Ultrasound Images. Frontiers in Oncology, 12, p.834028.
Mosavi, A.H., Mohammadzadeh, A., Rathinasamy, S., Zhang, C., Reuter, U., Levente, K. and Adeli, H., 2022. Deep learning fuzzy immersion and invariance control for type-I diabetes. Computers in Biology and Medicine, 149, p.105975.
Almutairi, K., Algarni, S., Alqahtani, T., Moayedi, H., 2022. A TLBO-Tuned Neural Processor for Predicting Heating Load in Residential Buildings. Sustainability, 14(10), p.5924.
Ahmad, Z., Zhong, H., 2020. Machine learning modeling of aerobic biodegradation for azo dyes and hexavalent chromium. Mathematics, 8(6), p.913.
Mosavi, A., Shokri, M., Mansor, Z., Qasem, S.N., Band, S.S. and Mohammadzadeh, A., 2020. Machine learning for modeling the singular multi-pantograph equations. Entropy, 22(9), p.1041.
Ardabili, S., 2019, September. Deep learning and machine learning in hydrological processes climate change and earth systems a systematic review. In International conference on global research and education (pp. 52-62). Springer, Cham.
Moayedi, H., (2021). Suggesting a stochastic fractal search paradigm in combination with artificial neural network for early prediction of cooling load in residential buildings. Energies, 14(6), 1649.
Rezakazemi, M., et al., 2019. ANFIS pattern for molecular membranes separation optimization. Journal of Molecular Liquids, 274, pp.470-476.
Mosavi, A., Faghan, Y., Ghamisi, P., Duan, P., Ardabili, S.F., Salwana, E. and Band, S.S., 2020. Comprehensive review of deep reinforcement learning methods and applications in economics. Mathematics, 8(10), p.1640.
Samadianfard, S., Jarhan, S., Salwana, E., 2019. Support vector regression integrated with fruit fly optimization algorithm for river flow forecasting in Lake Urmia Basin. Water, 11(9), p.1934.
Moayedi, H., (2021). Double-target based neural networks in predicting energy consumption in residential buildings. Energies, 14(5), 1331.
Mohammadzadeh S, D., Kazemi, S.F., 2019. Prediction of compression index of fine-grained soils using a gene expression programming model. Infrastructures, 4(2), p.26.
Karballaeezadeh, N., Mohammadzadeh S, D., Shamshirband, S., Hajikhodaverdikhan, P., 2019. Prediction of remaining service life of pavement using an optimized support vector machine (case study of Semnan–Firuzkuh road). Engineering Applications of Computational Fluid Mechanics, 13(1), pp.188-198.
Rezaei, M. Et al., (2022). Adaptation of A Real-Time Deep Learning Approach with An Analog Fault Detection Technique for Reliability Forecasting of Capacitor Banks Used in Mobile Vehicles. IEEE Access v. 21 pp. 89-99.
Khakian, R., et al., (2020). Modeling nearly zero energy buildings for sustainable development in rural areas. Energies, 13(10), 2593.
This paper presents the results of tests on the suitability of designed heads (impellers) for aluminum refining. The research was carried out on a physical model of the URO-200, followed by numerical simulations in the FLOW 3D program. Four design variants of impellers were used in the study. The degree of dispersion of the gas phase in the model liquid was used as a criterion for evaluating the performance of each solution using different process parameters, i.e., gas flow rate and impeller speed. Afterward, numerical simulations in Flow 3D software were conducted for the best solution. These simulations confirmed the results obtained with the water model and verified them.
Constantly increasing requirements concerning metallurgical purity in terms of hydrogen content and nonmetallic inclusions make casting manufacturers use effective refining techniques. The answer to this demand is the implementation of the aluminum refining technique making use of a rotor with an original design guaranteeing efficient refining [1,2,3,4]. The main task of the impeller (rotor) is to reduce the contamination of liquid metal (primary and recycled aluminum) with hydrogen and nonmetallic inclusions. An inert gas, mainly argon or a mixture of gases, is introduced through the rotor into the liquid metal to bring both hydrogen and nonmetallic inclusions to the metal surface through the flotation process. Appropriately and uniformly distributed gas bubbles in the liquid metal guarantee achieving the assumed level of contaminant removal economically. A very important factor in deciding about the obtained degassing effect is the optimal rotor design [5,6,7,8]. Thanks to the appropriate geometry of the rotor, gas bubbles introduced into the liquid metal are split into smaller ones, and the spinning movement of the rotor distributes them throughout the volume of the liquid metal bath. In this solution impurities in the liquid metal are removed both in the volume and from the upper surface of the metal. With a well-designed impeller, the costs of refining aluminum and its alloys can be lowered thanks to the reduced inert gas and energy consumption (optimal selection of rotor rotational speed). Shorter processing time and a high degree of dehydrogenation decrease the formation of dross on the metal surface (waste). A bigger produced dross leads to bigger process losses. Consequently, this means that the choice of rotor geometry has an indirect impact on the degree to which the generated waste is reduced [9,10].
Another equally important factor is the selection of process parameters such as gas flow rate and rotor speed [11,12]. A well-designed gas injection system for liquid metal meets two key requirements; it causes rapid mixing of the liquid metal to maintain a uniform temperature throughout the volume and during the entire process, to produce a chemically homogeneous metal composition. This solution ensures effective degassing of the metal bath. Therefore, the shape of the rotor, the arrangement of the nozzles, and their number are significant design parameters that guarantee the optimum course of the refining process. It is equally important to complete the mixing of the metal bath in a relatively short time, as this considerably shortens the refining process and, consequently, reduces the process costs. Another important criterion conditioning the implementation of the developed rotor is the generation of fine diffused gas bubbles which are distributed throughout the metal volume, and whose residence time will be sufficient for the bubbles to collide and adsorb the contaminants. The process of bubble formation by the spinning rotors differs from that in the nozzles or porous molders. In the case of a spinning rotor, the shear force generated by the rotor motion splits the bubbles into smaller ones. Here, the rotational speed, mixing force, surface tension, and fluid density have a key effect on the bubble size. The velocity of the bubbles, which depends mainly on their size and shape, determines their residence time in the reactor and is, therefore, very important for the refining process, especially since gas bubbles in liquid aluminum may remain steady only below a certain size [13,14,15].
The impeller designs presented in the article were developed to improve the efficiency of the process and reduce its costs. The impellers used so far have a complicated structure and are very pricey. The success of the conducted research will allow small companies to become independent of external supplies through the possibility of making simple and effective impellers on their own. The developed structures were tested on the water model. The results of this study can be considered as pilot.
Rotors were realized with the SolidWorks computer design technique and a 3D printer. The developed designs were tested on a water model. Afterward, the solution with the most advantageous refining parameters was selected and subjected to calculations with the Flow3D package. As a result, an impeller was designed for aluminum refining. Its principal lies in an even distribution of gas bubbles in the entire volume of liquid metal, with the largest possible participation of the bubble surface, without disturbing the metal surface. This procedure guarantees the removal of gaseous, as well as metallic and nonmetallic, impurities.
2.1. Rotor Designs
The developed impeller constructions, shown in Figure 1, Figure 2, Figure 3 and Figure 4, were printed on a 3D printer using the PLA (polylactide) material. The impeller design models differ in their shape and the number of holes through which the inert gas flows. Figure 1, Figure 2 and Figure 3 show the same impeller model but with a different number of gas outlets. The arrangement of four, eight, and 12 outlet holes was adopted in the developed design. A triangle-shaped structure equipped with three gas outlet holes is presented in Figure 4.
A schematic of the water model of reactor URO 200.
The URO 200 reactor can be classified as a cyclic reactor. The main element of the device is a rotor, which ends the impeller. The whole system is attached to a shaft via which the refining gas is supplied. Then, the shaft with the rotor is immersed in the liquid metal in the melting pot or the furnace chamber. In URO 200 reactors, the refining process lasts 600 s (10 min), the gas flow rate that can be obtained ranges from 5 to 20 dm3·min−1, and the speed at which the rotor can move is 0 to 400 rpm. The permissible quantity of liquid metal for barbotage refining is 300 kg or 700 kg [8,16,17]. The URO 200 has several design solutions which improve operation and can be adapted to the existing equipment in the foundry. These solutions include the following [8,16]:
URO-200XR—used for small crucible furnaces, the capacity of which does not exceed 250 kg, with no control system and no control of the refining process.
URO-200SA—used to service several crucible furnaces of capacity from 250 kg to 700 kg, fully automated and equipped with a mechanical rotor lift.
URO-200KA—used for refining processes in crucible furnaces and allows refining in a ladle. The process is fully automated, with a hydraulic rotor lift.
URO-200KX—a combination of the XR and KA models, designed for the ladle refining process. Additionally, refining in heated crucibles is possible. The unit is equipped with a manual hydraulic rotor lift.
URO-200PA—designed to cooperate with induction or crucible furnaces or intermediate chambers, the capacity of which does not exceed one ton. This unit is an integral part of the furnace. The rotor lift is equipped with a screw drive.
Studies making use of a physical model can be associated with the observation of the flow and circulation of gas bubbles. They require meeting several criteria regarding the similarity of the process and the object characteristics. The similarity conditions mainly include geometric, mechanical, chemical, thermal, and kinetic parameters. During simulation of aluminum refining with inert gas, it is necessary to maintain the geometric similarity between the model and the real object, as well as the similarity related to the flow of liquid metal and gas (hydrodynamic similarity). These quantities are characterized by the Reynolds, Weber, and Froude numbers. The Froude number is the most important parameter characterizing the process, its magnitude is the same for the physical model and the real object. Water was used as the medium in the physical modeling. The factors influencing the choice of water are its availability, relatively low cost, and kinematic viscosity at room temperature, which is very close to that of liquid aluminum.
The physical model studies focused on the flow of inert gas in the form of gas bubbles with varying degrees of dispersion, particularly with respect to some flow patterns such as flow in columns and geysers, as well as disturbance of the metal surface. The most important refining parameters are gas flow rate and rotor speed. The barbotage refining studies for the developed impeller (variants B4, B8, B12, and RT3) designs were conducted for the following process parameters:
Rotor speed: 200, 300, 400, and 500 rpm,
Ideal gas flow: 10, 20, and 30 dm3·min−1,
Temperature: 293 K (20 °C).
These studies were aimed at determining the most favorable variants of impellers, which were then verified using the numerical modeling methods in the Flow-3D program.
2.3. Numerical Simulations with Flow-3D Program
Testing different rotor impellers using a physical model allows for observing the phenomena taking place while refining. This is a very important step when testing new design solutions without using expensive industrial trials. Another solution is modeling by means of commercial simulation programs such as ANSYS Fluent or Flow-3D [18,19]. Unlike studies on a physical model, in a computer program, the parameters of the refining process and the object itself, including the impeller design, can be easily modified. The simulations were performed with the Flow-3D program version 12.03.02. A three-dimensional system with the same dimensions as in the physical modeling was used in the calculations. The isothermal flow of liquid–gas bubbles was analyzed. As in the physical model, three speeds were adopted in the numerical tests: 200, 300, and 500 rpm. During the initial phase of the simulations, the velocity field around the rotor generated an appropriate direction of motion for the newly produced bubbles. When the required speed was reached, the generation of randomly distributed bubbles around the rotor was started at a rate of 2000 per second. Table 1 lists the most important simulation parameters.
In the case of the CFD analysis, the numerical solutions require great care when generating the computational mesh. Therefore, computational mesh tests were performed prior to the CFD calculations. The effect of mesh density was evaluated by taking into account the velocity of water in the tested object on the measurement line A (height of 0.065 m from the bottom) in a characteristic cross-section passing through the object axis (see Figure 6). The mesh contained 3,207,600, 6,311,981, 7,889,512, 11,569,230, and 14,115,049 cells.
The velocity of the water depending on the size of the computational grid.
The quality of the generated computational meshes was checked using the criterion skewness angle QEAS [18]. This criterion is described by the following relationship:
QEAS=max{βmax−βeq180−βeq,βeq−βminβeq},
(1)
where βmax, βmin are the maximal and minimal angles (in degrees) between the edges of the cell, and βeq is the angle corresponding to an ideal cell, which for cubic cells is 90°.
Normalized in the interval [0;1], the value of QEAS should not exceed 0.75, which identifies the permissible skewness angle of the generated mesh. For the computed meshes, this value was equal to 0.55–0.65.
Moreover, when generating the computational grids in the studied facility, they were compacted in the areas of the highest gradients of the calculated values, where higher turbulence is to be expected (near the impeller). The obtained results of water velocity in the studied object at constant gas flow rate are shown in Figure 6.
The analysis of the obtained water velocity distributions (see Figure 6) along the line inside the object revealed that, with the density of the grid of nodal points, the velocity changed and its changes for the test cases of 7,889,512, 11,569,230, and 14,115,049 were insignificant. Therefore, it was assumed that a grid containing not less than 7,900,000 (7,889,512) cells would not affect the result of CFD calculations.
A single-block mesh of regular cells with a size of 0.0034 m was used in the numerical calculations. The total number of cells was approximately 7,900,000 (7,889,512). This grid resolution (see Figure 7) allowed the geometry of the system to be properly represented, maintaining acceptable computation time (about 3 days on a workstation with 2× CPU and 12 computing cores).
Structured equidistant mesh used in numerical calculations: (a) mesh with smoothed, surface cells (the so-called FAVOR method) used in Flow-3D; (b) visualization of the applied mesh resolution.
The calculations were conducted with an explicit scheme. The timestep was selected by the program automatically and controlled by stability and convergence. From the moment of the initial velocity field generation (start of particle generation), it was 0.0001 s.
When modeling the degassing process, three fluids are present in the system: water, gas supplied through the rotor head (impeller), and the surrounding air. Modeling such a multiphase flow is a numerically very complex issue. The necessity to overcome the liquid backpressure by the gas flowing out from the impeller leads to the formation of numerical instabilities in the volume of fluid (VOF)-based approach used by Flow-3D software. Therefore, a mixed description of the analyzed flow was used here. In this case, water was treated as a continuous medium, while, in the case of gas bubbles, the discrete phase model (DPM) model was applied. The way in which the air surrounding the system was taken into account is later described in detail.
The following additional assumptions were made in the modeling:
—The liquid phase was considered as an incompressible Newtonian fluid.
—The effect of chemical reactions during the refining process was neglected.
—The composition of each phase (gas and liquid) was considered homogeneous; therefore, the viscosity and surface tension were set as constants.
—Only full turbulence existed in the liquid, and the effect of molecular viscosity was neglected.
—The gas bubbles were shaped as perfect spheres.
—The mutual interaction between gas bubbles (particles) was neglected.
2.3.1. Modeling of Liquid Flow
The motion of the real fluid (continuous medium) is described by the Navier–Stokes Equation [20].
dudt=−1ρ∇p+ν∇2u+13ν∇(∇⋅ u)+F,
(2)
where du/dt is the time derivative, u is the velocity vector, t is the time, and F is the term accounting for external forces including gravity (unit components denoted by X, Y, Z).
In the simulations, the fluid flow was assumed to be incompressible, in which case the following equation is applicable:
∂u∂t+(u⋅∇)u=−1ρ∇p+ν∇2u+F.
(3)
Due to the large range of liquid velocities during flows, the turbulence formation process was included in the modeling. For this purpose, the k–ε model turbulence kinetic energy k and turbulence dissipation ε were the target parameters, as expressed by the following equations [21]:
where ρ is the gas density, σκ and σε are the Prandtl turbulence numbers, k and ε are constants of 1.0 and 1.3, and Gk and Gb are the kinetic energy of turbulence generated by the average velocity and buoyancy, respectively.
As mentioned earlier, there are two gas phases in the considered problem. In addition to the gas bubbles, which are treated here as particles, there is also air, which surrounds the system. The boundary of phase separation is in this case the free surface of the water. The shape of the free surface can change as a result of the forming velocity field in the liquid. Therefore, it is necessary to use an appropriate approach to free surface tracking. The most commonly used concept in liquid–gas flow modeling is the volume of fluid (VOF) method [22,23], and Flow-3D uses a modified version of this method called TrueVOF. It introduces the concept of the volume fraction of the liquid phase fl. This parameter can be used for classifying the cells of a discrete grid into areas filled with liquid phase (fl = 1), gaseous phase, or empty cells (fl = 0) and those through which the phase separation boundary (fl ∈ (0, 1)) passes (free surface). To determine the local variations of the liquid phase fraction, it is necessary to solve the following continuity equation:
dfldt=0.
(6)
Then, the fluid parameters in the region of coexistence of the two phases (the so-called interface) depend on the volume fraction of each phase.
ρ=flρl+(1−fl)ρg,
(7)
ν=flνl+(1−fl)νg,
(8)
where indices l and g refer to the liquid and gaseous phases, respectively.
The parameter of fluid velocity in cells containing both phases is also determined in the same way.
u=flul+(1−fl)ug.
(9)
Since the processes taking place in the surrounding air can be omitted, to speed up the calculations, a single-phase, free-surface model was used. This means that no calculations were performed in the gas cells (they were treated as empty cells). The liquid could fill them freely, and the air surrounding the system was considered by the atmospheric pressure exerted on the free surface. This approach is often used in modeling foundry and metallurgical processes [24].
2.3.2. Modeling of Gas Bubble Flow
As stated, a particle model was used to model bubble flow. Spherical particles (gas bubbles) of a given size were randomly generated in the area marked with green in Figure 7b. In the simulations, the gas bubbles were assumed to have diameters of 0.016 and 0.02 m corresponding to the gas flow rates of 10 and 30 dm3·min−1, respectively.
Experimental studies have shown that, as a result of turbulent fluid motion, some of the bubbles may burst, leading to the formation of smaller bubbles, although merging of bubbles into larger groupings may also occur. Therefore, to be able to observe the behavior of bubbles of different sizes (diameter), the calculations generated two additional particle types with diameters twice smaller and twice larger, respectively. The proportion of each species in the system was set to 33.33% (Table 2).
The velocity of the particle results from the generated velocity field (calculated from Equation (3) in the liquid ul around it and its velocity resulting from the buoyancy force ub. The effect of particle radius r on the terminal velocity associated with buoyancy force can be determined according to Stokes’ law.
ub=29 (ρg−ρl)μlgr2,
(10)
where g is the acceleration (9.81).
The DPM model was used for modeling the two-phase (water–air) flow. In this model, the fluid (water) is treated as a continuous phase and described by the Navier–Stokes equation, while gas bubbles are particles flowing in the model fluid (discrete phase). The trajectories of each bubble in the DPM system are calculated at each timestep taking into account the mass forces acting on it. Table 3 characterizes the DPM model used in our own research [18].
Table 3
Characteristic of the DPM model.
Method
Equations
Euler–Lagrange
Balance equation: dugdt=FD(u−ug)+g(ϱg−ϱ)ϱg+F. FD (u − up) denotes the drag forces per mass unit of a bubble, and the expression for the drag coefficient FD is of the form FD=18μCDReϱ⋅gd2g24. The relative Reynolds number has the form Re≡ρdg|ug−u|μ. On the other hand, the force resulting from the additional acceleration of the model fluid has the form F=12dρdtρg(u−ug), where ug is the gas bubble velocity, u is the liquid velocity, dg is the bubble diameter, and CD is the drag coefficient.
3.1. Calculations of Power and Mixing Time by the Flowing Gas Bubbles
One of the most important parameters of refining with a rotor is the mixing power induced by the spinning rotor and the outflowing gas bubbles (via impeller). The mixing power of liquid metal in a ladle of height (h) by gas injection can be determined from the following relation [15]:
pgVm=ρ⋅g⋅uB,
(11)
where pg is the mixing power, Vm is the volume of liquid metal in the reactor, ρ is the density of liquid aluminum, and uB is the average speed of bubbles, given below.
uB=n⋅R⋅TAc⋅Pm⋅t,
(12)
where n is the number of gas moles, R is the gas constant (8.314), Ac is the cross-sectional area of the reactor vessel, T is the temperature of liquid aluminum in the reactor, and Pm is the pressure at the middle tank level. The pressure at the middle level of the tank is calculated by a function of the mean logarithmic difference.
Pm=(Pa+ρ⋅g⋅h)−Paln(Pa+ρ⋅g⋅h)Pa,
(13)
where Pa is the atmospheric pressure, and h is the the height of metal in the reactor.
Themelis and Goyal [25] developed a model for calculating mixing power delivered by gas injection.
pg=2Q⋅R⋅T⋅ln(1+m⋅ρ⋅g⋅hP),
(14)
where Q is the gas flow, and m is the mass of liquid metal.
Zhang [26] proposed a model taking into account the temperature difference between gas and alloy (metal).
pg=QRTgVm[ln(1+ρ⋅g⋅hPa)+(1−TTg)],
(15)
where Tg is the gas temperature at the entry point.
Data for calculating the mixing power resulting from inert gas injection into liquid aluminum are given below in Table 4. The design parameters were adopted for the model, the parameters of which are shown in Figure 5.
Table 4
Data for calculating mixing power introduced by an inert gas.
Table 5 presents the results of mixing power calculations according to the models of Themelis and Goyal and of Zhang for inert gas flows of 10, 20, and 30 dm3·min−1. The obtained calculation results significantly differed from each other. The difference was an order of magnitude, which indicates that the model is highly inaccurate without considering the temperature of the injected gas. Moreover, the calculations apply to the case when the mixing was performed only by the flowing gas bubbles, without using a rotor, which is a great simplification of the phenomenon.
Table 5
Mixing power calculated from mathematical models.
Mathematical Model
Mixing Power (W·t−1) for a Given Inert Gas Flow (dm3·min−1)
The mixing time is defined as the time required to achieve 95% complete mixing of liquid metal in the ladle [27,28,29,30]. Table 6 groups together equations for the mixing time according to the models.
Figure 8 and Figure 9 show the mixing time as a function of gas flow rate for various heights of the liquid column in the ladle and mixing power values.
Mixing time as a function of mixing power (Szekly model).
3.2. Determining the Bubble Size
The mechanisms controlling bubble size and mass transfer in an alloy undergoing refining are complex. Strong mixing conditions in the reactor promote impurity mass transfer. In the case of a spinning rotor, the shear force generated by the rotor motion separates the bubbles into smaller bubbles. Rotational speed, mixing force, surface tension, and liquid density have a strong influence on the bubble size. To characterize the kinetic state of the refining process, parameters k and A were introduced. Parameters k, A, and uB can be calculated using the below equations [33].
k=2D⋅uBdB⋅π−−−−−−√,
(16)
A=6Q⋅hdB⋅uB,
(17)
uB=1.02g⋅dB,−−−−−√
(18)
where D is the diffusion coefficient, and dB is the bubble diameter.
After substituting appropriate values, we get
dB=3.03×104(πD)−2/5g−1/5h4/5Q0.344N−1.48.
(19)
According to the last equation, the size of the gas bubble decreases with the increasing rotational speed (see Figure 10).
Effect of rotational speed on the bubble diameter.
In a flow of given turbulence intensity, the diameter of the bubble does not exceed the maximum size dmax, which is inversely proportional to the rate of kinetic energy dissipation in a viscous flow ε. The size of the gas bubble diameter as a function of the mixing energy, also considering the Weber number and the mixing energy in the negative power, can be determined from the following equations [31,34]:
The first stage of experiments (using the URO-200 water model) included conducting experiments with impellers equipped with four, eight, and 12 gas outlets (variants B4, B8, B12). The tests were carried out for different process parameters. Selected results for these experiments are presented in Figure 11, Figure 12, Figure 13 and Figure 14.
Impeller variant B4—gas bubbles dispersion registered for a gas flow rate of 10 dm3·min−1 and rotor speed of (a) 200, (b) 300, (c) 400, and (d) 500 rpm.
Impeller variant B8—gas bubbles dispersion registered for a gas flow rate of 10 dm3·min−1 and rotor speed of (a) 200, (b) 300, (c) 400, and (d) 500 rpm.
Gas bubble dispersion registered for different processing parameters (impeller variant RT3).
The analysis of the refining variants presented in Figure 11, Figure 12, Figure 13 and Figure 14 reveals that the proposed impellers design model is not useful for the aluminum refining process. The number of gas outlet orifices, rotational speed, and flow did not affect the refining efficiency. In all the variants shown in the figures, very poor dispersion of gas bubbles was observed in the object. The gas bubble flow had a columnar character, and so-called dead zones, i.e., areas where no inert gas bubbles are present, were visible in the analyzed object. Such dead zones were located in the bottom and side zones of the ladle, while the flow of bubbles occurred near the turning rotor. Another negative phenomenon observed was a significant agitation of the water surface due to excessive (rotational) rotor speed and gas flow (see Figure 13, cases 20; 400, 30; 300, 30; 400, and 30; 500).
Research results for a ‘red triangle’ impeller equipped with three gas supply orifices (variant RT3) are presented in Figure 14.
In this impeller design, a uniform degree of bubble dispersion in the entire volume of the modeling fluid was achieved for most cases presented (see Figure 14). In all tested variants, single bubbles were observed in the area of the water surface in the vessel. For variants 20; 200, 30; 200, and 20; 300 shown in Figure 14, the bubble dispersion results were the worst as the so-called dead zones were identified in the area near the bottom and sidewalls of the vessel, which disqualifies these work parameters for further applications. Interestingly, areas where swirls and gas bubble chains formed were identified only for the inert gas flows of 20 and 30 dm3·min−1 and 200 rpm in the analyzed model. This means that the presented model had the best performance in terms of dispersion of gas bubbles in the model liquid. Its design with sharp edges also differed from previously analyzed models, which is beneficial for gas bubble dispersion, but may interfere with its suitability in industrial conditions due to possible premature wear.
3.4. Qualitative Comparison of Research Results (CFD and Physical Model)
The analysis (physical modeling) revealed that the best mixing efficiency results were obtained with the RT3 impeller variant. Therefore, numerical calculations were carried out for the impeller model with three outlet orifices (variant RT3). The CFD results are presented in Figure 15 and Figure 16.
Simulation results of the impeller RT3, for given flows and rotational speeds after a time of 1 s: simulation variants (a) A, (b) B, (c) C, (d) D, (e) E, and (f) F.
Simulation results of the impeller RT3, for given flows and rotational speeds after a time of 5.4 s.: simulation variants (a) A, (b) B, (c) C, (d) D, (e) E, and (f) F.
CFD results are presented for all analyzed variants (impeller RT3) at two selected calculation timesteps of 1 and 5.40 s. They show the velocity field of the medium (water) and the dispersion of gas bubbles.
Figure 15 shows the initial refining phase after 1 s of the process. In this case, the gas bubble formation and flow were observed in an area close to contact with the rotor. Figure 16 shows the phase when the dispersion and flow of gas bubbles were advanced in the reactor area of the URO-200 model.
The quantitative evaluation of the obtained results of physical and numerical model tests was based on the comparison of the degree of gas dispersion in the model liquid. The degree of gas bubble dispersion in the volume of the model liquid and the areas of strong turbulent zones formation were evaluated during the analysis of the results of visualization and numerical simulations. These two effects sufficiently characterize the required course of the process from the physical point of view. The known scheme of the below description was adopted as a basic criterion for the evaluation of the degree of dispersion of gas bubbles in the model liquid.
Minimal dispersion—single bubbles ascending in the region of their formation along the ladle axis; lack of mixing in the whole bath volume.
Accurate dispersion—single and well-mixed bubbles ascending toward the bath mirror in the region of the ladle axis; no dispersion near the walls and in the lower part of the ladle.
Uniform dispersion—most desirable; very good mixing of fine bubbles with model liquid.
Excessive dispersion—bubbles join together to form chains; large turbulence zones; uneven flow of gas.
The numerical simulation results give a good agreement with the experiments performed with the physical model. For all studied variants (used process parameters), the single bubbles were observed in the area of water surface in the vessel. For variants presented in Figure 13 (200 rpm, gas flow 20 and dm3·min−1) and relevant examples in numerical simulation Figure 16, the worst bubble dispersion results were obtained because the dead zones were identified in the area near the bottom and sidewalls of the vessel, which disqualifies these work parameters for further use. The areas where swirls and gas bubble chains formed were identified only for the inert gas flows of 20 and 30 dm3·min−1 and 200 rpm in the analyzed model (physical model). This means that the presented impeller model had the best performance in terms of dispersion of gas bubbles in the model liquid. The worst bubble dispersion results were obtained because the dead zones were identified in the area near the bottom and side walls of the vessel, which disqualifies these work parameters for further use.
Figure 17 presents exemplary results of model tests (CFD and physical model) with marked gas bubble dispersion zones. All variants of tests were analogously compared, and this comparison allowed validating the numerical model.
Compilations of model research results (CFD and physical): A—single gas bubbles formed on the surface of the modeling liquid, B—excessive formation of gas chains and swirls, C—uniform distribution of gas bubbles in the entire volume of the tank, and D—dead zones without gas bubbles, no dispersion. (a) Variant B; (b) variant F.
It should be mentioned here that, in numerical simulations, it is necessary to make certain assumptions and simplifications. The calculations assumed three particle size classes (Table 2), which represent the different gas bubbles that form due to different gas flow rates. The maximum number of particles/bubbles (Table 1) generated was assumed in advance and related to the computational capabilities of the computer. Too many particles can also make it difficult to visualize and analyze the results. The size of the particles, of course, affects their behavior during simulation, while, in the figures provided in the article, the bubbles are represented by spheres (visualization of the results) of the same size. Please note that, due to the adopted Lagrangian–Eulerian approach, the simulation did not take into account phenomena such as bubble collapse or fusion. However, the obtained results allow a comprehensive analysis of the behavior of gas bubbles in the system under consideration.
The comparative analysis of the visualization (quantitative) results obtained with the water model and CFD simulations (see Figure 17) generated a sufficient agreement from the point of view of the trends. A precise quantitative evaluation is difficult to perform because of the lack of a refraction compensating system in the water model. Furthermore, in numerical simulations, it is not possible to determine the geometry of the forming gas bubbles and their interaction with each other as opposed to the visualization in the water model. The use of both research methods is complementary. Thus, a direct comparison of images obtained by the two methods requires appropriate interpretation. However, such an assessment gives the possibility to qualitatively determine the types of the present gas bubble dispersion, thus ultimately validating the CFD results with the water model.
A summary of the visualization results for impellers RT3, i.e., analysis of the occurring gas bubble dispersion types, is presented in Table 8.
Table 8
Summary of visualization results (impeller RT3)—different types of gas bubble dispersion.
Tests carried out for impeller RT3 confirmed the high efficiency of gas bubble distribution in the volume of the tested object at a low inert gas flow rate of 10 dm3·min−1. The most optimal variant was variant B (300 rpm, 10 dm3·min−1). However, the other variants A and C (gas flow rate 10 dm3·min−1) seemed to be favorable for this type of impeller and are recommended for further testing. The above process parameters will be analyzed in detail in a quantitative analysis to be performed on the basis of the obtained efficiency curves of the degassing process (oxygen removal). This analysis will give an unambiguous answer as to which process parameters are the most optimal for this type of impeller; the results are planned for publication in the next article.
It should also be noted here that the high agreement between the results of numerical calculations and physical modelling prompts a conclusion that the proposed approach to the simulation of a degassing process which consists of a single-phase flow model with a free surface and a particle flow model is appropriate. The simulation results enable us to understand how the velocity field in the fluid is formed and to analyze the distribution of gas bubbles in the system. The simulations in Flow-3D software can, therefore, be useful for both the design of the impeller geometry and the selection of process parameters.
The results of experiments carried out on the physical model of the device for the simulation of barbotage refining of aluminum revealed that the worst results in terms of distribution and dispersion of gas bubbles in the studied object were obtained for the black impellers variants B4, B8, and B12 (multi-orifice impellers—four, eight, and 12 outlet holes, respectively).
