Rate of change of torque for gear tooth damage detection

This paper presents an investigation into the applicability of a magnetoelastic rate of change of torque (ROC) sensor to condition monitoring of gearbox tooth damage, using a recirculating torque test gearbox test facility. ROC sensors have significant potential in condition monitoring of rotating machinery, due to direct signal paths between the sensors and the damaged region. To assess this, data has been collected across a wide range of operating conditions for tooth bends as small as 2.4 µm. A number of industry-standard analysis techniques have then been applied. This includes a variety of algorithms (based around time-synchronous averaging), wavelet transforms, neural network classification, shapely additive explanations (SHAP) and deep learning. It is concluded that the ROC sensor demonstrates significant potential in condition monitoring of rotating machinery, worthy of further study.