基于小波神经网络的船舶冷却水系统的传感器故障诊断

船舶冷却水系统是保证船舶动力装置安全可靠运行的系统之一,是故障诊断系统的重要组成部分,具有十分重要的工程意义.以船舶冷却水系统传感器故障检测问题为目标,提出了基于小波神经网络预测器的传感器故障诊断方法.该方法避免了神经网络结构设计上的盲目性及训练过程易陷入局部最小等问题,同时采用同伦算法优化网络,解决小波神经网络对初始值敏感的问题.采集大连海事大学实验船的传感器数据,以温度传感器为例,对1℃偏差故障、0.03℃/s速率漂移故障及不同方差下的精度等级下降故障进行了仿真,结果表明同伦小波网络的诊断率高达95%,证明了该方法的有效性.

Fault Diagnosis of Ship Cooling Water Sensor System Based on Wavelet Neural Network

Ship cooling water system is to ensure the safety and reliability operation of the marine power plant system.To detect cool- ing water system,a method based on wavelet forecast for sensor fault diagnosis is proposed.WNN method can avoid the problem of a local minimum,and then homotopy network optimization algorithm is used to solve the BP algorithm slower and sensitive to initial values.Using the actual running data from Maritime University experimental ship sensor,it is proved that the method is feasible.