An analytical model of electronic fault diagnosis on extension of the dependency theory

• An improved dependency model is proposed considering the drawbacks of the traditional dependency model.
• Fault diagnosis is formulated as an optimization problem with binary logic operations
• The model can make estimations of the fault sections, as well as the malfunctioned switches and test points.
• The proposed method is demonstrated via an avionic processor case using the improved dependency model.

Based on the D-matrix model, the dependency theory is widely used in the field of fault diagnosis to model the fault flows in complex electronic systems. However, the traditional dependency model can only handle a single fault; it fails to recognize and diagnose multiple faults. In addition, it is not tolerant with system structural or functional changes. These inherent weaknesses of the traditional dependency theory may lead to unsatisfactory acquisition of the diagnosis results. To solve the problem, an improved dependency model is invented as novel analytic diagnosis model to better describe the relationships between faults and tests. The system fault diagnosis based on the improved dependency model is formulated as an optimization problem with binary logic operations where all the fault hypotheses are tested. The calculation process consists of three steps: establishment of the objective function, determination of the nominal states, and determination of the expected states. Finally, the proposed method is demonstrated via an avionic processor case using the improved dependency model. The optimization-based fault diagnosis problem is formulated and the optimal solution is obtained. The diagnosis result demonstrates that the proposed method is successful on performance assessment and fault diagnosis.