Perpustakaan Digital ITB

This researchaimstodesignandoptimiseanautonomoustram(AT)ve- hicle's crashenergymanagement(CEM)throughnumericalsimulationsand machinelearningprocessestoimproveitscrashworthiness.Inthiscase,the structure's performanceisevaluatedthroughspecicenergyabsorption(SEA) and eectivedeceleration.TheCEMsystemisequippedwithcrashboxesand a deformableupperbumper,whicharetheprimarystructuralloadpathsin absorbing energyduringacollision. The ATchassismodelisbasedona12-meterlow- oorplatformequipped with anadditionalCEMsystem.Thestructureismodelledintovariouscong- urations accordingtothedata-samplingprocess.Then,numericalsimulations with ascenariobasedonexistingcrashworthinessregulationsareperformedon all modelsusingFEM-basedsoftware.Thesimulationresultsarecollectedand used fortheneuralnetworktraining.Subsequently,theselectedneuralnet- workmodelsareemployedfortheNSGA-IIalgorithm,whichproducesaPareto frontcontainingpossibleoptimumsolutions.Lastly,theTOPSISalgorithm providesthedecision-makingprocesstoestimatetheoptimumconguration. The resultsshowthattheestimatedoptimumATstructureismadeof AA6061-T6 alloy,equippedwithanupperbumperwithathickness(tub) of1.1 mm andcircularcrashboxeswithseveralparameters:cross-sectionsize(C) of 100 mm,thickness(tc) of3.3mm,anddouble angecutat3H. Furthermore, this congurationimprovesupto193.109%fortheSEAand23.618%forthe eectivedecelerationofthebaselinemodel.