Uncertainty Optimization of Thin-walled Beam Crashworthiness Based on Approximate Model with Step Encryption Technology

Crashworthiness is one of the most important performances of vehicles, and the front rails are the main crash energy absorption parts during the frontal crashing process. In this paper, the front rail was simplified to a thin-walled beam with a cross section of single-hat which was made of steel and aluminum. And the two boards of it were connected by riveting without rivets. In order to optimize its crashworthiness, the thickness (t), radius (R) and the rivet spacing (d) were selected as three design variables, and its specific energy absorption was the objective while the average impact force was the constraint. Considering the error of manufacturing and measurements, the parameters σ_s and E_t of the steel were selected as the uncertainty variables to improve the design reliability. The algorithm IP-GA and the approximate model-RBF (Radial Basis Function) were applied in this nonlinear uncertainty optimization. In order to improve the accuracy of the RBF model, a new step-encryption technology was proposed, in which the encryption points will be added to the current sample points according to the results of each iteration. As a result, when the uncertainty level was 5%, the optimal design vector [t, R, d] was [1.75mm, 3.25mm, 29.48mm], and the possible interval of the specific energy absorption was [841J/kg, 1028J/kg] while the possible interval of constraint was [49.12KN, 71.39KN]. And the optimum was verified with the exact solver. Therefore, this study can provide important references for the crashworthiness design of the front rails.