In-situ fault detection for the spindle motor of CNC machines via multi-stage residual fusion convolution neural networks

Abstract

The faults from the spindle motor of CNC machines result in excessive vibration affecting the manufacturing quality. In-situ signals of intact motors are complex and nonlinear due to coupling of multiple subsystems. The manufacturing and assembly errors lead to individual differences resulting in more challenging fault detection of motor systems compared with detection of disassembly parts in an experimental environment. In this paper, an in-situ fault diagnosis method via the multi-stage residual fusion convolution neural networks (MSRFCNN) model is proposed. The MSRFCNN with excellent generalization is specially designed for the cross-individual detection of the motor systems, which can extract comprehensive individual-irrelevant features from in-situ multi-channel signals with industrial noise. Integrated experiments are performed on real industrial motor datasets for assessing and analyzing the effectiveness of the proposed method. The experiment results show that the MSRFCNN is superior to classical and some state-of-the-art DL methods.