Gear Reducer Fault Diagnosis Using Learning Model for Spectral Density of Acoustic Signal

Gear reducer is the main mechanical component used to regulate the speed of electric motors. This study seeks to develop an effective approach for detecting gear tooth defects in a gear reducer. The presented approach uses acoustic signals to detect cracks or wear defects in the gear teeth of the gear reducer. In order to analyze the captured acoustic signals, a feature extraction step using spectral density was developed. Subsequently, the classification step was performed using well-known supervised machine learning models including support vector machine and k-nearest neighbors algorithm. The experiments conducted show good results in the classification of gear reducer faults. The developed approach can be effectively applied for the early fault diagnosis of various transport devices, such as escalators and elevators, by monitoring acoustic signals.