A Novel Compound Neural Network for Fault Sources Recognition
Independent component analysis (ICA) is a powerful tool for redundancy reduction and nongaussian data analysis. And, artificial neural network (ANN), especially the selforganizing map (SOM) based on unsupervised learning is a kind of excellent method for pattern clustering and recognition. By combining ICA with ANN, a novel compound neural network for fault sources recognition was proposed. First, neural ICA algorithm was applied to fusion of multi-channel measurements by sensors. Moreover, further feature extraction was made. Thus, statistical features higher than second order were captured from the measurements. Second, a typical neural classifier such as the back-propagation (BP), the radial basis function (RBF) or the SOM network was trained for the final fault sources recognition. The results from contrast experiments in fault diagnosis of rotating machines show that the proposed compound neural network with ICA based feature extraction can recognize various fault sources at considerable accuracy, and be constructed in simpler way, both of which imply its great potential in fault diagnosis.
independent component analysis redundancy reduction compoun dneura inetwork faultdiagnosis document code: A CLC number: TH17 TP391
Jiao Weidong Qian Suxiang Lin Peng Ma Zewen Yuan Qingping
Department of Mechanical Engineer Jiaxing University Jiaxing, China 314001
国际会议
太原
英文
32-36
2010-10-22（万方平台首次上网日期，不代表论文的发表时间）