Application Of Fault Dictionary Method For Analog Circuit

In recent years, great progress has been made in IC industry. In order to improvethe security and reliability of the system, increasingly importance has been attached tofault diagnosis. Statistics show that80%of the fault in electronic equipment comesfrom analog circuit. So fault diagnosis of analog circuit has very high application value.But because of its continuous working state, tolerance parameter, nonlinear circuit andfew test points, leading to it has not formed a complete diagnosis system, the maturityand perfection are far behind digital circuit. Fault diagnosis of analog circuit hasbecome a new research hotspot.This thesis discusses deeply the approach for fault dictionary of fault diagnosistechnology. In this thesis, it introduces the concept of fuzzy set and fuzzy field for thecircuits with tolerance, uses the Bayesian theorem of pattern recognition knowledge toanalyse the error rate in fault diagnosis, and uses the minimum error rate to divide thefuzzy field. It also discusses the test point selection technology which removes theredundant test points by using the contains method, elimination method to obtain theeffective test points as little as possible. The thesis uses the co-simulation of Pspice andMatlab to collect and analyze the data of the amplifier circuit. Ultimately we establishfault dictionary with the traditional fuzzy set method and the Bayesian method, thencompare the accuracy of the two dictionaries. Experiments show that the twodictionaries can also analyse the circuits with tolerance.Using Bayesian theory to divide fuzzy sets can improve the accuracy in faultdiagnosis, but it requires large quantity of data before fault diagnosis and the division offuzzy set is also more complicated, so it needs to spend more time. Therefore, at the endof this thesis, it presents a new idea which can not only ensure the accuracy rate, butalso reduce the amount of calculation in fault diagnosis.