Detection and localization of turbine-generator bearing vibration using wavelet neural network

Due to rotating at high speed and operating under malcondition, the turbine-generator set rotor sometimes vibrates violently, which damages the other major components, and moreover the abnormal vibration would cause serious fault accident and economical loss. By means of condition monitoring and fault diagnosis technique, a novel approach using wavelet neural network is brought forward to transient vibration signal processing and fault pattern recognition. The feature extraction technique is needed for preliminary processing of recorded time-series signal over a long period of time to obtain suitable parameters which, in linear and nonlinear combination, reveal weather the fault is evolving. The transient signal can be decomposed into series of wavelet subspaces based on wavelet transformation, each of which covers specific frequency band in time-frequency domain. These feature vectors are input nodes to the wavelet neural network for fault pattern recognition, which operates on the feature vectors to produce recognition decisions based on previously accumulated knowledge. The experiment results demonstrate that the proposed approach combining wavelet transform and neural network is effective for fault diagnosis of turbine-generator set.