中文摘要：

提出一种新型的基于主成分分析（PCA）和支持向量机（SVM）的故障诊断方法。首先提取振动信号的多项时域指标,并利用小波包分解提取频域特征;再利用PCA从提取的时域、频域特征中选取敏感特征,实现降维处理,减小数据处理复杂度;最后利用SVM进行特征子集的训练和测试,实现故障分离。该方法在柴油机的失火、撞缸、小头瓦磨损等典型实际故障中的诊断准确率高达98%,证实了该方法的有效性。

英文摘要：

A new method was proposed based on PCA and SVM. First of all, the fault characteristics of vibration signals in time domain and frequency features were extracted by wavelet packet decomposition. Then the sensitive characteristics were selected with PCA to achieve dimensionality reduction and to decrease the complexity of data processing. Finally, SVM was used for training and testing of the feature subsets, and realizing the fault separation. Appling this method to typical faults of diesel engine such as misfire, cylinder collision and small head tile wear, the diagnosis accuracy rate is up to 98%, which confirmed the validity of this method.