The Application Of Machine Learning To Fault Diagnosis Of Analog Circuits

The fault diagnosis of analog circuits is difficult due to the tolerance of elements, the non-liner of circuits and the difficulty of modeling. It is one of the challenging topics. The progress in very deep submicron semiconductor technology prompts the advent of System-on-Chip and analog/digital mixed-signal integrated circuits. Many theoretical problems appear in the analog test area. Using general or traditional theories and methods of fault diagnosis, they are difficult to be solved. Machine Learning, as one branch of the Computational Intelligence, provides a powerful way to diagnose faults of analog circuits. It is interested to researchers at present.In this thesis, Machine Learning is used to fault diagnosis because it can solve the problems well due to the learning ability. A Pattern Recognition based fault diagnosis system is proposed in this thesis from the practice in projects, and then three kinds of fault diagnosis model are compared using a variety of boolean classification performance metrics.A practicable diagnosis system with flow is proposed based on pattern recognition. Each of the components is discussed in detail: the selection of testing circuit, the selection of faults set, solving tolerance using Monte Carlo simulation method, feature exaction based on Primary Component Analysis and the selection of machine learning methods. Then a systemic diagnosis flow is proposed based on the components listed above, and machine learning methods can be used in analog circuit fault diagnosis with this flow in industrialized automatic diagnosis.A variety of classification performance metrics are used to compare these machine learning methods. There are few comprehensive empirical comparing learning algorithms. Learning algorithms are now used in many domains, and different performance metrics are appropriate for each domain. In this thesis, those performance metrics in point are picked out. Based on these metrics, the application results of decision tree, neural networks and SVM are compared impersonally.The model performances of three machine learming algorithms above are also evaluated on the fault diagnosis results of two international standard circuits.