Shape optimization of Star Crash Box Using Particle Swarm Optimization and Bayesian Optimization

Abstract

Crash box is energy-absorbing component to ensure the passive safety of vehicles during frontal crash. For crash-boxes - lightweight design, safety requirements absorbing energy are relevant. This research aims to determine the value of geometric design parameters in design space of star crash box optimizing specific energy absorption (SEA). The geometric modelling, meshing and finally input file for LS dyna is created using python scripting. Crash simulation is performed in LS Dyna. The energy absorption is taken from glstat of binout file and mass is taken from massout file. The particle swarm optimization is done using skopt python module. The geometric design parameters used are height (a), width (b), x-intrusion (u), y- intrusion (v) and thickness (t). For each simulation reference material Mild steel with density 7830 kg/m3, Young’s modulus 200 GPa and cowper-symond parameters c = 40s-1 and p = 5 is used. The impactor of 250 kg mass with speed of 15 mm/ms is used. The values of geometric parameters in baseline geometry of star crash box were a= 90 mm, b = 90 mm, u = 15 mm, v = 15 mm and t = 1.85. From the simulation result, the SEA of 37948.57 J/Kg is obtained. Using PSO optimization algorithm, 480 simulations was run in batch mode. The maximum SEA of 63777.547 J/Kg was obtained at the values a = 60 mm, b = 112.420 mm, u = 13.876 mm, v = 0 mm and t = 0.987. In case of Bayesian optimization, the simulation is run for 129 in batch mode to obtain the optimized value SEA is obtained to be 68897.182 J/Kg at geometric parameter values of a= 72.291 mm, b=75.314 mm, u = 20.162 mm, v = 4.978 mm and t = 0.985. Both the optimization gives significantly more value of SEA than that of baseline geometry.