In this case, the control of flow, speed, and number of gas exit orifices did not improve the process efficiency, and the developed design did not meet the criteria for industrial tests. In the case of the ‘red triangle’ impeller (variant RT3), uniform gas bubble dispersion was achieved throughout the volume of the modeling fluid for most of the tested variants. The worst bubble dispersion results due to the occurrence of the so-called dead zones in the area near the bottom and sidewalls of the vessel were obtained for the flow variants of 20 dm3·min−1 and 200 rpm and 30 dm3·min−1 and 200 rpm. For the analyzed model, areas where swirls and gas bubble chains were formed were found only for the inert gas flow of 20 and 30 dm3·min−1 and 200 rpm. The model impeller (variant RT3) had the best performance compared to the previously presented impellers in terms of dispersion of gas bubbles in the model liquid. Moreover, its design differed from previously presented models because of its sharp edges. This can be advantageous for gas bubble dispersion, but may negatively affect its suitability in industrial conditions due to premature wearing.
The CFD simulation results confirmed the results obtained from the experiments performed on the physical model. The numerical simulation of the operation of the ‘red triangle’ impeller model (using Flow-3D software) gave good agreement with the experiments performed on the physical model. This means that the presented model impeller, as compared to other (analyzed) designs, had the best performance in terms of gas bubble dispersion in the model liquid.
In further work, the developed numerical model is planned to be used for CFD simulations of the gas bubble distribution process taking into account physicochemical parameters of liquid aluminum based on industrial tests. Consequently, the obtained results may be implemented in production practice.
This paper was created with the financial support grants from the AGH-UST, Faculty of Foundry Engineering, Poland (16.16.170.654 and 11/990/BK_22/0083) for the Faculty of Materials Engineering, Silesian University of Technology, Poland.
Conceptualization, K.K. and D.K.; methodology, J.P. and T.M.; validation, M.S. and S.G.; formal analysis, D.K. and T.M.; investigation, J.P., K.K. and S.G.; resources, M.S., J.P. and K.K.; writing—original draft preparation, D.K. and T.M.; writing—review and editing, D.K. and T.M.; visualization, J.P., K.K. and S.G.; supervision, D.K.; funding acquisition, D.K. and T.M. All authors have read and agreed to the published version of the manuscript.
1. Zhang L., Xuewei L., Torgerson A.T., Long M. Removal of Impurity Elements from Molten Aluminium: A Review. Miner. Process. Extr. Metall. Rev. 2011;32:150–228. doi: 10.1080/08827508.2010.483396. [CrossRef] [Google Scholar]
2. Saternus M. Impurities of liquid aluminium-methods on their estimation and removal. Met. Form. 2015;23:115–132. [Google Scholar]
3. Żak P.L., Kalisz D., Lelito J., Gracz B., Szucki M., Suchy J.S. Modelling of non-metallic particle motion process in foundry alloys. Metalurgija. 2015;54:357–360. [Google Scholar]
4. Kalisz D., Kuglin K. Efficiency of aluminum oxide inclusions rmoval from liquid steel as a result of collisions and agglomeration on ceramic filters. Arch. Foundry Eng. 2020;20:43–48. [Google Scholar]
5. Kuglin K., Kalisz D. Evaluation of the usefulness of rotors for aluminium refining. IOP Conf. Ser. Mater. Sci. Eng. 2021;1178:012036. doi: 10.1088/1757-899X/1178/1/012036. [CrossRef] [Google Scholar]
6. Saternus M., Merder T. Physical modeling of the impeller construction impact o the aluminium refining process. Materials. 2022;15:575. doi: 10.3390/ma15020575. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
7. Saternus M., Merder T. Physical modelling of aluminum refining process conducted in batch reactor with rotary impeller. Metals. 2018;8:726. doi: 10.3390/met8090726. [CrossRef] [Google Scholar]
8. Saternus M., Merder T., Pieprzyca J. The influence of impeller geometry on the gas bubbles dispersion in uro-200 reactor—RTD curves. Arch. Metall. Mater. 2015;60:2887–2893. doi: 10.1515/amm-2015-0461. [CrossRef] [Google Scholar]
9. Hernández-Hernández M., Camacho-Martínez J., González-Rivera C., Ramírez-Argáez M.A. Impeller design assisted by physical modeling and pilot plant trials. J. Mater. Process. Technol. 2016;236:1–8. doi: 10.1016/j.jmatprotec.2016.04.031. [CrossRef] [Google Scholar]
10. Mancilla E., Cruz-Méndez W., Garduño I.E., González-Rivera C., Ramírez-Argáez M.A., Ascanio G. Comparison of the hydrodynamic performance of rotor-injector devices in a water physical model of an aluminum degassing ladle. Chem. Eng. Res. Des. 2017;118:158–169. doi: 10.1016/j.cherd.2016.11.031. [CrossRef] [Google Scholar]
11. Michalek K., Socha L., Gryc K., Tkadleckova M., Saternus M., Pieprzyca J., Merder T. Modelling of technological parameters of aluminium melt refining in the ladle by blowing of inert gas through the rotating impeller. Arch. Metall. Mater. 2018;63:987–992. [Google Scholar]
12. Walek J., Michalek K., Tkadlecková M., Saternus M. Modelling of Technological Parameters of Aluminium Melt Refining in the Ladle by Blowing of Inert Gas through the Rotating Impeller. Metals. 2021;11:284. doi: 10.3390/met11020284. [CrossRef] [Google Scholar]
13. Michalek K., Gryc K., Moravka J. Physical modelling of bath homogenization in argon stirred ladle. Metalurgija. 2009;48:215–218. [Google Scholar]
14. Michalek K. The Use of Physical Modeling and Numerical Optimization for Metallurgical Processes. VSB; Ostrawa, Czech Republic: 2001. [Google Scholar]
15. Chen J., Zhao J. Light Metals. TMS; Warrendale, PA, USA: 1995. Bubble distribution in a melt treatment water model; pp. 1227–1231. [Google Scholar]
16. Saternus M. Model Matematyczny do Sterowania Procesem Rafinacji Ciekłych Stopów Aluminium Przy Zastosowaniu URO-200. Katowice, Poland: 2004. Research Project Nr 7 T08B 019 21. [Google Scholar]
17. Pietrewicz L., Wężyk W. Urządzenia do rafinacji gazowej typu URO-200 sześć lat produkcji i doświadczeń; Proceedings of the Aluminum Conference; Zakopane, Poland. 12–16 October 1998. [Google Scholar]
19. Sinelnikov V., Szucki M., Merder T., Pieprzyca J., Kalisz D. Physical and numerical modeling of the slag splashing process. Materials. 2021;14:2289. doi: 10.3390/ma14092289. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
20. White F. Fluid Mechanics. McGraw-Hill; New York, NY, USA: 2010. (McGraw-Hill Series in Mechanical Engineering). [Google Scholar]
21. Yang Z., Yang L., Cheng T., Chen F., Zheng F., Wang S., Guo Y. Fluid Flow Characteristic of EAF Molten Steel with Different Bottom-Blowing Gas Flow Rate Distributions. ISIJ. 2020;60:1957–1967. doi: 10.2355/isijinternational.ISIJINT-2019-794. [CrossRef] [Google Scholar]
22. Nichols B.D., Hirt C.W. Methods for calculating multi-dimensional, transient free surface flows past bodies; Proceedings of the First International Conference on Numerical Ship Hydrodynamics; Gaithersburg, MD, USA. 20–22 October 1975. [Google Scholar]
23. Hirt C.W., Nichols B.D. Volume of Fluid (VOF) Method for the Dynamics of Free Boundaries. J. Comput. Phys. 1981;39:201–255. doi: 10.1016/0021-9991(81)90145-5. [CrossRef] [Google Scholar]
24. Szucki M., Suchy J.S., Lelito J., Malinowski P., Sobczyk J. Application of the lattice Boltzmann method for simulation of the mold filling process in the casting industry. Heat Mass Transf. 2017;53:3421–3431. doi: 10.1007/s00231-017-2069-5. [CrossRef] [Google Scholar]
25. Themelis N.J., Goyal P. Gas injection in steelmaking. Candian Metall. Trans. 1983;22:313–320. [Google Scholar]
26. Zhang L., Jing X., Li Y., Xu Z., Cai K. Mathematical model of decarburization of ultralow carbon steel during RH treatment. J. Univ. Sci. Technol. Beijing. 1997;4:19–23. [Google Scholar]
27. Chiti F., Paglianti A., Bujalshi W. A mechanistic model to estimate powder consumption and mixing time in aluminium industries. Chem. Eng. Res. Des. 2004;82:1105–1111. doi: 10.1205/cerd.82.9.1105.44156. [CrossRef] [Google Scholar]
28. Bouaifi M., Roustan M. Power consumption, mixing time and homogenization energy in dual-impeller agitated gas-liquid reactors. Chem. Eng. Process. 2011;40:87–95. doi: 10.1016/S0255-2701(00)00128-8. [CrossRef] [Google Scholar]
29. Kang J., Lee C.H., Haam S., Koo K.K., Kim W.S. Studies on the overall oxygen transfer rate and mixing time in pilot-scale surface aeration vessel. Environ. Technol. 2001;22:1055–1068. doi: 10.1080/09593332208618215. [PubMed] [CrossRef] [Google Scholar]
30. Moucha T., Linek V., Prokopov E. Gas hold-up, mixing time and gas-liquid volumetric mass transfer coefficient of various multiple-impeller configurations: Rushton turbine, pitched blade and techmix impeller and their combinations. Chem. Eng. Sci. 2003;58:1839–1846. doi: 10.1016/S0009-2509(02)00682-6. [CrossRef] [Google Scholar]
31. Szekely J. Flow phenomena, mixing and mass transfer in argon-stirred ladles. Ironmak. Steelmak. 1979;6:285–293. [Google Scholar]
32. Iguchi M., Nakamura K., Tsujino R. Mixing time and fluid flow phenomena in liquids of varying kinematic viscosities agitated by bottom gas injection. Metall. Mat. Trans. 1998;29:569–575. doi: 10.1007/s11663-998-0091-1. [CrossRef] [Google Scholar]
33. Hjelle O., Engh T.A., Rasch B. Removal of Sodium from Aluminiummagnesium Alloys by Purging with Cl2. Aluminium-Verlag GmbH; Dusseldorf, Germany: 1985. pp. 343–360. [Google Scholar]
34. Zhang L., Taniguchi S. Fundamentals of inclusion removal from liquid steel by bubble flotation. Int. Mat. Rev. 2000;45:59–82. doi: 10.1179/095066000101528313. [CrossRef] [Google Scholar]
Dissertação de Mestrado Ciclo de Estudos Integrados Conducentes ao Grau de Mestre em Engenharia Mecânica Trabalho efectuado sob a orientação do Doutor Hélder de Jesus Fernades Puga Professor Doutor José Joaquim Carneiro Barbosa
ABSTRACT
논문의 일부로 튜터 선택 가능성과 해결해야 할 주제가 설정되는 매개변수를 염두에 두고 개발 주제 ‘Flow- 3D ®에 의한 저압 충전 시스템 최적화’가 선택되었습니다. 이를 위해서는 달성해야 할 목표와 이를 달성하기 위한 방법을 정의하는 것이 필요했습니다.
충전 시스템을 시뮬레이션하고 검증할 수 있는 광범위한 소프트웨어에도 불구하고 Flow-3D®는 시장에서 최고의 도구 중 하나로 표시되어 전체 충전 프로세스 및 행동 표현과 관련하여 탁월한 정확도로 시뮬레이션하는 능력을 입증했습니다.
이를 위해 관련 프로세스를 더 잘 이해하고 충진 시스템 시뮬레이션을 위한 탐색적 기반 역할을 하기 위해 이 도구를 탐색하는 것이 중요합니다. 지연 및 재료 낭비에 반영되는 실제적인 측면에서 충전 장치의 치수를 완벽하게 만드는 비용 및 시간 낭비. 이러한 방식으로 저압 주조 공정에서 충진 시스템을 설계하고 물리적 모델을 탐색하여 특성화하는 방법론을 검증하기 위한 것입니다.
이를 위해 다음 주요 단계를 고려하십시오.
시뮬레이션 소프트웨어 Flow 3D® 탐색; 충전 시스템 모델링; 모델의 매개변수를 탐색하여 모델링된 시스템의 시뮬레이션, 검증 및 최적화.
따라서 연구 중인 압력 곡선과 주조 분석에서 가장 관련성이 높은 정보의 최종 마이닝을 검증하기 위한 것입니다.
사용된 압력 곡선은 수집된 문헌과 이전에 수행된 실제 작업을 통해 얻었습니다. 결과를 통해 3단계 압력 곡선이 층류 충진 체계의 의도된 목적과 관련 속도가 0.5 𝑚/𝑠를 초과하지 않는다는 결론을 내릴 수 있었습니다.
충전 수준이 2인 압력 곡선은 0.5 𝑚/𝑠 이상의 속도로 영역을 채우는 더 난류 시스템을 갖습니다. 열전달 매개변수는 이전에 얻은 값이 주물에 대한 소산 거동을 확증하지 않았기 때문에 연구되었습니다.
이러한 방식으로 주조 공정에 더 부합하는 새로운 가치를 얻었습니다. 달성된 결과는 유사한 것으로 나타난 NovaFlow & Solid®에 의해 생성된 결과와 비교되어 시뮬레이션에서 설정된 매개변수를 검증했습니다. Flow 3D®는 주조 부품 시뮬레이션을 위한 강력한 도구로 입증되었습니다.
As part of the dissertation and bearing in mind the parameters in which the possibility of a choice of tutor and the subject to be addressed is established, the subject for development ’Optimization of filling systems for low pressure by Flow 3D ®’ was chosen. For this it was necessary to define the objectives to achieve and the methods to attain them. Despite the wide range of software able to simulate and validate filling systems, Flow 3D® has been shown as one of the best tools in the market, demonstrating its ability to simulate with distinctive accuracy with respect to the entire process of filling and the behavioral representation of the fluid obtained. To this end, it is important to explore this tool for a better understanding of the processes involved and to serve as an exploratory basis for the simulation of filling systems, simulation being one of the great strengths of the current industry due to the need to reduce costs and time waste, in practical terms, that lead to the perfecting of the dimensioning of filling devices, which are reflected in delays and wasted material. In this way it is intended to validate the methodology to design a filling system in lowpressure casting process, exploring their physical models and thus allowing for its characterization. For this, consider the following main phases: The exploration of the simulation software Flow 3D®; modeling of filling systems; simulation, validation and optimization of systems modeled by exploring the parameters of the models. Therefore, it is intended to validate the pressure curves under study and the eventual mining of the most relevant information in a casting analysis. The pressure curves that were used were obtained through the gathered literature and the practical work previously performed. Through the results it was possible to conclude that the pressure curve with 3 levels meets the intended purpose of a laminar filling regime and associated speeds never exceeding 0.5 𝑚/𝑠. The pressure curve with 2 filling levels has a more turbulent system, having filling areas with velocities above 0.5 𝑚/𝑠. The heat transfer parameter was studied due to the values previously obtained didn’t corroborate the behavior of dissipation regarding to the casting. In this way, new values, more in tune with the casting process, were obtained. The achieved results were compared with those generated by NovaFlow & Solid®, which were shown to be similar, validating the parameters established in the simulations. Flow 3D® was proven a powerful tool for the simulation of casting parts.
[1] E. Stanley and D. B. Sc, “Fluid Flow Aspects of Solidification Modelling : Simulation of Low Pressure Die Casting .” [2] Y. Sahin, “Computer aided foundry die-design,” Metallography, vol. 24, no. 8, pp. 671–679, 2003. [3] F. Bonollo, J. Urban, B. Bonatto, and M. Botter, “Gravity and low pressure die casting of aluminium alloys : a technical and economical benchmark,” La Metall. Ital., vol. 97, no. 6, pp. 23–32, 2005. [4] P. a and R. R, “Study of the effect of process parameters on the production of a nonsimmetric low pressure die casting part,” La Metall. Ital., pp. 57–63, 2009. [5] “Fundição em baixa pressão | Aluinfo.” [Online]. Available: http://www.aluinfo.com.br/novo/materiais/fundicao-em-baixa-pressao. [Accessed: 18- Sep-2015]. [6] “Low Pressure Sand Casting by Wolverine Bronze.” [Online]. Available: http://www.wolverinebronze.com/low-pressure-sand-casting.php. [Accessed: 18-Sep2015]. [7] A. Reikher, “Numerical Analysis of Die-Casting Process in Thin Cavities Using Lubrication Approximation,” no. December, 2012. [8] P. Fu, A. a. Luo, H. Jiang, L. Peng, Y. Yu, C. Zhai, and A. K. Sachdev, “Low-pressure die casting of magnesium alloy AM50: Response to process parameters,” J. Mater. Process. Technol., vol. 205, no. 1–3, pp. 224–234, 2008. [9] X. Li, Q. Hao, W. Jie, and Y. Zhou, “Development of pressure control system in counter gravity casting for large thin-walled A357 aluminum alloy components,” Trans. Nonferrous Met. Soc. China, vol. 18, no. 4, pp. 847–851, 2008. [10] J. a. Hines, “Determination of interfacial heat-transfer boundary conditions in an aluminum low-pressure permanent mold test casting,” Metall. Mater. Trans. B, vol. 35, no. 2, pp. 299–311, 2004. [11] A. Lima, A. Freitas, and P. Magalhães, “Processos de vazamento em moldações permanentes,” pp. 40–49, 2003. [12] Y. B. Choi, K. Matsugi, G. Sasaki, K. Arita, and O. Yanagisawa, “Analysis of Manufacturing Processes for Metal Fiber Reinforced Aluminum Alloy Composite Fabricated by Low-Pressure Casting,” Mater. Trans., vol. 47, no. 4, pp. 1227–1231, 68 2006. [13] G. Mi, X. Liu, K. Wang, and H. Fu, “Numerical simulation of low pressure die-casting aluminum wheel,” China Foundry, vol. 6, no. 1, pp. 48–52, 2009. [14] J. Kuo, F. Hsu, and W. Hwang, “ADVANCED Development of an interactive simulation system for the determination of the pressure ± time relationship during the ® lling in a low pressure casting process,” vol. 2, pp. 131–145, 2001. [15] S.-G. Liu, F.-Y. Cao, X.-Y. Zhao, Y.-D. Jia, Z.-L. Ning, and J.-F. Sun, “Characteristics of mold filling and entrainment of oxide film in low pressure casting of A356 alloy,” Mater. Sci. Eng. A, vol. 626, pp. 159–164, 2015. [16] “Casting Training Class – Lecture 10 – Solidification and Shrinkage-Casting.” FLOW3D®. [17] “UAB Casting Engineering Laboratory.” [Online]. Available: file:///C:/Users/Jos%C3%A9 Belo/Desktop/Artigo_Software/UAB Casting Engineering Laboratory.htm. [Accessed: 09-Nov-2015]. [18] A. Louvo, “Casting Simulation as a Tool in Concurrent Engineering,” pp. 1–12, 1997. [19] T. R. Vijayaram and P. Piccardo, “Computers in Foundries,” vol. 30, 2012. [20] M. Sadaiah, D. R. Yadav, P. V. Mohanram, and P. Radhakrishnan, “A generative computer-aided process planning system for prismatic components,” Int. J. Adv. Manuf. Technol., vol. 20, no. 10, pp. 709–719, 2002. [21] Ministry_of_Planning, “Digital Data,” vol. 67, pp. 1–6, 2004. [22] S. Shamasundar, D. Ramachandran, and N. S. Shrinivasan, “COMPUTER SIMULATION AND ANALYSIS OF INVESTMENTCASTING PROCESS.” [23] J. M. Siqueira and G. Motors, “Simulation applied to Aluminum High Pressure Die Casting,” pp. 1–5, 1998. [24] C. Fluid, COMPUTATIONAL FLUID DYNAMICS. Abdulnaser Sayma & Ventus Publishing ApS, 2009. [25] C. a. Felippa, “1 – Overview,” Adv. Finite Elem. Methods, pp. 1–9. [26] a. Meena and M. El Mansori, “Correlative thermal methodology for castability simulation of ductile iron in ADI production,” J. Mater. Process. Technol., vol. 212, no. 11, pp. 2484–2495, 2012. [27] T. R. Vijayaram, S. Sulaiman, a. M. S. Hamouda, and M. H. M. Ahmad, “Numerical simulation of casting solidification in permanent metallic molds,” J. Mater. Process. 69 Technol., vol. 178, pp. 29–33, 2006. [28] “General CFD FAQ — CFD-Wiki, the free CFD reference.” [Online]. Available: http://www.cfd-online.com/Wiki/General_CFD_FAQ. [Accessed: 10-Nov-2015]. [29] “FEM | FEA | CFD.” [Online]. Available: http://fem4analyze.blogspot.pt/. [Accessed: 09-Nov-2015]. [30] “Fundição; revista da Associação portuguesa de fundição,” Fundição, vol. N o 227. [31] “Casting Training Class – Lecture 1 – Introduction_to_FLOW-3D – Casting.” FLOW3D®. [32] F. Science, “FLOW-3D Cast Documentation,” no. 3.5, p. 80, 2012. [33] “Casting Training Class – Lecture 4 – Geometry Building – General.” FLOW-3D®. [34] F. Science, “FLOW-3D v11.0.3 User Manual,” pp. 1–132, 2015. [35] “Casting Training Class – Lecture 5 Meshing Concept – General.” FLOW-3D®. [36] “Casting Training Class – Lecture 6 – Boundary_Conditions – Casting.” FLOW-3D®. [37] “Casting Training Class – Lecture 9 – Physical Models-castings.” FLOW-3D®. [38] P. A. D. Jácome, M. C. Landim, A. Garcia, A. F. Furtado, and I. L. Ferreira, “The application of computational thermodynamics and a numerical model for the determination of surface tension and Gibbs–Thomson coefficient of aluminum based alloys,” Thermochim. Acta, vol. 523, no. 1–2, pp. 142–149, 2011. [39] J. P. Anson, R. A. L. Drew, and J. E. Gruzleski, “The surface tension of molten aluminum and Al-Si-Mg alloy under vacuum and hydrogen atmospheres,” Metall. Mater. Trans. B Process Metall. Mater. Process. Sci., vol. 30, no. 6, pp. XVI–1032, 1999.
Publication Date:2013-07-24 Research Org.: Los Alamos National Lab. (LANL), Los Alamos, NM (United States) Sponsoring Org.: DOE/LANL OSTI Identifier: 1088904 Report Number(s): LA-UR-13-25537 DOE Contract Number: AC52-06NA25396 Resource Type: Technical Report Country of Publication: United States Language: English Subject: Engineering(42); Materials Science(36); Radiation Chemistry, Radiochemistry, & Nuclear Chemistry(38)
Introduction
The plutonium foundry at Los Alamos National Laboratory casts products for various special nuclear applications. However, plutonium’s radioactivity, material properties, and security constraints complicate the ability to perform experimental analysis of mold behavior. The Manufacturing Engineering and Technologies (MET-2) group previously developed a graphite mold to vacuum cast small plutonium disks to be used by the Department of Homeland Security as point sources for radiation sensor testing.
A two-stage pouring basin consisting of a funnel and an angled cavity directs the liquid into a vertical runner. A stack of ten disk castings connect to the runner by horizontal gates. Volumetric flow rates were implemented to limit overflow into the funnel and minimize foundry returns. Models using Flow-3D computational fluid dynamics software are employed here to determine liquid Pu flow paths, optimal pour regimes, temperature changes, and pressure variations.
Setup
Hardcopy drawings provided necessary information to create 3D .stl models for import into Flow-3D (Figs. 1 and 2). The mesh was refined over several iterations to isolate the disk cavities, runner, angled cavity, funnel, and input pour. The final flow and mold-filling simulation utilizes a fine mesh with ~5.5 million total cells. For the temperature study, the mesh contained 1/8 as many cells to reduce computational time and set temperatures to 850 °C for the molten plutonium and 500 °C for the solid graphite mold components (Fig. 3).
Flow-3D solves mass continuity and Navier-Stokes momentum equations over the structured rectangular grid model using finite difference and finite volume numerical algorithms. The solver includes terms in the momentum equation for body and viscous accelerations and uses convective heat transfer.
Simulation settings enabled Flow-3D physics calculations for gravity at 980.665 cm/s 2 in the negative Z direction (top of mold to bottom); viscous, turbulent, incompressible flow using dynamically-computed Renormalized Group Model turbulence calculations and no-slip/partial slip wall shear, and; first order, full energy equation heat transfer.
Mesh boundaries were all set to symmetric boundary conditions except for the Zmin boundary set to outflow and the Zmax boundary set to a volume flow. Vacuum casting conditions and the high reactivity of remaining air molecules with Pu validate the assumption of an initially fluidless void.
Results
The flow follows a unique three-dimensional path. The mold fills upwards with two to three disks receiving fluid in a staggered sequence. Figures 5-9 show how the fluid fills the cavity, and Figure 7 includes the color scale for pressure levels in these four figures. The narrow gate causes a high pressure region which forces the fluid to flow down the cavity centerline.
It proceeds to splash against the far wall and then wrap around the circumference back to the gate (Figs. 5 and 6). Flow in the angled region of the pouring basin cascades over the bottom ledge and attaches to the far wall of the runner, as seen in Figure 7.
This channeling becomes less pronounced as fluid volume levels increase. Finally, two similar but non-uniform depressed regions form about the centerline. These regions fill from their perimeter and bottom until completion (Fig. 8). Such a pattern is counter, for example, to a steady scenario in which a circle of molten Pu encompassing the entire bottom surface rises as a growing cylinder.
Cavity pressure becomes uniform when the cavity is full. Pressure levels build in the rising well section of the runner, where impurities were found to settle in actual casting. Early test simulations optimized the flow as three pours so that the fluid would never overflow to the funnel, the cavities would all fill completely, and small amounts of fluid would remain as foundry returns in the angled cavity.
These rates and durations were translated to the single 2.7s pour at 100 cm 3 per second used here. Figure 9 shows anomalous pressure fluctuations which occurred as the cavities became completely filled. Multiple simulations exhibited a rapid change in pressure from positive to negative and back within the newly-full disk and surrounding, already-full disks.
The time required to completely fill each cavity is plotted in Figure 10. Results show negligible temperature change within the molten Pu during mold filling and, as seen in Figure 11, at fill completion.
Conclusions
Non-uniform cavity filling could cause crystal microstructure irregularities during solidification. However, the small temperature changes seen – due to large differences in specific heat between Pu and graphite – over a relatively short time make such problems unlikely in this case.
In the actual casting, cooling required approximately ten minutes. This large difference in time scales further reduces the chance for temperature effects in such a superheated scenario. Pouring basin emptying decreases pressure at the gate which extends fill time of the top two cavities.
The bottom cavity takes longer to fill because fluid must first enter the runner and fill the well. Fill times continue linearly until the top two cavities. The anomalous pressure fluctuations may be due to physical attempts by the system to reach equilibrium, but they are more likely due to numerical errors in the Flow3D solver.
Unsuccessful tests were performed to remove them by halving fluid viscosity. The fine mesh reduced, but did not eliminate, the extent of the fluctuations. Future work is planned to study induction and heat transfer in the full Pu furnace system, including quantifying temporal lag of the cavity void temperature to the mold wall temperature during pre-heat and comparing heat flux levels between furnace components during cool-down.
Thanks to Doug Kautz for the opportunity to work with MET-2 and for assigning an interesting unclassified project. Additional thanks to Mike Bange for CFD guidance, insight of the project’s history, and draft review.
The elimination of internal macro-defects is a key issue in Ti–6Al–4V alloys fabricated via powder bed fusion using electron beams (PBF-EB), wherein internal macro-defects mainly originate from the virgin powder and inappropriate printing parameters. This study compares different types powders by combining support vector machine techniques to determine the most suitable powder for PBF-EB and to predict the processing window for the printing parameters without internal macro-defects. The results show that powders fabricated via plasma rotating electrode process have the best sphericity, flowability, and minimal porosity and are most suitable for printing. Surface roughness criterion was also applied to determine the quality of the even surfaces, and support vector machine was used to construct processing maps capable of predicting a wide range of four-dimensional printing parameters to obtain macro-defect-free samples, offering the possibility of subsequent development of Ti–6Al–4V alloys with excellent properties. The macro-defect-free samples exhibited good elongation, with the best overall mechanical properties being the ultimate tensile strength and elongation of 934.7 MPa and 24.3%, respectively. The elongation of the three macro-defect-free samples was much higher than that previously reported for additively manufactured Ti–6Al–4V alloys. The high elongation of the samples in this work is mainly attributed to the elimination of internal macro-defects.
Introduction
Additive manufacturing (AM) technologies can rapidly manufacture complex or custom parts, reducing process steps and saving manufacturing time [[1], [2], [3], [4]], and are widely used in the aerospace, automotive, and other precision industries [5,6]. Powder bed fusion using an electron beam (PBF-EB) is an additive manufacturing method that uses a high-energy electron beam to melt metal powders layer by layer to produce parts. In contrast to selective laser melting, PBF-EB involves the preparation of samples in a high vacuum environment, which effectively prevents the introduction of impurities such as O and N. It also involves a preheating process for the print substrate and powder, which reduces residual thermal stress on the sample and subsequent heat treatment processes [[2], [3], [4],7]. Due to these features and advantages, PBF-EB technology is a very important AM technology with great potential in metallic materials. Moreover, PBF-EB is the ideal AM technology for the manufacture of complex components made of many alloys, such as titanium alloys, nickel-based superalloys, aluminum alloys and stainless steels [[2], [3], [4],8].
Ti–6Al–4V alloy is one of the prevalent commercial titanium alloys possessing high specific strength, excellent mechanical properties, excellent corrosion resistance, and good biocompatibility [9,10]. It is widely used in applications requiring low density and excellent corrosion resistance, such as the aerospace industry and biomechanical applications [11,12]. The mechanical properties of PBF-EB-processed Ti–6Al–4V alloys are superior to those fabricated by casting or forging, because the rapid cooling rate in PBF-EB results in finer grains [[12], [13], [14], [15], [16], [17], [18]]. However, the PBF-EB-fabricated parts often include internal macro-defects, which compromises their mechanical properties [[19], [20], [21], [22]]. This study focused on the elimination of macro-defects, such as porosity, lack of fusion, incomplete penetration and unmelted powders, which distinguishes them from micro-defects such as vacancies, dislocations, grain boundaries and secondary phases, etc. Large-sized fusion defects cause a severe reduction in mechanical strength. Smaller defects, such as pores and cracks, lead to the initiation of fatigue cracking and rapidly accelerate the cracking process [23]. The issue of internal macro-defects must be addressed to expand the application of the PBF-EB technology. The main studies for controlling internal macro-defects are online monitoring of defects, remelting and hot isostatic pressing (HIP). The literatures [24,25] report the use of infrared imaging or other imaging techniques to identify defects, but the monitoring of smaller sized defects is still not adequate. And in some cases remelting does not reduce the internal macro-defects of the part, but instead causes coarsening of the macrostructure and volatilization of some metal elements [23]. The HIP treatment does not completely eliminate the internal macro-defects, the original defect location may still act as a point of origin of the crack, and the subsequent treatment will consume more time and economic costs [23]. Therefore, optimizing suitable printing parameters to avoid internal macro-defects in printed parts at source is of great industrial value and research significance, and is an urgent issue in PBF-EB related technology.
