关键词: 稀疏表达, 特征选择, 故障诊断, 过程工业

Abstract: In this paper, a new sparse representation based feature selection method is proposed, in which the sample matrix is composed of the mean and variant of training sample, and testing sample is the index vector responding to sample matrix. Homotopy algorithm is used to solve the optimization problem. Traditional selecting methods based on wavelet package decomposition and Bhattacharyya distance methods, and recently used sparse methods, sparse representation classifier, sparsity preserving projection and sparse principal component analysis, are compared to the proposed method. Simulations show the proposed selecting method gives the improved performance on fault diagnosis with Tennessee Eastman Process data.

Key words: sparse representation, feature selection, fault diagnosis, process industry