Signal restoration for predictive maintenance of high-speed train bearings
The predictive maintenance project for high-speed train bearings won the Yann LeCun Prize for Best Innovation at the TechnInnov 2023 event. ( https://ai-africa.ecc-emines.ma/.). This initiative aims to improve the detection and prevention of bearing failures on high-speed trains using non-invasive methods, thereby reducing the costs associated with bearing failures.
This project was carried out by Hermann Agossou during his internship at SIANA, under the supervision of Mohammed El Rhabi (ECC) and Mohamed Sedki (SIANA).
Bearings are essential components for transmitting rotary motion from the axle to the wheels, but they can fail due to faults, resulting in significant costs. Our method is based on the analysis of vibratory or acoustic signals to identify bearing faults.
The method comprises two main stages: denoising of the observed signal and the simultaneous BSS (Blind Source Separation) de-noising procedure to extract the relevant information. Numerical results obtained from sample observations mixing two non-Gaussian and independent real sources demonstrate the effectiveness of our approach in terms of fault detection.