Application of neural network to faults diagnosis of nonlinear circuits

This paper studies the Counter-propagation Networks (CPN) to faults diagnosis of the circuit. Using the CPN to build center of information fusion and fuse the data of multi-sensor in order to reduce the uncertainty of the faults diagnosis. Reset rules to overcome a shortage which the input vector limit too tight by the improvement of CPN algorithm of initial weight; Optimize operation steps of algorithm to improve the operating effects of algorithm; The results show that it improves membership value of the actual faults components and enhances the object´s diagnosis analysis that faults diagnosis method of multi-sensor information fusion are based on the CPN and fuzzy mathematics. The experimental data shows that this method can accurately position the fault components of circuit, it performs advantage of fast speed training, high rate of diagnosis and wide suitability.