Fault Diagnosis of Analog Filter Circuit Based on Genetic Algorithm

Abstract

Hard (open and short) faults and discrete parameter faults (DPFs) are the mostly used fault models in the simulation-before-test (SBT) method. Because the parameter of the analog element is continuous, the DPF cannot elaborately characterize all possible continuous parameter faults (CPF) occurring in the analog circuit. To address such a problem, a genetic algorithm (GA)-based simulation after the test (SAT) fault diagnosis method is proposed in this paper. The fault diagnosis is transformed into an optimization problem. The genes represent the parameter values of potential faulty components. The faulty circuit response is the objective. Our target is to minimize the difference between the actual faulty response and the GA simulated response. The chromosome that minimizes the difference gives the solution. This method does not save all possible faults in advance while it can diagnosis all continuous fault values. The effectiveness of the proposed method is examined by using filter circuit examples.INDEX TERMS Fault diagnosis, genetic algorithm, optimization.

Genetic Algorithm Based Faulty Parameter Identification for Linear Analog Circuit

“…This partial CPS still cannot cover all the possible parameter shifting, viz., (0, ∞). The whole fault range (0,∞) are handled in literatures [8,[10][11][12][13][14]. The ratios of voltage increments (or slope model) in literature [10] can model and diagnose such fault.…”

“…Therefore, this model can diagnose CPS within the whole range of (0,∞). In literature [13], a genetic algorithm (GA) based simulation after test method is used to diagnose the CPS too. The parameter value is represented by gene.…”

“…In the analog fault diagnosis field, the GA [21][22][23] is mainly used to select the best test points or stimulus to improve the testability. There are few GA based methods [13,24] can locate fault sources. Only one method [25] can estimate the faulty parameter.…”

“…The slope [10] and circle [11] models and the GA based [13] method can accurately locate faulty component within the whole range (0,∞). Hence, we do not handle the fault location in this paper.…”

“…This is what we do in this paper. The basic estimation processes are (1) measuring the faulty response of the CUT; (2) finding the fault source c by any fault location method [10][11][12][13]; (3) estimating the upper and lower bounds of c than can generate the same response as . The should be as small as possible and should be as large as possible.…”