There are two causes of internal macro-defects in the AM process: gas pores trapped in the virgin powder and the inappropriate printing parameters [7,23]. Gui et al. [26] classify internal macro-defects during PBF-EB process according to their shape, such as spherical defects, elongated shape defects, flat shape defects and other irregular shape defects. Of these, spherical defects mainly originate from raw material powders. Other shape defects mainly originate from lack of fusion or unmelted powders caused by unsuitable printing parameters, etc. The PBF-EB process requires powders with good flowability, and spherical powders are typically chosen as raw materials. The prevalent techniques for the fabrication of pre-alloyed powders are gas atomization (GA), plasma atomization (PA), and the plasma rotating electrode process (PREP) [27,28]. These methods yield powders with different characteristics that affect the subsequent fabrication. The selection of a suitable powder for PBF-EB is particularly important to produce Ti–6Al–4V alloys without internal macro-defects. The need to optimize several printing parameters such as beam current, scan speed, line offset, and focus offset make it difficult to eliminate internal macro-defects that occur during printing [23]. Most of the studies [11,12,22,[29], [30], [31], [32], [33]] on the optimization of AM processes for Ti–6Al–4V alloys have focused on samples with a limited set of parameters (e.g., power–scan speed) and do not allow for the guidance and development of unknown process windows for macro-defect-free samples. In addition, process optimization remains a time-consuming problem, with the traditional ‘trial and error’ method demanding considerable time and economic costs. The development of a simple and efficient method to predict the processing window for alloys without internal macro-defects is a key issue. In recent years, machine learning techniques have increasingly been used in the field of additive manufacturing and materials development [[34], [35], [36], [37]]. Aoyagi et al. [38] recently proposed a novel and efficient method based on a support vector machine (SVM) to optimize the two-dimensional process parameters (current and scan speed) and obtain PBF-EB-processed CoCr alloys without internal macro-defects. The method is one of the potential approaches toward effective optimization of more than two process parameters and makes it possible for the machine learning techniques to accelerate the development of alloys without internal macro-defects.
Herein, we focus on the elimination of internal macro-defects, such as pores, lack of fusion, etc., caused by raw powders and printing parameters. The Ti–6Al–4V powders produced by three different methods were compared, and the powder with the best sphericity, flowability, and minimal porosity was selected as the feedstock for subsequent printing. The relationship between the surface roughness and internal macro-defects in the Ti–6Al–4V components was also investigated. The combination of SVM and surface roughness indices (Sdr) predicted a wider four-dimensional processing window for obtaining Ti–6Al–4V alloys without internal macro-defects. Finally, we investigated the tensile properties of Ti–6Al–4V alloys at room temperature with different printing parameters, as well as the corresponding microstructures and fracture types.
Section snippets
Starting materials
Three types of Ti–6Al–4V alloy powders, produced by GA, PA, and PREP, were compared. The particle size distribution of the powders was determined using a laser particle size analyzer (LS230, Beckman Coulter, USA), and the flowability was measured using a Hall flowmeter (JIS-Z2502, Tsutsui Scientific Instruments Co., Ltd., Japan), according to the ASTM B213 standard. The powder morphology and internal macro-defects were determined using scanning electron microscopy (SEM, JEOL JCM-6000) and X-ray
Comparison of the characteristics of GA, PA, and PREP Ti–6Al–4V powders
The particle size distributions (PSDs) and flowability of the three types of Ti–6Al–4V alloy powders produced by GA, PA, and PREP are shown in Fig. 2. Although the average particle sizes are similar (89.4 μm for GA, 82.5 μm for PA, and 86.1μm for PREP), the particle size range is different for the three types of powder (6.2–174.8 μm for GA, 27.3–139.2 μm for PA, and 39.4–133.9 μm for PREP). The flowability of the GA, PA, and PREP powders was 30.25 ± 0.98, 26.54 ± 0.37, and 25.03 ± 0.22 (s/50
Conclusions
The characteristics of the three types of Ti–6Al–4V alloy powders produced via GA, PA, and PREP were compared. The PREP powder with the best sphericity, flowability, and low porosity was found to be the most favorable powder for subsequent printing of Ti–6Al–4V alloys without internal macro-defects. The quantitative criterion of Sdr <0.015 for even surfaces was also found to be applicable to Ti–6Al–4V alloys. The process maps of Ti–6Al–4V alloys include two regions, high beam current/scan speed
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This study was based on the results obtained from project JPNP19007, commissioned by the New Energy and Industrial Technology Development Organization (NEDO). This work was also supported by JSPS KAKENHI (Proposal No. 21K03801) and the Inter-University Cooperative Research Program (Proposal nos. 18G0418, 19G0411, and 20G0418) of the Cooperative Research and Development Center for Advanced Materials, Institute for Materials Research, Tohoku University. It was also supported by the Council for
Aneesh Kumar J1, a, K. Krishnakumar1, b and S. Savithri2, c 1 Department of Mechanical Engineering, College of Engineering, Thiruvananthapuram, Kerala, 2 Computational Modelling& Simulation Division, Process Engineering & Environmental Technology Division CSIR-National Institute for Interdisciplinary Science & Technology Thiruvananthapuram, Kerala, India. a aneesh82kj@gmail.com, b kkk@cet.ac.in, c sivakumarsavi@gmail.com, ssavithri@niist.res.in Key words: Mold filling, centrifugal casting process, computer simulation, FLOW- 3D™
Abstract
원심 주조 공정은 기능적으로 등급이 지정된 재료, 즉 구성 요소 간에 밀도 차이가 큰 복합 재료 또는 금속 재료를 생산하는 데 사용되는 잠재적인 제조 기술 중 하나입니다. 이 공정에서 유체 흐름이 중요한 역할을 하며 복잡한 흐름 공정을 이해하는 것은 결함 없는 주물을 생산하는 데 필수입니다. 금형이 고속으로 회전하고 금형 벽이 불투명하기 때문에 흐름 패턴을 실시간으로 시각화하는 것은 불가능합니다. 따라서 현재 연구에서는 상용 CFD 코드 FLOW-3D™를 사용하여 수직 원심 주조 공정 중 단순 중공 원통형 주조에 대한 금형 충전 시퀀스를 시뮬레이션했습니다. 수직 원심주조 공정 중 다양한 방사 속도가 충전 패턴에 미치는 영향을 조사하고 있습니다.
Centrifugal casting process is one of the potential manufacturing techniques used for producing functionally graded materials viz., composite materials or metallic materials which have high differences of density among constituents. In this process, the fluid flow plays a major role and understanding the complex flow process is a must for the production of defect-free castings. Since the mold spins at a high velocity and the mold wall being opaque, it is impossible to visualise the flow patterns in real time. Hence, in the present work, the commercial CFD code FLOW-3D™, has been used to simulate the mold filling sequence for a simple hollow cylindrical casting during vertical centrifugal casting process. Effect of various spinning velocities on the fill pattern during vertical centrifugal casting process is being investigated.
References
[1] W. Shi-Ping, L. Chang-yun, G. Jing-jie, S. Yan-qing, L. Xiu-qiao, F. Heng-zhi, Numerical simulation and experimental investigation of two filling methods in vertical centrifugal casting, Trans. Nonferrous Met. Soc. China 16 (2006) 1035-1040. 10.1016/s1003-6326(06)60373-7 [2] G. Chirita, D. Soares, F.S. Silva, Advantages of the centrifugal casting technique for the production of structural components with Al-Si alloys, Mater. Des. 29 (2008) 20-27. 10.1016/j.matdes.2006.12.011 [3] A. Kermanpur, Sh. Mahmoudi, A. Hajipour, Numerical simulation of metal flow and solidification in the multi-cavity casting moulds of automotive components, J. Mater. Proc. Tech. 206 (208) 62-68. 10.1016/j.jmatprotec.2007.12.004 [4] D. McBride et. al. Complex free surface flows in centrifugal casting: Computational modelling and validation experiments, Computers & Fluids 82 (2013) 63-72. 10.1016/j.compfluid.2013.04.021
2014년 2월 영국 해협(영국)과 특히 Dawlish에 영향을 미친 온대 저기압 폭풍 사슬은 남서부 지역과 영국의 나머지 지역을 연결하는 주요 철도에 심각한 피해를 입혔습니다.
이 사건으로 라인이 두 달 동안 폐쇄되어 5천만 파운드의 피해와 12억 파운드의 경제적 손실이 발생했습니다. 이 연구에서는 폭풍의 파괴력을 해독하기 위해 목격자 계정을 수집하고 해수면 데이터를 분석하며 수치 모델링을 수행합니다.
우리의 분석에 따르면 이벤트의 재난 관리는 성공적이고 효율적이었으며 폭풍 전과 도중에 인명과 재산을 구하기 위해 즉각적인 조치를 취했습니다. 파도 부이 분석에 따르면 주기가 4–8, 8–12 및 20–25초인 복잡한 삼중 봉우리 바다 상태가 존재하는 반면, 조위계 기록에 따르면 최대 0.8m의 상당한 파도와 최대 1.5m의 파도 성분이 나타났습니다.
이벤트에서 가능한 기여 요인으로 결합된 진폭. 최대 286 KN의 상당한 임펄스 파동이 손상의 시작 원인일 가능성이 가장 높았습니다. 수직 벽의 반사는 파동 진폭의 보강 간섭을 일으켜 파고가 증가하고 최대 16.1m3/s/m(벽의 미터 너비당)의 상당한 오버탑핑을 초래했습니다.
이 정보와 우리의 공학적 판단을 통해 우리는 이 사고 동안 다중 위험 계단식 실패의 가장 가능성 있는 순서는 다음과 같다고 결론을 내립니다. 조적 파괴로 이어지는 파도 충격력, 충전물 손실 및 연속적인 조수에 따른 구조물 파괴.
The February 2014 extratropical cyclonic storm chain, which impacted the English Channel (UK) and Dawlish in particular, caused significant damage to the main railway connecting the south-west region to the rest of the UK. The incident caused the line to be closed for two months, £50 million of damage and an estimated £1.2bn of economic loss. In this study, we collate eyewitness accounts, analyse sea level data and conduct numerical modelling in order to decipher the destructive forces of the storm. Our analysis reveals that the disaster management of the event was successful and efficient with immediate actions taken to save lives and property before and during the storm. Wave buoy analysis showed that a complex triple peak sea state with periods at 4–8, 8–12 and 20–25 s was present, while tide gauge records indicated that significant surge of up to 0.8 m and wave components of up to 1.5 m amplitude combined as likely contributing factors in the event. Significant impulsive wave force of up to 286 KN was the most likely initiating cause of the damage. Reflections off the vertical wall caused constructive interference of the wave amplitudes that led to increased wave height and significant overtopping of up to 16.1 m3/s/m (per metre width of wall). With this information and our engineering judgement, we conclude that the most probable sequence of multi-hazard cascading failure during this incident was: wave impact force leading to masonry failure, loss of infill and failure of the structure following successive tides.
Introduction
The progress of climate change and increasing sea levels has started to have wide ranging effects on critical engineering infrastructure (Shakou et al. 2019). The meteorological effects of increased atmospheric instability linked to warming seas mean we may be experiencing more frequent extreme storm events and more frequent series or chains of events, as well as an increase in the force of these events, a phenomenon called storminess (Mölter et al. 2016; Feser et al. 2014). Features of more extreme weather events in extratropical latitudes (30°–60°, north and south of the equator) include increased gusting winds, more frequent storm squalls, increased prolonged precipitation and rapid changes in atmospheric pressure and more frequent and significant storm surges (Dacre and Pinto 2020). A recent example of these events impacting the UK with simultaneous significant damage to coastal infrastructure was the extratropical cyclonic storm chain of winter 2013/2014 (Masselink et al. 2016; Adams and Heidarzadeh 2021). The cluster of storms had a profound effect on both coastal and inland infrastructure, bringing widespread flooding events and large insurance claims (RMS 2014).
The extreme storms of February 2014, which had a catastrophic effect on the seawall of the south Devon stretch of the UK’s south-west mainline, caused a two-month closure of the line and significant disruption to the local and regional economy (Fig. 1b) (Network Rail 2014; Dawson et al. 2016; Adams and Heidarzadeh 2021). Restoration costs were £35 m, and economic effects to the south-west region of England were estimated up to £1.2bn (Peninsula Rail Taskforce 2016). Adams and Heidarzadeh (2021) investigated the disparate cascading failure mechanisms which played a part in the failure of the railway through Dawlish and attempted to put these in the context of the historical records of infrastructure damage on the line. Subsequent severe storms in 2016 in the region have continued to cause damage and disruption to the line in the years since 2014 (Met Office 2016). Following the events of 2014, Network Rail Footnote1 who owns the network has undertaken a resilience study. As a result, it has proposed a £400 m refurbishment of the civil engineering assets that support the railway (Fig. 1) (Network Rail 2014). The new seawall structure (Fig. 1a,c), which is constructed of pre-cast concrete sections, encases the existing Brunel seawall (named after the project lead engineer, Isambard Kingdom Brunel) and has been improved with piled reinforced concrete foundations. It is now over 2 m taller to increase the available crest freeboard and incorporates wave return features to minimise wave overtopping. The project aims to increase both the resilience of the assets to extreme weather events as well as maintain or improve amenity value of the coastline for residents and visitors.
In this work, we return to the Brunel seawall and the damage it sustained during the 2014 storms which affected the assets on the evening of the 4th and daytime of the 5th of February and eventually resulted in a prolonged closure of the line. The motivation for this research is to analyse and model the damage made to the seawall and explain the damage mechanisms in order to improve the resilience of many similar coastal structures in the UK and worldwide. The innovation of this work is the multidisciplinary approach that we take comprising a combination of analysis of eyewitness accounts (social science), sea level and wave data analysis (physical science) as well as numerical modelling and engineering judgement (engineering sciences). We investigate the contemporary wave climate and sea levels by interrogating the real-time tide gauge and wave buoys installed along the south-west coast of the English Channel. We then model a typical masonry seawall (Fig. 2), applying the computational fluid dynamics package FLOW3D-Hydro,Footnote2 to quantify the magnitude of impact forces that the seawall would have experienced leading to its failure. We triangulate this information to determine the probable sequence of failures that led to the disaster in 2014.
Data and methods
Our data comprise eyewitness accounts, sea level records from coastal tide gauges and offshore wave buoys as well as structural details of the seawall. As for methodology, we analyse eyewitness data, process and investigate sea level records through Fourier transform and conduct numerical simulations using the Flow3D-Hydro package (Flow Science 2022). Details of the data and methodology are provided in the following.
Eyewitness data
The scale of damage to the seawall and its effects led the local community to document the first-hand accounts of those most closely affected by the storms including residents, local businesses, emergency responders, politicians and engineering contractors involved in the post-storm restoration work. These records now form a permanent exhibition in the local museum in DawlishFootnote3, and some of these accounts have been transcribed into a DVD account of the disaster (Dawlish Museum 2015). We have gathered data from the Dawlish Museum, national and international news reports, social media tweets and videos. Table 1 provides a summary of the eyewitness accounts. Overall, 26 entries have been collected around the time of the incident. Our analysis of the eyewitness data is provided in the third column of Table 1 and is expanded in Sect. 3.Table 1 Eyewitness accounts of damage to the Dawlish railway due to the February 2014 storm and our interpretations
Our sea level data are a collection of three tide gauge stations (Newlyn, Devonport and Swanage Pier—Fig. 5a) owned and operated by the UK National Tide and Sea Level FacilityFootnote4 for the Environment Agency and four offshore wave buoys (Dawlish, West Bay, Torbay and Chesil Beach—Fig. 6a). The tide gauge sites are all fitted with POL-EKO (www.pol-eko.com.pl) data loggers. Newlyn has a Munro float gauge with one full tide and one mid-tide pneumatic bubbler system. Devonport has a three-channel data pneumatic bubbler system, and Swanage Pier consists of a pneumatic gauge. Each has a sampling interval of 15 min, except for Swanage Pier which has a sampling interval of 10 min. The tide gauges are located within the port areas, whereas the offshore wave buoys are situated approximately 2—3.3 km from the coast at water depths of 10–15 m. The wave buoys are all Datawell Wavemaker Mk III unitsFootnote5 and come with sampling interval of 0.78 s. The buoys have a maximum saturation amplitude of 20.5 m for recording the incident waves which implies that every wave larger than this threshold will be recorded at 20.5 m. The data are provided by the British Oceanographic Data CentreFootnote6 for tide gauges and the Channel Coastal ObservatoryFootnote7 for wave buoys.
Sea level analysis
The sea level data underwent quality control to remove outliers and spikes as well as gaps in data (e.g. Heidarzadeh et al. 2022; Heidarzadeh and Satake 2015). We processed the time series of the sea level data using the Matlab signal processing tool (MathWorks 2018). For calculations of the tidal signals, we applied the tidal package TIDALFIT (Grinsted 2008), which is based on fitting tidal harmonics to the observed sea level data. To calculate the surge signals, we applied a 30-min moving average filter to the de-tided data in order to remove all wind, swell and infra-gravity waves from the time series. Based on the surge analysis and the variations of the surge component before the time period of the incident, an error margin of approximately ± 10 cm is identified for our surge analysis. Spectral analysis of the wave buoy data is performed using the fast Fourier transform (FFT) of Matlab package (Mathworks 2018).
Numerical modelling
Numerical modelling of wave-structure interaction is conducted using the computational fluid dynamics package Flow3D-Hydro version 1.1 (Flow Science 2022). Flow3D-Hydro solves the transient Navier–Stokes equations of conservation of mass and momentum using a finite difference method and on Eulerian and Lagrangian frameworks (Flow Science 2022). The aforementioned governing equations are:
∇.u=0∇.u=0
(1)
∂u∂t+u.∇u=−∇Pρ+υ∇2u+g∂u∂t+u.∇u=−∇Pρ+υ∇2u+g
(2)
where uu is the velocity vector, PP is the pressure, ρρ is the water density, υυ is the kinematic viscosity and gg is the gravitational acceleration. A Fractional Area/Volume Obstacle Representation (FAVOR) is adapted in Flow3D-Hydro, which applies solid boundaries within the Eulerian grid and calculates the fraction of areas and volume in partially blocked volume in order to compute flows on corresponding boundaries (Hirt and Nichols 1981). We validated the numerical modelling through comparing the results with Sainflou’s analytical equation for the design of vertical seawalls (Sainflou 1928; Ackhurst 2020), which is as follows:
where pdpd is the hydrodynamic pressure, ρρ is the water density, gg is the gravitational acceleration, HH is the wave height, dd is the water depth, kk is the wavenumber, zz is the difference in still water level and mean water level, σσ is the angular frequency and tt is the time. The Sainflou’s equation (Eq. 3) is used to calculate the dynamic pressure from wave action, which is combined with static pressure on the seawall.
Using Flow3D-Hydro, a model of the Dawlish seawall was made with a computational domain which is 250.0 m in length, 15.0 m in height and 0.375 m in width (Fig. 3a). The computational domain was discretised using a single uniform grid with a mesh size of 0.125 m. The model has a wave boundary at the left side of the domain (x-min), an outflow boundary on the right side (x-max), a symmetry boundary at the bottom (z-min) and a wall boundary at the top (z-max). A wall boundary implies that water or waves are unable to pass through the boundary, whereas a symmetry boundary means that the two edges of the boundary are identical and therefore there is no flow through it. The water is considered incompressible in our model. For volume of fluid advection for the wave boundary (i.e. the left-side boundary) in our simulations, we utilised the “Split Lagrangian Method”, which guarantees the best accuracy (Flow Science, 2022).
The stability of the numerical scheme is controlled and maintained through checking the Courant number (CC) as given in the following:
C=VΔtΔxC=VΔtΔx
(4)
where VV is the velocity of the flow, ΔtΔt is the time step and ΔxΔx is the spatial step (i.e. grid size). For stability and convergence of the numerical simulations, the Courant number must be sufficiently below one (Courant et al. 1928). This is maintained by a careful adjustment of the ΔxΔx and ΔtΔt selections. Flow3D-Hydro applies a dynamic Courant number, meaning the program adjusts the value of time step (ΔtΔt) during the simulations to achieve a balance between accuracy of results and speed of simulation. In our simulation, the time step was in the range ΔtΔt = 0.0051—0.051 s.
In order to achieve the most efficient mesh resolution, we varied cell size for five values of ΔxΔx = 0.1 m, 0.125 m, 0.15 m, 0.175 m and 0.20 m. Simulations were performed for all mesh sizes, and the results were compared in terms of convergence, stability and speed of simulation (Fig. 3). A linear wave with an amplitude of 1.5 m and a period of 6 s was used for these optimisation simulations. We considered wave time histories at two gauges A and B and recorded the waves from simulations using different mesh sizes (Fig. 3). Although the results are close (Fig. 3), some limited deviations are observed for larger mesh sizes of 0.20 m and 0.175 m. We therefore selected mesh size of 0.125 m as the optimum, giving an extra safety margin as a conservative solution.
The pressure from the incident waves on the vertical wall is validated in our model by comparing them with the analytical equation of Sainflou (1928), Eq. (3), which is one of the most common set of equations for design of coastal structures (Fig. 4). The model was tested by running a linear wave of period 6 s and wave amplitude of 1.5 m against the wall, with a still water level of 4.5 m. It can be seen that the model results are very close to those from analytical equations of Sainflou (1928), indicating that our numerical model is accurately modelling the wave-structure interaction (Fig. 4).
Eyewitness account analysis
Contemporary reporting of the 4th and 5th February 2014 storms by the main national news outlets in the UK highlights the extreme nature of the events and the significant damage and disruption they were likely to have on the communities of the south-west of England. In interviews, this was reinforced by Network Rail engineers who, even at this early stage, were forecasting remedial engineering works to last for at least 6 weeks. One week later, following subsequent storms the cascading nature of the events was obvious. Multiple breaches of the seawall had taken place with up to 35 separate landslide events and significant damage to parapet walls along the coastal route also were reported. Residents of the area reported extreme effects of the storm, one likening it to an earthquake and reporting water ingress through doors windows and even through vertical chimneys (Table 1). This suggests extreme wave overtopping volumes and large wave impact forces. One resident described the structural effects as: “the house was jumping up and down on its footings”.
Disaster management plans were quickly and effectively put into action by the local council, police service and National Rail. A major incident was declared, and decisions regarding evacuation of the residents under threat were taken around 2100 h on the night of 4th February when reports of initial damage to the seawall were received (Table 1). Local hotels were asked to provide short-term refuge to residents while local leisure facilities were prepared to accept residents later that evening. Initial repair work to the railway line was hampered by successive high spring tides and storms in the following days although significant progress was still made when weather conditions permitted (Table 1).
Sea level observations and spectral analysis
The results of surge and wave analyses are presented in Figs. 5 and 6. A surge height of up to 0.8 m was recorded in the examined tide gauge stations (Fig. 5b-d). Two main episodes of high surge heights are identified: the first surge started on 3rd February 2014 at 03:00 (UTC) and lasted until 4th of February 2014 at 00:00; the second event occurred in the period 4th February 2014 15:00 to 5th February 2014 at 17:00 (Fig. 5b-d). These data imply surge durations of 21 h and 26 h for the first and the second events, respectively. Based on the surge data in Fig. 5, we note that the storm event of early February 2014 and the associated surges was a relatively powerful one, which impacted at least 230 km of the south coast of England, from Land’s End to Weymouth, with large surge heights.
Based on wave buoy records, the maximum recorded amplitudes are at least 20.5 m in Dawlish and West Bay, 1.9 m in Tor Bay and 4.9 m in Chesil (Fig. 6a-b). The buoys at Tor Bay and Chesil recorded dual peak period bands of 4–8 and 8–12 s, whereas at Dawlish and West Bay registered triple peak period bands at 4–8, 8–12 and 20–25 s (Fig. 6c, d). It is important to note that the long-period waves at 20–25 s occur with short durations (approximately 2 min) while the waves at the other two bands of 4–8 and 8–12 s appear to be present at all times during the storm event.
The wave component at the period band of 4–8 s can be most likely attributed to normal coastal waves while the one at 8–12 s, which is longer, is most likely the swell component of the storm. Regarding the third component of the waves with long period of 20 -25 s, which occurs with short durations of 2 min, there are two hypotheses; it is either the result of a local (port and harbour) and regional (the Lyme Bay) oscillations (eg. Rabinovich 1997; Heidarzadeh and Satake 2014; Wang et al. 1992), or due to an abnormally long swell. To test the first hypothesis, we consider various water bodies such as Lyme Bay (approximate dimensions of 70 km × 20 km with an average water depth of 30 m; Fig. 6), several local bays (approximate dimensions of 3.6 km × 0.6 km with an average water depth of 6 m) and harbours (approximate dimensions of 0.5 km × 0.5 km with an average water depth of 4 m). Their water depths are based on the online Marine navigation website.Footnote8 According to Rabinovich (2010), the oscillation modes of a semi-enclosed rectangle basin are given by the following equation:
where TmnTmn is the oscillation period, gg is the gravitational acceleration, dd is the water depth, LL is the length of the basin, WW is the width of the basin, m=1,2,3,…m=1,2,3,… and n=0,1,2,3,…n=0,1,2,3,…; mm and nn are the counters of the different modes. Applying Eq. (5) to the aforementioned water bodies results in oscillation modes of at least 5 min, which is far longer than the observed period of 20–25 s. Therefore, we rule out the first hypothesis and infer that the long period of 20–25 s is most likely a long swell wave coming from distant sources. As discussed by Rabinovich (1997) and Wang et al. (2022), comparison between sea level spectra before and after the incident is a useful method to distinguish the spectrum of the weather event. A visual inspection of Fig. 6 reveals that the forcing at the period band of 20–25 s is non-existent before the incident.
Numerical simulations of wave loading and overtopping
Based on the results of sea level data analyses in the previous section (Fig. 6), we use a dual peak wave spectrum with peak periods of 10.0 s and 25.0 s for numerical simulations because such a wave would be comprised of the most energetic signals of the storm. For variations of water depth (2.0–4.0 m), coastal wave amplitude (0.5–1.5 m) (Fig. 7) and storm surge height (0.5–0.8 m) (Fig. 5), we developed 20 scenarios (Scn) which we used in numerical simulations (Table 2). Data during the incident indicated that water depth was up to the crest level of the seawall (approximately 4 m water depth); therefore, we varied water depth from 2 to 4 m in our simulation scenarios. Regarding wave amplitudes, we referred to the variations at a nearby tide gauge station (West Bay) which showed wave amplitude up to 1.2 m (Fig. 7). Therefore, wave amplitude was varied from 0.5 m to 1.5 m by considering a factor a safety of 25% for the maximum wave amplitude. As for the storm surge component, time series of storm surges calculated at three coastal stations adjacent to Dawlish showed that it was in the range of 0.5 m to 0.8 m (Fig. 5). These 20 scenarios would help to study uncertainties associated with wave amplitudes and pressures. Figure 8 shows snapshots of wave propagation and impacts on the seawall at different times.
Table 2 The 20 scenarios considered for numerical simulations in this study
Large wave amplitudes can induce significant wave forcing on the structure and cause overtopping of the seawall, which could eventually cascade to other hazards such as erosion of the backfill and scour (Adams and Heidarzadeh, 2021). The first 10 scenarios of our modelling efforts are for the same incident wave amplitudes of 0.5 m, which occur at different water depths (2.0–4.0 m) and storm surge heights (0.5–0.8 m) (Table 2 and Fig. 9). This is because we aim at studying the impacts of effective water depth (deff—the sum of mean sea level and surge height) on the time histories of wave amplitudes as the storm evolves. As seen in Fig. 9a, by decreasing effective water depth, wave amplitude increases. For example, for Scn-1 with effective depth of 4.5 m, the maximum amplitude of the first wave is 1.6 m, whereas it is 2.9 m for Scn-2 with effective depth of 3.5 m. However, due to intensive reflections and interferences of the waves in front of the vertical seawall, such a relationship is barely seen for the second and the third wave peaks. It is important to note that the later peaks (second or third) produce the largest waves rather than the first wave. Extraordinary wave amplifications are seen for the Scn-2 (deff = 3.5 m) and Scn-7 (deff = 3.3 m), where the corresponding wave amplitudes are 4.5 m and 3.7 m, respectively. This may indicate that the effective water depth of deff = 3.3–3.5 m is possibly a critical water depth for this structure resulting in maximum wave amplitudes under similar storms. In the second wave impact, the combined wave height (i.e. the wave amplitude plus the effective water depth), which is ultimately an indicator of wave overtopping, shows that the largest wave heights are generated by Scn-2, 7 and 8 (Fig. 9a) with effective water depths of 3.5 m, 3.3 m and 3.8 m and combined heights of 8.0 m, 7.0 m and 6.9 m (Fig. 9b). Since the height of seawall is 5.4 m, the combined wave heights for Scn-2, 7 and 8 are greater than the crest height of the seawall by 2.6 m, 1.6 m and 1.5 m, respectively, which indicates wave overtopping.
For scenarios 11–20 (Fig. 10), with incident wave amplitudes of 1.5 m (Table 2), the largest wave amplitudes are produced by Scn-17 (deff = 3.3 m), Scn-13 (deff = 2.5 m) and Scn-12 (deff = 3.5 m), which are 5.6 m, 5.1 m and 4.5 m. The maximum combined wave heights belong to Scn-11 (deff = 4.5 m) and Scn-17 (deff = 3.3 m), with combined wave heights of 9.0 m and 8.9 m (Fig. 10b), which are greater than the crest height of the seawall by 4.6 m and 3.5 m, respectively.
Our simulations for all 20 scenarios reveal that the first wave is not always the largest and wave interactions, reflections and interferences play major roles in amplifying the waves in front of the seawall. This is primarily because the wall is fully vertical and therefore has a reflection coefficient of close to one (i.e. full reflection). Simulations show that the combined wave height is up to 4.6 m higher than the crest height of the wall, implying that severe overtopping would be expected.
Results of wave loading calculations
The pressure calculations for scenarios 1–10 are given in Fig. 11 and those of scenarios 11–20 in Fig. 12. The total pressure distribution in Figs. 11, 12 mostly follows a triangular shape with maximum pressure at the seafloor as expected from the Sainflou (1928) design equations. These pressure plots comprise both static (due to mean sea level in front of the wall) and dynamic (combined effects of surge and wave) pressures. For incident wave amplitudes of 0.5 m (Fig. 11), the maximum wave pressure varies in the range of 35–63 kPa. At the sea surface, it is in the range of 4–20 kPa (Fig. 11). For some scenarios (Scn-2 and 7), the pressure distribution deviates from a triangular shape and shows larger pressures at the top, which is attributed to the wave impacts and partial breaking at the sea surface. This adds an additional triangle-shaped pressure distribution at the sea surface elevation consistent with the design procedure developed by Goda (2000) for braking waves. The maximum force on the seawall due to scenarios 1–10, which is calculated by integrating the maximum pressure distribution over the wave-facing surface of the seawall, is in the range of 92–190 KN (Table 2).
For scenarios 11–20, with incident wave amplitude of 1.5 m, wave pressures of 45–78 kPa and 7–120 kPa, for the bottom and top of the wall, respectively, were observed (Fig. 12). Most of the plots show a triangular pressure distribution, except for Scn-11 and 15. A significant increase in wave impact pressure is seen for Scn-15 at the top of the structure, where a maximum pressure of approximately 120 kPa is produced while other scenarios give a pressure of 7–32 kPa for the sea surface. In other words, the pressure from Scn-15 is approximately four times larger than the other scenarios. Such a significant increase of the pressure at the top is most likely attributed to the breaking wave impact loads as detailed by Goda (2000) and Cuomo et al. (2010). The wave simulation snapshots in Fig. 8 show that the wave breaks before reaching the wall. The maximum force due to scenarios 11–20 is 120–286 KN.
