Anomaly Detection System for Stepper Motors

Abstract

Predictive maintenance (PdM) systems have the potential to autonomously detect underlying motor issues at early stages. Although many such systems have been proposed up to data, they have yet to be implemented. Most of these methods, which are based on supervised learning, require hours of manual data collection and annotation. Furthermore, they are mostly made to tackle a single instead of the multiple motor issues that may occur and are unable to adapt to varying motor speed and load conditions. Thus, they are not viable for industrial implementation. Therefore, this paper presents an unsupervised LSTM autoencoder-based anomaly detection system for electric motors. It analyzes the vibration and current consumption data from motors to detect anomalies, which is sufficient to account for the various motor defects. The system comes with a variety of features that allows users to autonomously collect data, train models and deploy models. In addition to that, users can remotely keep track of the motor’s conditions. To test the system, a hardware test bench using a stepper motor is made to simulate defective conditions. The LSTM Autoencoder-based anomaly detection system is described step-by-step in this paper.