基于小波包-AR谱和多核映射SVM气阀机构故障诊断

Fault Diagnosis of SVM Valve Mechanism Based on Wavelet Packet-AR Spectrum and Multi-Kernel Mapping

利用小波包对非平稳信号故障特征提取的优越性和支持向量机适用于小样本学习的特性,针对柴油机气阀机构故障提出了一种基于小波包-AR谱分析和多核映射支持向量机相结合进行柴油机气阀机构故障诊断的方法。该方法采用小波包-AR谱分析提取频带能量为特征向量,利用多核映射支持向量机故障分类器实现对柴油机气阀机构故障分类。试验结果表明,小波包-AR谱分析和多核映射支持向量机能对柴油机气阀机构故障进行有效诊断,故障预报正确率为100%。

Using wavelet packet to the superiority of non-stationary signal fault feature extraction and support vector machine is suitable for the characteristics of small sample learning,aims at the malfunctions of diesel engine valve mechanism is put forward based on the wavelet packet and the AR spectral analysis and multi-core mapping combination of support vector machine(SVM)method for diesel engine valve mechanism fault diagnosis.This method USES the wavelet packet-AR spectrum analysis to extract the energy of the frequency band as the characteristic vector,and USES the multi-core mapping support vector machine fault classifier to realize the fault classification of the gas valve mechanism of the diesel engine.The test results show that the wavelet packet-AR spectral analysis and multi-kernel mapping support vector function can effectively diagnose the fault of the diesel engine valve mechanism,and the fault prediction accuracy is 100%.