Machine learning-based multi-objective optimization of thermo-mechanical field of anisotropic plates

Designing plates is highly challenging due to the complex relationship between material layout and its properties. Traditional experimental, analytical, and computational methods suffer from problems such as long computation time and high cost, limiting their application in predicting and optimizing plates. In this paper, we propose an end-to-end prediction and multi-objective optimization framework integrating machine learning (ML) and non-dominated sorting genetic algorithm (NSGA-II). This framework can have high fidelity and simultaneously predict multiple thermo-mechanical fields of the anisotropic plate-heat source system. It can also perform multi-objective optimization for specific requirements by optimizing the distribution of materials and heat sources. To demonstrate the effectiveness of our approach, the maximum temperature and von Mises stress are considered as objectives, and the framework is applied to two different discrete multi-objective optimization problems, i.e., non-temperature constrained and temperature constrained optimization problem. By integrating ML and optimization algorithm, our framework offers a comprehensive solution for tackling complex optimization problems in the field of anisotropic plate-heat source system. © 2024 Elsevier Ltd.