Acoustic Emission Intelligent Prediction of Grinding Surface Roughness Based on BP Neural Network

Abstract

Aiming at the problem of low precision of on-line prediction of grinding surface roughness by back-propagation (BP) neural network, based on the acoustic emission (AE) prediction experiment of surface roughness of nodular cast iron QT700-2, 13 characteristic parameters which fully reflect the characteristics of grinding AE signal are extracted, including the correlation coefficients of four intrinsic mode functions of empirical mode decomposition of grinding AE signal. Genetic algorithm BP neural network GA-BP and particle swarm optimization BP neural network PSO-BP are established to improve the prediction accuracy. At the same time, the larger the entropy weight value, the greater the influence of the AE signal characteristic parameters on the prediction accuracy of the grinding surface roughness BP neural network. After the entropy weight value is obtained and multiplied by the original AE signal characteristic parameters, 13 AE signal characteristic parameters after pretreatment can be obtained to improve the prediction accuracy. Finally, 180 samples from 200 samples of AE prediction experiment of grinding surface roughness were randomly selected as BP neural network training set, and the remaining 20 samples were used as BP neural network test set; the prediction relative error was obtained by repeating the above process for 30 times, and the influence of different order of input data on the prediction accuracy of grinding surface roughness was considered, so as to improve the reliability of the prediction results. The results show that the improved GA-BP neural network has high prediction accuracy and the average relative error of 30 predictions can be controlled within 8.57%.