IG-GaussianNoiseU-UniformNoiseIV.RESULTSANDDISCUSSIONInthispaper,theauthorsevaluatedtheperformanceofthedevelopedANNbasedclassifiersfordetectionoffourfaultconditionsofthreephaseinductionmotorandexaminedtheresults.MLPNN,pletionofthetraining,workistestedtodetectdifferenttypesoffaults.SimilarlystepsizeisvariedinSVMand0.7stepsizeisfoundtobeoptimum.TheseconfirmourideathattheproposedfeatureselectionmethodbasedonthePCAcanselectthemostsuperiorfeaturesfromtheoriginalfeatureset,andtherefore,isapowerfulfeatureselectionmethod.Alsoproposedclassifierisenoughrobusttothenoise,inthesensethatclassifiergivessatisfactoryresultsforUniformandGaussiannoisewith14%varianceininputandwith20%parativeresultsareshowninFig.12andTableII.Fig.parativeanalysisofvariousclassifierw.r.t.uracy.PARATIVERESULTSOFNNBASEDCLASSIFIERS