在对旋转机械进行故障诊断时，通常要从时域、频域或时频域提取故障特征参数，组成原始的故障特征向量，然而在众多的故障特征当中并不是每个特征对于故障分类都是敏感且有效的。为此，本研究提出了基于ReliefF算法和相关度计算结合的故障特征降维方法。采用ReliefF加权特征选择算法对原始各特征的分类能力进行评价，选择出分类能力较强的特征；再通过特征相关度算法剔除其中分类能力相近的冗余特征，将剩余的分类能力较强的特征组成最终的降维特征向量用于故障分类和诊断，实现原始特征的降维。通过液压泵和滚动轴承的故障诊断实验，并与传统的主元分析(PCA)方法对比，结果表明该方法能够用较少的降维后的信号特征获得更高的故障正确识别率。

In the fault diagnosis of rotating machinery, the fault feature parameters are usually extracted from time domain, frequency domain or timefrequency domain. And the original fault feature vector is constituted by the extracted feature parameters. However, among the numerous fault features, not every feature is sensitive and effective to fault classification. Hence, a fault feature dimension reduction method based on ReliefF algorithm and correlation calculation was proposed. In the mothed, the weighted ReliefF feature selection algorithm was utilized to evaluate the classification ability of original features and choose the features with strong classification ability. Then, the redundant features possessing similar classification ability were eliminated by feature correlation algorithm. And the feature vector was composed by the remaining features with strong classification ability and used for fault classification and diagnosis. Through the above approach, the dimension of original features was reduced. Moreover, the proposed method was applied to the fault diagnosis for hydraulic pump and rolling bearing. Comparing with the traditional principal component analysis (PCA) method, the analysis results show that the proposed method can use fewer features after dimension reduction to obtain a higher correct recognition rate.