Reverse-Order Multi-Objective Evolution Algorithm for Multi-Objective Observer-Based Fault-Tolerant Control of T-S Fuzzy Systems

Bor-Sen Chen,Min-Yen Lee,Weihai Zhang,Wei-Yu Chen

doi:10.1109/access.2020.3047193

Abstract

In this study, a multi-objective $H_{2}/H_{\infty }$ observer-based fault-tolerant control (FTC) design with reverse-order multi-objective evolution algorithm (MOEA) is proposed to deal with the FTC problem of Takagi-Sugeno (T-S) fuzzy systems. To achieve the optimal robust FTC design for the T-S fuzzy systems under the sensor and actuator faults, as well as external disturbance and measurement noise, the multi-objective $H_{2} /H_{\infty }$ observer-based FTC scheme is proposed to efficiently estimate the system state and the fault signals based on a proposed smoothed fault signal model. Then, multi-objective $H_{2}/H_{\infty }$ FTC performance can be achieved by an estimated state and fault signal feedback scheme to efficiently compensate the effect of fault signals and attenuate the effect of external disturbance. By using the proposed indirect method, the multi-objective $H_{2}/H_{\infty }$ observer-based FTC design problem is transformed into linear matrix inequalities (LMIs)-constrained multi-objective optimization problem (MOP). Besides, to overcome the difficulties in searching large fuzzy parameters of observer-based FTC design for solving the LMIs-constrained MOP, a reverse-order MOEA is proposed to overcome the bottleneck to efficiently solve the MOP for multi-objective $H_{2}/H_{\infty }$ observer-based FTC of T-S fuzzy system by searching feasible objective vectors in the objective space instead of searching fuzzy design parameters in the parametric space. Two practical examples are considered for the performance validation with (i) $H_{2}/H_{\infty }$ observer-based FTC design for the missile guidance system with the actuator and sensor fault signals due to the sudden cheating side-step maneuvering and the hostile jamming interference and (ii) $H_{2}/H_{\infty }$ observer-based FTC design for inverted pendulum system which effected by the constant actuator fault.

Highlights

Along with the development of modern industrial production, due to the fact that the actuator and sensor in the control system become much vulnerable to the fault signal, higher requirements of safety and reliability for the control plant have put forward
Based on the proposed dynamic smoothed signal model, a novel reverse-order multi-objective evolution algorithm (MOEA) was proposed to solve a complex multi-objective observer-based faulttolerant control (FTC) of T-S fuzzy system, which could not be solved by the conventional MOEA due to large fuzzy control parameters and fuzzy observer parameters to be selected by evolution algorithm for multiobjective optimization
Through the proposed indirect suboptimal method, the multi-objective observer-based FTC design problem is reduced to an linear matrix inequalities (LMIs)-constrained multi-objective optimization problem (MOP), which could be efficiently solved by the proposed reverse-order MOEA with the help of LMI TOOLBOX in MATLAB via convex optimization algorithm such as interior point method

Summary

INTRODUCTION

Along with the development of modern industrial production, due to the fact that the actuator and sensor in the control system become much vulnerable to the fault signal, higher requirements of safety and reliability for the control plant have put forward. Since fuzzy controller and observer of complex fuzzy systems consist of a large number of local controllers and observers, respectively, it is almost impossible to employ EA for updating these fuzzy controller and observer parameters to achieve a MOP of complex T-S fuzzy systems, such as the multi-objective H2/H∞ observer-based FTC of T-S fuzzy systems with actuator and sensor faults This is why the conventional MOEAs have not been addressed on the MOP of nonlinear T-S fuzzy systems even they are very powerful in MOPs of other more simple designed systems. (II) Instead of using conventional MOEA to search the design parameters of fuzzy controller gain and observer gain for the multi-objective H2/H∞ observer-based FTC design problem, a reverse-order MOEA algorithm is proposed to directly search the optimal multi-objective vector and the corresponding design variables of controller and observer can be obtained by using MATLAB LMI TOOLBOX. S denotes the set of one-dimensional complex number. col[D] denotes the column space of matrix D

SYSTEM DESCRIPTION

SIMULATION EXAMPLE

CONCLUSION