Theory & modeling and realizing of integrated neural networks in fault diagnosis

The behavior of the equipment reflects the running condition from different aspects, which is the base to perform integrated diagnosis on equipment. By researching the information combination ability of single neural network, the author puts forward that using several neural networks, each of which diagnoses faults from one or many aspects, we can get the final result, which is the fused result from all sub-nets. In other word, that is to build integrated neural networks Thus the information provided by the equipment, such as vibration, temperature, pressure, etc., is efficiently used. The theory and modeling method of the integrated neural networks is explained in the paper, and the diagnosed example by integrated neural networks is also given. The result shows that when integrated neural networks is used on fault diagnosis, there are many advantages, such as being easy in adding and deleting network, being easy in training network, high diagnosis definition, etc. The idea of integrated neural networks is also very valuable for fault diagnosis on distributed system.