中文摘要：

针对风力发电机组轴承故障振动信号具有高噪声、非线性、非平稳的特性,提出了一种集成经验模态分解与峭度-相关系数准则和多特征量提取的风电机组轴承故障程度的诊断方法。利用集成经验模态将振动信号分解成若干个本征模态分量,采用峭度和相关系数准则选取一组包含信息量最丰富的分量,对该组分量从时域指标、自回归模型参数矩阵的奇异值和能量熵3个角度的变化中提取和构造多特征量矩阵,输入支持向量机建立故障程度不同的多分类预测模型,优化核参数和惩罚参数取得轴承故障程度最佳预测精度。通过实验室数据验证了该方法是一种可行的风电机组轴承故障诊断方法,可实现对风电机组轴承故障早期处于的轻度、中度和重度等3种相对故障程度的准确分类和识别。

英文摘要：

According to the characteristics of high noise,nonlinear and non-stationary of bearing of wind turbine,a method of bearing fault diagnosis based on ensemble empirical mode decomposition（EEMD）and kurtosis-correlation coefficients criterion as well as multiple features was proposed in this paper. Firstly,the original vibration signals were processed to several IMFs by using ensemble empirical mode decomposition. A set of IMFs were selected by the kurtosis-correlation coefficients criterion which contained the most mounts of information. Secondly,a matrix of multiply features was extracted and constructed from the set of IMFs through three aspects which included time domain,singular value of the auto-regressive mode parameters matrix and energy entropy. In order to set up a predictive model of multiply classification,the matrix was put into the support vector machine. To optimized the kernel parameters and penalty parameters for attaining the best accuracy rating of prediction, the data of laboratory were given to verify the method. The fault degrees of wind turbine＇s bearing which comes from actually monitoring can be accurately classified and diagnosed by using this method, and the results show that it is a feasible fault diagnosis method of the wind turbine＇s bearing.