Prediction of high-speed grinding temperature of titanium matrix composites using BP neural network based on PSO algorithm

High grinding temperature is the key reason of workpiece burnout, which hinders the improvement of the machining quality. In this work, the prediction of high-speed grinding temperature of titanium matrix composites is investigated using back propagation (BP) neural network based on particle swarm optimization (PSO) algorithm (also called as PSO-BP). Furthermore, a comparison has been carried out among GD-BP (the BP neural network trained with gradient descent method), LM-BP (the BP neural network trained with Levenberg-Marquardt (LM) algorithm), and PSO-BP. Results obtained show that the PSO-BP method has a more significant advantage in terms of convergence speed, fitting accuracy, and prediction accuracy than the other two methods (such as GD-BP and LM-BP) in predicting the grinding temperature. Accordingly, the grinding temperature is predicted by applying the PSO-BP method and the grinding parameters are optimized, which could avoid the burnout behavior of the titanium matrix composites.