The breaking wave impacts peaking at 286 KN in our simulations suggest destabilisation of the upper masonry blocks, probably by grout malfunction. This significant impact force initiated the failure of the seawall which in turn caused extensive ballast erosion. Wave impact damage was proposed by Adams and Heidarzadeh (2021) as one of the primary mechanisms in the 2014 Dawlish disaster. In the multi-hazard risk model proposed by these authors, damage mechanism III (failure pathway 5 in Adams and Heidarzadeh, 2021) was characterised by wave impact force causing damage to the masonry elements, leading to failure of the upper sections of the seawall and loss of infill material. As blocks were removed, access to the track bed was increased for inbound waves allowing infill material from behind the seawall to be fluidised and subsequently removed by backwash. The loss of infill material critically compromised the stability of the seawall and directly led to structural failure. In parallel, significant wave overtopping (discussed in the next section) led to ballast washout and cascaded, in combination with masonry damage, to catastrophic failure of the wall and suspension of the rails in mid-air (Fig. 1b), leaving the railway inoperable for two months.
Wave Overtopping
The two most important factors contributing to the 2014 Dawlish railway catastrophe were wave impact forces and overtopping. Figure 13 gives the instantaneous overtopping rates for different scenarios, which experienced overtopping. It can be seen that the overtopping rates range from 0.5 m3/s/m to 16.1 m3/s/m (Fig. 13). Time histories of the wave overtopping rates show that the phenomenon occurs intermittently, and each time lasts 1.0–7.0 s. It is clear that the longer the overtopping time, the larger the volume of the water poured on the structure. The largest wave overtopping rates of 16.1 m3/s/m and 14.4 m3/s/m belong to Scn-20 and 11, respectively. These are the two scenarios that also give the largest combined wave heights (Fig. 10b).
The cumulative overtopping curves (Figs. 14, 15) show the total water volume overtopped the structure during the entire simulation time. This is an important hazard factor as it determines the level of soil saturation, water pore pressure in the soil and soil erosion (Van der Meer et al. 2018). The maximum volume belongs to Scn-20, which is 65.0 m3/m (m-cubed of water per metre length of the wall). The overtopping volumes are 42.7 m3/m for Scn-11 and 28.8 m3/m for Scn-19. The overtopping volume is in the range of 0.7–65.0 m3/m for all scenarios.
For comparison, we compare our modelling results with those estimated using empirical equations. For the case of the Dawlish seawall, we apply the equation proposed by Van Der Meer et al. (2018) to estimate wave overtopping rates, based on a set of decision criteria which are the influence of foreshore, vertical wall, possible breaking waves and low freeboard:
where qq is the mean overtopping rate per metre length of the seawall (m3/s/m), gg is the acceleration due to gravity, HmHm is the incident wave height at the toe of the structure, RcRc is the wall crest height above mean sea level, hshs is the deep-water significant wave height and e(x)e(x) is the exponential function. It is noted that Eq. (6) is valid for 0.1<RcHm<1.350.1<RcHm<1.35. For the case of the Dawlish seawall and considering the scenarios with larger incident wave amplitude of 1.5 m (hshs= 1.5 m), the incident wave height at the toe of the structure is HmHm = 2.2—5.6 m, and the wall crest height above mean sea level is RcRc = 0.6–2.9 m. As a result, Eq. (6) gives mean overtopping rates up to approximately 2.9 m3/s/m. A visual inspection of simulated overtopping rates in Fig. 13 for Scn 11–20 shows that the mean value of the simulated overtopping rates (Fig. 13) is close to estimates using Eq. (6).
Discussion and conclusions
We applied a combination of eyewitness account analysis, sea level data analysis and numerical modelling in combination with our engineering judgement to explain the damage to the Dawlish railway seawall in February 2014. Main findings are:
Eyewitness data analysis showed that the extreme nature of the event was well forecasted in the hours prior to the storm impact; however, the magnitude of the risks to the structures was not well understood. Multiple hazards were activated simultaneously, and the effects cascaded to amplify the damage. Disaster management was effective, exemplified by the establishment of an emergency rendezvous point and temporary evacuation centre during the storm, indicating a high level of hazard awareness and preparedness.
Based on sea level data analysis, we identified triple peak period bands at 4–8, 8–12 and 20–25 s in the sea level data. Storm surge heights and wave oscillations were up to 0.8 m and 1.5 m, respectively.
Based on the numerical simulations of 20 scenarios with different water depths, incident wave amplitudes, surge heights and peak periods, we found that the wave oscillations at the foot of the seawall result in multiple wave interactions and interferences. Consequently, large wave amplitudes, up to 4.6 m higher than the height of the seawall, were generated and overtopped the wall. Extreme impulsive wave impact forces of up to 286 KN were generated by the waves interacting with the seawall.
We measured maximum wave overtopping rates of 0.5–16.1 m3/s/m for our scenarios. The cumulative overtopping water volumes per metre length of the wall were 0.7–65.0 m3/m.
Analysis of all the evidence combined with our engineering judgement suggests that the most likely initiating cause of the failure was impulsive wave impact forces destabilising one or more grouted joints between adjacent masonry blocks in the wall. Maximum observed pressures of 286 KN in our simulations are four times greater in magnitude than background pressures leading to block removal and initiating failure. Therefore, the sequence of cascading events was :1) impulsive wave impact force causing damage to masonry, 2) failure of the upper sections of the seawall, 3) loss of infill resulting in a reduction of structural strength in the landward direction, 4) ballast washout as wave overtopping and inbound wave activity increased and 5) progressive structural failure following successive tides.
From a risk mitigation point of view, the stability of the seawall in the face of future energetic cyclonic storm events and sea level rise will become a critical factor in protecting the rail network. Mitigation efforts will involve significant infrastructure investment to strengthen the civil engineering assets combined with improved hazard warning systems consisting of meteorological forecasting and real-time wave observations and instrumentation. These efforts must take into account the amenity value of coastal railway infrastructure to local communities and the significant number of tourists who visit every year. In this regard, public awareness and active engagement in the planning and execution of the project will be crucial in order to secure local stakeholder support for the significant infrastructure project that will be required for future resilience.
Ackhurst M (2020) Design of Vertical Gravity Sea and Quay Walls. ICE Publishing, Westminster, London
Adams K, Heidarzadeh M (2021) A multi-hazard risk model with cascading failure pathways for the Dawlish (UK) railway using historical and contemporary data. Int J Disaster Risk Reduc 56:102082. https://doi.org/10.1016/j.ijdrr.2021.102082ArticleGoogle Scholar
Courant R, Friedrichs K, Lewy H (1928) Über die partiellen Differenzengleichungen der mathematischen Physik. Math Ann 100(1):32–74ArticleGoogle Scholar
Cuomo G, Allsop W, Bruce T, Pearson J (2010) Breaking wave loads at vertical seawalls and breakwaters. Coastal Eng 57(4):424–439ArticleGoogle Scholar
Dawson D, Shaw J, Gehrels WR (2016) Sea-level rise impacts on transport infrastructure: the notorious case of the coastal railway line at Dawlish, England. J Transport Geog 51:97–109ArticleGoogle Scholar
Feser F, Barcikowska M, Krueger O, Schenk F, Weisse R, Xia L (2015) Storminess over the North Atlantic and northwestern Europe—A review. Quarter J R Meteorol Soc 141:350–382. https://doi.org/10.1002/qj.2364ArticleGoogle Scholar
Heidarzadeh M, Satake K (2015) Source properties of the 1998 July 17 Papua New Guinea tsunami based on tide gauge records. Geophys J Int 202(1):361–369ArticleGoogle Scholar
Heidarzadeh M, Gusman A, Ishibe T, Sabeti R, Šepić J (2022) Estimating the eruption-induced water displacement source of the 15 January 2022 Tonga volcanic tsunami from tsunami spectra and numerical modelling. Ocean Eng 261:112165. https://doi.org/10.1016/j.oceaneng.2022.112165ArticleGoogle Scholar
Masselink G, Castelle B, Scott T, Dodet G, Suanez S, Jackson D, Floc’h F (2016) Extreme wave activity during 2013/2014 winter and morphological impacts along the Atlantic coast of Europe. Geophys Res Lett 43(5):2135–2143. https://doi.org/10.1002/2015GL067492ArticleGoogle Scholar
Mathworks (2018) MATLAB, 2018. 9.7.0.1190202 (R2019b), Natick, Massachusetts: The MathWorks Inc.
Van der Meer JW, Allsop NWH, Bruce T, De Rouck J, Kortenhaus A, Pullen T, Schüttrumpf H, Troch P and Zanuttigh B (2018) EurOtop, Manual on wave overtopping of sea defences and related structures. An overtopping manual largely based on European research, but for worldwide application. Available online at: www.overtopping-manual.com.
Wang Y, Su HY, Ren Z, Ma Y (2022) Source properties and resonance characteristics of the tsunami generated by the 2021 M 8.2 Alaska earthquake. J Geophys Res Oceans, 127(3):e2021JC018308. https://doi.org/10.1029/2021JC018308ArticleGoogle Scholar
We are grateful to Brunel University London for administering the scholarship awarded to KA. The Flow3D-Hydro used in this research for numerical modelling is licenced to Brunel University London through an academic programme contract. We sincerely thank Prof Harsh Gupta (Editor-in-Chief) and two anonymous reviewers for their constructive review comments.
Funding
This project was funded by the UK Engineering and Physical Sciences Research Council (EPSRC) through a PhD scholarship to Keith Adams.
Author information
Authors and Affiliations
Department of Civil and Environmental Engineering, Brunel University London, Uxbridge, UB8 3PH, UKKeith Adams
Department of Architecture and Civil Engineering, University of Bath, Bath, BA2 7AY, UKMohammad Heidarzadeh
The authors have no relevant financial or non-financial interests to disclose.
Availability of data
All data used in this study are provided in the body of the article.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Adams, K., Heidarzadeh, M. Extratropical cyclone damage to the seawall in Dawlish, UK: eyewitness accounts, sea level analysis and numerical modelling. Nat Hazards (2022). https://doi.org/10.1007/s11069-022-05692-2
TianLiabJ.M.T.DaviesaXiangzhenZhuc aUniversity of Birmingham, Birmingham B15 2TT, United Kingdom bGrainger and Worrall Ltd, Bridgnorth WV15 5HP, United Kingdom cBrunel Centre for Advanced Solidification Technology, Brunel University London, Kingston Ln, London, Uxbridge UB8 3PH, United Kingdom
Abstract
An entrainment defect (also known as a double oxide film defect or bifilm) acts a void containing an entrapped gas when submerged into a light-alloy melt, thus reducing the quality and reproducibility of the final castings. Previous publications, carried out with Al-alloy castings, reported that this trapped gas could be subsequently consumed by the reaction with the surrounding melt, thus reducing the void volume and negative effect of entrainment defects. Compared with Al-alloys, the entrapped gas within Mg-alloy might be more efficiently consumed due to the relatively high reactivity of magnesium. However, research into the entrainment defects within Mg alloys has been significantly limited. In the present work, AZ91 alloy castings were produced under different carrier gas atmospheres (i.e., SF6/CO2, SF6/air). The evolution processes of the entrainment defects contained in AZ91 alloy were suggested according to the microstructure inspections and thermodynamic calculations. The defects formed in the different atmospheres have a similar sandwich-like structure, but their oxide films contained different combinations of compounds. The use of carrier gases, which were associated with different entrained-gas consumption rates, affected the reproducibility of AZ91 castings.
연행 결함(이중 산화막 결함 또는 이중막이라고도 함)은 경합금 용융물에 잠길 때 갇힌 가스를 포함하는 공극으로 작용하여 최종 주물의 품질과 재현성을 저하시킵니다. Al-합금 주물을 사용하여 수행된 이전 간행물에서는 이 갇힌 가스가 주변 용융물과의 반응에 의해 후속적으로 소모되어 공극 부피와 연행 결함의 부정적인 영향을 줄일 수 있다고 보고했습니다. Al-합금에 비해 마그네슘의 상대적으로 높은 반응성으로 인해 Mg-합금 내에 포집된 가스가 더 효율적으로 소모될 수 있습니다. 그러나 Mg 합금 내 연행 결함에 대한 연구는 상당히 제한적이었습니다. 현재 작업에서 AZ91 합금 주물은 다양한 캐리어 가스 분위기(즉, SF6/CO2, SF6/공기)에서 생산되었습니다. AZ91 합금에 포함된 연행 결함의 진화 과정은 미세 조직 검사 및 열역학 계산에 따라 제안되었습니다. 서로 다른 분위기에서 형성된 결함은 유사한 샌드위치 구조를 갖지만 산화막에는 서로 다른 화합물 조합이 포함되어 있습니다. 다른 동반 가스 소비율과 관련된 운반 가스의 사용은 AZ91 주물의 재현성에 영향을 미쳤습니다.
As the lightest structural metal available on Earth, magnesium became one of the most attractive light metals over the last few decades. The magnesium industry has consequently experienced a rapid development in the last 20 years [1,2], indicating a large growth in demand for Mg alloys all over the world. Nowadays, the use of Mg alloys can be found in the fields of automobiles, aerospace, electronics and etc.[3,4]. It has been predicted that the global consumption of Mg metals will further increase in the future, especially in the automotive industry, as the energy efficiency requirement of both traditional and electric vehicles further push manufactures lightweight their design [3,5,6].
The sustained growth in demand for Mg alloys motivated a wide interest in the improvement of the quality and mechanical properties of Mg-alloy castings. During a Mg-alloy casting process, surface turbulence of the melt can lead to the entrapment of a doubled-over surface film containing a small quantity of the surrounding atmosphere, thus forming an entrainment defect (also known as a double oxide film defect or bifilm) [7], [8], [9], [10]. The random size, quantity, orientation, and placement of entrainment defects are widely accepted to be significant factors linked to the variation of casting properties [7]. In addition, Peng et al. [11] found that entrained oxides films in AZ91 alloy melt acted as filters to Al8Mn5 particles, trapping them as they settle. Mackie et al. [12] further suggested that entrained oxide films can act to trawl the intermetallic particles, causing them to cluster and form extremely large defects. The clustering of intermetallic compounds made the entrainment defects more detrimental for the casting properties.
Most of the previous studies regarding entrainment defects were carried out on Al-alloys [7,[13], [14], [15], [16], [17], [18], and a few potential methods have been suggested for diminishing their negative effect on the quality of Al-alloy castings. Nyahumwa et al.,[16] shows that the void volume within entrainment defects could be reduced by a hot isostatic pressing (HIP) process. Campbell [7] suggested the entrained gas within the defects could be consumed due to reaction with the surrounding melt, which was further verified by Raiszedeh and Griffiths [19].The effect of the entrained gas consumption on the mechanical properties of Al-alloy castings has been investigated by [8,9], suggesting that the consumption of the entrained gas promoted the improvement of the casting reproducibility.
Compared with the investigation concerning the defects within Al-alloys, research into the entrainment defects within Mg-alloys has been significantly limited. The existence of entrainment defects has been demonstrated in Mg-alloy castings [20,21], but their behaviour, evolution, as well as entrained gas consumption are still not clear.
In a Mg-alloy casting process, the melt is usually protected by a cover gas to avoid magnesium ignition. The cavities of sand or investment moulds are accordingly required to be flushed with the cover gas prior to the melt pouring [22]. Therefore, the entrained gas within Mg-alloy castings should contain the cover gas used in the casting process, rather than air only, which may complicate the structure and evolution of the corresponding entrainment defects.
SF6 is a typical cover gas widely used for Mg-alloy casting processes [23], [24], [25]. Although this cover gas has been restricted to use in European Mg-alloy foundries, a commercial report has pointed out that this cover is still popular in global Mg-alloy industry, especially in the countries which dominated the global Mg-alloy production, such as China, Brazil, India, etc. [26]. In addition, a survey in academic publications also showed that this cover gas was widely used in recent Mg-alloy studies [27]. The protective mechanism of SF6 cover gas (i.e., the reaction between liquid Mg-alloy and SF6 cover gas) has been investigated by several previous researchers, but the formation process of the surface oxide film is still not clearly understood, and even some published results are conflicting with each other. In early 1970s, Fruehling [28] found that the surface film formed under SF6 was MgO mainly with traces of fluorides, and suggested that SF6 was absorbed in the Mg-alloy surface film. Couling [29] further noticed that the absorbed SF6 reacted with the Mg-alloy melt to form MgF2. In last 20 years, different structures of the Mg-alloy surface films have been reported, as detailed below.(1)
Single-layered film. Cashion [30,31] used X-ray Photoelectron Spectroscopy (XPS) and Auger Spectroscopy (AES) to identify the surface film as MgO and MgF2. He also found that composition of the film was constant throughout the thickness and the whole experimental holding time. The film observed by Cashion had a single-layered structure created from a holding time from 10 min to 100 min.(2)
Double-layered film. Aarstad et. al [32] reported a doubled-layered surface oxide film in 2003. They observed several well-distributed MgF2 particles attached to the preliminary MgO film and grew until they covered 25–50% of the total surface area. The inward diffusion of F through the outer MgO film was the driving force for the evolution process. This double-layered structure was also supported by Xiong’s group [25,33] and Shih et al. [34].(3)
Triple-layered film. The triple-layered film and its evolution process were reported in 2002 by Pettersen [35]. Pettersen found that the initial surface film was a MgO phase and then gradually evolved to the stable MgF2 phase by the inward diffusion of F. In the final stage, the film has a triple-layered structure with a thin O-rich interlayer between the thick top and bottom MgF2 layers.(4)
Oxide film consisted of discrete particles. Wang et al [36] stirred the Mg-alloy surface film into the melt under a SF6 cover gas, and then inspect the entrained surface film after the solidification. They found that the entrained surface films were not continues as the protective surface films reported by other researchers but composed of discrete particles. The young oxide film was composed of MgO nano-sized oxide particles, while the old oxide films consist of coarse particles (about 1 µm in average size) on one side that contained fluorides and nitrides.
The oxide films of a Mg-alloy melt surface or an entrained gas are both formed due to the reaction between liquid Mg-alloy and the cover gas, thus the above-mentioned research regarding the Mg-alloy surface film gives valuable insights into the evolution of entrainment defects. The protective mechanism of SF6 cover gas (i.e., formation of a Mg-alloy surface film) therefore indicated a potential complicated evolution process of the corresponding entrainment defects.
However, it should be noted that the formation of a surface film on a Mg-alloy melt is in a different situation to the consumption of an entrained gas that is submerged into the melt. For example, a sufficient amount of cover gas was supported during the surface film formation in the studies previously mentioned, which suppressed the depletion of the cover gas. In contrast, the amount of entrained gas within a Mg-alloy melt is finite, and the entrained gas may become fully depleted. Mirak [37] introduced 3.5%SF6/air bubbles into a pure Mg-alloy melt solidifying in a specially designed permanent mould. It was found that the gas bubbles were entirely consumed, and the corresponding oxide film was a mixture of MgO and MgF2. However, the nucleation sites (such as the MgF2 spots observed by Aarstad [32] and Xiong [25,33]) were not observed. Mirak also speculated that the MgF2 formed prior to MgO in the oxide film based on the composition analysis, which was opposite to the surface film formation process reported in previous literatures (i.e., MgO formed prior to MgF2). Mirak’s work indicated that the oxide-film formation of an entrained gas may be quite different from that of surface films, but he did not reveal the structure and evolution of the oxide films.
In addition, the use of carrier gas in the cover gases also influenced the reaction between the cover gas and the liquid Mg-alloy. SF6/air required a higher content of SF6 than did a SF6/CO2 carrier gas [38], to avoid the ignition of molten magnesium, revealing different gas-consumption rates. Liang et.al [39] suggested that carbon was formed in the surface film when CO2 was used as a carrier gas, which was different from the films formed in SF6/air. An investigation into Mg combustion [40] reported a detection of Mg2C3 in the Mg-alloy sample after burning in CO2, which not only supported Liang’s results, but also indicated a potential formation of Mg carbides in double oxide film defects.
The work reported here is an investigation into the behaviour and evolution of entrainment defects formed in AZ91 Mg-alloy castings, protected by different cover gases (i.e., SF6/air and SF6/CO2). These carrier gases have different protectability for liquid Mg alloy, which may be therefore associated with different consumption rates and evolution processes of the corresponding entrained gases. The effect of the entrained-gas consumption on the reproducibility of AZ91 castings was also studied.
2. Experiment
2.1. Melting and casting
Three kilograms AZ91 alloy was melted in a mild steel crucible at 700 ± 5 °C. The composition of the AZ91 alloy has been shown in Table 1. Prior to heating, all oxide scale on the ingot surface was removed by machining. The cover gases used were 0.5%SF6/air or 0.5%SF6/CO2 (vol.%) at a flow rate of 6 L/min for different castings. The melt was degassed by argon with a flow rate of 0.3 L/min for 15 min [41,42], and then poured into sand moulds. Prior to pouring, the sand mould cavity was flushed with the cover gas for 20 min [22]. The residual melt (around 1 kg) was solidified in the crucible.
Table 1. Composition (wt.%) of the AZ91 alloy used in this study.
Al
Zn
Mn
Si
Fe
Ni
Mg
9.4
0.61
0.15
0.02
0.005
0.0017
Residual
Fig. 1(a) shows the dimensions of the casting with runners. A top-filling system was deliberately used to generate entrainment defects in the final castings. Green and Campbell [7,43] suggested that a top-filling system caused more entrainment events (i.e., bifilms) during a casting process, compared with a bottom-filling system. A melt flow simulation (Flow-3D software) of this mould, using Reilly’s model [44] regarding the entrainment events, also predicted that a large amount of bifilms would be contained in the final casting (denoted by the black particles in Fig. 1b).
Shrinkage defects also affect the mechanical properties and reproducibility of castings. Since this study focused on the effect of bifilms on the casting quality, the mould has been deliberately designed to avoid generating shrinkage defects. A solidification simulation using ProCAST software showed that no shrinkage defect would be contained in the final casting, as shown in Fig. 1c. The casting soundness has also been confirmed using a real time X-ray prior to the test bar machining.
The sand moulds were made from resin-bonded silica sand, containing 1wt. % PEPSET 5230 resin and 1wt. % PEPSET 5112 catalyst. The sand also contained 2 wt.% Na2SiF6 to act as an inhibitor [45]. The pouring temperature was 700 ± 5 °C. After the solidification, a section of the runner bars was sent to the Sci-Lab Analytical Ltd for a H-content analysis (LECO analysis), and all the H-content measurements were carried out on the 5th day after the casting process. Each of the castings was machined into 40 test bars for a tensile strength test, using a Zwick 1484 tensile test machine with a clip extensometer. The fracture surfaces of the broken test bars were examined using Scanning Electron Microscope (SEM, Philips JEOL7000) with an accelerating voltage of 5–15 kV. The fractured test bars, residual Mg-alloy solidified in the crucible, and the casting runners were then sectioned, polished and also inspected using the same SEM. The cross-section of the oxide film found on the test-bar fracture surface was exposed by the Focused Ion Beam milling technique (FIB), using a CFEI Quanta 3D FEG FIB-SEM. The oxide film required to be analysed was coated with a platinum layer. Then, a gallium ion beam, accelerated to 30 kV, milled the material substrate surrounding the platinum coated area to expose the cross section of the oxide film. EDS analysis of the oxide film’s cross section was carried out using the FIB equipment at accelerating voltage of 30 kV.
2.2. Oxidation cell
As previously mentioned, several past researchers investigated the protective film formed on a Mg-alloy melt surface [38,39,[46], [47], [48], [49], [50], [51], [52]. During these experiments, the amount of cover gas used was sufficient, thus suppressing the depletion of fluorides in the cover gas. The experiment described in this section used a sealed oxidation cell, which limited the supply of cover gas, to study the evolution of the oxide films of entrainment defects. The cover gas contained in the oxidation cell was regarded as large-size “entrained bubble”.
As shown in Fig. 2, the main body of the oxidation cell was a closed-end mild steel tube which had an inner length of 400 mm, and an inner diameter of 32 mm. A water-cooled copper tube was wrapped around the upper section of the cell. When the tube was heated, the cooling system created a temperature difference between the upper and lower sections, causing the interior gas to convect within the tube. The temperature was monitored by a type-K thermocouple located at the top of the crucible. Nie et al. [53] suggested that the SF6 cover gas would react with the steel wall of the holding furnace when they investigated the surface film of a Mg-alloy melt. To avoid this reaction, the interior surface of the steel oxidation cell (shown in Fig. 2) and the upper half section of the thermocouple were coated with boron nitride (the Mg-alloy was not in contact with boron nitride).
During the experiment, a block of solid AZ91 alloy was placed in a magnesia crucible located at the bottom of the oxidation cell. The cell was heated to 100 °C in an electric resistance furnace under a gas flow rate of 1 L/min. The cell was held at this temperature for 20 min, to replace the original trapped atmosphere (i.e. air). Then, the oxidation cell was further heated to 700 °C, melting the AZ91 sample. The gas inlet and exit valves were then closed, creating a sealed environment for oxidation under a limited supply of cover gas. The oxidation cell was then held at 700 ± 10 °C for periods of time from 5 min to 30 min in 5-min intervals. At the end of each holding time, the cell was quenched in water. After cooling to room temperature, the oxidised sample was sectioned, polished, and subsequently examined by SEM.
3. Results
3.1. Structure and composition of the entrainment defects formed in SF6/air
The structure and composition of the entrainment defect formed in the AZ91 castings under a cover gas of 0.5%SF6/air was observed by SEM and EDS. The results indicate that there exist two types of entrainment defects which are sketched in Fig. 3: (1) Type A defect whose oxide film has a traditional single-layered structure and (2) Type B defect, whose oxide film has two layers. The details of these defects were introduced in the following. Here it should be noticed that, as the entrainment defects are also known as biofilms or double oxide film, the oxide films of Type B defect were referred to as “multi-layered oxide film” or “multi-layered structure” in the present work to avoid a confusing description such as “the double-layered oxide film of a double oxide film defect”.
Fig. 4(a-b) shows a Type A defect having a compact single-layered oxide film with about 0.4 µm thickness. Oxygen, fluorine, magnesium and aluminium were detected in this film (Fig. 4c). It is speculated that oxide film is the mixture of fluoride and oxide of magnesium and aluminium. The detection of fluorine revealed that an entrained cover gas was contained in the formation of this defect. That is to say that the pores shown in Fig. 4(a) were not shrinkage defects or hydrogen porosity, but entrainment defects. The detection of aluminium was different with Xiong and Wang’s previous study [47,48], which showed that no aluminium was contained in their surface film of an AZ91 melt protected by a SF6 cover gas. Sulphur could not be clearly recognized in the element map, but there was a S-peak in the corresponding ESD spectrum.
Fig. 5(a-b) shows a Type B entrainment defect having a multi-layered oxide film. The compact outer layers of the oxide films were enriched with fluorine and oxygen (Fig. 5c), while their relatively porous inner layers were only enriched with oxygen (i.e., poor in fluorine) and partly grew together, thus forming a sandwich-like structure. Therefore, it is speculated that the outer layer is the mixture of fluoride and oxide, while the inner layer is mainly oxide. Sulphur could only be recognized in the EDX spectrum and could not be clearly identified in the element map, which might be due to the small S-content in the cover gas (i.e., 0.5% volume content of SF6 in the cover gas). In this oxide film, aluminium was contained in the outer layer of this oxide film but could not be clearly detected in the inner layer. Moreover, the distribution of Al seems to be uneven. It can be found that, in the right side of the defect, aluminium exists in the film but its concentration can not be identified to be higher than the matrix. However, there is a small area with much higher aluminium concentration in the left side of the defect. Such an uneven distribution of aluminium was also observed in other defects (shown in the following), and it is the result of the formation of some oxide particles in or under the film.
Figs. 4 and 5 show cross sectional observations of the entrainment defects formed in the AZ91 alloy sample cast under a cover gas of SF6/air. It is not sufficient to characterize the entrainment defects only by the figures observed from the two-dimensional section. To have a further understanding, the surface of the entrainment defects (i.e. the oxide film) was further studied by observing the fracture surface of the test bars.
Fig. 6(a) shows fracture surfaces of an AZ91 alloy tensile test bar produced in SF6/air. Symmetrical dark regions can be seen on both sides of the fracture surfaces. Fig. 6(b) shows boundaries between the dark and bright regions. The bright region consisted of jagged and broken features, while the surface of the dark region was relatively smooth and flat. In addition, the EDS results (Fig. 6c-d and Table 2) show that fluorine, oxygen, sulphur, and nitrogen were only detected in the dark regions, indicating that the dark regions were surface protective films entrained into the melt. Therefore, it could be suggested that the dark regions were an entrainment defect with consideration of their symmetrical nature. Similar defects on fracture surfaces of Al-alloy castings have been previously reported [7]. Nitrides were only found in the oxide films on the test-bar fracture surfaces but never detected in the cross-sectional samples shown in Figs. 4 and 5. An underlying reason is that the nitrides contained in these samples may have hydrolysed during the sample polishing process [54].
Table 2. EDS results (wt.%) corresponding to the regions shown in Fig. 6 (cover gas: SF6/air).
In conjunction with the cross-sectional observation of the defects shown in Figs. 4 and 5, the structure of an entrainment defect contained in a tensile test bar was sketched as shown in Fig. 6(e). The defect contained an entrained gas enclosed by its oxide film, creating a void section inside the test bar. When the tensile force applied on the defect during the fracture process, the crack was initiated at the void section and propagated along the entrainment defect, since cracks would be propagated along the weakest path [55]. Therefore, when the test bar was finally fractured, the oxide films of entrainment defect appeared on both fracture surfaces of the test bar, as shown in Fig. 6(a).
3.2. Structure and composition of the entrainment defects formed in SF6/CO2
Similar to the entrainment defect formed in SF6/air, the defects formed under a cover gas of 0.5%SF6/CO2 also had two types of oxide films (i.e., single-layered and multi-layered types). Fig. 7(a) shows an example of the entrainment defects containing a multi-layered oxide film. A magnified observation to the defect (Fig. 7b) shows that the inner layers of the oxide films had grown together, presenting a sandwich-like structure, which was similar to the defects formed in an atmosphere of SF6/air (Fig. 5b). An EDS spectrum (Fig. 7c) revealed that the joint area (inner layer) of this sandwich-like structure mainly contained magnesium oxides. Peaks of fluorine, sulphur, and aluminium were recognized in this EDS spectrum, but their amount was relatively small. In contrast, the outer layers of the oxide films were compact and composed of a mixture of fluorides and oxides (Fig. 7d-e).
Fig. 8(a) shows an entrainment defect on the fracture surfaces of an AZ91 alloy tensile test bar, which was produced in an atmosphere of 0.5%SF6/CO2. The corresponding EDS results (Table 3) showed that oxide film contained fluorides and oxides. Sulphur and nitrogen were not detected. Besides, a magnified observation (Fig. 8b) indicated spots on the oxide film surface. The diameter of the spots ranged from hundreds of nanometres to a few micron meters.
To further reveal the structure and composition of the oxide film clearly, the cross-section of the oxide film on a test-bar fracture surface was onsite exposed using the FIB technique (Fig. 9). As shown in Fig. 9a, a continuous oxide film was found between the platinum coating layer and the Mg-Al alloy substrate. Fig. 9 (b-c) shows a magnified observation to oxide films, indicating a multi-layered structure (denoted by the red box in Fig. 9c). The bottom layer was enriched with fluorine and oxygen and should be the mixture of fluoride and oxide, which was similar to the “outer layer” shown in Figs. 5 and 7, while the only-oxygen-enriched top layer was similar to the “inner layer” shown in Figs. 5 and 7.
