Tool Wear Recognition Based on Deep Kernel Autoencoder With Multichannel Signals Fusion

Abstract

Intelligent tool wear recognition techniques have great significances to guide the cutting process in automated manufacturing systems. Traditional methods heavily rely on human experience and fail to consider the impact on processing. In this article, a novel deep kernel autoencoder (DKAE) feature learning method optimized by the gray wolf optimizer (GWO) is proposed for tool wear state intelligent recognition. The Gaussian kernel function is used to construct the loss function of the neural networks to enhance the feature learning ability. In addition, the current sensors are used to collect the multichannel spindle motor current signals in three directions of X-, Y-, and Z-axes. Compressed sensing technology is adopted to fuse and reduce the dimension of massive multichannel current signals into a single sample signal. A series of experiments with numerical control machines in the real manufacturing process of the impeller is run to test the superiority of tool wear recognition with this method. The results indicate that this work can be applied for real-time tool wear monitoring and greatly improved recognition accuracy.