Multiple manifolds analysis and its application to fault diagnosis

Blind vibration component separation and nonlinear feature extraction applied to the nonstationary vibration signals for the gearbox multi-fault diagnosis

“…Yang et al [27] proposed a method for nonlinear time series noise reduction based on principal manifold learning applied to the analysis of gearbox vibration signal with tooth broken, but only for signal denoising. Li et al [28] proposed the multiple manifolds analysis (MMA) approach to extract manifold information from the bearing vibration signals with different faults and Wang et al [29] combined locally linear embedding (LLE) and kernel fisher discriminant analysis (KFDA) to detect rolling bearing fault. In the previous work we also adopted the LLE algorithm for the feature reduction of the gear crack level identification [3].…”

“…A method is proposed based on the KICA, LLE and fuzzy knearest neighbor (FKNN). In comparison with the fault diagnosis method based on manifold learning reported in [3,28,29], the proposed technique in this work adopts not only nonlinear dimensionality reduction algorithm, but also KICA for nonlinear BSS problem. Thus, it possesses a more powerful fault diagnosis capability than existing approaches.…”

Envelope Extraction Algorithm of the Vehicle Acoustic Signal Based on Step-Size Transformation

“…Acoustic signal analysis-based traffic information detection technology has the advantages of low cost, small information redundancy and small external interference [1][2][3]. It is obviously different in the waveform of the acoustic signal when the car passing the deceleration belt and driving on the normal highway [4][5][6]. Meanwhile, for this signal, its envelope curve can better reflect the variation of the feature compared with the original signal.…”

“…Signal envelope extraction is a technique which can find the waveform edge using a certain signal processing method [5]. And the envelope contains some time-domain characteristic parameters which can reflect some properties and features of the signal to some extent.…”

Life grade recognition of rotating machinery based on Supervised Orthogonal Linear Local Tangent Space Alignment and Optimal Supervised Fuzzy C-Means Clustering

“…Recently, a novel theory of nonlinear dimension reduction called manifold learning is becoming the research focus. It can discover the intrinsic feature of nonlinear high-dimensional data by projecting 4 them into a low-dimensional feature space and preserving the local neighborhood structure [15][16][17][18]. Typical manifold learning methods mainly include Linear Discriminate Analysis (LDA) [19], Neighborhood Preserving Embedding (NPE) [20], Locality Preserving Projection (LPP) [21], Linear Local Tangent Space Alignment (LLTSA) [22], etc.…”