A Computer Vision Deep Learning Tool for Automatic Recognition of Bearing Failure Modes

We introduce an object detection model specifically designed to identify failure modes in images of bearing components, including the inner ring, outer ring, and rolling elements. The method effectively detects and pinpoints failure features, subsequently determining the associated failure mode within the image. With images sourced from real-world bearing applications, our model can recognize various ISO-failure modes such as surface-initiated fatigue, abrasive wear, adhesive wear, moisture corrosion, fretting corrosion, current leakage erosion, and indents from particles. The proposed model could be used in an assistive tool where failure modes give insights on how to prolong average future bearing life in an asset and therefore reduce related costs and environmental impacts.