基于CNN-SVM的水电机组智能故障诊断方法研究

Research on Intelligent Fault Diagnosis Method of Hydroelectric Generating Unit Based on CNN-SVM

在当前水电机组智能故障诊断的方法中,人为选择故障分类特征的主观性及故障小样本数据的局限性对故障诊断结果的准确性具有重要影响。对此,结合卷积神经网络(CNN)的特征提取优势和支持向量机(SVM)优良的小样本处理能力,提出了一种针对水电机组振动故障诊断的CNN-SVM方法。该方法以水电机组振动信号的时域波形图作为模型输入,然后利用CNN提取信号特征并导入SVM实现机组故障诊断。最后,通过具体的实例分析,验证了所提诊断方法的优势。

In previous researches on intelligent fault diagnosis methods of the hydroelectric generating unit,the subjectivity of the artificial selection of the fault classification characteristics and the limitations of small sample data have important impacts on the accuracy of fault diagnosis results.To solve this problem,a CNN-SVM method for the fault diagnosis of the hydroelectric generating unit was proposed by combining with the feature extraction advantages of convolutional neural network(CNN)and the excellent ability of support vector machine(SVM)in processing small sample.In this method,the time-domain diagram of the vibration signal of the hydroelectric generating unit was used as the model input,and the CNN method was employed to extract the signal features.Then,the extracted feature vector was input to the SVM method to realize the final fault diagnosis of the unit.Finally,the advantages of the diagnosis method proposed in this paper were verified through a specific example analysis.