Method for Automatically Recognizing Various Operation Statuses of Legacy Machines

Abstract

Monitoring the operation status of machine tools using IoT is widely carried out in order to improve productivity. However, at manufacturing sites, a lot of legacy machines, which are old and lack the capability of sending data on their operation status to networks, are still in use because the average useful life of machine tools is more than 20 years. Therefore, we developed method for recognizing the operation status of machine tools using a spindle motor current acquired by a current sensor. Because this current is in proportion to the spindle torque, a conventional method recognises the operation status when the current amplitude is above a threshold value. However, in a high-mix low-volume factory, the threshold value must be reset frequently because a drill and the cutting process change per order so the spindle torque varies. In contrast, we propose an automatic recognition method that can learn the variance in spindle motor current by unsupervised learning and with labelling scheme based on prior knowledge. We applied the proposed method to eight kinds of machine tools in a real factory, and the accuracy rate of the operation status estimation was more than 80% for all the machines. This result shows that all machine tools could be monitored by using this method.