Performance Prediction of Abrasive Water Jet Machining Composite Laminates Using Artificial Neural Networks and Regression Analysis Method

The machining of composite materials possesses several challenges due to their thermal sensitivity and highly heterogeneous nature. The abrasive water jet (AWJ) has proven to be a suitable process to overcome this problem. Nevertheless, there are limitations associated with AWJ cutting like the development of a kerf taper angle, delamination, and surface defects. Therefore, an optimal selection of process parameters remains important to achieve maximum productivity and good cutting quality. In this context, the present study reports an experimental investigation to model the kerf taper angle θ and surface roughness Ra of E glass/Vinylester 411 resin laminates machined with AWJ. A full factorial experimental design is carried out by varying the water pressure, traverse speed, and standoff distance. Furthermore, artificial neural networks (ANN) and regression analysis (RA) approaches are used to model θ and Ra. Multiple regression and neural network-based models are compared using statistical methods. It is found that the proposed models are capable to predict the cutting performances to high desirable accuracy. However, the ANN models produce better results compared to those with multiple regression. Finally, the optimal machining conditions to minimize θ and Ra were identified using a genetic algorithm.