Analog Circuit Fault Diagnosis Using Multiwavelet Transform And Support Vector Machine

Since in the1960s，analo g c ircuit fault dia gnos is has been a hot topic in manyareas, because of the components parameters of ana log c ircuit with tolerance, faultmode having the characteristics of the comp le xity and diversity, make the traditio na ldiagnostic methods and practica l application ha ving certain gap. This artic le usewavelet decompos ition as the fault signa l processing tools, co mbined with intellige ntoptimizatio n algorithm and support vector machine, for better so lving the proble msencountered in the analog fault diagnosis provides a possibility.The paper conc lude on the basis of the existing fa ult diagnosis theory andmethod, aiming at the access issues of optima l samp le in fa ult d iag nosis, it ma inlystudies the application of mult iwave let transform in fa ult diagnos is. the methodobtains the optima l samp le in the feature space b y multiwave let transform, andcarries on the norma lization processin g, then get the optimal vector. The n it is inputto the trained support machine for fault dia gnos is. Multiwave let transform canovercome the d imens ion d isaster proble m s uch as wa velet tra ns form, reduce thevector dimens io ns. Simulation showed the imp le mentation of the process and theeffect of diagnosis.Aiming at the diffic ulties encountered in ana log fault dia gnos is patternrecognition, this paper proposes a partic le swarm optimization a lgorithm supportvector machine method for ana log c ircuit d iagnos is. In the method multiwave lettransform was used to extract fa ult characteristics of the circuit, and use partic leswarm algorithm parameters to optimize the structure of the support vector machine,which can determine the optima l network structure, its collection is input to theoptima l s upport vector machine network, which can get the optima l c lassificationresults. Partic le swarm a lgorithm has stronger optima l ability, it can improves thegenera lizatio n ability of network quickly. Simulatio n examp les show that thismethod can carry on the diagnosis for the fault effectively.In view of the support vector machine network structure parameters in theprocess of selection is random, and ma y affe ct the effic ienc y of fault diagnos is, thispaper introduces ant colo ny a lgorithm optimization of support vector machine tofault diagnosis. And comparied with partic le swarm optimization, partic le swarmoptimizatio n has better dia gnos is accuracy and test a ccuracy than ant colonyoptimization.