Research On The Hierarchical Diagnosis Of Analog Circuit Fault Based On Neural Network Group

In recent years,electronic systems have been widely used in various fields,and more stringent requirements have been put forward for their operational stability.As an important mean to ensure the safe and reliable operation of electronic system,fault diagnosis technology can quickly diagnose the fault type,locate the fault location and judge the fault degree,greatly reduce the economic loss caused by circuit fault,and improve the enterprise efficiency.Electronic systems are usually composed of analog-digital hybrid circuits.Although analog circuits occupy a small proportion,their continuity,non-linearity,tolerance,and susceptibility to environmental interference have caused the analog circuit to have a higher frequency of failures in the system circuit.Therefore,the study of analog circuit fault diagnosis technology has important practical engineering significance.At present,the research in the field of analog circuit fault diagnosis focuses on the optimization of feature extraction and neural network algorithm,but most of these methods can only locate the fault components,and can not identify the parameter interval of the fault components.Therefore,the hierarchical diagnosis of analog circuit fault is studied.The existing hierarchical diagnosis methods of analog circuit fault mainly include hierarchical fault dictionary method,network tearing method and hierarchical multi-level decision tree method.Although these methods can identify the parameter interval of faulty components,the process of diagnosis is more complicated,the amount of calculation is large,and the engineering practicability and generality are also low.In addition,through further research on the existing diagnosis methods of analog circuit fault,it is found that most methods use multiple test nodes of the circuit to diagnose some components with higher sensitivity,and the characteristic parameters of fault are only the output voltage of the circuit.To solve the above problems,a hierarchical diagnosis method of analog circuit faults based on BP neural network group is proposed.This method simultaneously performs transient analysis and AC analysis on the only test node at the output of the analog circuit,in which the signal after transient analysis is subjected to FFT transformation,and DC component,fundamental amplitude,fundamental phase and distortion value of the signal are extracted as fault characteristic parameters.At the same time,the voltage amplitudes under different frequencies after AC analysis are also extracted as characteristic parameters,and then the extracted fault characteristic parameters of the BP are sent to the BP neural network group to realize the automatic hierarchical diagnosis of the faults of all components of the circuit,and finally identify the specific parameter interval of the fault components.Subsequently,in order to verify the feasibility of the diagnosis method,the classic triode amplifier circuits in analog circuit are simulated and analyzed by Pspice and Matlab.On the basis of feasible simulation,a hardware test platform based on FPGA and DSP is designed.The platform includes signal conditioning,high-speed A/D,FPGA and DSP module circuits,and mainly completes the task of parameter collection,storage and processing of analog circuits.After that,the hardware platform transmits the processed data to Matlab through the serial port.Matlab uses the data for neural network group training and fault hierarchical diagnosis.And the diagnosis results are displayed in the PC interface.Finally,the actual test has been carried out through the system hardware test platform.The actual test results show that the diagnosis method can indeed perform automatic hierarchical diagnosis of all components of the circuit through an output test node,and identify the parameter interval of the faulty component with a minimum recognition rate of 30%.