Nonlinear manifold learning based fault classification method

In this paper, nonlinear manifold learning is introduced into equipment fault diagnosis field and solving fault pattern classification problem, a novel method of fault classification based on manifold learning is proposed. Two nonlinear manifold learning algorithms based on LLE and Laplacian Eigenmaps are utilized to learn original fault signal directly and extract intrinsic manifold feature in data set. The new approach can greatly hold the global geometry structure information embedded in the signal, and availably overcome the flaw of general feature extraction methods which only obtain data´s local linear structure, obviously improve classify performance of fault classification. The experimentations on simulation and engineering instance show that, the presented method excels compared to PCA, and is a feasibly and effective technique for fault classification.