Physics-informed neural networks for predicting the surface temperature of carbon fiber reinforced polymers under laser irradiation.

Laser technology is indispensable in the processing and inspection of materials, including carbon fiber reinforced polymers (CFRP). Upon exposure to laser irradiation, CFRP undergoes complex physical and chemical transformations, with temperature increases potentially compromising material performance. Therefore, precise temperature control is essential for optimizing the quality and efficiency of laser material processing. Traditional numerical modeling of temperature rise requires extensive computational resources, whereas purely data-driven approaches necessitate large-scale datasets. Physics-informed neural networks (PINNs) provide a hybrid approach that integrates physical laws with data-driven characteristics, offering significant advantages in computational efficiency and addressing challenges associated with limited sample sizes. In this work, we developed a PINNs-based method to predict the surface temperature of CFRP under laser irradiation. The model incorporates the principles of heat transfer directly into its loss function. We validated this model using comprehensive datasets derived from both experimental measurements and simulations, which included temperature profiles of CFRP subjected to various laser irradiation conditions. The results demonstrate that our proposed method enables highly accurate temperature predictions, thereby enhancing temperature regulation during laser processing and ultimately improving the overall performance of CFRP products. We believe that our work provides a versatile method for improving the effectiveness of accurate laser processing by applying the PINN method.