Prediction of the damage threshold of CFRP irradiated by millisecond laser based on back propagation neural network

Abstract Determining the damage threshold is fundamental to conducting laser processing of carbon fiber reinforced polymer (CFRP). However, traditional methods for acquiring this threshold are costly and experimentally challenging. Therefore, this study proposes a back-propagation neural network (BPNN)-based prediction model for estimating the damage threshold of CFRP under millisecond laser irradiation. The model is trained on a dataset derived from a multi-angle joint measurement method, using laser pulse length and target material thickness as input features, and the corresponding damage threshold as the output. By improving the traditional BPNN structure, the model’s predictive accuracy was significantly enhanced. The mean absolute error (MAE), root mean square error (RMSE), and R 2 value between predicted and experimental results were 4.08%, 5.08%, and 95.32%, respectively. The strong agreement between experimental and predicted values demonstrates the high reliability of the improved BPNN model in predicting the damage threshold of CFRP under millisecond pulsed laser irradiation.