基于多小波和PSO-RBF神经网络的水电机组振动故障诊断

发布时间：2018-03-21 18:23

本文选题：水电机组　切入点：故障诊断　出处：《西北农林科技大学学报(自然科学版)》2017年02期 　论文类型：期刊论文

【摘要】：【目的】研究水电机组振动故障诊断的方法,为水电机组状态监测提供一种新的信号处理方法。【方法】对水电机组的振动信号进行多小波变换,提取振动信号的特征向量,将此特征向量作为学习样本输入到经过粒子群优化的径向基神经网络,通过训练后建立频谱特征向量和故障类型的映射关系,然后以测试样本和多故障测试样本为例进行应用检验。【结果】优化后的神经网络在第30次迭代时就达到了目标值,而优化前则需要56次迭代才能达到目标值。测试样本的诊断结果和测试样本的多故障诊断结果显示,期望输出与实际输出基本一致,故障识别的正确率达到100%。【结论】多小波-能量和经过粒子群优化的RBF神经网络结合的方法适用于水电机组的振动故障诊断,其诊断精度高,具有工程应用价值。
[Abstract]:[purpose] to study the method of vibration fault diagnosis of hydropower unit, and to provide a new signal processing method for condition monitoring of hydropower unit. [method] the vibration signal of hydropower unit is transformed by multi-wavelet transform, and the characteristic vector of vibration signal is extracted. The feature vector is input into the radial basis function neural network which is optimized by particle swarm optimization as a learning sample, and the mapping relationship between the spectrum eigenvector and the fault type is established after training. Then take test samples and multi-fault test samples for application test. [results] the optimized neural network reached the target value at the 30th iteration. Before optimization, 56 iterations are required to reach the target value. The diagnostic results of the test samples and the multi-fault diagnosis results of the test samples show that the expected output is basically the same as the actual output. [conclusion] the method of combining multi-wavelet energy with RBF neural network based on particle swarm optimization is suitable for vibration fault diagnosis of hydroelectric generating sets, which has high diagnostic accuracy and engineering application value.
【作者单位】：西安理工大学水利水电学院;中国电建集团西北勘测设计研究院有限公司;
【基金】：国家自然科学基金项目(51209172,51279161) 陕西省自然科学基础研究计划项目(2010JK730)
【分类号】：TV738;TP183

【相似文献】