Time Series Prediction of Machine Tool Sensor Data with Deep Echo State Network for Anomaly Detection

Abstract

In this presentation, we present Deep Echo State Network （DeepESN）, which is suitable for time series prediction of sensor data during machining of machine tools. The recursive structure is built into each reservoir layer of DeepESN to improve the prediction accuracy of time series. In the experiment, we used Sunspot and MackeyGrass instead of machine tool sensor data, and compared the prediction accuracy of DeepESN with recursion structure, DeepESN without recursion structure, and LSTM.