Nonlinear Extended-state ARX-Laguerre PI Observer Fault Diagnosis of Bearings

Abstract

This paper proposes an extended-state ARX-Laguerre proportional integral observer (PIO) for fault detection and diagnosis (FDD) in bearings. The proposed FDD technique improves fault estimation using a nonlinear function while generating a robust residual signal using the sliding mode technique, which can indirectly improve the performance of FDD. Experimental results indicate that the system modeling error in a healthy condition is less than 2.5 × 10-10 N.m. In the next step, the ARX-Laguerre PIO is designed to define the state and output of the system observer. The high gain extended-state observer is designed in the third step to estimate the mechanical (bearing) faults based on the nonlinear function. In the last step, robust residual signals are generated based on the sliding mode algorithm for accurate fault identification. This approach improves the performance of an ARX-Laguerre linear PIO method. Employing the proposed method, we demonstrate that in the presence of uncertainties and disturbances, the ball, inner, outer, inner-ball, outer-ball, inner-outer, and inner-outer-ball failures with various motor torque speeds (300 RPM, 400 RPM, 450 RPM, and 500 RPM) and crack sizes (3 mm and 6 mm) are detected, identified, and estimated efficiently. The effectiveness of the proposed technique is compared with an ARX-Laguerre proportional integral observation (ALPIO). Experimental results indicate that the proposed technique outperforms the ALPIO technique, yielding 17.82% and 16.625% performance improvements for crack sizes of 3 mm and 6 mm, respectively.

Keywords

• fault diagnosis
• extended-state observation technique
• proportional integral observer
• ARX-Laguerre algorithm
• bearings
• sliding mode algorithm