Fault Diagnosis Of Analog Circuit Based On Improved Neural Network

Nowadays,electronic products are widely used in various fields.It is of great significance to find out the potential safety problems in electronic products in time.Analog circuit is an indispensable part of electronic products,and analog circuit has the characteristics of complex coupling,non-linearity,tolerance,etc.,which leads to the slow development of analog circuit fault diagnosis technology,unable to meet the current electronic industry demand for the safety,reliability and testability of electronic products.Therefore,it is very important for the development of analog circuit fault diagnosis technology to find an intelligent diagnosis method suitable for modern analog circuit.This paper focuses on three problems: fault data acquisition,fault feature extraction,fault recognition and classification.(1)Aiming at the low efficiency and low fault coverage of traditional fault data acquisition methods,an automatic fault data generation platform based on SLPS is designed.First of all,the SLPS module is used to build the joint simulation model of Pspice and Matlab,which realizes the function of Matlab to call the circuit model in Pspice.Then,the strategy of fault auto implantation is studied,and the goal of fault auto implantation is achieved by modifying the information in net file.Secondly,the complex function code is encapsulated into exe file,which makes the file have the characteristics of independent operation and simple interface.Finally,the simulation experiment can generate circuit fault data quickly after setting the parameters of fault type and fault number.The operation result of the platform is the same as the result of running Pspice software alone,and its operation efficiency is much higher than the traditional method,which has a certain practical value.(2)Aiming at the problem that it is difficult to extract the fault features of analog circuits effectively,a multi information fusion fault feature extraction method based on principal component analysis is proposed.Firstly,wavelet packet algorithm is used to eliminate the noise signal in the original fault data,and more precise available timefrequency information is extracted,which is sorted into wavelet packet energy coefficient features.Then,using the theory of information statistics,we can fully mine the fault information in the original data and extract the statistical characteristics of information.Finally,the wavelet packet energy coefficient features and information statistical features are composed of multi information fusion fault features,and the principal component analysis method is used to reduce the dimension of the fusion fault features,reduce the fault feature data dimension,save most of the fault feature information,and reduce the calculation of fault identification and classification in the later stage.(3)Aiming at the problem of low accuracy of analog circuit fault classification,a fault diagnosis method based on optimized neural network is proposed.Firstly,a single BP neural network is trained by using several different fault features.Experiments show that the fusion features after dimension reduction are excellent in accuracy and timeconsuming.Secondly,in view of the defects of single BP neural network,such as easy to fall into local optimal value,genetic algorithm is used to optimize the initial weight and threshold value of neural network.Experiments show that the diagnosis accuracy of GA-BP is higher than that of single BP.Finally,in view of the shortcomings of the traditional genetic algorithm,such as precocity,quantum genetic algorithm is used to optimize the neural network.Experiments show that QGA-BP can achieve more accurate diagnosis results than GA-BP.