Surrogate-based multi-objective optimization of multi-cell thin-walled beams under three-point bending for battery enclosure applications

Abstract

The side-impact safety of lithium-ion battery packs in electric vehicles (EVs) is a critical design challenge owing to the limited deformation space and the high energy density of the cells. In this study, the crashworthiness behaviour of multi-cell thin-walled square tubes under quasi-static three-point bending is investigated using a nonlinear explicit finite element approach. Seven internal cross-sectional geometries combined with six wall thicknesses (0.75–2.00 mm) yield 42 design cases, which are evaluated through five crashworthiness indicators. The finite element model is validated against experimental data from the literature, and relative errors of 2.94% in peak crushing force (PCF) and 3.26% in mean crushing force (MCF) are obtained. Surrogate models based on polynomial regression are constructed, and the categorical geometric variable is incorporated using one-hot encoding. The non-dominated sorting genetic algorithm II (NSGA-II) is applied to construct the Pareto front in the energy absorption (EA)–PCF design space, and the technique for order of preference by similarity to ideal solution (TOPSIS) is utilized to select the most balanced design. The TOPSIS-selected optimum is the G5 geometry at a wall thickness of 1.83 mm, which delivers 233.80 J of EA at a PCF of 5.12 kN, corresponding to a 22.3% EA gain over the unreinforced reference at the same PCF. It can be concluded that the internal geometry and the wall thickness should be considered jointly to improve the side-impact safety of EV battery enclosure structures.