Research For Fault Diagnosis Of Analog Circuits Based On Intelligent Information Fusion

Since 1960's, while the analog circuit fault diagnosis has been the field of circuits and systems research priorities, the research has made a lot of valuable theoretical results, but the overall development has been relatively slow. Because of the problems, such as the nonlinear effects, the complex and diversity of analog circuit components fault diagnosing make the existing theories and methods have a certain distance away from the practical applications. Information fusion technology is a powerful tool for information processing. It provides a theoretical support for the solutions of problems in the analog circuits fault diagnosis. With the development of the multi-functional and large-scale analog circuit, the need of automatic test and diagnosis is growing. The core of this paper is exploring information fusion based on intelligent fault diagnosis method of analog circuits, and analyzing circuits and more information extraction and integration of decision-making intelligent diagnosis method.Against the shortcomings of BP neural network, which include the slow convergence, that easily falling into local optimal results much and the difficulty to determine the structure of network, the paper introduces three kinds of improved algorithms, and the practical tests show that: BFGS-BP network algorithm for small-scale circuits is fast, and accurate, especially have a higher recognition rate for soft faults; flexible BP network algorithm(RPROP) has good diagnostic results for large circuits.In order to solve the problems that the analog circuit faults diagnosis usually has unitary information to select, this chapter proposes a solution that we can get the electricity information by extracting the voltage and current when the circuit is in three different excitations and fault modes. To extract the current of the nodes, we can calculate out the current of the nodes indirectly with the help of the around circuits'voltage and the circuit characteristics.This chapter elaborates the basic concept of Information Fusion, and the classical model and algorithm of the fusion diagnosis. Then we propose a decision-making integration programme with D-S evidence theoryFinally, this paper designs and implements a diagnosis system for the application of the analog circuit fault diagnosis with the core of BFGS-BP network, multi-information fusion fault diagnosis algorithm and D-S evidence theory. First, BFGS-BP network is trained by using the voltage and current which under three different excitation voltages and diagnoses three groups of practical test information for the preliminary diagnosis. Then with the results of the initial diagnosis, the basic probability assignment can be calculated. After that, we process the fusion diagnosis by using decision-making criteria of D-S evidence theory. Through the actual test, using the neural network information fusion method can improve the fault diagnosis performance.