CNC Interpolator Parameter Optimization using Deep Learning

Abstract

A CNC parameter optimization approach is presented to predict machining quality based on deep learning. The approach aims to optimize tracking error, contouring error, and cycle time simultaneously. CNC interpolator parameters including the limit of velocity, acceleration, jerk and corner tolerance are regarded as experimental factors. The standard test toolpath KANINO is adopted to collect signals of motion axes in various combinations of interpolation parameters. The back propagation neural network (BPNN) is utilized to establish the predicted model between the interpolation parameters and machining performance index. The parameter combination is optimized by the trained BPNN model with the non-dominated sorting genetic algorithm II (NSGA II). Finally, experimental validations are provided to demonstrate effectiveness of the proposed method in improvement of machining quality.