简介:Thispaperexaminesacomputerprogramdevelopedtoanalyzethevibrationofrotatingmachineriesbasedontheoriesofvibrationandmultibodydynamics(MBD).Bendingvibrationproblemsofrotatingmachinerieshavegenerallybeencategorizedaseitherlinearornonlinear.LinearproblemscanbeformulatedbystandardmethodsandnonlinearproblemscanbeformulatedbyMBDmethods.Inourstudy,nonlinearproblemsaretreatedbytheuseofageneral-purposecomputerprogram,RecurDyn(RD).Intheprogramwedeveloped,rotorbendingvibrationanalysis(RotB)structuralpropertiessuchasshafts,rotatingrotarydisks,unbalancedmassesandfoundationstructuresaremodeledasmultibodyelements.Also,nonlinearitiessuchascontact,non-symmetricalshafteffects,bearingcharacteristics,nonlinearrestoringanddampingcharacteristicsinthebearingsaretakenintoaccount.ThecomputationalresultsdemonstratethevalidityofRotB.
简介:Inthepresentpaper,twomodelsbasedonartificialneuralnetworksandgeneticprogrammingforpredictingsplittensilestrengthandpercentageofwaterabsorptionofconcretescontainingZrO2nanoparticleshavebeendevelopedatdifferentagesofcuring.Forbuildingthesemodels,trainingandtestingusingexperimentalresultsfor144specimensproducedwith16differentmixtureproportionswereconducted.Thedatausedinthemultilayerfeedforwardneuralnetworksmodelsandinputvariablesofgeneticprogrammingmodelswerearrangedinaformatofeightinputparametersthatcoverthecementcontent,nanoparticlecontent,aggregatetype,watercontent,theamountofsuperplasticizer,thetypeofcuringmedium,ageofcuringandnumberoftestingtry.Accordingtotheseinputparameters,intheneuralnetworksandgeneticprogrammingmodels,thesplittensilestrengthandpercentageofwaterabsorptionvaluesofconcretescontainingZrO2nanoparticleswerepredicted.ThetrainingandtestingresultsintheneuralnetworkandgeneticprogrammingmodelshaveshownthattwomodelshavestrongpotentialforpredictingthesplittensilestrengthandpercentageofwaterabsorptionvaluesofconcretescontainingZrO2nanoparticles.Ithasbeenfoundthatneuralnetwork(NN)andgeneexpressionprogramming(GEP)modelswillbevalidwithintherangesofvariables.Inneuralnetworksmodel,asthetrainingandtestingendedwhenminimumerrornormofnetworkgained,thebestresultswereobtainedandingeneticprogrammingmodel,when4geneswereselectedtoconstructthemodel,thebestresultswereacquired.Althoughneuralnetworkhavepredictedbetterresults,geneticprogrammingisabletopredictreasonablevalueswithasimplermethodratherthanneuralnetwork.