A Machine Learning Technique for Intelligent Manufacturing: Integrating Wavelet Packet Transform (WPT) Into Neural Network Modeling

With the rapid development of Industry 4.0, intelligent manufacturing has received growing attention in recent years. Consequently, a variety of machine learning techniques are being developed to enable and empower intelligent manufacturing. This paper presents a machine learning technique to support intelligent machining, one of the major operations widely employed in modern manufacturing. This technique integrates wavelet packet transform (WPT) into neural network modeling to predict the machined surface roughness. The inputs of the predictive model developed in the present study include eight variables: the cutting speed, the fed rate, three dominant wavelet packets for three-dimensional cutting forces, and three dominant wavelet packets for three-dimensional cutting vibrations. The output of the model is the dominant wavelet packet for the peak roughness of machined surface (Rp). A total of 45 computer numerically control (CNC) bar turning experiments were conducted on aluminum alloy 2024-T351 with coated carbide inserts TPG 432 KC 8050. Experimental measurements were made to generate 45 sets of experimental data to train and test the model. The results show that the model developed in the present study has good accuracy for predicting peak roughness Rp, with an average mean square error of 7.3% using test data.