针对非线性系统故障诊断难以解决的问题,提出了一种基于扩展局部线性嵌入映射(Locally Linear Embedding,LLE)的故障诊断方法.通过引入切空间距离代替欧氏距离,可以更加科学的满足算法近邻点局部线性的要求,从而可以更好的保留原始数据的局部流形特征.另外,将故障状态与高维空间分布结合起来,通过确定数据点在空间超球内的分布完成故障的检测,在这个过程中将超球的确定与LLE算法中基于核函数的样本外数据扩展相结合,减少了计算量,提高了算法的实时性,从而为复杂非线性系统的故障诊断提供了一种新的有效的方法.
It is quite a difficult problem of fault detection in nonlinear system. In this paper, a new method based on LLE is proposed to solve the problem. Firstly, tangent space distance is introduced to LLE (Locally Linear Embedding) algorithm, which can satisfy the hypothesis of locally linear patch of LLE, and so it can preserve the local manifold features of the original data better. And then, fault state is combined with special distribution to complete the fault detection. Out of sample extension is also considered, and combined with fault detection algorithm, which can reduce computation obviously and improves real time capability of the Thereby algorithm. provides a new effective method for the fault diagnosis of complicated nonlinear system.