Neural network-based state estimation of nonlinear systems. Application to fault detection and isolation (Q1046954)

The objective of the book is to address the problem of state estimation, system identification and observer-based fault detection and isolation for nonlinear systems. The book consists of the following chapters: (1) Introduction; (2) Neural Network-Based State Estimation Schemes; (3) Neural Network-Based System Identification Schemes; (4) An Actuator Fault Detection and Isolation Scheme: Experiments in Robotic Manipulators; (5) A Robust Actuator Gain Fault Detection and Isolation Scheme; (6) A Robust Sensor and Actuator Fault Detection and Estimation Approach. Chapter 1 introduces several important concepts concerning design and implementation of neural network observers. Chapter 2 presents a stable neural network-based observer for general multivariable nonlinear systems. In the first step, a linear-in-parameter neural network is used to approximate the nonlinear dynamics. The proposed neural network observer is extended to a nonlinear-in-parameter neural network structure which can be applied to systems with higher degrees of nonlinearity without any a-priori knowledge about the system dynamics. The learning rule for neural network is a novel approach based on the modified backpropagation algorithm. An e-modification term is added to guarantee robustness of the observer. The stability of the recurrent neural network observer is shown by Lyapunov's direct method. Simulation results for a flexible-joint manipulator are presented to demonstrate the enhanced performance capabilities that can be achieved by utilizing the proposed neural network observers. The linear-in-parameter and nonlinear-in-parameter neural networks introduced in Chapter 2 are applied in Chapter 3 to design a stable identifier. By using Lyapunov's direct method, a novel and a more efficient approach is introduced to demonstrate the stability of the overall system. To show the performance of the proposed algorithm, an experimental set-up consisting of a three-link macro-micro manipulator is considered. The proposed approach is applied to identify the dynamics of the experimental robot. Chapter 4 focuses on an actuator fault detection and identification scheme for nonlinear systems. A state space approach is used in which the states are assumed to be available. Moreover, the actuators are assumed to be faulty with bias faults. The Nonlinear-in-Parameters Neural Network (NLPNN) identifier introduced in Chapter 3 is employed to identify the unknown fault. Chapter 5 addresses the problem of actuator gain fault detection and isolation scheme for nonlinear systems. The proposed Fault Detection and Isolation (FDI) approach employs a neural network-based observer to detect and identify the severity of actuator gain faults in the presence of disturbances and uncertainties in model and sensory measurements. The neural network weights are updated based on a modified dynamic backpropagation scheme. Chapter 6 focuses on a robust actuator and sensor fault detection and isolation scheme for a general nonlinear system using a neural network-based observer. The nonlinear system is subject to state and sensor uncertainties and disturbances. Two recurrent neural networks are employed to identify general unknown actuator and sensor faults. The neural network weights are updated based on a modified backpropagation scheme. Unlike many previous methods presented in the literature, the proposed FDI scheme does not rely on the availability of all the state measurements. The effectiveness of the proposed fault diagnosis strategy is demonstrated through extensive simulation studies.