Except the continuous film, some individual particles were also observed in or below the continuous film, as shown in Fig. 9. An Al-enriched particle was detected in the left side of the oxide film shown in Fig. 9b and might be speculated to be spinel Mg2AlO4 because it also contains abundant magnesium and oxygen elements. The existing of such Mg2AlO4 particles is responsible for the high concentration of aluminium in small areas of the observed film and the uneven distribution of aluminium, as shown in Fig. 5(c). Here it should be emphasized that, although the other part of the bottom layer of the continuous oxide film contains less aluminium than this Al-enriched particle, the Fig. 9c indicated that the amount of aluminium in this bottom layer was still non-negligible, especially when comparing with the outer layer of the film. Below the right side of the oxide film shown in Fig. 9b, a particle was detected and speculated to be MgO because it is rich in Mg and O. According to Wang’s result [56], lots of discrete MgO particles can be formed on the surface of the Mg melt by the oxidation of Mg melt and Mg vapor. The MgO particles observed in our present work may be formed due to the same reasons. While, due to the differences in experimental conditions, less Mg melt can be vapored or react with O2, thus only a few of MgO particles formed in our work. An enrichment of carbon was also found in the film, revealing that CO2 was able to react with the melt, thus forming carbon or carbides. This carbon concentration was consistent with the relatively high carbon content of the oxide film shown in Table 3 (i.e., the dark region). In the area next to the oxide film.
Table 3. EDS results (wt.%) corresponding to the regions shown in Fig. 8 (cover gas: SF6/ CO2).
This cross-sectional observation of the oxide film on a test bar fracture surface (Fig. 9) further verified the schematic of the entrainment defect shown in Fig. 6(e). The entrainment defects formed in different atmospheres of SF6/CO2 and SF6/air had similar structures, but their compositions were different.
3.3. Evolution of the oxide films in the oxidation cell
The results in Section 3.1 and 3.2 have shown the structures and compositions of entrainment defects formed in AZ91 castings under cover gases of SF6/air and SF6/CO2. Different stages of the oxidation reaction may lead to the different structures and compositions of entrainment defects. Although Campbell has conjectured that an entrained gas may react with the surrounding melt, it is rarely reported that the reaction occurring between the Mg-alloy melt and entrapped cover gas. Previous researchers normally focus on the reaction between a Mg-alloy melt and the cover gas in an open environment [38,39,[46], [47], [48], [49], [50], [51], [52], which was different from the situation of a cover gas trapped into the melt. To further understand the formation of the entrainment defect in an AZ91 alloy, the evolution process of oxide films of the entrainment defect was further studied using an oxidation cell.
Fig. 10 (a and d) shows a surface film held for 5 min in the oxidation cell, protected by 0.5%SF6/air. There was only one single layer consisting of fluoride and oxide (MgF2 and MgO). In this surface film. Sulphur was detected in the EDS spectrum, but its amount was too small to be recognized in the element map. The structure and composition of this oxide film was similar to the single-layered films of entrainment defects shown in Fig. 4.
After a holding time of 10 min, a thin (O, S)-enriched top layer (around 700 nm) appeared upon the preliminary F-enriched film, forming a multi-layered structure, as shown in Fig. 10(b and e). The thickness of the (O, S)-enriched top layer increased with increased holding time. As shown in Fig. 10(c and f), the oxide film held for 30 min also had a multi-layered structure, but the thickness of its (O, S)-enriched top layer (around 2.5 µm) was higher than the that of the 10-min oxide film. The multi-layered oxide films shown in Fig. 10(b-c) presented a similar appearance to the films of the sandwich-like defect shown in Fig. 5.
The different structures of the oxide films shown in Fig. 10 indicated that fluorides in the cover gas would be preferentially consumed due to the reaction with the AZ91 alloy melt. After the depletion of fluorides, the residual cover gas reacted further with the liquid AZ91 alloy, forming the top (O, S)-enriched layer in the oxide film. Therefore, the different structures and compositions of entrainment defects shown in Figs. 4 and 5 may be due to an ongoing oxidation reaction between melt and entrapped cover gas.
This multi-layered structure has not been reported in previous publications concerning the protective surface film formed on a Mg-alloy melt [38,[46], [47], [48], [49], [50], [51]. This may be due to the fact that previous researchers carried out their experiments with an un-limited amount of cover gas, creating a situation where the fluorides in the cover gas were not able to become depleted. Therefore, the oxide film of an entrainment defect had behaviour traits similar to the oxide films shown in Fig. 10, but different from the oxide films formed on the Mg-alloy melt surface reported in [38,[46], [47], [48], [49], [50], [51].
Similar with the oxide films held in SF6/air, the oxide films formed in SF6/CO2 also had different structures with different holding times in the oxidation cell. Fig. 11(a) shows an oxide film, held on an AZ91 melt surface under a cover gas of 0.5%SF6/CO2 for 5 min. This film had a single-layered structure consisting of MgF2. The existence of MgO could not be confirmed in this film. After the holding time of 30 min, the film had a multi-layered structure; the inner layer was of a compact and uniform appearance and composed of MgF2, while the outer layer is the mixture of MgF2 and MgO. Sulphur was not detected in this film, which was different from the surface film formed in 0.5%SF6/air. Therefore, fluorides in the cover gas of 0.5%SF6/CO2 were also preferentially consumed at an early stage of the film growth process. Compared with the film formed in SF6/air, the MgO in film formed in SF6/CO2 appeared later and sulphide did not appear within 30 min. It may mean that the formation and evolution of film in SF6/air is faster than SF6/CO2. CO2 may have subsequently reacted with the melt to form MgO, while sulphur-containing compounds accumulated in the cover gas and reacted to form sulphide in very late stage (may after 30 min in oxidation cell).
4. Discussion
4.1. Evolution of entrainment defects formed in SF6/air
HSC software from Outokumpu HSC Chemistry for Windows (http://www.hsc-chemistry.net/) was used to carry out thermodynamic calculations needed to explore the reactions which might occur between the trapped gases and liquid AZ91 alloy. The solutions to the calculations suggest which products are most likely to form in the reaction process between a small amount of cover gas (i.e., the amount within a trapped bubble) and the AZ91-alloy melt.
In the trials, the pressure was set to 1 atm, and the temperature set to 700 °C. The amount of the cover gas was assumed to be 7 × 10−7 kg, with a volume of approximately 0.57 cm3 (3.14 × 10−8 kmol) for 0.5%SF6/air, and 0.35 cm3 (3.12 × 10−8 kmol) for 0.5%SF6/CO2. The amount of the AZ91 alloy melt in contact with the trapped gas was assumed to be sufficient to complete all reactions. The decomposition products of SF6 were SF5, SF4, SF3, SF2, F2, S(g), S2(g) and F(g) [57], [58], [59], [60].
Fig. 12 shows the equilibrium diagram of the thermodynamic calculation of the reaction between the AZ91 alloy and 0.5%SF6/air. In the diagram, the reactants and products with less than 10−15 kmol have not been shown, as this was 5 orders of magnitude less than the amount of SF6 present (≈ 1.57 × 10−10 kmol) and therefore would not affect the observed process in a practical way.
This reaction process could be divided into 3 stages.
Stage 1: The formation of fluorides. the AZ91 melt preferentially reacted with SF6 and its decomposition products, producing MgF2, AlF3, and ZnF2. However, the amount of ZnF2 may have been too small to be detected practically (1.25 × 10−12 kmol of ZnF2 compared with 3 × 10−10 kmol of MgF2), which may be the reason why Zn was not detected in any the oxide films shown in Sections 3.1–3.3. Meanwhile, sulphur accumulated in the residual gas as SO2.
Stage 2: The formation of oxides. After the liquid AZ91 alloy had depleted all the available fluorides in the entrapped gas, the amount of AlF3 and ZnF2 quickly reduced due to a reaction with Mg. O2(g) and SO2 reacted with the AZ91 melt, forming MgO, Al2O3, MgAl2O4, ZnO, ZnSO4 and MgSO4. However, the amount of ZnO and ZnSO4 would have been too small to be found practically by EDS (e.g. 9.5 × 10−12 kmol of ZnO,1.38 × 10−14 kmol of ZnSO4, in contrast to 4.68 × 10−10 kmol of MgF2, when the amount of AZ91 on the X-axis is 2.5 × 10−9 kmol). In the experimental cases, the concentration of F in the cover gas is very low, whole the concentration f O is much higher. Therefore, the stage 1 and 2, i.e, the formation of fluoride and oxide may happen simultaneously at the beginning of the reaction, resulting in the formation of a singer-layered mixture of fluoride and oxide, as shown in Figs. 4 and 10(a). While an inner layer consisted of oxides but fluorides could form after the complete depletion of F element in the cover gas.
Stages 1- 2 theoretically verified the formation process of the multi-layered structure shown in Fig. 10.
The amount of MgAl2O4 and Al2O3 in the oxide film was of a sufficient amount to be detected, which was consistent with the oxide films shown in Fig. 4. However, the existence of aluminium could not be recognized in the oxide films grown in the oxidation cell, as shown in Fig. 10. This absence of Al may be due to the following reactions between the surface film and AZ91 alloy melt:(1)
Mg + MgAl2O4 = MgO + Al, △G(700 °C) =-106.34 kJ/molwhich could not be simulated by the HSC software since the thermodynamic calculation was carried out under an assumption that the reactants were in full contact with each other. However, in a practical process, the AZ91 melt and the cover gas would not be able to be in contact with each other completely, due to the existence of the protective surface film.
Stage 3: The formation of Sulphide and nitride. After a holding time of 30 min, the gas-phase fluorides and oxides in the oxidation cell had become depleted, allowing the melt reaction with the residual gas, forming an additional sulphur-enriched layer upon the initial F-enriched or (F, O)-enriched surface film, thus resulting in the observed multi-layered structure shown in Fig. 10 (b and c). Besides, nitrogen reacted with the AZ91 melt until all reactions were completed. The oxide film shown in Fig. 6 may correspond to this reaction stage due to its nitride content. However, the results shows that the nitrides were not detected in the polished samples shown in Figs. 4 and 5, but only found on the test bar fracture surfaces. The nitrides may have hydrolysed during the sample preparation process, as follows [54]:(3)
Mg3N2 + 6H2O =3Mg(OH)2 + 2NH3↑(4)
AlN+ 3H2O =Al(OH)3 + NH3↑
In addition, Schmidt et al. [61] found that Mg3N2 and AlN could react to form ternary nitrides (Mg3AlnNn+2, n= 1, 2, 3…). HSC software did not contain the database of ternary nitrides, and it could not be added into the calculation. The oxide films in this stage may also contain ternary nitrides.
4.2. Evolution of entrainment defects formed in SF6/CO2
Fig. 13 shows the results of the thermodynamic calculation between AZ91 alloy and 0.5%SF6/CO2. This reaction processes can also be divided into three stages.
Stage 1: The formation of fluorides. SF6 and its decomposition products were consumed by the AZ91 melt, forming MgF2, AlF3, and ZnF2. As in the reaction of AZ91 in 0.5%SF6/air, the amount of ZnF2 was too small to be detected practically (1.51 × 10−13 kmol of ZnF2 compared with 2.67 × 10−10 kmol of MgF2). Sulphur accumulated in the residual trapped gas as S2(g) and a portion of the S2(g) reacted with CO2, to form SO2 and CO. The products in this reaction stage were consistent with the film shown in Fig. 11(a), which had a single layer structure that contained fluorides only.
Stage 2: The formation of oxides. AlF3 and ZnF2 reacted with the Mg in the AZ91 melt, forming MgF2, Al and Zn. The SO2 began to be consumed, producing oxides in the surface film and S2(g) in the cover gas. Meanwhile, the CO2 directly reacted with the AZ91 melt, forming CO, MgO, ZnO, and Al2O3. The oxide films shown in Figs. 9 and 11(b) may correspond to this reaction stage due to their oxygen-enriched layer and multi-layered structure.
The CO in the cover gas could further react with the AZ91 melt, producing C. This carbon may further react with Mg to form Mg carbides, when the temperature reduced (during solidification period) [62]. This may be the reason for the high carbon content in the oxide film shown in Figs. 8–9. Liang et al. [39] also reported carbon-detection in an AZ91 alloy surface film protected by SO2/CO2. The produced Al2O3 may be further combined with MgO, forming MgAl2O4[63]. As discussed in Section 4.1, the alumina and spinel can react with Mg, causing an absence of aluminium in the surface films, as shown in Fig. 11.
Stage 3: The formation of Sulphide. the AZ91 melt began to consume S2(g) in the residual entrapped gas, forming ZnS and MgS. These reactions did not occur until the last stage of the reaction process, which could be the reason why the S-content in the defect shown Fig. 7(c) was small.
In summary, thermodynamic calculations indicate that the AZ91 melt will react with the cover gas to form fluorides firstly, then oxides and sulphides in the last. The oxide film in the different reaction stages would have different structures and compositions.
4.3. Effect of the carrier gases on consumption of the entrained gas and the reproducibility of AZ91 castings
The evolution processes of entrainment defects, formed in SF6/air and SF6/CO2, have been suggested in Sections 4.1 and 4.2. The theoretical calculations were verified with respect to the corresponding oxide films found in practical samples. The atmosphere within an entrainment defect could be efficiently consumed due to the reaction with liquid Mg-alloy, in a scenario dissimilar to the Al-alloy system (i.e., nitrogen in an entrained air bubble would not efficiently react with Al-alloy melt [64,65], however, nitrogen would be more readily consumed in liquid Mg alloys, commonly referred to as “nitrogen burning” [66]).
The reaction between the entrained gas and the surrounding liquid Mg-alloy converted the entrained gas into solid compounds (e.g. MgO) within the oxide film, thus reducing the void volume of the entrainment defect and hence probably causing a collapse of the defect (e.g., if an entrained gas of air was depleted by the surrounding liquid Mg-alloy, under an assumption that the melt temperature is 700 °C and the depth of liquid Mg-alloy is 10 cm, the total volume of the final solid products would be 0.044% of the initial volume taken by the entrapped air).
The relationship between the void volume reduction of entrainment defects and the corresponding casting properties has been widely studied in Al-alloy castings. Nyahumwa and Campbell [16] reported that the Hot Isostatic Pressing (HIP) process caused the entrainment defects in Al-alloy castings to collapse and their oxide surfaces forced into contact. The fatigue lives of their castings were improved after HIP. Nyahumwa and Campbell [16] also suggested a potential bonding of the double oxide films that were in contact with each other, but there was no direct evidence to support this. This binding phenomenon was further investigated by Aryafar et.al.[8], who re-melted two Al-alloy bars with oxide skins in a steel tube and then carried out a tensile strength test on the solidified sample. They found that the oxide skins of the Al-alloy bars strongly bonded with each other and became even stronger with an extension of the melt holding time, indicating a potential “healing” phenomenon due to the consumption of the entrained gas within the double oxide film structure. In addition, Raidszadeh and Griffiths [9,19] successfully reduced the negative effect of entrainment defects on the reproducibility of Al-alloy castings, by extending the melt holding time before solidification, which allowed the entrained gas to have a longer time to react with the surrounding melt.
With consideration of the previous work mentioned, the consumption of the entrained gas in Mg-alloy castings may diminish the negative effect of entrainment defects in the following two ways.
(1) Bonding phenomenon of the double oxide films. The sandwich-like structure shown in Fig. 5 and 7 indicated a potential bonding of the double oxide film structure. However, more evidence is required to quantify the increase in strength due to the bonding of the oxide films.
(2) Void volume reduction of entrainment defects. The positive effect of void-volume reduction on the quality of castings has been widely demonstrated by the HIP process [67]. As the evolution processes discussed in Section 4.1–4.2, the oxide films of entrainment defects can grow together due to an ongoing reaction between the entrained gas and surrounding AZ91 alloy melt. The volume of the final solid products was significant small compared with the entrained gas (i.e., 0.044% as previously mentioned).
Therefore, the consumption rate of the entrained gas (i.e., the growth rate of oxide films) may be a critical parameter for improving the quality of AZ91 alloy castings. The oxide film growth rate in the oxidization cell was accordingly further investigated.
Fig. 14 shows a comparison of the surface film growth rates in different cover gases (i.e., 0.5%SF6/air and 0.5%SF6/CO2). 15 random points on each sample were selected for film thickness measurements. The 95% confidence interval (95%CI) was computed under an assumption that the variation of the film thickness followed a Gaussian distribution. It can be seen that all the surface films formed in 0.5%SF6/air grew faster than those formed in 0.5%SF6/CO2. The different growth rates suggested that the entrained-gas consumption rate of 0.5%SF6/air was higher than that of 0.5%SF6/CO2, which was more beneficial for the consumption of the entrained gas.
It should be noted that, in the oxidation cell, the contact area of liquid AZ91 alloy and cover gas (i.e. the size of the crucible) was relatively small with consideration of the large volume of melt and gas. Consequently, the holding time for the oxide film growth within the oxidation cell was comparatively long (i.e., 5–30 min). However, the entrainment defects contained in a real casting are comparatively very small (i.e., a few microns size as shown in Figs. 3–6, and [7]), and the entrained gas is fully enclosed by the surrounding melt, creating a relatively large contact area. Hence the reaction time for cover gas and the AZ91 alloy melt may be comparatively short. In addition, the solidification time of real Mg-alloy sand castings can be a few minutes (e.g. Guo [68] reported that a Mg-alloy sand casting with 60 mm diameter required 4 min to be solidified). Therefore, it can be expected that an entrained gas trapped during an Mg-alloy melt pouring process will be readily consumed by the surrounding melt, especially for sand castings and large-size castings, where solidification times are long.
Therefore, the different cover gases (0.5%SF6/air and 0.5%SF6/CO2) associated with different consumption rates of the entrained gases may affect the reproducibility of the final castings. To verify this assumption, the AZ91 castings produced in 0.5%SF6/air and 0.5%SF6/CO2 were machined into test bars for mechanical evaluation. A Weibull analysis was carried out using both linear least square (LLS) method and non-linear least square (non-LLS) method [69].
Fig. 15(a-b) shows a traditional 2-p linearized Weibull plot of the UTS and elongation of the AZ91 alloy castings, obtained by the LLS method. The estimator used is P= (i-0.5)/N, which was suggested to cause the lowest bias among all the popular estimators [69,70]. The casting produced in SF6/air has an UTS Weibull moduli of 16.9, and an elongation Weibull moduli of 5.0. In contrast, the UTS and elongation Weibull modulus of the casting produced in SF6/CO2 are 7.7 and 2.7 respectively, suggesting that the reproducibility of the casting protected by SF6/CO2 were much lower than that produced in SF6/air.
In addition, the author’s previous publication [69] demonstrated a shortcoming of the linearized Weibull plots, which may cause a higher bias and incorrect R2 interruption of the Weibull estimation. A Non-LLS Weibull estimation was therefore carried out, as shown in Fig. 15 (c-d). The UTS Weibull modulus of the SF6/air casting was 20.8, while the casting produced under SF6/CO2 had a lower UTS Weibull modulus of 11.4, showing a clear difference in their reproducibility. In addition, the SF6/air elongation (El%) dataset also had a Weibull modulus (shape = 5.8) higher than the elongation dataset of SF6/CO2 (shape = 3.1). Therefore, both the LLS and Non-LLS estimations suggested that the SF6/air casting has a higher reproducibility than the SF6/CO2 casting. It supports the method that the use of air instead of CO2 contributes to a quicker consumption of the entrained gas, which may reduce the void volume within the defects. Therefore, the use of 0.5%SF6/air instead of 0.5%SF6/CO2 (which increased the consumption rate of the entrained gas) improved the reproducibility of the AZ91 castings.
However, it should be noted that not all the Mg-alloy foundries followed the casting process used in present work. The Mg-alloy melt in present work was degassed, thus reducing the effect of hydrogen on the consumption of the entrained gas (i.e., hydrogen could diffuse into the entrained gas, potentially suppressing the depletion of the entrained gas [7,71,72]). In contrast, in Mg-alloy foundries, the Mg-alloy melt is not normally degassed, since it was widely believed that there is not a ‘gas problem’ when casting magnesium and hence no significant change in tensile properties[73]. Although studies have shown the negative effect of hydrogen on the mechanical properties of Mg-alloy castings [41,42,73], a degassing process is still not very popular in Mg-alloy foundries.
Moreover, in present work, the sand mould cavity was flushed with the SF6 cover gas prior to pouring [22]. However, not all the Mg-alloy foundries flushed the mould cavity in this way. For example, the Stone Foundry Ltd (UK) used sulphur powder instead of the cover-gas flushing. The entrained gas within their castings may be SO2/air, rather than the protective gas.
Therefore, although the results in present work have shown that using air instead of CO2 improved the reproducibility of the final casting, it still requires further investigations to confirm the effect of carrier gases with respect to different industrial Mg-alloy casting processes.
7. Conclusion
Entrainment defects formed in an AZ91 alloy were observed. Their oxide films had two types of structure: single-layered and multi-layered. The multi-layered oxide film can grow together forming a sandwich-like structure in the final casting.2.
Both the experimental results and the theoretical thermodynamic calculations demonstrated that fluorides in the trapped gas were depleted prior to the consumption of sulphur. A three-stage evolution process of the double oxide film defects has been suggested. The oxide films contained different combinations of compounds, depending on the evolution stage. The defects formed in SF6/air had a similar structure to those formed in SF6/CO2, but the compositions of their oxide films were different. The oxide-film formation and evolution process of the entrainment defects were different from that of the Mg-alloy surface films previous reported (i.e., MgO formed prior to MgF2).3.
The growth rate of the oxide film was demonstrated to be greater under SF6/air than SF6/CO2, contributing to a quicker consumption of the damaging entrapped gas. The reproducibility of an AZ91 alloy casting improved when using SF6/air instead of SF6/CO2.
Acknowledgements
The authors acknowledge funding from the EPSRC LiME grant EP/H026177/1, and the help from Dr W.D. Griffiths and Mr. Adrian Carden (University of Birmingham). The casting work was carried out in University of Birmingham.
N.M.T De Silva University of Colombo School of Computing 2018
Abstract
산사태가 발생하기 쉬운 구릉 지역 근처에서 발생하는 최근 인구 증가 및 개발은 취약성을 증가시킵니다. 기후 변화의 영향은 산사태 위험의 가능성을 더욱 높입니다. 따라서 인명 및 재산 피해를 방지하기 위해서는 불안정한 경사면 거동에 대한 적절한 관찰과 분석이 중요합니다.
산사태 흐름 경로 예측은 산사태 흐름 경로를 결정하는 데 중요하며 위험 매핑의 필수 요소입니다. 그러나 현상의 복잡한 특성과 관련 매개변수의 불확실성으로 인해 흐름 경로 예측은 어려운 작업입니다. 이 작업에서는 Kegalle 지역의 Aranayaka 지역의 주요 산사태 사고를 흐름 경로를 모델링하기 위한 사례 연구로 사용합니다.
위치에서 디지털 고도 모델을 기반으로 잠재적 소스 영역이 식별되었습니다. 확산 영역 평가는 D8 및 다중 방향 흐름 알고리즘이라는 두 가지 흐름 방향 알고리즘을 기반으로 했습니다. 이 프로토타입 모델을 사용하여 사용자는 슬라이드의 최대 너비, 런아웃 거리 및 슬립 표면적과 같은 산사태 관련 통계를 대화식으로 얻을 수 있습니다.
모델에서 얻은 결과는 실제 Aranayaka 산사태 데이터 세트와 해당 지역의 산사태 위험 지도와 비교되었습니다. D8 알고리즘을 사용하여 구현된 도구에서 생성된 산사태 흐름 경로는 65% 이상의 일치를 나타내고 다중 방향 흐름 알고리즘은 실제 흐름 경로 및 기타 관련 통계와 69% 이상의 일치를 나타냅니다.
또한, 생성된 유동 경로 방향과 예상되는 산사태 시작 지점이 실제 산사태 경계 내부에 잘 일치합니다.
Recent population growth and developments taking place close to landslides prone hilly areas increase their vulnerability. Climate change impacts further raise the potential of landslide hazard. Therefore, to prevent loss of lives and damage to property, proper observation and analysis of unstable slope behavior is crucial. Landslide flow path forecasting is important for determining a landslide flow route and it is an essential element in hazard mapping. However, due to the complex nature of the phenomenon and the uncertainties of associated parameters flow path prediction is a challenging task. In this work, the major landslide incident at Aranayaka area in Kegalle district is taken as the case study to model the flow path. At the location, potential source areas were identified on the basis of the Digital Elevation Model. Spreading area assessment was based on two flow directional algorithms namely D8 and Multiple Direction Flow Algorithm. Using this prototype model, a user can interactively get landslide specific statistics such as the maximum width of the slide, runout distance, and slip surface area. Results obtained by the model were compared with the actual Aranayaka landslide data set the landslide hazard map of the area. Landslide flow paths generated from the implemented tool using D8 algorithm shows more than 65% agreement and Multiple Direction Flow Algorithm shows more than 69% agreement with the actual flow paths and other related statistics. Also, the generated flow path directions and predicted possible landslide initiation points fit inside the actual landslide boundary with good agreement.
References
[1] na, “Landslides 101,” [Online]. Available: https://landslides.usgs.gov/learn/ls101.php. [Accessed 20 01 2017]. [2] U.S. Department of the Interior, “Landslide Types and Processes,” July 2004. [Online]. Available: https://pubs.usgs.gov/fs/2004/3072/fs-2004-3072.html. [Accessed 20 July 2017]. [3] “Department of conservation,” 2016. [Online]. Available: http://www.conservation.ca.gov/cgs/geologic_hazards/landslides/Pages/LandslideTypes.aspx. [Accessed 20 July 2017]. [4] 22nd May 2016, Aerial Survey report on inundation damages and sediment disasters, 2016. [5] Peter V. Gorsevski, Paul Gessler and Randy B. Foltz, “Spatial Prediction of Landslide Hazard Using Discriminant Analysis and GIS,” in GIS in the Rockies 2000 Conference and Workshop Applications for the 21st Century, Denver, Colorado. , 2000. [6] M. Casadei, W. E. Dietrich and N. L. Miller, “Testing A Model For Predicting The Timing And Location Of Shallow Landslide Initiation In Soil-Mantled Landscapes,” Earth Surface Processes and Landforms, vol. 28, p. 925–950, 2003. [7] Roberto Arnaldo Trancoso Gomes, Renato Fontes, Osmar Abílio de Carvalho Júnior, Nelson Ferreira and Eurípedes Vargas do Amaral, “Combining Spatial Models for Shallow Landslides and Debris-Flows Prediction,” Remote Sensing, no. 5, pp. 2219-2237, 2013. [8] P. Bertolo and G. F. Wieczorek, “Calibration of numerical models for small debris flow in Yosemite Valley, California, USA,” Documentation and monitoring of landslides and debris flows for mathematical modelling and design of mitigation measures, 13 December 2005. [9] T. A. Gebreslassie, “Dynamic simulations of landslide runout in cohesive Soils,” Oslo, 2015. [10] P. Tarolli and D. G. Tarboton, “A new method for determination of most likely landslide initiation points and the evaluation of digital terrain model scale in terrain stability mapping,” Hydrology and Earth System Sciences, no. 10, p. 663–677, 2006. [11] G.-B. Kim, “Numerical Simulation Of Three-Dimensional Tsunami Generation By Subaerial Landslides,” 2013.
[12] Giuseppe Formetta, Giovanna Capparelli, and Pasqua, “Evaluating performance of simplified physically based models for shallow landslide susceptibility,” Hydrol. Earth System, no. 20, p. 4585–4603, 2016. [13] F. Dai and C. Lee, “Landslide characteristics and slope instability modeling using GIS, Lantau Island, Hong Kong,” 2001. [14] S. McDougall, “Landslide runout analysis — current practice and challenges,” Canadian Geotechnical Colloquium, vol. 57, pp. 605-620, 2017. [15] P. Quinn, K. Beven, P. Chevallier And O. Planchon, “The Prediction Of Hillslope Flow Paths For Distributed Hydrological Modelling Using Digital Terrain Models,” Hydrological Processes, Vol. Vol. 5, Pp. 59-79, 1991. [16] V. B., “Comparison of Single and Multiple Flow Direction Algorithm for Computing Topographic Parameters in Topmodel,” 2000. [17] J.-f. G. P. T. K. T. Guang-ju ZHAO, “Comparison of two different methods for determining flow direction in catchment hydrological modeling,” Water Science and Engineering, vol. 2, no. 4, pp. 1-15, 2009. [18] D. G. Tarboton, “A New Method For The Determination Of Flow Directions And Upslope Areas In Grid Digital Elevation Models,” Water Resources Research, Vol. 33, No. 2, P. 33, 309-319. [19] V. Baumann, “Debris flow susceptibility mapping at a regional scale along the National Road N7, Argentina,” in CGS Geotechnical Conference, Argentina, 2011. [20] Q. Z. Petter Pilesjö, “Theoretical Estimation Of Flow Accumulation From A Grid-Based Digital Elevation Model,” In Proceedings Of Gis Am/Fm Asia’97 And Geoinformatics’97 Conference, Taipei, 1997. [21] G. A. D. Y. a. C. S. L. John P. WILSON, “Water in the Landscape: A Review of Contemporary Flow Routing Algorithms,” pp. 213-236. [22] “MathWorks,” The MathWorks, Inc., [Online]. Available: https://in.mathworks.com/. [Accessed 1 10 2017]. [23] “What is GIS?,” Esri, [Online]. Available: http://www.esri.com/what-is-gis. [Accessed 12 11 2017]. [24] P. Barrett, “Paul Barrett,” Wikimedia Foundation, Inc, [Online]. Available: http://www.pbarrett.net/techpapers/euclid.pdf. [Accessed 05 01 2018].
[25] S. S. Gruber, Land-surface parameters and objects in hydrology., Elsevier, 2009. [26] “Landscape Evolution Modeling with CHILD,” Community Surface Dynamics Modeling System, [Online]. Available: http://csdms.colorado.edu/wiki/Labs_Landscape_Evolution_Modeling_With_Child_Part_2. [Accessed 13 01 2018]. [27] M. Cooper, “Depth Recovery through Linear Algebra,” in Line Drawing Interpretation, Springer Science & Business Media., p. 118. [28] A. P. Nicholas, “Cellular modelling in fluvial geomorphology,” in Earth Surface Processes and Landforms, 2005, p. 645–649. [29] “Making Successful Maps,” DroneDeploy, [Online]. Available: https://support.dronedeploy.com/v1.0/docs/making-successful-maps. [Accessed 15 10 2017]. [30] “M_Map:A mapping package for Matlab,” rich@eos.ubc.ca, [Online]. Available: https://www.eoas.ubc.ca/~rich/map.html. [Accessed 15 01 2018]. [31] P. Dulanjalee, “Landslide Flow Path Assessment for Susceptibility Mapping at a Regional Scale,” in ‘Investing in Disaster Risk Reduction for Resilience’’- NBRO holds 8th Annual Symposium, Colombo, 2017. [32] L. K. a. D. M. E. Boyagoda, ” Subsurface Geotechnical Characterization,” in International Symposium, NBRO, Colombo, Sri Lanka, 2016. [33] E. E. Duncan and A. A. Rahman, “An Amalgamation Of 3d Spatial Data Model For Surface And Subsurface Spatial Objects.,” in Knowing To Manage The Territory, Protect The Environment, Evaluate The Cultural Heritage, Rome, Italy, 2012.
2020년 12월 22일 접수, 2021년 5월 1일 수정, 2021년 7월 15일 수락, 2021년 7월 21일 온라인 사용 가능, 기록 버전 2021년 8월 17일 .
