PIO Output Fault Diagnosis by ARX-Laguerre Model Applied to 2 ^nd Order Electrical System

Abstract

The novelty of this work consists in the synthesis of a new structure of proportional-integral observer (PIO) reformulated from the new linear ARX-Laguerre representation with filters on system input and output. This is in order to estimate the unknown outputs presented as faults and to detect the time instant corresponding to the system malfunction. The stability and the convergence properties of the proposed PIO are ensured by using Linear Matrix Inequality. Furthermore an optimal identification of both Laguerre poles is achieved by a genetic algorithm approach where a parametric significant reduction is ensured to guarantee a reduced observer. The performances of the identification approach and the resulting PIO are tested on an experimental 2 nd order electrical system. INDEX TERMS ARX-Laguerre model, genetic algorithm, proportional-integral observer.

Deep Learning-Based Adaptive Neural-Fuzzy Structure Scheme for Bearing Fault Pattern Recognition and Crack Size Identification

“…The combination of a data-driven technique and control algorithm for fault diagnosis were presented in [ 16 ]. In this work, the system was modeled by the ARX–Laguerre technique and the PI observer was recommended for signal estimation.…”

“…Mathematical-based system modeling (such as five degrees of freedom vibration bearing modeling) is reliable but has some drawbacks, such as the lack of complexity and uncertainty related to modeling [ 13 , 14 , 15 ]. Linear-based system identification techniques (such as the combination of autoregressive with external inputs, and autoregressive with external inputs and Laguerre technique) have been used to address the above challenges [ 15 , 16 , 17 , 18 , 19 ]. Artificial intelligence methods (such as various kinds of neural networks and fuzzy logic procedures) have been used for nonlinear-based system modeling.…”

“…Therefore, the SVAL technique is introduced for signal approximation. First, the autoregressive algorithm is used to approximate the normal RAW signal of the bearing [ 16 , 19 , 47 ]. …”

“…To detect anomalies using the model-based technique, modeling is the critical step. System modeling is categorized into two principal groups: (a) mathematical-based system modeling, such as Newton–Euler, Lagrange, and finite element methods, and (b) system identification approaches, such as autoregressive (AR), autoregressive with external input (ARX), ARX–Laguerre techniques, neural network approach, and fuzzy logic methods [ 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ]. To improve the accuracy of signal approximation in nonlinear and nonstationary signals, we propose a combination of autoregressive techniques, namely the Laguerre technique with support vector regression (SVR), which will henceforth be called the support vector autoregressive–Laguerre (SVAL).…”

“…After approximating the vibration signal of the bearing in the normal state, an estimator is designed to solve the signal estimation problems in the state-space modeling technique. Estimators can be categorized into linear (such as proportional integral (PI) [ 16 , 19 ] and proportional multi-integral [ 26 ]) and nonlinear (such as high gain [ 27 ], feedback linearization [ 28 ], neural network [ 29 ], fuzzy logic [ 30 , 31 ], and sliding mode [ 32 , 33 , 34 , 35 , 36 ]) observers. In this work the high order variable structure observer is recommended for pattern recognition and crack size identification.…”