基于LLTSA算法维数约简的滚动轴承故障诊断

Rolling Bearing Fault Diagnosis Model Based on Dimension Reduction Using Linear Local Tangent Space Alignment

针对滚动轴承高维故障特征集识别精度低的问题,提出基于线性局部切空间排列(Linear Local Tangent Space Alignment,LLTSA)算法的维数约简故障诊断模型。首先结合小波包分解、时域、频域及时频域统计方法构造全面表征轴承不同故障特性的混合域特征集,通过敏感度的特征选取方法,从混合特征集中选取轴承故障的敏感特征集,再利用LLTSA算法将高维敏感特征集约简为故障区分度更好的低维特征矢量,并用模糊C均值(Fuzzy C-means,FCM)聚类算法进行故障模式识别,本研究方法能够突出不同特征对分类的贡献率,强化敏感特征,弱化不相关特征,提升了分类精度。最后用深沟球轴承不同部位故障诊断实例验证该模型的有效性。

Aiming at the problem of low accuracy of high dimensional tault teature set of rolhng bearing, a fault diagnosis model of dimension reduction based on linear local tangent space alignment is proposed. Firstly, the wavelet packet decomposition, time domain, frequency domain and time domain frequency domain method are used to construct the feature set of the hybrid domain, sensitivity characteristic of the bearing fault is constructed by the method of feature selection. Then, the LLTSA algorithm is used to＇ simplify the feature set of the high dimension mixed domain to the lower dimension feature vector, and Fuzzy C- means clustering algorithm is used to identify the fault pattern, the method Can highlight the contribution of different features to the classification, strengthen the sensitive features, weaken the irrelevant features, and improve the classification accuracy. Finally, the effectiveness of the model is verified by the example of deep groove ball bearing fault diagnosis which in different parts.