一种基于子空间法的机械故障欠定盲分离新算法

针对以往稀疏盲信号分离算法中恢复源信号时所采用的线性规划或最短路径法计算相对复杂,提出了一种基于子空间法的机械故障欠定盲源信号恢复方法。该算法假设源信号由两个正交向量构成:其中一个向量位于混叠矩阵A的行空间中,另一个位于A的零空间中,位于行空间中的向量可以通过A的Moore-Penrose伪逆得到,位于零空间中的向量通过贝叶斯估计得到。新算法容易实现,分离速度快,能够很好地满足盲分离对速度的要求。将其用于实测齿轮故障信号的盲分离,研究表明该方法能够分离齿轮系统的典型故障,取得了较好效果。

A New Algorithm of Underdetermined Blind Source Separation for Machine Fault Diagnosis based on Subspace Method

In the traditional Underdetermined BSS algorithms,sources recovering by linear programming or shortest-path technique and these two methods are with complex computation and hard to be operated,a new algorithm based on subspace method is proposed.In the new algorithm,sources are assumed to be composed of two orthogonal components,one lying in the rowspace and the other in the nullspace of the mixing matrix A.The component lying in the rowspace is obtained by Moore-Penrose psedudo-inverse of A,the component lying in the nullspace is obtained by Bayes estimator.The new algorithm is easy to realize with fast speed.Experiment shows that this method can separate compound faults of gearbox successfully.