A comparison of simplex and simulated annealing for optimization of a new rear underrun protective device | www.esteco.com

A comparison of simplex and simulated annealing for optimization of a new rear underrun protective device

Tommaso Ingrassia, Vincenzo Nigrelli, Rosario Buttitta (Universita` degli Studi di Palermo)

In this paper, through the analysis of a case study, concerning the designing of a new High Energy Absorption Rear Underrun Protective Device (HEARUPD), two different optimization approaches (simplex and simulated annealing) have been compared. The objective of the optimization process has been to minimize the decelerations measured on a car colliding with the rear part of a heavy vehicle equipped with the new rear underrun protective device (RPUD).
In the implemented optimization processes, the crash between an economy car and the rear part of a truck has been simulated by dynamic numerical (FEM) analyses. The impacts between the vehicles have been simulated using explicit nonlinear FEM analyses through the LS-DYNA code, whereas the modeFRONTIER software has been used as optimization package. The use of the simplex and simulated annealing optimization techniques has allowed to optimize the HEARUPD crashing performances but, above all, to gather useful information about their main advantages, limits and better application fields.








Moreover, authors have proposed the use of a suitable linear function of four variables with the purpose of reducing the multi-objective optimization processes to mono-objective ones. That has been made to simplify the analysis procedures without affecting the quality and the completeness of the optimization processes. The obtained results, as well as showing the high effectiveness of the integrated use of numerical crash analyses and optimization methods, demonstrate that simplex method is more effective than simulated annealing one for optimization problems where the single analysis loop requires much time. Even if the solutions are quite similar in terms of calculated values of the objective function, design and state variables, simplex method needs shorter computational time than simulated annealing to obtain an optimized solution.