Bayesian Mesh Optimization for Graph Neural Networks to Enhance Engineering Performance Prediction

Abstract

Abstract In the field of engineering design, surrogate models for 3D computer-aided design (CAD) have been widely used to replace computationally expensive simulations. However, the conventional surrogate modeling process, which relies on the geometric parameters (or design variables) of CAD, has limitations when dealing with complex structural shapes commonly found in industry datasets. These limitations include information loss in low dimensions and difficulty in parametrization. This study proposes a Bayesian graph neural network (GNN) framework for a 3D deep-learning-based surrogate model that predicts engineering performance by directly learning the geometric features of CAD with mesh representation. Our proposed framework derives the optimal size of the mesh elements, creating a high-accuracy surrogate model with Bayesian optimization. It also solves the heterogeneity problem of 3D CAD data in that 2D images have regular pixel structures, whereas 3D CADs have irregular structures. From the experimental results, the mesh quality is highly correlated with the prediction accuracy of the surrogate model, and there exists an optimal mesh size that satisfies the high-performance requirements of the surrogate model. We expect that our proposed framework has the potential to be applied to mesh-based simulations in various engineering fields, reflecting the physics-based information widely used in computer-aided engineering.