Design of an Intelligent Grinding Parameter Selection Assistance System

Abstract

In this study, an intelligent grinding parameter selection assistance system (IGPSAS) that can be used by operators for grinding was designed. In the data collection stage, an ESG-1020 surface grinder and aluminum were used for grinding experiments. The proposed IGPSAS consists of two parts: A Taguchi-based convolutional neural network (TCNN) and a differential evolution algorithm. First, the proposed TCNN was used to establish a surface roughness prediction model. Then, the proposed differential evolution algorithm was used to determine the best processing parameters. To achieve better surface smoothness prediction capabilities in the CNN model, the Taguchi method was used to optimize the parameters of the network model architecture. The effect of each factor was analyzed, and a network with stable parameters was selected for machine processing. The performance of the proposed TCNN was verified experimentally. The mean average percentage error (MAPE) of the proposed TCNN's surface roughness prediction in the measurement of a NewView 8300 optical surface profile was 15.65%. In addition, the differential evolution algorithm was used to select the best processing parameters and perform actual processing. The MAPE of the surface roughness prediction of the proposed IGPSAS was experimentally determined to be 10.97%, demonstrating that the system effectively provides the user with the ability to operate the machine with the parameters set according to the desired processing quality.