Anomaly detection using Convolution Neural Network during shaping processing

Abstract

Currently, detection of anomalies in cutting tools is often performed by an experienced person by checking sound or visually. From the viewpoint of improving manufacturing efficiency, automatic detection of anomalies such as tool wear and damage during machining, without human intervention, is recognized as an important technological issue. For this purpose, machine learning using features obtained from acceleration sensors, microphones, AE sensors, etc., has been studied for tool condition monitoring. In this paper, we report on the use of a Convolution Neural Network, a type of machine learning, for the purpose of detecting anomalies in the machining process of a seperator. We then investigated whether the CNN could classify normal chips from anomalous chips by using the frequency bin spectrum calculated from the measured time history waveforms as input.