Towards Machine Learning Applications for Computational Fluid Dynamics Modeling in Chemical Engineering

Abstract

Computational Fluid Dynamics (CFD) simulation of multiphase industrial flows is a significant research concern for studying the performance and efficiency of chemical processes. Within the last years, CFD had growth interest from researchers with the significant increase in computational resources capacities and high-performance calculations. The trend toward focusing on machine learning (ML) techniques to solve complex industrial problems is observed. In contrast, ML has found encouraging and promising applications in that research field by offering a wealth of techniques to extract unreachable knowledge from data that can improve the chemical processes understanding and allows efficient optimization and intensification. This study aims to present the different uses of ML in the chemical processes’ field, particularly the CFD modeling, and highlight the main and variate use of ML for complex geometries design and mesh optimization. We have paid particular attention to the ML trend in turbulence modeling to generate data-driven physical models from CFD simulations at different fidelity levels. Then, the extended application of ML to accelerate CFD calculations, and the development of surrogate models are provided. This work reveals that the application of ML to chemical engineering is a promising way for developing this research area.