Colloquium: Tadeusz Kaniewski (FPT)

The computational cost of high-fidelity engineering simulations, for example CFD, is prohibitive if the application requires frequent design iterations or even fully fledged optimization. A popular way to cut down the computational costs and enable fast iteration cycles is to use surrogate models that are trained to predict simulation results from historical simulation data. While most traditional methods are parametric, ANNs are able to process geometries directly and are thus agnostic to the parametrization of the geometric models, which makes them appealing when working on multiple design programs. However, ANNs may fail to transfer the learned knowledge when used on new design programs that are significantly different from those used to train the model or when the size of the training dataset is too small. The goal of this project is to increase the reliability of ANNs-based surrogate models on new design programs and on small datasets. One main direction to achieve this goal is to incorporate prior physical knowledge into the learning process. Methods of supplementing the training data with potential flow solution, physical meaningful scaling, and governing-equation-based losses were used. The data set used for this study was based on a real-life-inspired complex 3D parametrized geometry of an automotive HVAC system. The methods were tested in generalization, transfer learning, and single-program inference tasks. Moreover, physics-informed losses were tested in an ill-posed setting as a prediction outlier correction tool. The results show that the addition of prior physical knowledge improves the model performance, especially in the low-data regimes. Preliminary results of applying physics-informed losses as a correction methods provide inconclusive results and require further experimentation in the future. 

Supervisor: Dr. Anh Khoa Doan