In this study a general evolutionary algorithm is used to minimise the mass of a bumper beam according to FMVSS pendulum test specification (repair test). In contrast to most of the studies realised in the past, the focus is here on true shape optimisation with variation of the cross-sectional parameters. Special interest is spent in this paper on evaluating the ability of the algorithmic approach to obtain the global optimum. For this, a combined, i.e. sequential, global and local search strategy is employed. Finally, a sensitivity analysis of the optimum is performed again investigating the influence of the new type of shape optimisation parameters. In addition, the representation of different crash scenarios by the bumper legislation test is investigated. The results show (through a routine) that the identified optimum was with high probability the global minimum. The optimisation result is obtained even by repeated realisations and it is robust, i.e. the influence of small variations in noise variables (e.g. impactor shape and orientation) on the outputs is small. However if other crash scenarios are investigated represented by larger variations of impactor shape and orientation, the identified optimal design loses its quality and especially violates the deflection constraint. This indicates, that designing for a pre-defined repair test does not guarantee low repair costs in all relevant cases. Here, it is recommended to establish a hybrid approach for repair cost assessment based on partially virtual and partially physical testing. Hence, this paper concludes with few proposals for changes or addition in repair and legislation tests.     «

In this study a general evolutionary algorithm is used to minimise the mass of a bumper beam according to FMVSS pendulum test specification (repair test). In contrast to most of the studies realised in the past, the focus is here on true shape optimisation with variation of the cross-sectional parameters. Special interest is spent in this paper on evaluating the ability of the algorithmic approach to obtain the global optimum. For this, a combined, i.e. sequential, global and local search stra...     »