Innovation Series: Advanced Science (ISSN 2938-9933, CNKI Indexed)

Abrasive Wear Characteristics and Failure Diagnosis Method of Mechanical Seal Faces

Jing Wang^1,*, Zhonghua Zhou²

¹ Chongqing Technology and Business Institute, Chongqing Open University, Chongqing 401520, China

² China Oilfield Services Limited, Sanhe 065200, Hebei, China

Corresponding Author: Jing Wang (1065560315@qq.com)

Funding: Scientific and Technological Research Project of Chongqing Municipal Education Commission (No. KJQN202204018): Research on Sealing Mechanism of Mechanical Equipment Lip Seal

Abstract: A failure diagnosis method based on multi-source input fusion was proposed to solve the problems of increased leakage, abnormal friction temperature rise and failure discrimination lag caused by abrasive wear of mechanical seal end face. The feature vectors of scratch density, wear area proportion, texture contrast, friction torque, temperature rise and vibration are constructed by taking the end face morphology image and the running state signal as input. The gated fusion module is used to complete the unified expression of image features and signal features, and the weighted cross-entropy loss is used to optimize the classification boundary of different wear levels. In the experiment, 1200 groups of effective samples were collected and divided into training set, validation set and test set according to 840/180/180. The test results show that the diagnostic accuracy of the proposed model reaches 95.3%, the F1-score reaches 94.6%, and the average inference time of a single sample is 12 ms, which is 3.9 percentage points higher than that of the CNN model, and it can realize the stable recognition of the end surface wear state.

Keywords: Mechanical seal; Abrasive wear; Feature fusion; Failure diagnosis

References