Computational efficiency enhancement of fluid flow simulations using Machine Learning (ML) driven reduced-form models

Computational fluid dynamics (CFD) simulations of fluid flows often involve tradeoff between accuracy and computational efficiency. It is often difficult to obtain simultaneous improvement in both without providing immense computational resources. Machine learning (ML) based reduced-form models offer new opportunities for tackling this tradeoff. An ML model trained by either experimental and/or simulation data has the potential to significantly improve computational efficiency without major sacrifice in accuracy. In this project, we would like to develop an ML-based CFD model. Training and testing datasets will be obtained using CFD simulations performed in an open source CFD code OpenFoam. The project will provide experience at the intersection of CFD and machine learning, an emerging field in mechanical engineering.