Sensor Fault Diagnosis and Fault-Tolerant Control Base on Two-level Neural Network

This paper presents a structure of two-level neural network which consists of a main neural network and some sub neural networks. The main neural network is used to detect sensor faults, and the sub neural networks are used to locate the fault sensors and recover their signals. Based on the analytical redundancy technology, the sensors in the control system of underwater vehicle and their analytical relations are analyzed, and the active fault-tolerant approach of the two-level neural network for the control system is investigated and simulated. The simulation results show that the presented approach is effective in sensors fault detection and active fault-tolerant control.