2023 Volume 64 Issue 10 Pages 2542-2546
“The International Symposium on Innovation in Materials Processing (ISIMP)” was held in Jeju, Korea, from 26th to 29th October 2021. The proceedings for the session on “Integrated Computer-Aided Process Engineering (ICAPE)” were published in October 2022 as a special issue of Materials Transactions (Vol. 63, No. 10). The primary purpose of the ICAPE session was to address the recent advances in scale-bridging simulations and characterization to understand, describe, and predict the microstructure–property relationship of newly developed materials in a lab to industrial-level processes. Among the papers presented at the symposium, this article briefly reviews the following topics: macroscale numerical analysis, such as finite element methods (FEM), microstructure simulations such as phase-field modelling (PFM) and molecular dynamics (MD), and optimization techniques such as machine learning (ML) and design of experiments (DOE).
From developing new materials to their commercialization, migrating a process from the laboratory to pilot plant or commercial scale is the most time-consuming and costly step. Despite decades of efforts devoted to describing material behaviors during manufacturing processes,1–21) simulating “microstructural evolutions” of materials under various processing conditions has received less attention than continuum mechanics modeling, which aims to describe the mechanical behavior of materials during processing. This lack of attention can be attributed to the complex nature of microstructural evolution, which involves co-occurring multiple physical phenomena, including heat transfer, fluid flow, phase transformation, and elastic/plastic deformations. Consequently, integrating material simulation models across multiple length scales is crucial for capturing the process–structure–property (PSP) relationships of materials during manufacturing.
Recent advancements in computational methods, such as phase-field modeling (PFM) and molecular dynamics (MD) simulations, have enabled researchers to study microstructural evolution with high accuracy and efficiency.21–42) In addition, optimization techniques such as machine learning (ML),29–36) artificial intelligence (AI),37–39) and design of experiments (DoE)43–47) may significantly accelerate process development by rapidly exploring and identifying optimal process conditions based on computational and/or experimental data, reducing the need for costly and time-consuming trial-and-error approaches. Hence, integrating these optimization techniques with advanced simulation methods can further enhance the accuracy and efficiency of material property prediction and process optimization.
The Integrated Computer-Aided Process Engineering (ICAPE) session at the International Symposium on Innovation in Materials Processing (ISIMP), which will be held in Jeju, Korea, from 26th to 29th October 2021, aimed to connect experts in the detailed analysis, characterization, and computational modeling of materials across various length scales ranging from atomistic to macroscopic. The session focused on recent breakthroughs in scale-bridging simulations and characterization techniques for understanding, characterizing, and predicting the PSP relationships of newly developed materials at the laboratory and industrial levels. Eleven selected papers47–57) were published in a special issue of Materials Transactions (Vol. 63, No. 10), and this article will review the three main topics covered in the session.
Macroscale numerical modeling can simulate the macroscopic deformation and behavior of materials during various manufacturing processes, including liquid-state (e.g., casting12–15)) and solid-state (e.g., rolling,16–18) forging,18–20) extrusion, and powder57–60)) processes, to predict the resulting microstructure and properties. Large-scale continuum models enforce several simplifying assumptions so that the model can resolve process evolution at the product scale.
Kim et al.50) employed finite element methods (FEM) to investigate the structure–property relationship of lattice truss metals by correlating effective properties with quantitative structural parameters, such as Maxwell stability parameter, truss thickness, and coordination number, excluding porosity. Even at the same porosity, the effective properties can vary by many folds owing to the structural difference. The calculation with nine representative lattice trusses (Fig. 1) demonstrated that the structural dependence of the elastic modulus is higher than that of the thermal conductivity. Among the elastic modulus vs. thermal conductivity plot (Fig. 2) of the nine calculated models, the octa-cross structure had the highest E/k value (where E is the elastic modulus and k is the thermal conductivity), leading to high stiffness merit and effective thermal conductivity reduction. These results could be used for controlling the properties of lattice trusses and designing new lattice structures.
Lattice trusses investigated in Ref. 50). All the lattices have the same porosity of 0.75.
Young’s modulus vs. thermal conductivity for the lattice trusses with the same porosity of 0.75.50)
A discrete element method (DEM) simulation is a powerful tool for predicting powder packing behavior by numerically tracking the movement of each particle within a population of independent particles. This involves a series of calculations that account for individual particle interactions and geometrical properties, such as particle shape, size, and density. DEM simulations can provide valuable insights into powder behavior by accurately reproducing powder interactions and surface features.57,58,60) Kim et al.57) combined DEM simulation and ML techniques to optimize the densification of amorphous powder using three types of powders with different sizes, as shown in Fig. 3. Because DEM simulation incorporates the cohesive and van der Waals forces, it predicts the powder packing behavior more accurately than the analytical model (i.e., the Desmond model). Finally, a packing fraction of 94.14% with an R-squared value for the fit of 0.96 was achieved by combining the DEM simulation and an ML optimization.
