Analog Circuit Fault Diagnosis Based On Sensitivity Analysis Method

With the increase of complexity and density of analog circuits, test methods are demanded to locate faults and substitute fault components in order to guarantee reliable operation of circuits. However, the test cost is heavy because of the complexity of analog circuits. Therefore it is of practical significance to improve the level of test and fault diagnosis.This thesis discussed approaches of fault diagnosis based on the sensitivity analysis which is a new test method of analog circuits. Main objects are described as follows:First, there exists a problem which the existence of solutions can not intuitively be predicted by choosing test frequency in the first-order sensitivity method on fault diagnosis. On the basis of the sensitivity analysis and its related theories, the sensitivity matrix analysis approach is presented to achieve a good predictability and controllability of solutions. This approach has been demonstrated to be distinct for predicting solutions of equations in simulation experiments.However, the accuracy of solutions is affected because the above diagnosis method is based on first-order sensitivity equations. Thus the second-order diagnosis equations are introduced. It is necessary to propose the second-order sensitivity with higher accuracy than first-order sensitivity, although the basic thought and sensitivity matrix analysis approach, similar to the first-order sensitivity, are used. After an error analysis is made in detail for second-order equations and some essential points of genetic algorithm in fault diagnosis are also illustrated, an improved genetic algorithm is adopted to choose an optimal test frequency. The resulting optimized-frequency diagnosis algorithm, based on the second-order sensitivity, helps get a solution with better accuracy. When the tolerances of normal components are ignored and partly ignored, experiment results show that the accuracy of solutions, compared to the first-order sensitivity, is improved and is further improved via combining the improved genetic algorithm with the second-order sensitivity diagnosis method.