The Fault Diagnosis Research Based on SOM-BP Composite Neural Network Learning Algorithm

Introduce the basic principle and learning algorithm of the SOM network and BP network. The diagnosis mode is established with the common breakdown condition and the related parameters of the gear boxes used as the training sample. Due to the complex nonlinear relation between breakdown mode and characteristic parameters of gear-boxes, the SOM-BP composite neural net work is used. First have a preliminary pattern recognition classification for training samples by SOM network and details of fault classification by BP network under the MATLAB 7.1 environment, through the simulation test and comparison with BP network, reliability of the composite neural network for gear box failure diagnosis are verified.

[1]  Tetsuo Furukawa,et al.  SOM of SOMs , 2009, Neural Networks.

[2]  Hong Wang,et al.  Soft Fault Diagnosis for Analog Circuits Based on Slope Fault Feature and BP Neural Networks , 2007 .

[3]  Yap Keem Siah,et al.  Improvement of ANN-BP by data pre-segregation using SOM , 2009, 2009 IEEE International Conference on Computational Intelligence for Measurement Systems and Applications.

[4]  Mohamed Benbouzid,et al.  A Brief Status on Condition Monitoring and Fault Diagnosis in Wind Energy Conversion Systems , 2009 .