Simulation result images of (a) the highest packing fraction condition, (b) a relatively low packing fraction, and (c) a magnified image of the marked segment in (b).57)
Large-scale atomistic simulations, such as MD and Monte Carlo simulations, may predict the behavior of many atoms at a time by solving the equations of motion governing their behavior under various conditions, such as changes in temperature, pressure, and other external factors.17,18,33–36,61–65) Although large-scale atomistic simulations enable the prediction of various properties of materials at the atomic and molecular scales, they are limited by the absence of proper (semi-)empirical interatomic potentials. Jang et al.55) developed an interatomic potential for the Mg–Mn binary system based on the second nearest-neighbor-modified embedded-atom method formalism. The Mg–Mn potential effectively replicates the structural, elastic, and thermodynamic properties of the compound and solution phases of its associated alloy system, consistent with experimental data and first-principles calculations. The developed potential could investigate the dislocation behavior of the Mg–Mn alloy and elucidate that Mn affects the formability and strength of Mg alloys.
A phase-field model (PMF) has recently attracted considerable attention for describing the microstructural changes during various processes by integrating a set of partial differential equations for the whole system, avoiding the explicit treatment of the boundary conditions at the interface.6–8,24,25,52,59) A multiphase PFM model was developed by Lee et al.52) to concurrently consider spinodal decomposition or nucleation/growth in the Fe–Cr system. This modeling describes the phase separation behavior of Fe alloys containing 40 and 55 at% Cr at 700 and 1000 K, respectively, as displayed in Fig. 4. The PFM results were consistent with the experimentally derived phase diagram of Fe–Cr.
The morphological variation for precipitate with different Cr content and temperature.52)
Park et al.54) employed a multiscale modeling technique by coupling mesoscale modeling, such as the crystal plasticity–FEM (CP–FEM) and discontinuous dynamic recrystallization (DDRX)-based cellular automata model, to describe the DDRX behavior in 304LN stainless steel. The three orientation selection schemes of the nucleus (i.e., random orientation (Case R), inheritance of the orientation of parent-deformed grain (Case I), and generalized strain energy release maximization theory (Case G)) are particularly exploited in the simulation. The DDRX characteristics predicted by the three schemes are compared with the experimental observations. Figure 5 shows the comparison of DDRX textures with the experimental results for Cases R, I, and G at a strain of 0.7 at 0.01/s and 1000°C. The DDRX characteristics, including flow stress, DDRX volume fraction, and grain size evolution, were all different from each other because of differences in deformation behavior and grain growth owing to the orientation given to a nucleus, and the random orientation (Case R) was found to be the most reasonable.
Comparison of DDRX textures at a strain of 0.7 under 1000°C and 0.01/s between (a) experimental results, (b) Case R, (c) Case I, and (d) Case G.54)
Recently, many researchers have been using AI to discover new materials or optimize the composition and process conditions to achieve desirable performances, from DoE43–47) to ML.36,38,56) Lee et al.47) optimized the parameters for synthesizing nickel phosphide as a catalyst using the Taguchi method in the DoE and established correlations between the modeling results and those of the electrochemical reactions. The Taguchi method finds the optimal point with minimum experiments43,44,47) and can solve various distributions such as experimental equipment and worker proficiency. Using this method, the authors analyzed the magnitude of the influence of input parameters on electrochemical properties, which enables robust synthesis process design and optimization to reduce cost and time in the future.
In addition, Jeon et al. constructed an ML model to predict the transformation temperature from γ to γ + θ phases (Acm) in low-alloy steels.56) Understanding these phases is key for controlling the overall properties of steel; therefore, predicting phase transformation temperatures is very important. Although an analytical equation can calculate the temperature,66) several experimental data sets are required to accurately predict the phase transformation temperature. Hence, the authors applied data analysis, hyperparameter adjustment, ML training, comparison of ML with the existing equation, and the Shapley additive explanation (SHAP) method to understand the prediction mechanisms of the Acm temperature. Thus, the artificial neural network model developed by the authors was more accurate than the empirical equation in predicting the Acm temperature. Furthermore, the authors found that the C content exhibited the highest effect on the Acm temperature, followed by Cr, Mn, Si, Ni, Cu, and Mo.
Herein, the papers published in the ICAPE special issue of Materials Transactions (Vol. 63, No. 10) are presented. Computer-aided engineering for materials and processes was found to hold significant potential for advancing materials science and engineering and fostering innovation across various industries. With the increasing availability of materials data and computational power, material modeling and ML/AI techniques can be successfully applied in materials science and engineering by understanding PSP relations.
This study was supported by the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (No. NRF-2021M3H4A6A01045764, 2022M3H4A6A01037255, and 2020M3H4A3106736).
