A fault feature extraction model based on nonlinear manifold learningJIANG Quan-sheng1 LI Hua-rong1 HUANG Peng2
JIANG Quan-sheng1 LI Hua-rong1 HUANG Peng2
1. School of Electrical Automation, Chaohu University, Chaohu 238000, China; 2. School of Mechanical Engineering, Southeast University, Nanjing 211189, China
Abstract: Manifold learning is one of the effective methods as obtaining the geometric distribution within the high-dimensional nonlinear dataset, which can be used for fault signal feature extraction and diagnosis. To carry out the diagnostic problem of nonlinear, complex failure symptom in the mechanical fault diagnosis, we propose a feature extraction model based on manifold learning method. In the model, aiming to different processing conditions of the collected sample, the Laplacian Eigenmaps and its incremental, supervision algorithm are applied to implementing the feature extraction and classification to fault sample. As a result of the non-linear dimension reduction method, the model greatly retained the overall geometry information in the fault signal, which significantly enhanced the classification performance of fault pattern recognition. The experimental results for fault diagnosis of air compressor demonstrate the feasibility and effectiveness of the proposed model.