Perpustakaan Digital ITB

This study aims to design and optimize the cross-car structure of an autonomous tram vehicle (AT) using numerical simulations and machine learning to improve its crashworthiness. This study explores different dimensions and geometries by evaluating two performance parameters of structural behavior, namely specific energy absorption (SEA) and maximum stress experienced by the battery. In this research, a finite element simulation will be carried out on the AT chassis structure with side impact cases. Finite element simulations for side crashes for various configurations are made as input material for the optimization process using machine learning methods. The optimization process is carried out through the stages of ANN, NSGA-II, and TOPSIS. ANN is performed to produce a regression model for each parameter. NSGA-II is performed to estimate possible solutions for predetermined limits and objectives, while TOPSIS is used to select the most optimal solution from these possible solutions. The results of machine learning estimation for the optimum design are tested using the finite element method as validation. The optimization results show that the optimum cross-car structure configuration estimate is made using AA6061-T6 material, section C with a cross-sectional size of 80mm, thickness of 3mm, and half of the distance between the batteries is 5mm. This configuration increases performance compared to the initial model at the resulting SEA value of 44.4% with stress experienced by the battery of 61.7 MPa.