Manifold learning, also known as nonlinear dimensionality reduction or nonlinear embedding, is a set of techniques in machine learning and data analysis that aims to uncover the underlying structure or geometry of high-dimensional data in a lower-dimensional space. In mechanical field, faults often introduce subtle changes in the data patterns, making it difficult to identify them using traditional methods. Manifold learning can help capture these subtle changes by representing the data in a lower-dimensional space that emphasizes the fault-related variations and become a valuable technique in fault diagnosis. This chapter addresses a variety of intelligent fault diagnosis and prognosis methods based on manifold learning, including spectral clustering manifold-based fault feature selection, locally linear embedding (LLE)-based fault recognition, and distance-preserving projection-based fault classification. These methods are applied to the fault diagnosis and prognosis of gears, rolling bearings, engines, etc., and the effectiveness is verified by case studies.
Manifold Learning Based Intelligent Fault Diagnosis and Prognosis
Intelligent Fault Diagnosis and Health Assessment for Complex Electro-Mechanical Systems ; Kapitel : 4 ; 193-271
11.09.2023
79 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Semi-supervised Learning Based Intelligent Fault Diagnosis Methods
Springer Verlag | 2023
|British Library Online Contents | 2010
|British Library Online Contents | 2011
|Supervised SVM Based Intelligent Fault Diagnosis Methods
Springer Verlag | 2023
|