约束独立分量分析(CICA)算法在故障诊断中的应用

Constraint Independent Component Analysis (CICA) Algorithm in Fault Diagnosis

盲源分离(BSS)是利用源信号间的统计独立性,在源信号和混合通道都是未知的条件下,仅由观测信号分离出各源信号的过程,也称独立分量分析(ICA)。经典的ICA仅仅用到数据的统计信息特征,即统计独立性。然而,机械故障存在其他如频率特征等已知的先验知识,将主要利用这些先验信息进行故障诊断。提出一种带参考信号约束的ICA算法(CICA)进行盲源信号的分离,选取与待提取信号频率相同的脉冲信号作为参考信号,以均方误差作为相似性测度的方法进行了实验仿真,仿真实验表明CICA算法能够很好地分离出待提取信号。

Overview the defects of the ICA,and to overcome them,this paper presents a kind of ICA algorithm that contains reference signal constraints,and the basic mathematical model and principle of constraint independent component analysis(CICA) algorithm.This paper selects pulse signal,whose frequency is the same to the extracted signals' s,and the mean square error as the method of similarity measure to do experimental simulation.