| The traditional "Scheduled maintenance" and "breakdown maintenance" measures of analog circuits have many shortcomings,which cannot meet the needs of health management for analog circuits in the new era.The incipient fault diagnosis and fault prognostic methods of analog circuits are studied based on the framework of machine learning method and fault prognostic and health management technology so as to decrease the failure rate and cut down the maintenance cost of analog circuits.To solve the problem of low discrimination of fault features in incipient fault diagnosis of analog circuits,the stacked auto-encoding network was used to extract the in-depth fault features of analog circuit,and then the support vector machine was used as fault classifier to locate and classify the components of analog circuits with fault tendency.A fault prognostic model based on multi-feature vector extraction and adaptive particle swarm optimization(APSO)optimized support vector regression(SVR)was established to solve the lack of research on fault prognostic of analog circuits.Firstly,the feature extraction of analog circuit is carried out,and the multi-feature fusion vector is constructed by combining the statistical feature with the wavelet packet energy feature to solve the difficulty in feature extraction of analog circuits.Then the Euclidean distance of the feature vectors is calculated to quantify the decay state of the components in the analog circuit,and the fault indicator and fault threshold of the analog circuit are obtained accordingly.Finally,APSO is used to optimize the SVR to construct the prognostic model to improve the accuracy of the remaining useful life prediction in the analog circuit.In the incipient fault diagnosis experiment of analog circuit,it can be seen from the diagnosis result that this model can obtain higher fault identification accuracy.In the fault prognostic experiment of analog circuit,it can be seen from the prediction results show that the prognostic model based on APSO-SVR has a higher accuracy for analog circuit fault prognostic. |