Abstract
이 문서는 재료 압출 적층 제조 에서 여러 레이어를 인쇄하는 동안 증착 흐름의 전산 유체 역학 시뮬레이션 을 제공합니다 . 개발된 모델은 증착된 레이어의 형태를 예측하고 점소성 재료 를 인쇄하는 동안 레이어 변형을 캡처합니다 . 물리학은 일반화된 뉴턴 유체 로 공식화된 Bingham 구성 모델의 연속성 및 운동량 방정식에 의해 제어됩니다. . 증착된 층의 단면 모양이 예측되고 재료의 다양한 구성 매개변수에 대해 층의 변형이 연구됩니다. 층의 변형은 인쇄물의 정수압과 압출시 압출압력으로 인한 것임을 알 수 있다. 시뮬레이션에 따르면 항복 응력이 높을수록 변형이 적은 인쇄물이 생성되는 반면 플라스틱 점도 가 높을수록 증착된 레이어에서변형이 커 집니다 . 또한, 인쇄 속도, 압출 속도 의 영향, 층 높이 및 인쇄된 층의 변형에 대한 노즐 직경을 조사합니다. 마지막으로, 이 모델은 후속 인쇄된 레이어의 정수압 및 압출 압력을 지원하기 위해 증착 후 점소성 재료가 요구하는 항복 응력의 필요한 증가에 대한 보수적인 추정치를 제공합니다.
This paper presents computational fluid dynamics simulations of the deposition flow during printing of multiple layers in material extrusionadditive manufacturing. The developed model predicts the morphology of the deposited layers and captures the layer deformations during the printing of viscoplastic materials. The physics is governed by the continuity and momentum equations with the Bingham constitutive model, formulated as a generalized Newtonian fluid. The cross-sectional shapes of the deposited layers are predicted, and the deformation of layers is studied for different constitutive parameters of the material. It is shown that the deformation of layers is due to the hydrostatic pressure of the printed material, as well as the extrusion pressure during the extrusion. The simulations show that a higher yield stress results in prints with less deformations, while a higher plastic viscosity leads to larger deformations in the deposited layers. Moreover, the influence of the printing speed, extrusion speed, layer height, and nozzle diameter on the deformation of the printed layers is investigated. Finally, the model provides a conservative estimate of the required increase in yield stress that a viscoplastic material demands after deposition in order to support the hydrostatic and extrusion pressure of the subsequently printed layers.
키워드
점성 플라스틱 재료, 재료 압출 적층 제조(MEX-AM), 다층 증착, 전산유체역학(CFD), 변형 제어 Viscoplastic Materials, Material Extrusion Additive Manufacturing (MEX-AM), Multiple-Layers Deposition, Computational Fluid Dynamics (CFD), Deformation Control
Introduction
Three-dimensional printing of viscoplastic materials has grown in popularity over the recent years, due to the success of Material Extrusion Additive Manufacturing (MEX-AM) [1]. Viscoplastic materials, such as ceramic pastes [2,3], hydrogels [4], thermosets [5], and concrete [6], behave like solids when the applied load is below their yield stress, and like a fluid when the applied load exceeds their yield stress [7]. Viscoplastic materials are typically used in MEX-AM techniques such as Robocasting [8], and 3D concrete printing [9,10]. The differences between these technologies lie in the processing of the material before the extrusion and in the printing scale (from microscale to big area additive manufacturing). In these extrusion-based technologies, the structure is fabricated in a layer-by-layer approach onto a solid surface/support [11, 12]. During the process, the material is typically deposited on top of the previously printed layers that may be already solidified (wet-on-dry printing) or still deformable (wet-on-wet printing) [1]. In wet-on-wet printing, control over the deformation of layers is important for the stability and geometrical accuracy of the prints. If the material is too liquid after the deposition, it cannot support the pressure of the subsequently deposited layers. On the other hand, the material flowability is a necessity during extrusion through the nozzle. Several experimental studies have been performed to analyze the physics of the extrusion and deposition of viscoplastic materials, as reviewed in Refs. [13–16]. The experimental measurements can be supplemented with Computational Fluid Dynamics (CFD) simulations to gain a more complete picture of MEX-AM. A review of the CFD studies within the material processing and deposition in 3D concrete printing was presented by Roussel et al. [17]. Wolfs et al. [18] predicted numerically the failure-deformation of a cylindrical structure due to the self-weight by calculating the stiffness and strength of the individual layers. It was found that the deformations can take place in all layers, however the most critical deformation occurs in the bottom layer. Comminal et al. [19,20] presented three-dimensional simulations of the material deposition in MEX-AM, where the fluid was approximated as Newtonian. Subsequently, the model was experimentally validated in Ref. [21] for polymer-based MEX-AM, and extended to simulate the deposition of multiple layers in Ref. [22], where the previously printed material was assumed solid. Xia et al. [23] simulated the influence of the viscoelastic effects on the shape of deposited layers in MEX-AM. A numerical model for simulating the deposition of a viscoplastic material was recently presented and experimentally validated in Refs. [24] and [25]. These studies focused on predicting the cross-sectional shape of a single printed layer for different processing conditions (relative printing speed, and layer height). Despite these research efforts, a limited number of studies have focused on investigating the material deformations in wet-on-wet printing when multiple layers are deposited on top of each other. This paper presents CFD simulations of the extrusion-deposition flow of a viscoplastic material for several subsequent layers (viz. three- and five-layers). The material is continuously printed one layer over another on a fixed solid surface. The rheology of the viscoplastic material is approximated by the Bingham constitutive equation that is formulated using the Generalized Newtonian Fluid (GNF) model. The CFD model is used to predict the cross-sectional shapes of the layers and their deformations while printing the next layers on top. Moreover, the simulations are used to quantify the extrusion pressure applied by the deposited material on the substrate, and the previously printed layers. Numerically, it is investigated how the process parameters (i.e., the extrusion speed, printing speed, nozzle diameter, and layer height) and the material rheology affect the deformations of the deposited layers. Section 2 describes the methodology of the study. Section 3 presents and discusses the results. The study is summarized and concluded in Section 4.
References
[1] R.A. Buswell, W.R. Leal De Silva, S.Z. Jones, J. Dirrenberger, 3D printing using concrete extrusion: a roadmap for research, Cem. Concr. Res. 112 (2018) 37–49. [2] Z. Chen, Z. Li, J. Li, C. Liu, C. Lao, Y. Fu, C. Liu, Y. Li, P. Wang, Y. He, 3D printing of ceramics: a review, J. Eur. Ceram. Soc. 39 (4) (2019) 661–687. [3] A. Bellini, L. Shor, S.I. Guceri, New developments in fused deposition modeling of ceramics, Rapid Prototyp. J. 11 (4) (2005) 214–220. [4] S. Aktas, D.M. Kalyon, B.M. Marín-Santib´ anez, ˜ J. P´erez-Gonzalez, ´ Shear viscosity and wall slip behavior of a viscoplastic hydrogel, J. Rheol. 58 (2) (2014) 513–535. [5] J. Lindahl, A. Hassen, S. Romberg, B. Hedger, P. Hedger Jr., M. Walch, T. Deluca, W. Morrison, P. Kim, A. Roschli, D. Nuttall, Large-scale Additive Manufacturing with Reactive Polymers, Oak Ridge National Lab.(ORNL), Oak Ridge, TN (United States), 2018. [6] V.N. Nerella, V. Mechtcherine, Studying the printability of fresh concrete for formwork-free Concrete onsite 3D Printing Technology (CONPrint3D), 3D Concr. Print. Technol. (2019) 333–347. [7] C. Tiu, J. Guo, P.H.T. Uhlherr, Yielding behaviour of viscoplastic materials, J. Ind. Eng. Chem. 12 (5) (2006) 653–662. [8] B. Dietemann, F. Bosna, M. Lorenz, N. Travitzky, H. Kruggel-Emden, T. Kraft, C. Bierwisch, Modeling robocasting with smoothed particle hydrodynamics: printing gapspanning filaments, Addit. Manuf. 36 (2020), 101488. [9] B. Khoshnevis, R. Russell, H. Kwon, S. Bukkapatnam, Contour crafting – a layered fabrication, Spec. Issue IEEE Robot. Autom. Mag. 8 (3) (2001) 33–42. [10] D. Asprone, F. Auricchio, C. Menna, V. Mercuri, 3D printing of reinforced concrete elements: technology and design approach, Constr. Build. Mater. 165 (2018) 218–231. [11] J. Jiang, Y. Ma, Path planning strategies to optimize accuracy, quality, build time and material use in additive manufacturing: a review, Micromachines 11 (7) (2020) 633. [12] J. Jiang, A novel fabrication strategy for additive manufacturing processes, J. Clean. Prod. 272 (2020), 122916. [13] F. Bos, R. Wolfs, Z. Ahmed, T. Salet, Additive manufacturing of concrete in construction: potentials and challenges, Virtual Phys. Prototyp. 11 (3) (2016) 209–225. [14] P. Wu, J. Wang, X. Wang, A critical review of the use of 3-D printing in the construction industry, Autom. Constr. 68 (2016) 21–31. [15] T.D. Ngo, A. Kashani, G. Imbalzano, K.T. Nguyen, D. Hui, Additive manufacturing (3D printing): a review of materials, methods, applications and challenges, Compos. Part B: Eng. 143 (2018) 172–196. [16] M. Valente, A. Sibai, M. Sambucci, Extrusion-based additive manufacturing of concrete products: revolutionizing and remodeling the construction industry, J. Compos. Sci. 3 (3) (2019) 88. [17] N. Roussel, J. Spangenberg, J. Wallevik, R. Wolfs, Numerical simulations of concrete processing: from standard formative casting to additive manufacturing, Cem. Concr. Res. 135 (2020), 106075. [18] R.J.M. Wolfs, F.P. Bos, T.A.M. Salet, Early age mechanical behaviour of 3D printed concrete: numerical modelling and experimental testing, Cem. Concr. Res. 106 (2018) 103–116. [19] R. Comminal, M.P. Serdeczny, D.B. Pedersen, J. Spangenberg, Numerical modeling of the strand deposition flow in extrusion-based additive manufacturing, Addit. Manuf. 20 (2018) 68–76. [20] R. Comminal, M.P. Serdeczny, D.B. Pedersen, J. Spangenberg, Numerical modeling of the material deposition and contouring precision in fused deposition modeling, in Proceedings of the 29th Annual International Solid Freeform Fabrication Symposium, Austin, TX, USA, 2018, pp. 1855–1864. [21] M.P. Serdeczny, R. Comminal, D.B. Pedersen, J. Spangenberg, Experimental validation of a numerical model for the strand shape in material extrusion additive manufacturing, Addit. Manuf. 24 (2018) 145–153. [22] M.P. Serdeczny, R. Comminal, D.B. Pedersen, J. Spangenberg, Numerical simulations of the mesostructure formation in material extrusion additive manufacturing, Addit. Manuf. 28 (2019) 419–429. [23] H. Xia, J. Lu, G. Tryggvason, A numerical study of the effect of viscoelastic stresses in fused filament fabrication, Comput. Methods Appl. Mech. Eng. 346 (2019) 242–259. [24] R. Comminal, W.R.L. da Silva, T.J. Andersen, H. Stang, J. Spangenberg, Influence of processing parameters on the layer geometry in 3D concrete printing: experiments and modelling, in: Proceedings of the Second RILEM International Conference on Concrete and Digital Fabrication, vol. 28, 2020, pp. 852–862. [25] R. Comminal, W.R.L. da Silva, T.J. Andersen, H. Stang, J. Spangenberg, Modelling of 3D concrete printing based on computational fluid dynamics, Cem. Concr. Res. 38 (2020), 106256. [26] E.C. Bingham, An investigation of the laws of plastic flow, US Bur. Stand. Bull. 13 (1916) 309–352. [27] N. Casson, A flow equation for pigment-oil suspensions of the printing ink type, Rheol. Disperse Syst. (1959) 84–104. [28] W.H. Herschel, R. Bulkley, Konsistenzmessungen von Gummi-Benzollosungen, ¨ Kolloid Z. 39 (1926) 291–300. [29] “FLOW-3D | We solve The World’s Toughest CFD Problems,” FLOW SCIENCE. 〈https://www.flow3d.com/〉. (Accessed 27 June 2020). [30] S. Jacobsen, R. Cepuritis, Y. Peng, M.R. Geiker, J. Spangenberg, Visualizing and simulating flow conditions in concrete form filling using pigments, Constr. Build. Mater. 49 (2013) 328–342. [31] E.J. O’Donovan, R.I. Tanner, Numerical study of the Bingham squeeze film problem, J. Non-Newton. Fluid Mech. 15 (1) (1984) 75–83. [32] C.W. Hirt, B.D. Nichols, Volume of fluid (VOF) method for the dynamics of free boundaries, J. Comput. Phys. 39 (1) (1981) 201–225. [33] R. Comminal, J. Spangenberg, J.H. Hattel, Cellwise conservative unsplit advection for the volume of fluid method, J. Comput. Phys. 283 (2015) 582–608. [34] A. Negar, S. Nazarian, N.A. Meisel, J.P. Duarte, Experimental prediction of material deformation in large-scale additive manufacturing of concrete, Addit. Manuf. 37 (2021), 101656.
In this study a gating system including sprue, runner and overflows for semi-solid rheocasting of aluminum alloy was designed by means of numerical simulations with a commercial software. The effects of pouring temperature, mold temperature and injection speed on the filling process performance of semi-solid die casting were studied. Based on orthogonal test analysis, the optimal die casting process parameters were selected, which were metal pouring temperature 590 °C, mold temperature 260 °C and injection velocity 0.5 m/s. Semi-solid slurry preparation process of Swirled Enthalpy Equilibration Device (SEED) was used for die casting production experiment. Aluminum alloy semi-solid bracket components were successfully produced with the key die casting process parameters selected, which was consistent with the simulation result. The design of semi-solid gating system was further verified by observing and analyzing the microstructure of different zones of the casting. The characteristic parameters, particle size and shape factor of microstructure of the produced semi-solid casting showed that the semi-solid aluminum alloy components are of good quality.
이 연구에서 알루미늄 합금의 반고체 레오캐스팅을 위한 스프루, 러너 및 오버플로를 포함하는 게이팅 시스템은 상용 소프트웨어를 사용한 수치 시뮬레이션을 통해 설계되었습니다. 주입 온도, 금형 온도 및 사출 속도가 반고체 다이캐스팅의 충전 공정 성능에 미치는 영향을 연구했습니다. 직교 테스트 분석을 기반으로 금속 주입 온도 590°C, 금형 온도 260°C 및 사출 속도 0.5m/s인 최적의 다이 캐스팅 공정 매개변수가 선택되었습니다. Swirled Enthalpy Equilibration Device(SEED)의 반고체 슬러리 제조 공정을 다이캐스팅 생산 실험에 사용하였다. 알루미늄 합금 반고체 브래킷 구성 요소는 시뮬레이션 결과와 일치하는 주요 다이 캐스팅 공정 매개변수를 선택하여 성공적으로 생산되었습니다. 반고체 게이팅 시스템의 설계는 주조의 다른 영역의 미세 구조를 관찰하고 분석하여 추가로 검증되었습니다. 생산된 반고체 주조물의 특성 매개변수, 입자 크기 및 미세 구조의 형상 계수는 반고체 알루미늄 합금 부품의 품질이 양호함을 보여주었습니다.
References
G. Li, H. Lu, X. Hu et al., Current progress in rheoforming of wrought aluminum alloys: a review. Met. Open Access Metall. J. 10(2), 238 (2020)CASGoogle Scholar
C. Xghab, D. Qza, E. Spma et al., Blistering in semi-solid die casting of aluminium alloys and its avoidance. Acta Mater. 124, 446–455 (2017)ArticleGoogle Scholar
M. Modigell, J. Koke, Rheological modelling on semi-solid metal alloys and simulation of thixocasting processes. J. Mater. Process. Technol. 111(1–3), 53–58 (2001)CASArticleGoogle Scholar
A. Pola, M. Tocci, P. Kapranos, Microstructure and properties of semi-solid aluminum alloys: a literature review. Met. Open Access Metall. J. 8(3), 181 (2018)Google Scholar
Q. Zhu, Semi-solid moulding: competition to cast and machine from forging in making automotive complex components. Trans. Nonferrous Met. Soc. China 20, 1042–1047 (2010)ArticleGoogle Scholar
B. Zhou, S. Lu, K. Xu et al., Microstructure and simulation of semisolid aluminum alloy castings in the process of stirring integrated transfer-heat (SIT) with water cooling. Int. J. Metalcast. 14(2), 396–408 (2019). https://doi.org/10.1007/s40962-019-00357-6CASArticleGoogle Scholar
S. Ji, Z. Fan, Solidification behavior of Sn–15 wt Pct Pb alloy under a high shear rate and high intensity of turbulence during semisolid processing. Metall. Mater. Trans. A. 33(11), 3511–3520 (2002). https://doi.org/10.1007/s11661-002-0338-4ArticleGoogle Scholar
H.V. Atkinson, Alloys for semi-solid processing. Solid State Phenom. 192–193, 16–27 (2013)Google Scholar
L. Rogal, Critical assessment: opportunities in developing semi-solid processing: aluminium, magnesium, and high-temperature alloys. Mater. Sci. Technol. Mst A Publ. Inst. Met. 33, 759–764 (2017)CASArticleGoogle Scholar
H. Guo, Rheo-diecasting process for semi-solid aluminum alloys. J. Wuhan Univ. Technol. Mater. Sci. Ed. 22(004), 590–595 (2007)CASArticleGoogle Scholar
T. Chucheep, J. Wannasin, R. Canyook, T. Rattanochaikul, S. Janudom, S. Wisutmethangoon, M.C. Flemings, Characterization of flow behavior of semi-solid slurries with low solid fractions. Metall. Mater. Trans. A 44(10), 4754–4763 (2013)CASArticleGoogle Scholar
M. Li, Y.D. Li, W.L. Yang et al., Effects of forming processes on microstructures and mechanical properties of A356 aluminum alloy prepared by self-inoculation method. Mater. Res. 22(3) (2019)
P. Côté, M.E. Larouche, X.G. Chen et al., New developments with the SEED technology. Solid State Phenom. 192(3), 373–378 (2012)ArticleGoogle Scholar
I. Dumanić, S. Jozić, D. Bajić et al., Optimization of semi-solid high-pressure die casting process by computer simulation, Taguchi method and grey relational analysis. Inter Metalcast. 15, 108–118 (2021). https://doi.org/10.1007/s40962-020-00422-5ArticleGoogle Scholar
A. Guo, J. Zhao, C. Xu et al., Effects of pouring temperature and electromagnetic stirring on porosity and mechanical properties of A357 aluminum alloy rheo-diecasting. J. Mater. Eng. Perform. (2018). https://doi.org/10.1007/s11665-018-3310-1ArticleGoogle Scholar
C.G. Kang, S.M. Lee, B.M. Kim, A study of die design of semi-solid die casting according to gate shape and solid fraction. J. Mater. Process. Technol. 204(1–3), 8–21 (2008)CASArticleGoogle Scholar
Z.Y. Liu, W.M. Mao, W.P. Wang et al., Investigation of rheo-diecasting mold filling of semi-solid A380 aluminum alloy slurry. Int. J. Miner. Metall. Mater. 24(006), 691–700 (2017)CASArticleGoogle Scholar
M. Arif, M.Z. Omar, N. Muhamad et al., Microstructural evolution of solid-solution-treated Zn–22Al in the semisolid state. J. Mater. Sci. Technol. 29(008), 765–774 (2013)CASArticleGoogle Scholar
aUniversity of Padova, Department of Management and Engineering, Vicenza, Italy bUniversity of Bayburt, Mechanical Engineering, Bayburt, Turkey cAtilim University, Metallurgical and Materials Engineering, Ankara, Turkey dIstanbul Technical University, Metallurgical and Materials Engineering, Istanbul, Turkey eCenter for Critical and Functional Materials, ITU, Istanbul, Turkey
ABSTRACT
The demand for lighter weight decreased thickness and higher strength has become the focal point in the automotive industry. In order to meet such requirements, the addition of several alloying elements has been started to be investigated. In this work, the additions of V, B, and Sr on feedability and tensile properties of A360 has been studied. A mold design that consisted of test bars has been produced. Initially, a simulation was carried out to optimize the runners, filling, and solidification parameters. Following the tests, it was found that V addition revealed the highest UTS but low elongation at fracture, while B addition exhibited visa verse. On the other hand, impact energy was higher with B additions.
더 가벼운 무게의 감소된 두께와 더 높은 강도에 대한 요구는 자동차 산업의 초점이 되었습니다. 이러한 요구 사항을 충족하기 위해 여러 합금 원소의 추가가 조사되기 시작했습니다. 이 연구에서는 A360의 이송성 및 인장 특성에 대한 V, B 및 Sr의 첨가가 연구되었습니다. 시험봉으로 구성된 금형 설계가 제작되었습니다. 처음에는 러너, 충전 및 응고 매개변수를 최적화하기 위해 시뮬레이션이 수행되었습니다. 시험 결과, V 첨가는 UTS가 가장 높지만 파단 연신율은 낮았고, B 첨가는 visa verse를 나타냈다. 반면에 충격 에너지는 B 첨가에서 더 높았다.
References
[1] A. Johanson, Effect of Vanadium on Grain Refinement of Aluminium, Institutt for materialteknologi, 2013. [2] D.G. McCartney, Grain refining of aluminium and its alloys using inoculants, Int. Mater. Rev. 34 (1) (1989) 247–260. [3] M.T. Di Giovanni, The Influence of Ni and V Trace Elements on the High Temperature Tensile Properties of A356 Aluminium Foundry Alloy, Institutt for materialteknologi, 2014. [4] D. Casari, T.H. Ludwig, M. Merlin, L. Arnberg, G.L. Garagnani, The effect of Ni and V trace elements on the mechanical properties of A356 aluminium foundry alloy in as-cast and T6 heat treated conditions, Mater. Sci. Eng., A 610 (2014) 414–426. [5] D. Casari, T.H. Ludwig, M. Merlin, L. Arnberg, G.L. Garagnani, Impact behavior of A356 foundry alloys in the presence of trace elements Ni and V, J. Mater. Eng. Perform. 24 (2) (2015) 894–908. [6] T.H. Ludwig, P.L. Schaffer, L. Arnberg, Influence of some trace elements on solidification path and microstructure of Al-Si foundry alloys, Metall. Mater. Trans. 44 (8) (2013) 3783–3796. [7] H.A. Elhadari, H.A. Patel, D.L. Chen, W. Kasprzak, Tensile and fatigue properties of a cast aluminum alloy with Ti, Zr and V additions, Mater. Sci. Eng., A 528 (28) (2011) 8128–8138. [8] Y. Wu, H. Liao, K. Zhou, “Effect of minor addition of vanadium on mechanical properties and microstructures of as-extruded near eutectic Al–Si–Mg alloy, Mater. Sci. Eng., A 602 (2014) 41–48. [9] E.S. Dæhlen, The Effect of Vanadium on AlFeSi-Intermetallic Phases in a Hypoeutectic Al-Si Foundry Alloy, Institutt for materialteknologi, 2013. [10] B. Lin, H. Li, R. Xu, H. Xiao, W. Zhang, S. Li, Effects of vanadium on modification of iron-rich intermetallics and mechanical properties in A356 cast alloys with 1.5 wt. % Fe, J. Mater. Eng. Perform. 28 (1) (2019) 475–484. [11] P.A. Tøndel, G. Halvorsen, L. Arnberg, Grain refinement of hypoeutectic Al-Si foundry alloys by addition of boron containing silicon metal, Light Met. (1993) 783. [12] Z. Chen, et al., Grain refinement of hypoeutectic Al-Si alloys with B, Acta Mater. 120 (2016) 168–178. [13] T. Wang, Z. Chen, H. Fu, J. Xu, Y. Fu, T. Li, “Grain refining potency of Al–B master alloy on pure aluminum, Scripta Mater. 64 (12) (2011) 1121–1124. [14] M. Gorny, ´ G. Sikora, M. Kawalec, Effect of titanium and boron on the stability of grain refinement of Al-Cu alloy, Arch. Foundry Eng. 16 (2016). [15] O. ¨ Gürsoy, E. Erzi, D. Dıs¸pınar, Ti grain refinement myth and cleanliness of A356 melt, in: Shape Casting, Springer, 2019, pp. 125–130. [16] D. Dispinar, A. Nordmark, J. Voje, L. Arnberg, Influence of hydrogen content and bi-film index on feeding behaviour of Al-7Si, in: 138th TMS Annual Meeting, Shape Casting, 3rd International Symposium, San Francisco, California, USA, 2009, pp. 63–70. February 2009. [17] M. Uludag, ˘ R. Çetin, D. Dıs¸pınar, Observation of hot tearing in Sr-B modified A356 alloy, Arch. Foundry Eng. 17 (2017). [18] X.L. Cui, Y.Y. Wu, T. Gao, X.F. Liu, “Preparation of a novel Al–3B–5Sr master alloy and its modification and refinement performance on A356 alloy, J. Alloys Compd. 615 (2014) 906–911. [19] F. Wang, Z. Liu, D. Qiu, J.A. Taylor, M.A. Easton, M.-X. Zhang, Revisiting the role of peritectics in grain refinement of Al alloys, Acta Mater. 61 (1) (2013) 360–370. [20] M. Akhtar, A. Khajuria, Effects of prior austenite grain size on impression creep and microstructure in simulated heat affected zones of boron modified P91 steels, Mater. Chem. Phys. 249 (2020) 122847. [21] M. Akhtar, A. Khajuria, Probing true creep-hardening interaction in weld simulated heat affected zone of P91 steels, J. Manuf. Process. 46 (2019) 345–356. [22] E.M. Schulson, T.P. Weihs, I. Baker, H.J. Frost, J.A. Horton, Grain boundary accommodation of slip in Ni3Al containing boron, Acta Metall. 34 (7) (1986) 1395–1399. [23] I. Baker, E.M. Schulson, J.R. Michael, The effect of boron on the chemistry of grain boundaries in stoichiometric Ni3Al, Philos. Mag. A B 57 (3) (Mar. 1988) 379–385. [24] S. Zhu, et al., Influences of nickel and vanadium impurities on microstructure of aluminum alloys, JOM (J. Occup. Med.) 65 (5) (2013) 584–592. [25] D.J. Beerntsen, Effect of vanadium and zirconium on the formation of CrAI 7 primary crystals in 7075 aluminum alloy, Metall. Mater. Trans. B 8 (3) (1977) 687–688. [26] G. Timelli, A. Fabrizi, S. Capuzzi, F. Bonollo, S. Ferraro, The role of Cr additions and Fe-rich compounds on microstructural features and impact toughness of AlSi9Cu3 (Fe) diecasting alloys, Mater. Sci. Eng., A 603 (2014) 58–68. [27] S. Kirtay, D. Dispinar, Effect of ranking selection on the Weibull modulus estimation, Gazi Univ. J. Sci. 25 (1) (2012) 175–187. [28] J. Rakhmonov, G. Timelli, F. Bonollo, “The effect of transition elements on hightemperature mechanical properties of Al–Si foundry alloys–A review, Adv. Eng. Mater. 18 (7) (2016) 1096–1105.
Investigating the breach outflow hydrograph is an essential task to conduct mitigation plans and flood warnings. In the present study, the spatial dam breach is simulated by using a three-dimensional computational fluid dynamics model, FLOW-3D. The model parameters were adjusted by making a comparison with a previous experimental model. The different parameters (initial breach shape, dimensions, location, and dam slopes) are studied to investigate their effects on dam breaching. The results indicate that these parameters have a significant impact. The maximum erosion rate and peak outflow for the rectangular shape are higher than those for the V-notch by 8.85% and 5%, respectively. Increasing breach width or decreasing depth by 5% leads to increasing maximum erosion rate by 11% and 15%, respectively. Increasing the downstream slope angle by 4° leads to an increase in both peak outflow and maximum erosion rate by 2.0% and 6.0%, respectively.
유출 유출 수문곡선을 조사하는 것은 완화 계획 및 홍수 경보를 수행하는 데 필수적인 작업입니다. 본 연구에서는 3차원 전산유체역학 모델인 FLOW-3D를 사용하여 공간 댐 붕괴를 시뮬레이션합니다. 이전 실험 모델과 비교하여 모델 매개변수를 조정했습니다.
다양한 매개변수(초기 붕괴 형태, 치수, 위치 및 댐 경사)가 댐 붕괴에 미치는 영향을 조사하기 위해 연구됩니다. 결과는 이러한 매개변수가 상당한 영향을 미친다는 것을 나타냅니다. 직사각형 형태의 최대 침식율과 최대 유출량은 V-notch보다 각각 8.85%, 5% 높게 나타났습니다.
위반 폭을 늘리거나 깊이를 5% 줄이면 최대 침식률이 각각 11% 및 15% 증가합니다. 하류 경사각을 4° 증가시키면 최대 유출량과 최대 침식률이 각각 2.0% 및 6.0% 증가합니다.
Keywords
Spatial dam breach; FLOW-3D; Overtopping erosion; Computational fluid dynamics (CFD)
1. Introduction
There are many purposes for dam construction, such as protection from flood disasters, water storage, and power generation. Embankment failures may have a catastrophic impact on lives and infrastructure in the downstream regions. One of the most common causes of embankment dam failure is overtopping. Once the overtopping of the dam begins, the breach formation will start in the dam body then end with the dam failure. This failure occurs within a very short time, which threatens to be very dangerous. Therefore, understanding and modeling the embankment breaching processes is essential for conducting mitigation plans, flood warnings, and forecasting flood damage.
The analysis of the dam breaching process is implemented by different techniques: comparative methods, empirical models with dimensional and dimensionless solutions, physical-based models, and parametric models. These models were described in detail [1]. Parametric modeling is commonly used to simulate breach growth as a time-dependent linear process and calculate outflow discharge from the breach using hydraulics principles [2]. Alhasan et al. [3] presented a simple one-dimensional mathematical model and a computer code to simulate the dam breaching process. These models were validated by small dams breaching during the floods in 2002 in the Czech Republic. Fread [4] developed an erosion model (BREACH) based on hydraulics principles, sediment transport, and soil mechanics to estimate breach size, time of formation, and outflow discharge. Říha et al. [5] investigated the dam break process for a cascade of small dams using a simple parametric model for piping and overtopping erosion, as well as a 2D shallow-water flow model for the flood in downstream areas. Goodarzi et al. [6] implemented mathematical and statistical methods to assess the effect of inflows and wind speeds on the dam’s overtopping failure.
Dam breaching studies can be divided into two main modes of erosion. The first mode is called “planar dam breach” where the flow overtops the whole dam width. While the second mode is called “spatial dam breach” where the flow overtops through the initial pilot channel (i.e., a channel created in the dam body). Therefore, the erosion will be in both vertical and horizontal directions [7].
The erosion process through the embankment dams occurs due to the shear stress applied by water flows. The dam breaching evolution can be divided into three stages [8], [9], but Y. Yang et al. [10] divided the breach development into five stages: Stage I, the seepage erosion; Stage II, the initial breach formation; Stage III, the head erosion; Stage IV, the breach expansion; and Stage V, the re-equilibrium of the river channel through the breach. Many experimental tests have been carried out on non-cohesive embankment dams with an initial breach to examine the effect of upstream inflow discharges on the longitudinal profile evolution and the time to inflection point[11].
Zhang et al. [12] studied the effect of changing downstream slope angle, sediment grain size, and dam crest length on erosion rates. They noticed that increasing dam crest length and decreasing downstream slope angle lead to decreasing sediment transport rate. While the increase in sediment grain size leads to an increased sediment transport rate at the initial stages. Höeg et al. [13] presented a series of field tests to investigate the stability of embankment dams made of various materials. Overtopping and piping were among the failure tests carried out for the dams composed of homogeneous rock-fill, clay, or gravel with a height of up to 6.0 m. Hakimzadeh et al. [14] constructed 40 homogeneous cohesive and non-cohesive embankment dams to study the effect of changing sediment diameter and dam height on the breaching process. They also used genetic programming (GP) to estimate the breach outflow. Refaiy et al. [15] studied different scenarios for the downstream drain geometry, such as length, height, and angle, to minimize the effect of piping phenomena and therefore increase dam safety.
