Abstract: One method of fast feature selection is used in the high dimensional and noisy data of fault diagnosis to reduce the number of its attributes，according to the mean and the square of each feature of data.In this paper，a multi-fault classifier is designed using the technology of Support Vector Machine（SVM） which has the function of pattern classification，based on feature extraction.The example shows that dimensionality reduction and computing complexity reduction are realized by using this method on the basis of achieving good performance of the multi-fault classifier.

Key words: feature selection, fault diagnosis, Support Vector Machine（SVM）, multi-fault classifier

摘要： 针对故障诊断中数据存在噪声和高维的缺点，使用一种快速特征提取方法对故障数据进行降维，该方法以特征信号的均值和方差作为其权重衡量的依据。利用支持向量机的模式分类功能，构造了基于特征提取的多故障分类器。实例表明，在保证诊断效果的情况下，该方法实现了数据降维，降低了运算复杂度。

关键词: 特征提取, 故障诊断, 支持向量机, 多类分类