Research On Fault Dignostics And Prognostics For Complex Electronic System

Traditional subsequent maintenance and scheduled maintenance have been unable to meet the equipment maintenance support needs. Therefore, the prognostic and health management(PHM) technology is produced. Owing to the many advantages of the PHM technologies, the PHM is the focus of the current research field. Based on the above reasons, the main research in this paper are as follows:1.Study on electronic systems PHM management strategy. First, start failure prediction framework to predict the detection signal. If the predicted signal exceeds the signal fault tolerance, start the module level fault diagnosis framework based on the multi-signal model diagnostic reasoning. In the multi-signal model diagnostic reasoning,if the speed has priority, the real-time diagnostic reasoning is choosed; if the precision has priority, Lagrangian relaxation diagnostic method is selected, to identify potential failures modules. Then, start the component-level fault diagnosis method based on least squares support vector machines(LS-SVM) to identify the specific failure component,repairing, replacing and re-testing the faulty modules or components.2.Study on analog circuits fault prediction method based on particle filter(PF).This paper mainly uses the analog circuits as the main object, proposing the analog circuit failure prediction framework. First, input the test stimulus signals. Then, extract the feature. Feature extraction is a key part of the failure prediction, and it will have a direct impact on the circuit degradation trend. After the feature extraction, the extracted features often require a preprocessing, to reflect the trend of circuit performance degradation, which is named after Fault Indictor(FI). Finally, based on the failure threshold, use PF prediction algorithm to predict the remaining life of the circuit under test, providing an effective reference standard for the entire service life of the electronic system.3.Study on multiple fault diagnosis methods based on multi-signal model.Complex electronic systems exists many fault modes, and is prone to occur multiple faults. System-level fault diagnosis problem has always been a more difficult problem.Therefore, this paper researches the system-level multiple fault diagnosis method based on multi-signal model in depth. Compare the performance of the real-time diagnosis with the fault diagnosis method based on multi-signal model. The experimental resultsshow that the multi-signal model is more suitable for complex electronic systems the fault diagnosis. Moreover, a quick and easy unreliable warning algorithm is presented for the test unreliable problem.4.Study on analog circuits fault diagnosis method based on multiple features. For the data-driven fault diagnosis methods, such as support vector machine(SVM), the key is feature extraction, which designs the best feature vector for different fault types to achieve the optimal diagnostic results. This paper researches the feature vectors for analog circuit in depth, including wavelet features, statistical signal features, the conventional time-domain features and frequency features. After the feature data is obtained, the least SVM(LS-SVM) multi-classifier is used for fault diagnosis, which includes training and testing classification stages. According to the diagnostic results based on the test data, the best diagnostic feature vector is selected as the optimal feature vectors, and use the optimal feature vector in on-line monitoring, and diagnostic reasoning to improve the diagnostic accuracy and efficiency of the online system.