Ferrographic Wear Particle Identification of Marine Lubricating Oil Based on YOLO v5s and Attention Mechanism

Due to the friction side interaction of mechanical equipment, different types of wear can produce different wear particles, which contains a large numbers of important state information about the operation of the mechanical device. In this paper, the image of worn particles obtained from the detection of lubricating oils of mechanical equipment such as ship tail bearings and gearboxes are the object of the introduction of the Coordinate Attention (CA) mechanism with lightweight characteristics in the front of the SPPF of the YOLO v5s main roof network, which forms a YOLO V5s-CA model. Built with YOLO v5 detection software, the detection of grain images obtained by iron spectrum technology. Meanwhile, a variety of evaluation indicators and methods are used. The results showed that during the training process, the model's precision is 0.97, mAP is 0.99 and the convergence step is 32. In wear particle recognition, YOLO v5s-CA is superior to YOLO V5s in terms of reliability and leak detection. YOLO v5s-CA provides a new method for more accurate wear particle identification, which can accurately reverse the reasoning of the type of friction wear occurring in mechanical equipment, providing new methods for ship lubrication oil online monitoring and equipment failure diagnosis.