Research On Analog Circuit Fault Diagnosis Based On Neural Network

With the rapid development of modern information technology and integrated electronic technology,the wide application of electronic devices in various fields has made the development of related technologies in the field of electronic devices closely linked to the country's overall national strength and economic stability.While bringing vitality to social and economic development,the serious consequences caused by the failure of electronic equipment also make people pay more attention to the maintenance and management of their use.However,due to the complex and diverse structure and function of electronic devices,the difficulty of detecting faults is increased.The existing analog circuit fault diagnosis technology mainly realizes the fault mode positioning and classification,but can not reflect the fault degree of the circuit in detail.In view of this situation,this paper studies the fault diagnosis method of analog circuit based on neural network and simulates the fault status of the circuit.The combination of classification and health status assessment can realize the classification of failure modes of analog circuits,evaluate the health status of analog circuits,and provide reference for analog circuit maintenance and management.The main research contents of the paper are as follows:Firstly,a fault feature extraction method based on multi-feature information fusion is designed.The feature extraction method based on statistical information and wavelet packet decomposition is used to extract faults from analog circuits.The fault feature information of dimensionality reduction is obtained through principal component analysis.It is helpful to reduce the computational complexity and optimize the structure of the separator.The simulation experiments show that this method can obtain comprehensive and accurate fault feature information;then,the fault classification method of analog circuit based on compact wavelet neural network is studied and replaced by wavelet function.The hidden layer excitation function of BP neural network adopts the weight adjustment method of variation factor and adaptive rate to improve the BP learning algorithm.The Sallen-Key band-pass filter circuit simulation experiment shows that the compact wavelet neural network is given.Compared with the improved BP neural network method with the same structure,the method is more effective in identifying and locating analog circuit faults.Finally,an analog circuit health status assessment method based on sample similarity is designed,and a weighted horse based on information entropy is used.The distance between the circuit to be measured and the standard state feature vector The simulated health status of the analog circuit was evaluated by translating the weighted Mahalanobis distance into the failure rate through health thresholds and fault thresholds.The simulation results show that this method can accurately assess the health status of analog circuits and apply to the monitoring of early soft faults in analog circuits.