Overview of Wavelet/Neural Network Fault Diagnostic Methods Applied to RotatingMachinery

摘要

New technology in the form of wavelet-based methods coupled with intelligentclassification schemes built around neural networks, can drive the development of substantially improved fault detection and identification (FDI) methods. Such systems represent important next generation FDI kernels for integration into advanced condition based maintenance systems for rotating machinery. This paper presents an overview of the results obtained by ALPHATECH in a program aimed at developing wavelet/neural network based FDI systems for vibrating machinery. The paper presents the performance results of these methods applied to a range of platforms including helicopter transmissions, turbopumps, and gas turbines. In addition, enhancements to the basic fault detection and identification system are presented and include overviews of multi-sensor wavelet-based differential features and improved FDI performance through classification fusing using hierarchical neural networks.