Gearbox Fault Diagnosis based on Time-frequency Combination Feature Extraction and Manifold Learning of MS-LTSA

Abstract

A fault diagnosis method of gearbox based on time frequency union （TFC） feature extraction and manifold learning of improved supervised local tangent space arrangement （MS-LTSA） is presented. Firstly，a feature extraction method combining time domain，frequency domain and HHT time-frequency domain is proposed to obtain the comprehensive feature vector information of vibration signals. Then，the singular values of high-dimensional feature vectors are extracted and the singular value matrix is denoised by manifold learning theory. Finally，an efficient and accurate fault identification of the gearbox is realized by the feature vector after noise reduction. The proposed MS-LTSA method realizes the combination of the internal structure information and the class discrimination information of the data set，and improves the clustering effect of the extracted low dimensional features. Through analysis of experimental data，the excellent performance and application value of the proposed method in gearbox diagnosis are verified.

Keywords

• Gearbox;Feature extraction;Manifold learning;Fault diagnosis