自适应局部独立分量分析

提出一种自适应局部独立分量分析降噪算法。该算法先将一维时间序列重构到高维相空间,用聚集模糊K均值聚类和聚类评价函数求取高维数据集的聚类个数和聚类中心位置,然后利用K均值聚类寻找局部投影区间,对每个聚类进行独立分量分析并投影到低维空间,将低维空间数据排列并重构成一维时间序列。与使用聚类的局部独立分量分析相比,该算法具有自适应性和稳定性。使用数值仿真试验和齿轮故障信号对该算法进行验证,结果表明该算法对此类信号具有良好的降噪效果。

Adaptive local independent component analysis

An adaptive local independent component analysis algorithm was proposed.In the algorithm,one dimensional time series was reconstructed into a high dimensional trajectory matrix,and then clustered into several clusters using K-means algorithm,in which the number of clusters and the location of each cluster were determined by agglomerative fuzzy K-means clustering.The independent component analysis was applied to each cluster and the trajectory matrix was projected into low-dimensional phase space to obtain low-dimensional data,which were sorted and put in the original order.The final noise reduction result was achieved after reconstructing the one dimensional time series from the sorted data.Comparing with the common method of local independent component analysis using clustering,the proposed algorithm is more adaptive and robust.The algorithm was validated to be effective by numerical simulations and its application in gear fault diagnosis.The results show that the algorithm is capable of dealing with this class of signals.