Estimationofcollapsemomentforthewall-thinnedpipebendsusingfuzzymodelidentificationOriginalResearchArticle

Inthiswork,thecollapsemomentduetowall-thinneddefectsisestimatedthroughfuzzymodelidentification.Asubtractiveclusteringmethodisusedasthebasisofafastandrobustalgorithmforidentifyingthefuzzymodel.Thefuzzymodelisoptimizedbyageneticalgorithmcombinedwithaleastsquaresmethod.Thedevelopedfuzzymodelhasbeenappliedtothenumericaldataobtainedfromthefiniteelementanalysis.Principalcomponentanalysisisusedtopreprocesstheinputsignalsintothefuzzymodeltoreducethesensitivitytotheinputchangeandthefuzzymodelaretrainedbyusingthedatasetpreparedfortraining(trainingdata)andverifiedbyusinganotherdatasetdifferent(independent)fromthetrainingdata.Also,threefuzzymodelsaretrained,respectively,forthreedatasetsdividedintothethreeclassesofextrados,intrados,andcrowndefects,whichisbecausetheyhavedifferentcharacteristics.Therelativerootmeansquare(RMS)errorsoftheestimatedcollapsemomentare0.5397%forthetrainingdataand0.8673%forthetestdata.Itisknownfromthisresultthatthefuzzymodelsaresufficientlyaccuratetobeusedintheintegrityevaluationofwall-thinnedpipebendsandelbows.