CHANGE DETECTION IN DRILLING PROCESS BASED ON TEMPERATURE NEARBY CUTTING EDGE BY LSTM NEURAL NETWORK

This study considers to efficiently collect "highly value-added" data for in-process anomaly detection of cutting tools in machining processes, and focuses on collection of time-series data of temperature nearby the tool cutting-edge by using a wireless tool holder system composed of an internal temperature measuring device and a wireless transmitter, which is connected with a thermocouple built-in the cutting tool. We then propose a method to detect a change of tool performance based on a recurrent neural network (RNN) with a long short-term memory (LSTM) structure. The capability of the proposed RNN system with LSTM is demonstrated through computational experiments, and demonstrate the time-series data of temperature nearby cutting tool tip is applicable for change detection of cutting tools status.