(IJCSIS) International Journal of Computer Science and Information Security,Vol. 8, No. 2, 2010
Time Domain Analysis based Fault DiagnosisMethodology for Analog Circuits-A ComparativeStudy of Fuzzy and Neural Classifier Performance
, R. Bharat Ram
, V. Manikandan
, N. Devarajan
Department of Electrical Engineering, Government College of TechnologyCoimbatore, India
Department of Electrical Engineering, Coimbatore Institute of TechnologyCoimbatore, India
— In this paper, we attempt to diagnose the occurrenceof faults in analog electronic circuits based upon variations intime domain specifications corresponding to the circuit conditionunder consideration relative to the fault free circuit. To achievethis, both a fuzzy as well as a neural classifier have been utilizedto operate with the fault dictionary data as base. Through thisprocess, a general comparison is drawn out between theperformance of either route in dealing with fault diagnosis of circuits. An illustrative example is considered, on which both thefuzzy and neural algorithms are tested, and their performance infault diagnosis is compared. Further, the suitability of the fuzzyand neural techniques to various kinds of diagnosis problemsdepending upon the nature of data available is also discussed.
Keywords—Fault diagnosis, fuzzy logic system, neural networks,Sallen-key Bandpass filter.
The identification of faults in any analog circuit is veryuseful and, in a few instances, an inevitable measure inensuring competent performance of the circuit. In general, theanalog diagnosis approaches can be categorized into two ,namely-simulation-after-test (SAT) and simulation-before-test(SBT). The simulation-after-test - approach involves thecomputation of various circuit parameters from the operationalcircuit and fault identification is carried out using these parameters, assuming that each measurement is independent of the other. This method is avoided due to the increase in processtime with increase in the size of the circuit, in addition to issuesconcerning non-linear circuits. On the other hand, a usefulalternative is found in the simulation-before-test approachwhich appreciably reduces the time taken for fault diagnosis.In the SBT approach -, a predefined set of test stimuli areused to extract certain signatures from the Circuit-Under-Test(CUT) that are unique to each faulty condition. Thesesignatures can then be suitably systematized to create a “faultdictionary”, which is then checked for redundancies that mayresult in masking of certain faults. Evidently, the parameterschosen to pose as signatures must be quantities that areobservable for all conditions of the circuit.Both the above-mentioned approaches are fairly proceduralin nature and do not necessitate the prerequisite of anintuitional knowledge of the functioning of the CUT. Constantsupervision of the circuit is entailed to ensure stable performance over an extended period of time. Theidentification of faults in systems is often a combination of fault detection and isolation, necessarily in the same order,which is commonly known as FDI -. Early detection of faults in a circuit can greatly assist in maintenance of thesystem by avoiding possibly harmful damage borne out of thefault. Occasionally, a circuit may so damaged that it mightassume an unstable state, making it impossible to extractsignatures from it that might help in identifying the fault. Inother cases, a fault might just be too critical or dangerous to be provoked for the sake of obtaining a signature.Analog fault diagnosis is inherently complicated by poor mathematical models, component tolerances, nonlinear behaviour of components, and limited accessibility to internalnodes of the circuit under test. In this paper, we state the resultsof a comparative study of the performance of fuzzy and neuralroutes in the detection and identification of faults in an analogelectronic circuit using the Simulation-Before-Test approach.This was achieved by taking into consideration the variations intime-domain response parameters pertaining to the transient of the CUT for a step input. A comprehensive fault dictionary was prepared from all the possible values of the parameterscorresponding to each state of the circuit, which was theneffectively utilized to construct a classifier capable of identifying the various faulty configurations of the CUT.II.
The fault diagnosis methodology, involving either a fuzzyor a neural system, may be divided into five distinct steps asfollows:Step I: Formulation of transfer function of the circuit under test assuming nominal values of all components in the circuit.Step II: Simulation of time-domain response of the circuitwhen a unit step signal is given as input, for possiblecombinations of component faults.Step III: Determination of time-domain response parameters,namely, settling time and peak amplitude for each timeresponse plot.