| 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。 |
| 关键词: 水电机组 故障诊断 多小波 神经网络 特征向量提取 |
| DOI： |
| 分类号: |
| 基金项目: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） |
|
| Vibration fault diagnosis for hydropower generating unit based on multiwavelet and PSO-RBF neural network |
|
LI Hui,WANG Yi,YANG Xiaoping,et al
|
| Abstract: |
| 【Objective】This paper proposed a vibration fault diagnosis method for the monitoring of hydropower generating unit operation.【Method】The vibration signals of power unit were transformed by multiwavelet.The feature vectors of the vibration signals were extracted and applied as learning samples for the PSO-RBF neural network.The mapping relation between spectrum feature vectors and fault types was established.The multi-fault test and measured test data were used for application verification.【Result】The target value of the optimized neural network was obtained at the 30th iteration while it cannot be realized until the 56th iteration without optimization.The expected output and the actual output were consistent.The accuracy of the fault recognition was 100%.【Conclusion】The proposed vibration fault diagnosis method based on multiwavelet-energy and RBF neural network is applicable and has high diagnosis accuracy and engineering popularization value. |
| Key words: hydropower generating unit fault diagnosis multiwavelet neural network extraction of eigenvector |
