基于ADE-WNN的水电机组振动故障诊断方法

摘要

There is a complex non-linear relationship between vibration characteristics and fault types of Hydro-generation. A novel vibration fault diagnosis method was proposed in this paper, which combined the wavelet neural network (WNN) and the adaptive differential evolution method (ADE). The proposed method combined characteristics of evolutionary computing with swarm intelligence. It could adaptively adjust crossover probability factors and scaling factors according to the status of the individual. ADE algorithm sped up the training speed of WNN and improved the network training precision. Experimental results showed that the proposed method had higher accuracy and faster speed of diagnosis compared with the traditional methods based on BP neural network and wavelet neural network.%水电机组振动特征和故障类型之间存在复杂的非线性关系,结合小波神经网络和自适应差分进化法,提出一种新型水电机组振动故障诊断方法.该算法具有进化计算和群体智能的特点,能够根据个体的状态自适应调节交叉概率因子和缩放因子;自适应差分进化算法应用于小波神经网络的参数搜索中,加快了小波神经网络的训练速度,提高了网络训练精度.实验结果表明:该方法比传统的基于BP神经网络和小波神经网络的故障诊断方法,具有更高的准确度和更快的诊断速度.