Experimental Study of the Cutting Force During Laser-Assisted Machining of Fused Silica Based on Artificial Neural Network and Response Surface Methodology

Abstract

Laser-assisted machining (LAM) is considered an efficient method for the processing of fused silica. In this study, an analysis model based on artificial neural network (ANN) with Bayesian regularization algorithm (BR) was used to investigate the effects of the machine parameters (rotation speed, feed rate, cutting depth, and pulse duty ratio) on the resultant cutting force during the LAM of fused silica. Its prediction capability was validated experimentally and evaluated quantitatively. The optimal combination of machine parameters corresponding to the minimum resultant cutting force was then studied using the genetic algorithm (GA) coupled with the established ANN model. Moreover, the optimal numerical solution was verified experimentally, and the processing quality under optimal machine parameters was characterized through analyzing the surface morphology and roughness. In addition, the performances of prediction and optimization of ANN model were compared with the model based on response surface methodology (RSM). And the mean absolute error in prediction and the optimal cutting force are reduced by 34.47% and 19.11% respectively, compared to RSM. The results clearly show that the ANN model achieves a better behavior in studying the influence of the machine parameters during the LAM of fused silica.