Fault detection in traction motor systems using machine learning an‎d audio signals

10/22/2026 12:00:00 AM

In recent years, rail systems have gained increasing importance due to their high capacity an‎d remarkable speed in transportation . Traction motors, as the core of these systems, play a key role in their optimal performance an‎d any malfunction in them can lead to costly stops an‎d unfortunate incidents . Early detection of bearing failures, as one of the critical components of traction motors, is essential to prevent unexpected stops an‎d reduce the downtime of industrial equipment . Analysis of acoustic signals an‎d their spectral characteristics provides a noninvasive an‎d effective method for identifying bearing defects . So far, numerous condition monitoring methods using acoustic an‎d vibration sensors have been developed, among which human hearing has often been used as a primary tool for detecting abnormal bearing noise . In this study, first, acoustic signals of traction motors in healthy an‎d faulty conditions ( with bearing failure ) were collected . Then, using geometric dimensions, frequencies related to bearing failure were extracted an‎d validated through frequency domain analysis . In the next step, using signal processing methods, key acoustic features were extracted an‎d using classical machine learning algorithms, an intelligent model was developed to classify an‎d detect types of bearing failure . Also, a new approach based on extracting time-frequency domain images from acoustic signals an‎d using pre-trained image classification models was proposed . The results of this study showed that the proposed model is able to identify types of bearing failures with high accuracy solely through the analysis of acoustic signals . This method, as an efficient, non-invasive an‎d cost-effective tool, can contribute significantly to improving safety an‎d reducing maintenance costs of traction motors .

تشخيص خطا , يادگيري ماشين

Fault detection , Machine learning

Erfan Madadi Mahani

Ahmmad Mirabadi