Self-validating Pneumatic Actuator Fault Diagnosis Based on Relevance Vector Machine

Abstract

In order to solve the fault diagnosis problem of self-validating (SEVA) pneumatic actuator, an actuator fault diagnosis approach based on relevance vector machine (RVM) regression modeling and relevance vector machine (RVM) multi-classifier is proposed. The RVM regression is used to establish the normal models of the SEVA pneumatic actuator. The residuals generated by comparing the output of the models and the actual SEVA actuator are used as the nonlinear features. Then, the structure of the RVM for multi-classification is designed using k-meaning clustering methods, which is used as fault classifier to identify the condition and fault pattern of the SEVA actuator. The proposed approach is verified using fault data generated by DABLib model and actuator data from Lublin Sugar Factory and compared with support vector machine (SVM) fault diagnosis approach. The results indicate that the proposed approach overcomes the drawbacks of SVM and resolves the small sample and nonlinear problem in SEVA pneumatic actuator fault diagnosis.