New method of fault feature extraction based on supervised LLE

Abstract

The Locally Linear Embedding (LLE) is one of the efficient nonlinear dimensionality reduction techniques, which can be used to fault feature extraction. But it is not taking the class information of the data into account. In this paper, we propose a novel approach of feature extraction based on supervised LLE algorithm. Via utilizing class information to guide the procedure of nonlinear mapping, the Supervised LLE enhances local within-class relations and help to classification. The approach uses the Supervised LLE to extract feature for class labels data, and utilizes RBF network to map the unlabeled data to the feature space, which easily implement fault pattern classification. The experiments on benchmark dataset and engineering instance demonstrate that, the proposed approach excels compared to PCA and LLE, and it is an accurate technique for classification.