基于小波网络的水下机器人推进器故障诊断  

Thruster Fault Diagnosis of Autonomous Underwater Vehicle Based on Least Disturbance Wavelet Neural Network

作　　者：庞永杰[1] 方少吉[1] 王丽容[1] 

机构地区：[1]哈尔滨工程大学船舶工程学院,黑龙江哈尔滨150001

出　　处：《中国造船》2008年第2期94-100,共7页Shipbuilding of China

摘　　要：针对小波神经网络的隐层小波函数可以调节伸缩因子与平移因子影响网络输出的特点,将其伸缩因子与平移因子引入到最小扰动动态学习率的学习算法中。此算法通过计算动态学习率,使得小波函数的伸缩因子与平移因子以及网络连接权值的变化最小,这样便可提高小波神经网络的稳定性和收敛速度。使用这种小波神经网络对机器人建模,通过比较模型的输出(运动状态估计值)与实际测量值可得到残差,并分析残差提取故障判断准则,从而进行推进器故障诊断。仿真试验验证了该方法的有效性。Based on the character that hidden layer wavelet function of wavelet networks can adjust scale factor and shift factor to affect output of networks, the least disturbance algorithm adding scale factor and shift factor is proposed and it can calculate the dynamical learning ratio to improve the stability and convergence of wavelet networks. This algorithm is applied to build the dynamical model of autonomous underwater vehicle （AUV） and via comparing the output of neural network with the real state value in the condition of thrust fault, fault detection rules are distilled to execute thruster fault diagnosis. The result of simulation proves the effectiveness of this algorithm.

关 键 词：船舶 船舶工程 水下机器人 故障诊断 小波网络 最小扰动 

分 类 号：U674.941[交通运输工程—船舶及航道工程] TP206.3[交通运输工程—船舶与海洋工程]