SVM-Based Hybrid Robust PIO Fault Diagnosis for Bearing

Abstract

Inner, outer, and ball faults are complex non-stationary and nonlinear faults that occurs in rotating machinery such as bearings. Designing an effective procedure for fault diagnosis (FD) is essential to safe operation of bearings. To address fault diagnosis issue, a robust, hybrid technique based on the ARX-Laguerre fuzzy-sliding proportional integral observer (ALFSPIO) for rolling element bearing (REB) is presented. The main important challenges in the ARX-Laguerre PI observer are robustness, and estimation accuracy. To address the robustness issue, sliding observation technique is introduced. Moreover, to increase the signal estimation accuracy, the fuzzy technique is used in parallel with ARX-Laguerre sliding PIO. Furthermore, using the ALFSPIO, the residual energy signals showed more differentiable for fault diagnosis. Beyond the above, the support vector machine (SVM) is used to fault detection and classification. The vibration dataset of Case Western Reverse University (CWRU) is used to validate the effectiveness of the proposed algorithm.