Prediction and optimization of global temperature field of composite materials under multiple heat sources

The property measurement and structure optimization of composite materials are difficult topics due to the diversity of combinations of composite constituents and complexity of their application environments. The distribution of composite constituents and heat sources can cause heat aggregation phenomenon which may lead to failure of materials. In this paper, we first propose a deep learning-based surrogate model (DLBSM) which can quickly and accurately achieve the end-to-end prediction between the layout of composite materials under multiple heat sources and its temperature field. The prediction result depicts that the coefficient of determination for the maximum and average temperature of all cases exceeds 0.996. Then, the layout optimization is transformed into a combinatorial optimization problem, and the DLBSM is combined with optimization algorithm to optimize the maximum temperature, temperature gradient, and uniformity of the temperature field. The optimized maximum temperature and temperature gradient are significantly reduced, while the temperature uniformity is improved. These enhancements effectively reduce the probability of failure in composites. This approach can significantly improve the efficiency of thermal behavior prediction of composite and its layout optimization compared with finite element method (FEM).