基于GA-BP神经网络的UUV航向容错控制

为提高UUV航向控制系统的可靠性，本文对传统的BP神经网络利用遗传算法进行了优化设计，并将优化的BP神经网络应用于UUV的航向容错控制中。首先利用改进的二阶灰色预测模型对光纤罗经进行故障诊断，若UUV罗经故障，根据UUV当前的舵角、航速与加速度信息重构航向信息，利用重构的航向数据代替光纤罗经的航向输出，实现对光纤罗经的航向容错控制。本文详细阐述了容错控制方法的实现步骤，并利用海试实验数据进行了仿真验证，仿真结果表明所提出的方法能快速、准确的诊断出光纤罗经的故障，并能较好的实现光纤罗经的航向容错。

Heading Fault-tolerant Control for UUV Based on GA-BP neural network

To improve the reliability of UUV heading control system, BP neural network was improved by Genetic Algorithm and applied in the UUV heading fault-tolerant. The Improved GM(2,1) model is applied to diagnosis the fault of fiber compass. While the fiber compass fault, the improved BP neural network learning model reconstruct heading data based on the current rudder angle, velocity and acceleration message. The heading fault-tolerant of fiber compass realized with the reconstructed heading data instead of the output of fiber compass. The manipulate steps of fault-tolerant control is described in detail, and the method has been simulated with trial data. The results show that the methods proposed can diagnose the faults quickly and accurately and can achieve the fault-tolerant of fiber compass well.