Research on a Pattern Recognition Method of Cyclic GMM-FCM Based on Joint Time-Domain Features

Abstract

The safety and reliability of the mechanical system in the industrial process determines the quality of products. Whether the fault can be identified and classified in time is the key to ensure the safe operation of the system and arrange the appropriate maintenance plan to restrain the deterioration of the fault. However, with the rapid development of manufacturing digitization, how to process large amounts of data quickly and accurately is faced with many problems. In this paper, a pattern recognition method of cyclic GMM-FCM (CGF) based on joint time-domain features is proposed. Firstly, the concept of joint time-domain features based on Vold-Kalman filter (VKF) is proposed. It retains the integrity of the signal components and avoids the problem of dimension disaster caused by anomaly detection, which laid a foundation for the accurate classification of sensitive feature sets. Secondly, a pattern recognition method of cyclic GMM-FCM is proposed. It can eliminate global and local outliers in sensitive feature sets and determine the number of FCM categories adaptively. It makes the classification result more reasonable and accurate. Finally, the effectiveness and superiority of the pattern recognition algorithm are verified by the gearbox vibration experiments in various states. The result shows that the method is feasible in engineering practice.

Keywords

• Joint time-domain features
• anomaly detection
• GMM
• FCM
• pattern recognition