Zhu et al. [16] examined the effect of headcut erosion on dam breach growth, especially in the case of cohesive dams. They found that the breach growth in non-cohesive embankments is slower than cohesive embankments due to the little effect of headcut. Schmocker and Hager [7] proposed a relationship for estimating peak outflow from the dam breach process.(1)QpQin-1=1.7exp-20hc23d5013H0
where: Qp = peak outflow discharge.
Qin = inflow discharge.
hc = critical flow depth.
d50 = mean sediment diameter.
Ho = initial dam height.
Yu et al. [17] carried out an experimental study for homogeneous non-cohesive embankment dams in a 180° bending rectangular flume to determine the effect of overtopping flows on breaching formation. They found that the main factors influencing breach formation are water level, river discharge, and embankment material diameter.
Wu et al. [18] carried out a series of experiments to investigate the effect of breaching geometry on both non-cohesive and cohesive embankment dams in a U-bend flume due to overtopping flows. In the case of non-cohesive embankments, the non-symmetrical lateral expansion was noticed during the breach formation. This expansion was described by a coefficient ranging from 2.7 to 3.3.
The numerical models of the dam breach can be categorized according to different parameters, such as flow dimensions (1D, 2D, or 3D), flow governing equations, and solution methods. The 1D models are mainly used to predict the outflow hydrograph from the dam breach. Saberi et al. [19] applied the 1D Saint-Venant equation, which is solved by the finite difference method to investigate the outflow hydrograph during dam overtopping failure. Because of the ability to study dam profile evolution and breach formation, 2D models are more applicable than 1D models. Guan et al. [20] and Wu et al. [21] employed both 2D shallow water equations (SWEs) and sediment erosion equations, which are solved by the finite volume method to study the effect of the dam’s geometry parameters on outflow hydrograph and dam profile evolution. Wang et al. [22] also proposed a second-order hybrid-type of total variation diminishing (TVD) finite-difference to estimate the breach outflow by solving the 2D (SWEs). The accuracy of (SWEs) for both vertical flow contraction and surface roughness has been assessed [23]. They noted that the accuracy of (SWEs) is acceptable for milder slopes, but in the case of steeper slopes, modelers should be more careful. Generally, the accuracy of 2D models is still low, especially with velocity distribution over the flow depth, lateral momentum exchange, density-driven flows, and bottom friction[24]. Therefore, 3D models are preferred. Larocque et al. [25] and Yang et al. [26] started to use three-dimensional (3D) models that depend on the Reynolds-averaged Navier-Stokes (RANS) equations.
Previous experimental studies concluded that there is no clear relationship between the peak outflow from the dam breach and the initial breach characteristics. Some of these studies depend on the sharp-crested weir fixed at the end of the flume to determine the peak outflow from the breach, which leads to a decrease in the accuracy of outflow calculations at the microscale. The main goals of this study are to carry out a numerical simulation for a spatial dam breach due to overtopping flows by using (FLOW-3D) software to find an empirical equation for the peak outflow discharge from the breach and determine the worst-case that leads to accelerating the dam breaching process.
2. Numerical simulation
The current study for spatial dam breach is simulated by using (FLOW-3D) software [27], which is a powerful computational fluid dynamics (CFD) program.
2.1. Geometric presentations
A stereolithographic (STL) file is prepared for each change in the initial breach geometry and dimensions. The CAD program is useful for creating solid objects and converting them to STL format, as shown in Fig. 1.
2.2. Governing equations
The governing equations for water flow are three-dimensional Reynolds Averaged Navier-Stokes equations (RANS).
The momentum equation:(3)∂ui∂t+1VFuj∂ui∂xj=1ρ∂∂xj-pδij+ν∂ui∂xj+∂uj∂xi-ρu`iu`j¯
where u is time-averaged velocity,ν is kinematic viscosity, VF is fractional volume open to flow, p is averaged pressure and -u`iu`j¯ are components of Reynold’s stress. The Volume of Fluid (VOF) technique is used to simulate the free surface profile. Hirt et al. [28] presented the VOF algorithm, which employs the function (F) to express the occupancy of each grid cell with fluid. The value of (F) varies from zero to unity. Zero value refers to no fluid in the grid cell, while the unity value refers to the grid cell being fully occupied with fluid. The free surface is formed in the grid cells having (F) values between zero and unity.(4)∂F∂t+1VF∂∂xFAxu+∂∂yFAyv+∂∂zFAzw=0
where (u, v, w) are the velocity components in (x, y, z) coordinates, respectively, and (Ax, Ay, Az) are the area fractions.
2.3. Boundary and initial conditions
To improve the accuracy of the results, the boundary conditions should be carefully determined. In this study, two mesh blocks are used to minimize the time consumed in the simulation. The boundary conditions for mesh block 1 are as follows: The inlet and sides boundaries are defined as a wall boundary condition (wall boundary condition is usually used for bound fluid by solid regions. In the case of viscous flows, no-slip means that the tangential velocity is equal to the wall velocity and the normal velocity is zero), the outlet is defined as a symmetry boundary condition (symmetry boundary condition is usually used to reduce computational effort during CFD simulation. This condition allows the flow to be transferred from one mesh block to another. No inputs are required for this boundary condition except that its location should be defined accurately), the bottom boundary is defined as a uniform flow rate boundary condition, and the top boundary is defined as a specific pressure boundary condition with assigned atmospheric pressure. The boundary conditions for mesh block 2 are as follows: The inlet is defined as a symmetry boundary condition, the outlet is defined as a free flow boundary condition, the bottom and sides boundaries are defined as a wall boundary condition, and the top boundary is defined as a specific pressure boundary condition with assigned atmospheric pressure as shown in Fig. 2. The initial conditions required to be set for the fluid (i.e., water) inside of the domain include configuration, temperature, velocities, and pressure distribution. The configuration of water depends on the dimensions and shape of the dam reservoir. While the other conditions have been assigned as follows: temperature is normal water temperature (25 °c) and pressure distribution is hydrostatic with no initial velocity.
2.4. Numerical method
FLOW-3D uses the finite volume method (FVM) to solve the governing equation (Reynolds-averaged Navier-Stokes) over the computational domain. A finite-volume method is an Eulerian approach for representing and evaluating partial differential equations in algebraic equations form [29]. At discrete points on the mesh geometry, values are determined. Finite volume expresses a small volume surrounding each node point on a mesh. In this method, the divergence theorem is used to convert volume integrals with a divergence term to surface integrals. After that, these terms are evaluated as fluxes at each finite volume’s surfaces.
2.5. Turbulent models
Turbulence is the chaotic, unstable motion of fluids that occurs when there are insufficient stabilizing viscous forces. In FLOW-3D, there are six turbulence models available: the Prandtl mixing length model, the one-equation turbulent energy model, the two-equation (k – ε) model, the Renormalization-Group (RNG) model, the two-equation (k – ω) models, and a large eddy simulation (LES) model. For simulating flow motion, the RNG model is adopted to simulate the motion behavior better than the k – ε and k – ω.
models [30]. The RNG model consists of two main equations for the turbulent kinetic energy KT and its dissipation.εT(5)∂kT∂t+1VFuAx∂kT∂x+vAy∂kT∂y+wAz∂kT∂z=PT+GT+DiffKT-εT(6)∂εT∂t+1VFuAx∂εT∂x+vAy∂εT∂y+wAz∂εT∂z=C1.εTKTPT+c3.GT+Diffε-c2εT2kT
where KT is the turbulent kinetic energy, PT is the turbulent kinetic energy production, GT is the buoyancy turbulence energy, εT is the turbulent energy dissipation rate, DiffKT and Diffε are terms of diffusion, c1, c2 and c3 are dimensionless parameters, in which c1 and c3 have a constant value of 1.42 and 0.2, respectively, c2 is computed from the turbulent kinetic energy (KT) and turbulent production (PT) terms.
2.6. Sediment scour model
The sediment scour model available in FLOW-3D can calculate all the sediment transport processes including Entrainment transport, Bedload transport, Suspended transport, and Deposition. The erosion process starts once the water flows remove the grains from the packed bed and carry them into suspension. It happens when the applied shear stress by water flows exceeds critical shear stress. This process is represented by entrainment transport in the numerical model. After entrained, the grains carried by water flow are represented by suspended load transport. After that, some suspended grains resort to settling because of the combined effect of gravity, buoyancy, and friction. This process is described through a deposition. Finally, the grains sliding motions are represented by bedload transport in the model. For the entrainment process, the shear stress applied by the fluid motion on the packed bed surface is calculated using the standard wall function as shown in Eq.7.(7)ks,i=Cs,i∗d50
where ks,i is the Nikuradse roughness and Cs,i is a user-defined coefficient. The critical bed shear stress is defined by a dimensionless parameter called the critical shields number as expressed in Eq.8.(8)θcr,i=τcr,i‖g‖diρi-ρf
where θcr,i is the critical shields number, τcr,i is the critical bed shear stress, g is the absolute value of gravity acceleration, di is the diameter of the sediment grain, ρi is the density of the sediment species (i) and ρf is the density of the fluid. The value of the critical shields number is determined according to the Soulsby-Whitehouse equation.(9)θcr,i=0.31+1.2d∗,i+0.0551-exp-0.02d∗,i
where d∗,i is the dimensionless diameter of the sediment, given by Eq.10.(10)d∗,i=diρfρi-ρf‖g‖μf213
where μf is the fluid dynamic viscosity. For the sloping bed interface, the value of the critical shields number is modified according to Eq.11.(11)θ`cr,i=θcr,icosψsinβ+cos2βtan2φi-sin2ψsin2βtanφi
where θ`cr,i is the modified critical shields number, φi is the angle of repose for the sediment, β is the angle of bed slope and ψ is the angle between the flow and the upslope direction. The effects of the rolling, hopping, and sliding motions of grains along the packed bed surface are taken by the bedload transport process. The volumetric bedload transport rate (qb,i) per width of the bed is expressed in Eq.12.(12)qb,i=Φi‖g‖ρi-ρfρfdi312
where Φi is the dimensionless bedload transport rate is calculated by using Meyer Peter and Müller equation.(13)Φi=βMPM,iθi-θ`cr,i1.5cb,i
where βMPM,i is the Meyer Peter and Müller user-defined coefficient and cb,i is the volume fraction of species i in the bed material. The suspended load transport is calculated as shown in Eq.14.(14)∂Cs,i∂t+∇∙Cs,ius,i=∇∙∇DCs,i
where Cs,i is the suspended sediment mass concentration, D is the diffusivity, and us,i is the grain velocity of species i. Entrainment and deposition are two opposing processes that take place at the same time. The lifting and settling velocities for both entrainment and deposition processes are calculated according to Eq.15 and Eq.16, respectively.(15)ulifting,i=αid∗,i0.3θi-θ`cr,igdiρiρf-1(16)usettling,i=υfdi10.362+1.049d∗,i3-10.36
where αi is the entrainment coefficient of species i and υf is the kinematic viscosity of the fluid.
2.7. Grid type
Using simple rectangular orthogonal elements in planes and hexahedral in volumes in the (FLOW-3D) program makes the mesh generation process easier, decreases the required memory, and improves numerical accuracy. Two mesh blocks were used in a joined form with a size ratio of 2:1. The first mesh block is coarser, which contains the reservoir water, and the second mesh block is finer, which contains the dam. For achieving accuracy and efficiency in results, the mesh size is determined by using a grid convergence test. The optimum uniform cell size for the first mesh block is 0.012 m and for the second mesh block is 0.006 m.
2.8. Time step
The maximum time step size is determined by using a Courant number, which controls the distance that the flow will travel during the simulation time step. In this study, the Courant number was taken equal to 0.25 to prevent the flow from traveling through more than one cell in the time step. Based on the Courant number, a maximum time step value of 0.00075 s was determined.
2.9. Numerical model validation
The numerical model accuracy was achieved by comparing the numerical model results with previous experimental results. The experimental study of Schmocker and Hager [7] was based on 31 tests with changes in six parameters (d50, Ho, Bo, Lk, XD, and Qin). All experimental tests were conducted in a straight open glass-sided flume. The horizontal flume has a rectangular cross-section with a width of 0.4 m and a height of 0.7 m. The flume was provided with a flow straightener and an intake with a length of 0.66 m. All tested dams were inserted at various distances (XD) from the intake. Test No.1 from this experimental program was chosen to validate the numerical model. The different parameters used in test No.1 are as follows:
(1) uniform sediment with a mean diameter (d50 = 0.31 mm), (2) Ho = 0.2 m, (3) Bo = 0.2 m, (4) Lk = 0.1 m,
(5) XD = 1.0 m, (6) Qin = 6.0 lit/s, (7) Su and Sd = 2:1, (8) mass density (ρs = 2650 kg/m3) (9) Homogenous and non-cohesive embankment dam. As shown in Fig. 2, the simulation is contained within a rectangular grid with dimensions: 3.56 m in the x-direction (where 0.66 m is used as inlet, 0.9 m as dam base width, and 1.0 m as outlet), in y-direction 0.2 m (dam length), and in the z-direction 0.3 m, which represents the dam height (0.2 m) with a free distance (0.1 m) above the dam. There are two main reasons that this experimental program is preferred for the validation process. The first reason is that this program deals with homogenous, non-cohesive soil, which is available in FLOW-3D. The second reason is that this program deals with small-scale models which saves time for numerical simulation. Finally, some important assumptions were considered during the validation process. The flow is assumed to be incompressible, viscous, turbulent, and three-dimensional.
By comparing dam profiles at different time instants for the experimental test with the current numerical model, it appears that the numerical model gives good agreement as shown in Fig. 3 and Fig. 4, with an average error percentage of 9% between the experimental results and the numerical model.
3. Analysis and discussions
The current model is used to study the effects of different parameters such as (initial breach shapes, dimensions, locations, upstream and downstream dam slopes) on the peak outflow discharge, QP, time of peak outflow, tP, and rate of erosion, E.
This study consists of a group of scenarios. The first scenario is changing the shapes of the initial breach according to Singh [1], the most predicted shapes are rectangular and V-notch as shown in Fig. 5. The second scenario is changing the initial breach dimensions (i.e., width and depth). While the third scenario is changing the location of the initial breach. Eventually, the last scenario is changing the upstream and downstream dam slopes.
All scenarios of this study were carried out under the same conditions such as inflow discharge value (Qin=1.0lit/s), dimensions of the tested dam, where dam height (Ho=0.20m), crest width.
(Lk=0.1m), dam length (Bo=0.20m), and homogenous & non-cohesive soil with a mean diameter (d50=0.31mm).
3.1. Dam breaching process evolution
The dam breaching process is a very complex process due to the quick changes in hydrodynamic conditions during dam failure. The dam breaching process starts once water flows reach the downstream face of the dam. During the initial stage of dam breaching, the erosion process is relatively quiet due to low velocities of flow. As water flows continuously, erosion rates increase, especially in two main zones: the crest and the downstream face. As soon as the dam crest is totally eroded, the water levels in the dam reservoir decrease rapidly, accompanied by excessive erosion in the dam body. The erosion process continues until the water levels in the dam reservoir equal the remaining height of the dam.
According to Zhou et al. [11], the breaching process consists of three main stages. The first stage starts with beginning overtopping flow, then ends when the erosion point directed upstream and reached the inflection point at the inflection time (ti). The second stage starts from the end of the stage1 until the occurrence of peak outflow discharge at the peak outflow time (tP). The third stage starts from the end of the stage2 until the value of outflow discharge becomes the same as the value of inflow discharge at the final time (tf). The outflow discharge from the dam breach increases rapidly during stage1 and stage2 because of the large dam storage capacity (i.e., the dam reservoir is totally full of water) and excessive erosion. While at stage3, the outflow values start to decrease slowly because most of the dam’s storage capacity was run out. The end of stage3 indicates that the dam storage capacity was totally run out, so the outflow equalized with the inflow discharge as shown in Fig. 6 and Fig. 7.
3.2. The effect of initial breach shape
To identify the effect of the initial breach shape on the evolution of the dam breaching process. Three tests were carried out with different cross-section areas for each shape. The initial breach is created at the center of the dam crest. Each test had an ID to make the process of arranging data easier. The rectangular shape had an ID (Rec5h & 5b), which means that its depth and width are equal to 5% of the dam height, and the V-notch shape had an ID (V-noch5h & 1:1) which means that its depth is equal to 5% of the dam height and its side slope is equal to 1:1. The comparison between rectangular and V-notch shapes is done by calculating the ratio between maximum dam height at different times (ZMax) to the initial dam height (Ho), rate of erosion, and hydrograph of outflow discharge for each test. The rectangular shape achieves maximum erosion rate and minimum inflection time, in addition to a rapid decrease in the dam reservoir levels. Therefore, the dam breaching is faster in the case of a rectangular shape than in a V-notch shape, which has the same cross-section area as shown in Fig. 8.
Also, by comparing the hydrograph for each test, the peak outflow discharge value in the case of a rectangular shape is higher than the V-notch shape by 5% and the time of peak outflow for the rectangular shape is shorter than the V-notch shape by 9% as shown in Fig. 9.
3.3. The effect of initial breach dimensions
The results of the comparison between the different initial breach shapes indicate that the worst initial breach shape is rectangular, so the second scenario from this study concentrated on studying the effect of a change in the initial rectangular breach dimensions. Groups of tests were carried out with different depths and widths for the rectangular initial breach. The first group had a depth of 5% from the dam height and with three different widths of 5,10, and 15% from the dam height, the second group had a depth of 10% with three different widths of 5,10, and 15%, the third group had a depth of 15% with three different widths of 5,10, and 15% and the final group had a width of 15% with three different heights of 5, 10, and 15% for a rectangular breach shape. The comparison was made as in the previous section to determine the worst case that leads to the quick dam failure as shown in Fig. 10.
The results show that the (Rec 5 h&15b) test achieves a maximum erosion rate for a shorter period of time and a minimum ratio for (Zmax / Ho) as shown in Fig. 10, which leads to accelerating the dam failure process. The dam breaching process is faster with the minimum initial breach depth and maximum initial breach width. In the case of a minimum initial breach depth, the retained head of water in the dam reservoir is high and the crest width at the bottom of the initial breach (L`K) is small, so the erosion point reaches the inflection point rapidly. While in the case of the maximum initial breach width, the erosion perimeter is large.
3.4. The effect of initial breach location
The results of the comparison between the different initial rectangular breach dimensions indicate that the worst initial breach dimension is (Rec 5 h&15b), so the third scenario from this study concentrated on studying the effect of a change in the initial breach location. Three locations were checked to determine the worst case for the dam failure process. The first location is at the center of the dam crest, which was named “Center”, the second location is at mid-distance between the dam center and dam edge, which was named “Mid”, and the third location is at the dam edge, which was named “Edge” as shown in Fig. 11. According to this scenario, the results indicate that the time of peak outflow discharge (tP) is the same in the three cases, but the maximum value of the peak outflow discharge occurs at the center location. The difference in the peak outflow values between the three cases is relatively small as shown in Fig. 12.
The rates of erosion were also studied for the three cases. The results show that the maximum erosion rate occurs at the center location as shown in Fig. 13. By making a comparison between the three cases for the dam storage volume. The results show that the center location had the minimum values for the dam storage volume, which means that a large amount of water has passed to the downstream area as shown in Fig. 14. According to these results, the center location leads to increased erosion rate and accelerated dam failure process compared with the two other cases. Because the erosion occurs on both sides, but in the case of edge location, the erosion occurs on one side.
3.5. The effect of upstream and downstream dam slopes
The results of the comparison between the different initial rectangular breach locations indicate that the worst initial breach location is the center location, so the fourth scenario from this study concentrated on studying the effect of a change in the upstream (Su) and downstream (Sd) dam slopes. Three slopes were checked individually for both upstream and downstream slopes to determine the worst case for the dam failure process. The first slope value is (2H:1V), the second slope value is (2.5H:1V), and the third slope value is (3H:1V). According to this scenario, the results show that the decreasing downstream slope angle leads to increasing time of peak outflow discharge (tP) and decreasing value of peak outflow discharge. The difference in the peak outflow values between the three cases for the downstream slope is 2%, as shown in Fig. 15, but changing the upstream slope has a negligible impact on the peak outflow discharge and its time as shown in Fig. 16.
The rates of erosion were also studied in the three cases for both upstream and downstream slopes. The results show that the maximum erosion rate increases by 6.0% with an increasing downstream slope angle by 4°, as shown in Fig. 17. The results also indicate that the erosion rates aren’t affected by increasing or decreasing the upstream slope angle, as shown in Fig. 18. According to these results, increasing the downstream slope angle leads to increased erosion rate and accelerated dam failure process compared with the upstream slope angle. Because of increasing shear stress applied by water flows in case of increasing downstream slope.
According to all previous scenarios, the dimensionless peak outflow discharge QPQin is presented for a fixed dam height (Ho) and inflow discharge (Qin). Fig. 19 illustrates the relationship between QP∗=QPQin and.
Lr=ho2/3∗bo2/3Ho. The deduced relationship achieves R2=0.96.(17)QP∗=2.2807exp-2.804∗Lr
4. Conclusions
A spatial dam breaching process was simulated by using FLOW-3D Software. The validation process was performed by making a comparison between the simulated results of dam profiles and the dam profiles obtained by Schmocker and Hager [7] in their experimental study. And also, the peak outflow value recorded an error percentage of 12% between the numerical model and the experimental study. This model was used to study the effect of initial breach shape, dimensions, location, and dam slopes on peak outflow discharge, time of peak outflow, and the erosion process. By using the parameters obtained from the validation process, the results of this study can be summarized in eight points as follows.1.
The rectangular initial breach shape leads to an accelerating dam failure process compared with the V-notch.2.
The value of peak outflow discharge in the case of a rectangular initial breach is higher than the V-notch shape by 5%.3.
The time of peak outflow discharge for a rectangular initial breach is shorter than the V-notch shape by 9%.4.
The minimum depth and maximum width for the initial breach achieve maximum erosion rates (increasing breach width, b0, or decreasing breach depth, h0, by 5% from the dam height leads to an increase in the maximum rate of erosion by 11% and 15%, respectively), so the dam failure is rapid.5.
The center location of the initial breach leads to an accelerating dam failure compared with the edge location.6.
The initial breach location has a negligible effect on the peak outflow discharge value and its time.7.
Increasing the downstream slope angle by 4° leads to an increase in both peak outflow discharge and maximum rate of erosion by 2.0% and 6.0%, respectively.8.
The upstream slope has a negligible effect on the dam breaching process.
References
[1]V. SinghDam breach modeling technologySpringer Science & Business Media (1996)Google Scholar[2]Wahl TL. Prediction of embankment dam breach parameters: a literature review and needs assessment. 1998.Google Scholar[3]Z. Alhasan, J. Jandora, J. ŘíhaStudy of dam-break due to overtopping of four small dams in the Czech RepublicActa Universitatis Agriculturae et Silviculturae Mendelianae Brunensis, 63 (3) (2015), pp. 717-729 View PDFCrossRefView Record in ScopusGoogle Scholar[4]D. FreadBREACH, an erosion model for earthen dam failures: Hydrologic Research LaboratoryNOAA, National Weather Service (1988)Google Scholar[5]J. Říha, S. Kotaška, L. PetrulaDam Break Modeling in a Cascade of Small Earthen Dams: Case Study of the Čižina River in the Czech RepublicWater, 12 (8) (2020), p. 2309, 10.3390/w12082309 View PDFView Record in ScopusGoogle Scholar[6]E. Goodarzi, L. Teang Shui, M. ZiaeiDam overtopping risk using probabilistic concepts–Case study: The Meijaran DamIran Ain Shams Eng J, 4 (2) (2013), pp. 185-197ArticleDownload PDFView Record in ScopusGoogle Scholar[7]L. Schmocker, W.H. HagerPlane dike-breach due to overtopping: effects of sediment, dike height and dischargeJ Hydraul Res, 50 (6) (2012), pp. 576-586 View PDFCrossRefView Record in ScopusGoogle Scholar[8]J.S. Walder, R.M. Iverson, J.W. Godt, M. Logan, S.A. SolovitzControls on the breach geometry and flood hydrograph during overtopping of noncohesive earthen damsWater Resour Res, 51 (8) (2015), pp. 6701-6724View Record in ScopusGoogle Scholar[9]H. Wei, M. Yu, D. Wang, Y. LiOvertopping breaching of river levees constructed with cohesive sedimentsNat Hazards Earth Syst Sci, 16 (7) (2016), pp. 1541-1551 View PDFCrossRefView Record in ScopusGoogle Scholar[10]Y. Yang, S.-Y. Cao, K.-J. Yang, W.-P. LiYang K-j, Li W-p. Experimental study of breach process of landslide dams by overtopping and its initiation mechanismsJ Hydrodynamics, 27 (6) (2015), pp. 872-883ArticleDownload PDFCrossRefView Record in ScopusGoogle Scholar[11]G.G.D. Zhou, M. Zhou, M.S. Shrestha, D. Song, C.E. Choi, K.F.E. Cui, et al.Experimental investigation on the longitudinal evolution of landslide dam breaching and outburst floodsGeomorphology, 334 (2019), pp. 29-43ArticleDownload PDFView Record in ScopusGoogle Scholar[12]J. Zhang, Z.-x. Guo, S.-y. CaoYang F-g. Experimental study on scour and erosion of blocked damWater Sci Eng, 5 (2012), pp. 219-229ArticleDownload PDFView Record in ScopusGoogle Scholar[13]K. Höeg, A. Løvoll, K. VaskinnStability and breaching of embankment dams: Field tests on 6 m high damsInt J Hydropower Dams, 11 (2004), pp. 88-92View Record in ScopusGoogle Scholar[14]H. Hakimzadeh, V. Nourani, A.B. AminiGenetic programming simulation of dam breach hydrograph and peak outflow dischargeJ Hydrol Eng, 19 (4) (2014), pp. 757-768View Record in ScopusGoogle Scholar[15]A.R. Refaiy, N.M. AboulAtta, N.Y. Saad, D.A. El-MollaModeling the effect of downstream drain geometry on seepage through earth damsAin Shams Eng J, 12 (3) (2021), pp. 2511-2531ArticleDownload PDFView Record in ScopusGoogle Scholar[16]Y. Zhu, P.J. Visser, J.K. Vrijling, G. WangExperimental investigation on breaching of embankmentsScience China Technological Sci, 54 (1) (2011), pp. 148-155 View PDFCrossRefView Record in ScopusGoogle Scholar[17]M.-H. Yu, H.-Y. Wei, Y.-J. Liang, Y. ZhaoInvestigation of non-cohesive levee breach by overtopping flowJ Hydrodyn, 25 (4) (2013), pp. 572-579ArticleDownload PDFCrossRefView Record in ScopusGoogle Scholar[18]S. Wu, M. Yu, H. Wei, Y. Liang, J. ZengNon-symmetrical levee breaching processes in a channel bend due to overtoppingInt J Sedim Res, 33 (2) (2018), pp. 208-215ArticleDownload PDFView Record in ScopusGoogle Scholar[19]O. Saberi, G. ZenzNumerical investigation on 1D and 2D embankment dams failure due to overtopping flowInt J Hydraulic Engineering, 5 (2016), pp. 9-18View Record in ScopusGoogle Scholar[20]M. Guan, N.G. Wright, P.A. Sleigh2D Process-Based Morphodynamic Model for Flooding by Noncohesive Dyke BreachJ Hydraul Eng, 140 (7) (2014), p. 04014022, 10.1061/(ASCE)HY.1943-7900.0000861 View PDFView Record in ScopusGoogle Scholar[21]W. Wu, R. Marsooli, Z. HeDepth-Averaged Two-Dimensional Model of Unsteady Flow and Sediment Transport due to Noncohesive Embankment Break/BreachingJ Hydraul Eng, 138 (6) (2012), pp. 503-516View Record in ScopusGoogle Scholar[22]Z. Wang, D.S. BowlesThree-dimensional non-cohesive earthen dam breach model. Part 1: Theory and methodologyAdv Water Resour, 29 (10) (2006), pp. 1528-1545ArticleDownload PDFView Record in ScopusGoogle Scholar[23]Říha J, Duchan D, Zachoval Z, Erpicum S, Archambeau P, Pirotton M, et al. Performance of a shallow-water model for simulating flow over trapezoidal broad-crested weirs. J Hydrology Hydromechanics. 2019;67:322-8.Google Scholar[24]C.B. VreugdenhilNumerical methods for shallow-water flowSpringer Science & Business Media (1994)Google Scholar[25]L.A. Larocque, J. Imran, M.H. Chaudhry3D numerical simulation of partial breach dam-break flow using the LES and k–∊ turbulence modelsJ Hydraul Res, 51 (2) (2013), pp. 145-157 View PDFCrossRefView Record in ScopusGoogle Scholar[26]C. Yang, B. Lin, C. Jiang, Y. LiuPredicting near-field dam-break flow and impact force using a 3D modelJ Hydraul Res, 48 (6) (2010), pp. 784-792 View PDFCrossRefView Record in ScopusGoogle Scholar[27]FLOW-3D. Version 11.1.1 Flow Science, Inc., Santa Fe, NM. https://wwwflow3dcom.Google Scholar[28]C.W. Hirt, B.D. NicholsVolume of fluid (VOF) method for the dynamics of free boundariesJ Comput Phys, 39 (1) (1981), pp. 201-225ArticleDownload PDFGoogle Scholar[29]S.V. PatankarNumerical heat transfer and fluid flow, Hemisphere PublCorp, New York, 58 (1980), p. 288View Record in ScopusGoogle Scholar[30]M. Alemi, R. MaiaNumerical simulation of the flow and local scour process around single and complex bridge piersInt J Civil Eng, 16 (5) (2018), pp. 475-487 View PDFCrossRefView Record in ScopusGoogle Scholar
Using an improved Carreau constitutive model, a numerical simulation of the casting process of a type of solid propellant slurry vacuum plate casting was carried out using the Flow3D software. Through the flow process in the orifice flow channel and the combustion chamber, the flow velocity of the slurry passing through the plate flow channel was quantitatively analyzed, and the viscosity, shear rate, and leveling characteristics of the slurry in the combustion chamber were qualitatively analyzed and predicted. The pouring time, pouring quality, and flow state predicted by the numerical simulation were verified using a visual tester consisting of a vacuum plate casting system in which a pouring experiment was carried out. Studies have shown that HTPB three-component propellant slurry is a typical yielding pseudoplastic fluid. When the slurry flows through the flower plate and the airfoil, the fluid shear rate reaches its maximum value and the viscosity of the slurry decreases. The visual pouring platform was built and the experiment was controlled according to the numerically-calculated parameters, ensuring the same casting speed. The comparison between the predicted casting quality and the one obtained in the verification test resulted in an error less than 10 %. Moreover, the error between the simulated casting completion time and the process verification test result was also no more than 10 %. Last, the flow state of the slurry during the simulation was consistent with the one during the experimental test. The overall leveling of the slurry in the combustion chamber was adequate and no relatively large holes and flaws developed during the pouring process.
개선된 Carreau 구성 모델을 사용하여 FLOW-3D 소프트웨어를 사용하여 고체 추진제 슬러리 진공판 유형의 Casting Process에 대한 수치 시뮬레이션을 수행했습니다. 오리피스 유로와 연소실에서의 유동과정을 통해 판 유로를 통과하는 슬러리의 유속을 정량적으로 분석하고, 연소실에서 슬러리의 점도, 전단율, 레벨링 특성을 정성적으로 분석하하고, 예측하였습니다.
