Research On Fault Diagnosis Method And Application On Electronic System Based On Graph Model

In the modern large-scale systems, electronic equipment often takes the criticalposition, and serious consequences will occur due to its failure in the using process.However, the difficulty of diagnosis is increased by complex structure, strongcoupling relationship, signal discreteness, fault propagation uncertainty ofelectronic equipment. So it is difficult to meet the system level fault diagnosisrequirements of modern complex electronic equipment with traditional faultdiagnosis methods. In this thesis, to solve the above problems, fault diagnosismethods suit for large-scale complex electronic systems will be researched based onpractical application background.To meet the system level fault diagnosis requirements of complex electronicequipment, this thesis focuses on the fault diagnosis methods based on graph model,and researches the modeling methods for laws of fault generation, spread andinfluence. In this thesis, HSDG (Hierarchical Signed Directed Graph) model basedon SDG (Signed Directed Graph) is adopted to establish fault diagnosis model forthe electronic systems with enough test information. HSDG is proposed to solve theproblems in SDG, such as complicated calculation process, large faults search space,by emerging the hierarchical and structural characteristics of large-scale complexelectronic systems into SDG, which can simplify the calculation process andimprove the efficiency of fault diagnosis effectively. For the fault diagnosisproblems of electronic systems without enough test information and uncertainty offault propagation, considering the probability distribution characteristics of testinformation, failure mode and failure propagation, fault diagnosis method ofelectronic system based on probability graph model is researched, to establish amore accurate fault diagnosis model. In this thesis, the diagnosis model based onprobability graph with BN (Bayesian Network) is established. BN model has theintuition of structure expression with graph model and stringency of reasoningprocess with probability theory, so it can describe the fault propagationcharacteristics of system accurately and improve the accuracy and efficiency offault diagnosis.The simulation platform based on practical background, a satellite electronicequipment, is established to verify the validity of the proposed HSDG and BN methods in this thesis. And fault injection and samples can be realized with thisplatform rapidly to verify the proposed fault diagnosis methods. The experimentresults show that: Compared with traditional SDG method, HSDG model canimprove the efficiency of fault diagnosis dramatically. Fault diagnosis method withBN model can achieve the desired accuracy, and it can satisfy the demand of faultdiagnosis of complex electronic systems.