Fault Detection and Diagnosis of Gear Train Based on Motor Current Signature Analysis

Abstract

Abstract Gear fault diagnosis generally uses oscillation signals to extract fault features, but the oscillation identify technique is inconvenient to set up sensors and is easily affected by the surrounding and noise. The motor itself has the property of a sensor, and signals such as motor current is able to reflect the change of the load torque. This paper proposes a gear fault detection approach, which applies the Motor Current Signature Analysis (MCSA) to exploit spectral characteristics to recognize malfunction in servo motor drive gears. First, we calculate the nominal revolutions per minute (rpm) to explore frequencies of interest. Second, frequency bands where faulty signals would be appear are constructed. Next, we extract power spectral density (PSD) feature data. Finally, the paper computes spectral metrics for the frequency band of interest. By using two spectral metrics, gear health and failure data are clearly grouped in different areas of the scatter chart. The experimental results give evidence of the effectiveness of analyzing current characteristics of servo motors to classify fault and health data.