Optimization of processing parameters for waterjet-guided laser machining of SiC/SiC composites

Interactions between light and matter during short-pulse water-jet guided laser materials processing are highly nonlinear, and acutely sensitive to laser machining parameters. Traditionally, the physical simulation calculation methods based on laser, water and composite materials are complicated. This work combines neural networks and physical simulation models in the understanding of laser drilling of composite materials. Neural networks are used to predict SiC/SiC composites laser drilling results by using processing parameters (average power, scanning speed, and filling spacing) as input parameters, optimal combinations of processing parameters based on the neural network are identified, and the effectiveness of the learned knowledge is validated using a physical simulation model. The results show that the neural network can identify the nonlinear effect of processing parameters on machining quality with the MAE of 0.054 and the RMSE of 0.067. The physical simulation model could explain why this nonlinear effect exists. This method can be applied to a wide range of fields. In the face of unknown material and physical processing processes, the approach of combining neural networks and physical simulation models has the potential to significantly reduce the optimization time and deepen the understanding of laser processing.