타설시간, 타설품질, 수치해석으로 예측된 유동상태는 타설실험을 수행한 진공판주조시스템으로 구성된 비주얼 테스터를 이용하여 검증하였습니다.
연구에 따르면 HTPB 3성분 추진제 슬러리는 전형적인 생성 가소성 유체입니다. 슬러리가 플라워 플레이트와 에어포일을 통과할 때 유체 전단율이 최대값에 도달하고 슬러리의 점도가 감소합니다.
시각적 주입 플랫폼이 구축되었고 동일한 주조 속도를 보장하기 위해 수치적으로 계산된 매개변수에 따라 실험이 제어되었습니다. 예측된 casting 품질과 검증 테스트에서 얻은 품질을 비교한 결과 10 % 미만의 오류가 발생했습니다.
또한 모의 casting 완료시간과 공정검증시험 결과의 오차도 10 % 이하로 나타났습니다.
마지막으로 시뮬레이션 중 슬러리의 흐름 상태는 실험 테스트 시와 일치하였다. 연소실에서 슬러리의 전체 레벨링은 적절했으며 주입 과정에서 상대적으로 큰 구멍과 결함이 발생하지 않았습니다.
References
[1] B. M. Bandgar, V. N. Krishnamurthy, T. Mukundan, K. C. Sharma, Mathematical Modeling of Rheological Properties of HydroxylTerminated Polybutadiene Binder and Dioctyl Adipate Plasticizer, J. Appl. Polym. Sci. 2002, 85, 1002–1007. [2] B. Thiyyarkandy, M. Jain, G. S. Dombe, M. Mehilal, P. P. Singh, B. Bhattacharya, Numerical Studies on Flow Behavior of Composite Propellant Slurry during Vacuum Casting, J.Aerosp.Technol. Manage. 2012, 4, 197–203. [3] T. Shimada, H. Habu, Y. Seike, S. Ooya, H. Miyachi, M. Ishikawa, X-Ray Visualization Measurement of Slurry Flow in Solid Propellant Casting, Flow Meas. Instrum. 2007, 18, 235–240. [4] Y. Damianou, G. C. Georgiou, On Poiseuille Flows of a Bingham Plastic with Pressure-Dependent Rheological Parameters, J. Non-Newtonian Fluid Mech. 2017, 250, 1–7. [5] S. Sadasivan, S. K. Arumugam, M. Aggarwal, Numerical Simulation of Diffuser of a Gas Turbine using the Actuator Disc Model, J.Appl. Fluid Mech. 2019, 12, 77–84. [6] M. Acosta, V. L. Wiesner, C. J. Martinez, R. W. Trice, J. P. Youngblood, Effect of Polyvinylpyrrolidone Additions on the Rheology of Aqueous, Highly Loaded Alumina Suspensions, J. Am. Ceram. Soc. 2013, 96, 1372–1382. [7] Y. Wu, Numerical Simulation and Experiment Study of Flower Plate Pouring System for Solid Propellant, Chin. J. Expl. Propell. 2017, 41, 506–511. [8] T. M. G. Chu, J. W. Halloran, High-Temperature Flow Behavior of Ceramic Suspensions, J. Am. Ceram. Soc. 2004, 83, 2189– 2195. [9] T. Kaully, A. Siegmann, D. Shacham, Rheology of Highly Filled Natural CaCO3 Composites. I. Effects of Solid Loading and Particle Size Distribution on Capillary Rheometry, Polym. Compos. 2007, 28, 512–523. [10] M. M. Rueda, M.-C. Auscher, R. Fulchiron, T. Périé, G. Martin, P. Sonntag, P. Cassagnau, Rheology and Applications of Highly Filled Polymers: A Review of Current Understanding, Prog. Polym. Sci. 2017, 66, 22–53. [11] F. Soltani, Ü. Yilmazer, Slip Velocity and Slip Layer Thickness in Flow of Concentrated Suspensions, J. Appl. Polym. Sci. 1998, 70, 515–522.
[12] E. Landsem, T. L. Jensen, F. K. Hansen. E. Unneberg, T. E. Kristensen, Neutral Polymeric Bonding Agents (NPBA) and Their Use in Smokeless Composite Rocket Propellants Based on HMX-GAP-BuNENA. Propellants, Explos., Pyrotech.. 2012, 37, 581–589. [13] J. Mewis, N. J. Wagner, Colloidal Suspension Rheology, Cambridge University Press, 2011. [14] D. M. Kalyon, An Overview of the Rheological Behavior and Characterization of Energetic Formulations: Ramifications on Safety and Product Quality, J. Energ. Mater. 2006, 24, 213–245. [15] H. Ohshima, Effective Viscosity of a Concentrated Suspension of Uncharged Spherical Soft Particles, Langmuir 2010, 26, 6287–6294.
곡면에서 GMAW 기반 적층 가공의 용접 성형 특성은 중력의 영향을 크게 받습니다. 성형면의 경사각이 크면 혹 비드(hump bead)와 같은 심각한 결함이 발생합니다.
본 논문에서는 양생면에서 용접 비드 형성의 형성 특성과 제어 방법을 연구하기 위해 용접 용융 풀 유동 역학의 전산 모델을 수립하고 제안된 모델을 검증하기 위해 증착 실험을 수행하였습니다.
결과는 용접 비드 경사각(α)이 증가함에 따라 역류의 속도가 증가하고 상향 용접의 경우 α > 60°일 때 불규칙한 험프 결함이 나타나는 것으로 나타났습니다.
상부 과잉 액체의 하향 압착력과 하부 상향 유동의 반동력과 표면장력 사이의 상호작용은 용접 혹 형성의 주요 요인이었다. 하향 용접의 경우 양호한 형태를 얻을 수 있었으며, 용접 비드 경사각이 증가함에 따라 용접 높이는 감소하고 용접 폭은 증가하였습니다.
하향 및 상향 용접을 위한 곡면의 용융 거동 및 성형 특성을 기반으로 험프 결함을 제어하기 위해 위브 용접을 통한 증착 방법을 제안하였습니다.
성형 궤적의 변화로 인해 용접 방향의 중력 성분이 크게 감소하여 용융 풀 흐름의 안정성이 향상되었으며 복잡한 표면에서 안정적이고 일관된 용접 비드를 얻는 데 유리했습니다.
하향 용접과 상향 용접 사이의 단일 비드의 치수 편차는 7% 이내였으며 하향 및 상향 혼합 혼합 비드 중첩 증착에서 비드의 변동 편차는 0.45로 GMAW 기반 적층 제조 공정에서 허용될 수 있었습니다.
이러한 발견은 GMAW를 기반으로 하는 곡선 적층 적층 제조의 용접 비드 형성 제어에 기여했습니다.
The weld forming characteristics of GMAW-based additive manufacturing on curved surface are dramatically influenced by gravity. Large inclined angle of the forming surface would lead to severe defects such as hump bead. In this paper, a computational model of welding molten pool flow dynamics was established to research the forming characteristic and control method of weld bead forming on cured surface, and deposition experiments were conducted to verify the proposed model. Results indicated that the velocity of backward flows increased with the increase of weld bead tilt angle (α) and irregular hump defects appeared when α > 60° for upward welding. The interaction between the downward squeezing force of the excess liquid at the top and the recoil force of the upward flow at the bottom and the surface tension were primary factors for welding hump formation. For downward welding, a good morphology shape could be obtained, and the weld height decreased and the weld width increased with the increase of weld bead tilt angle. Based on the molten behaviors and forming characteristics on curved surface for downward and upward welding, the method of deposition with weave welding was proposed to control hump defects. Gravity component in the welding direction was significantly reduced due to the change of forming trajectory, which improved the stability of the molten pool flow and was beneficial to obtain stable and consistent weld bead on complex surface. The dimensional deviations of the single bead between downward and upward welding were within 7% and the fluctuation deviation of the bead in multi-bead overlapping deposition with mixing downward and upward welding was 0.45, which could be acceptable in GMAW-based additive manufacturing process. These findings contributed to the weld bead forming control of curve layered additive manufacturing based on GMAW.
7.Xie FB, Chen LF, Li ZY, Tang K (2020) Path smoothing and feed rate planning for robotic curved layer additive manufacturing. Robot Comput Integr Manuf 65. https://doi.org/10.1016/j.rcim.2020.101967
8.Ding YY, Dwivedi R, Kovacevic R (2017) Process planning for 8-axis robotized laser-based direct metal deposition system: a case on building revolved part. Robot Comput Integr Manuf 44:67–76. https://doi.org/10.1016/j.rcim.2016.08.008ArticleGoogle Scholar
12.Yuan L, Pan ZX, Ding DH, He FY, Duin SV, Li HJ, Li WH (2020) Investigation of humping phenomenon for the multi-directional robotic wire and arc additive manufacturing. Robot Comput Integr Manuf 63. https://doi.org/10.1016/j.rcim.2019.101916
13.Nguyen MC, Medale M, Asserin O, Gounand S, Gilles P (2017) Sensitivity to welding positions and parameters in GTA welding with a 3D multiphysics numerical model. Numer Heat Transf Part A Appl 71:233–249. https://doi.org/10.1080/10407782.2016.1264747ArticleGoogle Scholar
17.Philip Y, Xu ZY, Wang Y, Wang R, Ye X (2019) Investigation of humping defect formation in a lap joint at a high-speed hybrid laser-GMA welding. Results Phys 13. https://doi.org/10.1016/j.rinp.2019.102341
18.Hu ZQ, Qin XP, Shao T, Liu HM (2018) Understanding and overcoming of abnormity at start and end of the weld bead in additive manufacturing with GMAW. Int J Adv Manuf Technol 95:2357–2368. https://doi.org/10.1007/s00170-017-1392-9ArticleGoogle Scholar
19.Tang SY, Wang GL, Huang C, Li RS, Zhou SY, Zhang HO (2020) Investigation, modeling and optimization of abnormal areas of weld beads in wire and arc additive manufacturing. Rapid Prototyp J 26:1183–1195. https://doi.org/10.1108/RPJ-08-2019-0229ArticleGoogle Scholar
20.Bai X, Colegrove P, Ding J, Zhou XM, Diao CL, Bridgeman P, Honnige JR, Zhang HO, Williams S (2018) Numerical analysis of heat transfer and fluid flow in multilayer deposition of PAW-based wire and arc additive manufacturing. Int J Heat Mass Transf 124:504–516. https://doi.org/10.1016/j.ijheatmasstransfer.2018.03.085ArticleGoogle Scholar
21.Siewert E, Schein J, Forster G (2013) Determination of enthalpy, temperature, surface tension and geometry of the material transfer in PGMAW for the system argon-iron. J Phys D Appl Phys 46. https://doi.org/10.1088/0022-3727/46/22/224008
23.Fachinotti VD, Cardona A (2008) Semi-analytical solution of the thermal field induced by a moving double-ellipsoidal welding heat source in a semi-infinite body. Mec Comput XXVII:1519–1530
24.Nguyen NT, Mai YW, Simpson S, Ohta A (2004) Analytical approximate solution for double ellipsoidal heat source in finite thick plate. Weld J 83:82–93Google Scholar
25.Goldak J, Chakravarti A, Bibby M (1985) A double ellipsoid finite element model for welding heat sources. IIW Doc. No. 212-603-85
29.Zhan XH, Zhang D, Liu XB, Chen J, Wei YH, Liu RP (2017) Comparison between weave bead welding and multi-layer multi-pass welding for thick plate Invar steel. Int J Adv Manuf Technol 88:2211–2225. https://doi.org/10.1007/s00170-016-8926-4ArticleGoogle Scholar
30.Xu GX, Li L, Wang JY, Zhu J, Li PF (2018) Study of weld formation in swing arc narrow gap vertical GMA welding by numerical modeling and experiment. Int J Adv Manuf Technol 96:1905–1917. https://doi.org/10.1007/s00170-018-1729-zArticleGoogle Scholar
Flood Inundation Evolution of Barrier Lake and Evaluation of Regional Ecological Spatiotemporal Response — A Case Study of Sichuan-Tibet Region
Abstract
중국 쓰촨-티베트 지역은 댐 호수의 발생과 붕괴를 동반한 지진 재해가 빈번한 지역이었습니다. 댐 호수의 붕괴는 하류 직원의 생명과 재산 안전을 심각하게 위협합니다.
동시에 국내외 학자들은 주변의 댐 호수에 대해 우려하고 있으며 호수에 대한 생태 연구는 거의 없으며 댐 호수가 생태에 미치는 영향은 우리 호수 건설 프로젝트에서 매우 중요한 계몽 의의를 가지고 있습니다.
이 기사의 목적은 방벽호의 댐 붕괴 위험을 과학적으로 예측하고 생태 환경에 대한 영향을 조사하며 통제 조치를 제시하는 것입니다. 본 논문은 쓰촨-티베트 지역의 Diexihaizi, Tangjiashan 댐호, Hongshihe 댐의 4대 댐 호수 사건을 기반으로 원격 감지 이미지에서 수역을 추출하고 HEC-RAS 모델을 사용하여 위험이 있는지 여부를 결정합니다.
댐 파손 여부 및 댐의 경로 예측; InVEST 모델을 이용하여 1990년부터 2020년까지 가장 작은 행정 구역(군/구)이 위치한 서식지를 평가 및 분석하고, 홍수 침수 결과를 기반으로 평가합니다. 결과는 공학적 처리 후 안정적인 댐 호수(Diexi Haizi)가 서식지 품질 지수에 안정화 효과가 있음을 보여줍니다.
댐 호수의 형성은 인근 토지 이용 유형과 지역 경관 생태 패턴을 변화 시켰습니다. 서식지 품질 지수는 사이 호수 주변 1km 지역에서 약간 감소하지만 3km 지역과 5km 지역에서 서식지 품질이 향상됩니다. 인공 홍수 방류 및 장벽 호수의 공학적 보강이 필요합니다.
이 논문에서 인간의 통제가 강한 지역은 다른 지역의 서식지 질 지수보다 더 잘 회복될 것입니다.
The Sichuan-Tibet region of China has always been an area with frequent earthquake disasters, accompanied by the occurrence and collapse of dammed lakes. The collapse of dammed lakes seriously threatens the lives and property safety of downstream personnel.
At the same time, domestic and foreign scholars are concerned about the surrounding dammed lake there are few ecological studies on the lake, and the impact of the dammed lake on the ecology has very important enlightenment significance for our lake construction project. It is the purpose of this article to scientifically predict the risk of dam break in a barrier lake, explore its impact on the ecological environment and put forward control measures.
Based on the four major dammed lake events of Diexihaizi, Tangjiashan dammed lake, and Hongshihe dammed lake in the Sichuan-Tibet area, this paper extracts water bodies from remote sensing images and uses the HEC-RAS model to determine whether there is a risk of the dam break and whether Forecast the route of the dam; and use the InVEST model to evaluate and analyze the habitat of the smallest administrative district (county/district) where it is located from 1990 to 2020 and make an evaluation based on the results of flood inundation.
The results show that the stable dammed lake (Diexi Haizi) after engineering treatment has a stabilizing effect on the habitat quality index. The formation of the dammed lake has changed the nearby land-use types and the regional landscape ecological pattern.
The habitat quality index will decrease slightly in the 1 km area around Sai Lake, but the habitat quality will increase in the 3 km area and the 5 km area. Artificial flood discharge and engineering reinforcement of barrier lakes are necessary. In this paper, the areas with strong human control will recover better than other regions’ habitat quality index.
Fengshan Jiang ( florachaing@mail.ynu.edu.cn ) Yunnan University https://orcid.org/0000-0001-6231-6180 Xiaoai Dai Chengdu University of Technology https://orcid.org/0000-0003-1342-6417 Zhiqiang Xie Yunnan University Tong Xu Yunnan University Siqiao Yin Yunnan University Ge Qu Chengdu University of Technology Shouquan Yang Yunnan University Yangbin Zhang Yunnan University Zhibing Yang Yunnan University Jiarui Xu Yunnan University Zhiqun Hou Kunming institute of surveying and mapping
Keywords
dammed lake, regional ecology, flood simulation, habitat quality
References
Chaoying Hu H S, Tianming Zhang. 2017. Environmental impact assessment of barrier lake treatment project based on ecological footprint[J]. People’s Yangtze River, 48: 30-32
Dai F C, Lee C F, Deng J H, et al. 2004. The 1786 earthquake-triggered landslide dam and subsequent dam-break flood on the Dadu River, southwestern China[J]. Geomorphology, 65.
Dongjing Chen Z X 2002. Research on Ecological Security Evaluation of Inland River Basin in Northwest China——A Case Study of Zhangye Region in the Middle Reaches of Heihe River Basin[J]. Arid zone geography: 219-224
Dongsheng Chang L Z, Yao Xu, Runqiu Huang. 2009. Risk Assessment of Overtopping Dam Burst in Hongshi River Barrier Lake[J]. Journal of Engineering Geology, 17: 50-55
Fan X, Yunus Ali P, Jansen John D, et al. 2019. Comment on ‘Gigantic rockslides induced by fluvial incision in the Diexi area along the eastern margin of the Tibetan Plateau’ by Zhao et al. (2019) Geomorphology 338, 27–42[J]. Geomorphology.
Feng Yu X L, Hong Wang, Hongjing Yu. 2006. Land Use Change and Ecological Security Evaluation in Huangfuchuan Watershed[J]. Acta Geographica Sinica: 645-653.
Hafiyyan Q, Adityawan M B, Harlan D, et al. 2021. Comparison of Taylor Galerkin and FTCS models for dam-break simulation[J]. IOP Conference Series: Earth and Environmental Science, 737.
Haiwen Li X B 2020. Comprehensive Evaluation of the Restoration Status of Damaged Ecological Space along the Plateau Fragile Area of the Sichuan-Tibet Railway[J]. Journal of Railway Science and Engineering, 17: 2412-2422.
Haohao Li X R, Huabin Yang. 2008. Rescue construction and thinking of Hongshihe dammed lake in Qingchuan County[J]. Water Conservancy and Hydropower Technology (Chinese and English): 50-51+62
Hejun Chai, Runqiu Huang, Hanchao Liu I O E G, Chengdu University of Technology 1997. Analysis and Evaluation of the Dangerous Degree of Landslide Blocking the River[J]. Chinese Journal of Geological Hazard and Control: 2-8+16
Hong Wang Y L, Lili Song, Yun Chen. 2020. Comparison of characteristics of thunderstorm and gale activity and environmental factors in Sichuan-Tibet area[J]. Journal of Applied Meteorology, 31: 435-446.
Hongyan X, Xu H, Jiang H, et al. 2020. Potential pollen evidence for the 1933 M 7.5 Diexi earthquake and implications for post-seismic landscape recovery[J]. Environmental Research Letters, 15.
Hui Xu J C, Zhijiu Cui, Pei Guo. 2019. Analysis of Grain Size Characteristics of Sediment in Dammed Lake——Taking Diexi Ancient Dammed Lake in the Upper Minjiang River as an Example[J]. Acta Sedimentologica Sinica, 37: 51-61
Jian Yang B P, Min Zhao. 2014. Research on Ecological Restoration Technology in Wenchuan Earthquake-Stricken Area ——Taking Tangjiashan Barrier Lake Area as an Example[J]. Sichuan Building Science Research, 40: 164-167.
Jian Yang B P 2017. Evaluation of Ecological Quality of Tangjiashan Dammed Lake Region in Beichuan County[J]. People’s Yangtze River, 48: 27-32
Jianfeng Chen Y W, Yang Li. 2006. Application of HEC-RAS model in flood simulation[J]. Northeast Water Resources and Hydropower: 12-13+42+71.
Jiankang Liu Z C, Tao Yu. 2016. Dam failure risk and its impact of Hongshiyan dammed lake in Ludian, Yunnan[J]. Journal of Mountain Science, 34: 208-215
Jianrong Fan B T, Genwei Cheng, Heping Tao, Jianqiang Zhang,Dong Yan, Fenghuan Su. 2008. Information extraction of dammed bodies induced by the May 12 Wenchuan earthquake based on multi-source remote sensing data[J]. Journal of Mountain Science: 257-262.
Jinghuan Tian K Z, Meng Chen, Fuxin Chai. 2012. Research on the application of HEC-RAS model in flood risk analysis and assessment[J]. Hydropower Energy Science, 30: 23-25
Juan He X W 2015. Dam-break flood analysis based on HEC-RAS and HEC-GeoRAS[J]. Journal of Water Resources and Water Transport Engineering: 112-116
Junwei Gan L Y, Jinjun Li. 2017. Research on the Influencing Factors of Sichuan-Tibet Tourism Industry Competitiveness Based on DEMATEL[J]. Arid Land Resources and Environment, 31: 197-202
Lansheng Wang L Y, Xiaoqun Wang, Liping Duan 2005. Discovery of the ancient dammed lake in Diexi, Minjiang River[J]. Journal of Chengdu University of Technology (Natural Science Edition): 1-11
Ma S, Zhu J, Ya. H. Year. Construction of Risk Assessment System of Dam-break in Barrier Lake Based on Collaborative Workflow: 9.
Ming Zeng Y C, Bingyu Zou. 2019. Discussion on the Method of Forecasting the Flood Evolution of Barrier Lake Burst—— Taking “11·3” Jinsha River Baige Barrier Lake as an Example[J]. Water Resources and Hydropower Express, 40: 11-14
Ouyang C, An H, Zhou S, et al. 2019. Insights from the failure and dynamic characteristics of two sequential landslides at Baige village along the Jinsha River, China Landslides[J]. 16.
Peng M, Zhang L M 2012. Analysis of human risks due to dam-break floods—part 1: a new model based on Bayesian networks[J]. Natural Hazards, 64.
Qianfeng Li Y L, Gang Liu, Zhiyun Ouyang, Hua Zheng. 2013. The Impact of Land Use Change on Ecosystem Service Function——Taking Miyun Reservoir Watershed as an Example[J]. Acta Ecologica Sinica, 33: 726-736.
Qiang Xu G Z, Weile Li, Zhaoyang He, Xiujun Dong, Chen Guo, Wenkai Feng. 2018. Analysis and study of two landslides and dams blocking the river in Baige on the Jinsha River in October and November 2018[J]. Journal of Engineering Geology, 26: 1534-1551
Qin Ji J Y, Hongju Chen, Man Li. 2019. Analysis of Economic Differences Along the Sichuan-Tibet Railway from the Perspective of Spatial and Industrial Decomposition[J]. Glacier permafrost: 1-14
Qingchun Li Y H, Yubing Shi. 2020. Study on the stability of the residual dam in Tangjiashan dammed lake[J]. Journal of Underground Space and Engineering, 16: 993-998
Qiwen Xiang J P, Guangze Zhang, Zhengxuan Xu, Dingkai Zhang, Wenli Tu. 2020. Monitoring and Analysis of Surface Deformation in Zheduo Mountain Area of Sichuan-Tibet Railway Based on SBAS Technology[J]. Surveying Engineering, 29: 48-54+59
Shangfu Kuang X W, Jinchi Huang, Yinqi Wei 2008. Analysis and Evaluation of Dam-Break Risk of Barred Lake and Its Influence[J]. China Water Resources: 17-21.
Sheng-Hsueh Y, Yii-Wen P, Jia-Jyun D, et al. 2013. A systematic approach for the assessment of flooding hazard and risk associated with a landslide dam[J]. Natural Hazards, 65.
Sun L 2021. Research on Fast Perception and Simulation Calculation Method of Landslide Dam in Alpine and Gorge Area: Taking Baige Dammed Lake as an Example[J]. Water Conservancy and Hydropower Technology (Chinese and English), 52: 44-52
Tamiru H, O. D M 2021. Application of ANN and HEC-RAS model for flood inundation mapping in lower Baro Akobo River Basin, Ethiopia[J]. Journal of Hydrology: Regional Studies, 36.
Tao Pan S W, Erfu Dai, Yujie Liu. 2013. Spatio-temporal changes of water supply services in the ecosystem of the Three Rivers Source Region based on InVEST model[J]. Journal of Applied Ecology, 24: 183-189
Vera K, Sergey C, Inna K, et al. 2017. Modeling potential scenarios of the Tangjiashan Lake outburst and risk assessment in the downstream valley[J]. Frontiers of Earth Science, 11.
Wang Z 1985. Preliminary Discussion on the Evaluation of Ecological Environment Quality in Minjiang River Basin[J]. Journal of ecology: 29-32
Wei Chen Z S, Hui Guo,Hao Wang, Ting Wei, Nan Li, Kaiyi Zhang Shuxiang Yang, Kaijia Dai. 2007. Analysis of Bird Resources and Habitats in Wuhan Urban Lakes and Urban Wetlands in Winter[J]. Forestry Investigation and Planning: 46- 50
Wei G, Gaohong X, Jun S, et al. 2020. Simulation of Flood Process Based on the Model of Improved Barrier Lake’s Gradual Dam Break Model %J Journal of Coastal Research[J]. 104.
Wei X, Jiang H, Xu H, et al. 2021. Response of sedimentary and pollen records to the 1933 Diexi earthquake on the eastern Tibetan Plateau[J]. Ecological Indicators, 129.
Wei Xu M L, Jie Yang, Chunzhi Li, Xiaojuan Shang. 2011. Risk Analysis of Flood Overflow in Huainan Section of Huaihe River Based on HEC-RAS[J]. Journal of Yangtze River Scientific Research Institute, 28: 13-18
Weiwei Zhan R H, Xiangjun Pei, Weile Li. 2017. Research on empirical prediction model of channel type landslide-debris flow movement distance[J]. Journal of Engineering Geology, 25: 154-163
Xianju Zheng H L, Wenhai Huang. 2015. Numerical Simulation of Reconstruction of Natural Dams Induced by Heavy Rain ——An Example of Tangjiashan Dammed Lake[J]. Business story: 62-63
Xiao-Qun W, Xin H, Man S, et al. 2020. Possible relatedness between the outburst of the Diexi ancient dammed lake and ancient Chengdu’s cultural change[J]. Journal of Mountain Science, 17: 2497-2511.
Xingbo Zhou X D, Yu Yao. 2019. Analysis of the dam-break flood of the Baige dammed lake on the Jinsha River[J]. Hydroelectric Power, 45: 8-12+32
Xinhua Zhang R X, Ming Wang, Zhiqiu Yu, Bingdong Li, Bo Wang. 2020. Investigation and analysis of flood disaster caused by dam break of Baige landslide on Jinsha River[J]. Engineering Science and Technology, 52: 89-100
Xinxiao Yu B Z, Xizhi Lv, Zhige Yang. 2012. Evaluation of Forest Water Conservation Function of Beijing Mountainous Area Based on InVEST Model[J]. Forestry Science, 48: 1-5
Xu J, Guo J, Zhang J, et al. 2021. Route choice model based on cellular automata and cumulative prospect theory: Case analysis of transportation network in Sichuan-Tibet region[J]. Journal of Intelligent & Fuzzy Systems, 40.
Xuan Liang Z Z 2021. Research on the Influence of Numerical Simulation of Tailings Pond Based on FLOW-3D on Downstream[J]. Jiangxi Water Conservancy Science and Technology, 47: 11-20
Yu Zheng P Z, Feng Tang, Li Zhao, Xu Zhao. 2018. Research on the Impact of Land Use Change on Habitat Quality in Changli County Based on InVEST Model[J]. China’s Agricultural Resources and Regionalization, 39: 121-128
Yuanyuan Yang E D, Hua Fu. 2012. Research Framework of Value Evaluation of Ecosystem Service Function Based on InVEST Model[J]. Journal of Capital Normal University (Natural Science Edition), 33: 41-47
Yunfei Ma T L, Jinbiao Xiong. 2021. Numerical simulation of dam-break flow based on VOF method and DFBI model[J]. Applied Technology, 48: 23-28
Zhe Wu X C, Beibei Liu, Jinfeng Chu, Lixu Peng. 2013. Research progress of InVEST model and its application[J]. Tropical Agriculture Science, 33: 58-62
Zhengpeng Li Y H, Yilun Li, Yuehong Ying, Zehua Huangfu. 2021. Numerical simulation of dam-break flood in Qianping Reservoir based on BIM+GIS technology[J]. People’s Yellow River, 43: 160-164
Zhenming Shi X X, Ming Peng, Minglang Lin. 2015. Analysis of Seepage Stability of Barrier Dam with High Permeability Area——Taking Hongshihe Barrier Dam as an Example[J]. Journal of Hydraulic Engineering, 46: 1162-1171.
Zhu J, Qi H, Hu Y, et al. 2012. A DVGE service system for risk assessment of dam-break in barrier lake[J]. International Conference on Automatic Control and Artificial Intelligence (ACAI 2012).
Zhu Y, Peng M, Cai S, et al. 2021. Risk-Based Warning Decision Making of Cascade Breaching of the Tangjiashan Landslide Dam and Two Smaller Downstream Landslide Dams[J]. Frontiers in Earth Science.
Zuyu Chen G H, Qiang Zhang, Shuaifeng Wu. 2020. Disaster Mitigation Analysis of Cascade Hydropower Stations on the Jinsha River in “11.03” Baige Barrier Lake Emergency Treatment[J]. Hydropower, 46: 59-63
Zuyu Chen S C, Lin Wang, Qiming Zhong, Qiang Zhang, Songli Jin. 2020. Inversion analysis of the “11.03” Baige barrier lake burst flood in the upper reaches of the Jinsha River[J]. Science in China: Technological Science, 50: 763-774.
대용량 배출구가 있는 수중 여수로는 일반적으로 홍수 처리 및 침전물 세척의 이중 기능을 수행하기 위해 댐 정상 아래에 제공됩니다. 이 방수로를 통과하는 홍수 물은 난류 거동을 나타냅니다.
게다가 이러한 난류의 수력학적 분석은 어려운 작업입니다.
따라서 본 연구는 파키스탄 Mangla Dam에 건설된 수중 여수로의 수리학적 거동을 수치해석을 통해 조사하는 것을 목적으로 한다. 또한 다양한 작동 조건에서 화기의 유압 성능을 평가했습니다.
Mangla Spillway의 흐름을 수치적으로 모델링하는 데 전산 유체 역학 코드 FLOW 3D가 사용되었습니다. 레이놀즈 평균 Navier-Stokes 방정식은 난류 흐름을 수치적으로 모델링하기 위해 FLOW 3D에서 사용됩니다.
연구 결과에 따르면 개발된 모델은 최대 6%의 허용 오차로 흐름 매개변수를 계산하므로 수중 여수로 흐름을 시뮬레이션할 수 있습니다.
또한, 여수로 슈트 베드 주변 모델에 의해 계산된 공기 농도는 폭기 장치에 램프를 설치한 후 6% 이상으로 상승한 3%로 개발된 모델도 침수형 폭기 장치의 성능을 평가할 수 있음을 보여주었습니다.
Submerged spillways with large capacity outlets are generally provided below the dam crest to perform the dual functions of flood disposal and sediment flushing. Flood water passing through these spillways exhibits turbulent behavior. Moreover; hydraulic analysis of such turbulent flows is a challenging task. Therefore, the present study aims to use numerical simulations to examine the hydraulic behavior of submerged spillways constructed at Mangla Dam, Pakistan. Besides, the hydraulic performance of aerator was also evaluated at different operating conditions. Computational fluid dynamics code FLOW 3D was used to numerically model the flows of Mangla Spillway. Reynolds-averaged Navier–Stokes equations are used in FLOW 3D to numerically model the turbulent flows. The study results indicated that the developed model can simulate the submerged spillway flows as it computed the flow parameters with an acceptable error of up to 6%. Moreover, air concentration computed by model near spillway chute bed was 3% which raised to more than 6% after the installation of ramp on aerator which showed that developed model is also capable of evaluating the performance of submerged spillway aerator.