Based On Genetic Neural Network For Fault Diagnosis Of Switch Current Circuit

With the continuous development of science and technology, Monolithic mixedwith a single integrated circuit module has been a trend of the development ofintegrated circuit design.90%modern hybrid integrated circuits belong to digitalcircuit, which means only a small fraction belongs to analogue circuit, but digitalCMOS technology cannot match with switched capacitor completely. So thedevelopment of the mathematical model of the integration technology is limited,switch current technology is brought out in this context. As the mixed signal systemand the important module interface of chip module system, switch current technologyhas got more and more attention, in recent years, the switch current circuit is lifted anew study boom by the international academic frontiers of circuit.Switch current technology has been developed, and the design of the relatedcircuit is also improved step by step, some practical circuits have been designed onthe basis of switch current. However, at present, the testing and fault diagnosis ofswitch current circuit is rather less involved. Based on the previous research about thefault diagnosis of switch current, this paper presents a new switch current faultdiagnosis method.This paper introduces the background of the development of switch currentcircuit, the present situation, and the fundamental storage unit and basic modules ofhigh performance switch current at first. And then introduces the CMOS1~3grademodels and CMOS hard fault model, and the big and small signal CMOS signal modelin detail. Finally, the application of the optimized BP neural network by the geneticalgorithm in switch current circuit fault diagnosis.At present, the fault diagnosis of switch current is still in the lever of hard faultdiagnosis and the simple circuit. For those who are difficult to distinguish between thesubtle fault and large quantity of data acquisition circuit are difficult to realize thediagnosis.and switch current circuit parameters fault is difficult to realize thediagnosis. And the simulation of the switch current circuit is without a goodsimulation platform. At present，the mainly simulation software of switch currentcircuit is Asize, this software’s function is simple, only can be used to the ideal switchcurrent circuits. It can only do some simple fault contrast and simple analysis if it isused for fault diagnosis. On this condition, according to the characteristics of the switch current circuit, this paper proposes a hard fault model in chapter4to simulatethe simulation of hard fault. For The soft fault simulation, this paper uses CMOS bigand small signal models to switch current circuit into analog circuits, and then test it.After using Pspice to simulate the circuit for extracting detailed data information,then uses the wavelet decomposition to extract data characteristics to simplify thedata structure. Finally, the neural network is used to distinguish the fault state toachieve the purpose of the fault diagnosis.