Crashworthiness optimization of combined straight-tapered tubes using genetic algorithm and neural networks

This paper was aimed to evaluate crashworthiness capability of new designed multi-cell structures with different cross-sectional shapes (i.e. square, hexagonal, octagonal, decagonal and circular) to dissipate collision energy. The multi-cell structures included outer and inner tubes connected together by four stiffening plates. Two dimensional parameters (i.e. S1 and S2) were defined to describe cross sectional configuration of the structures. Indeed, S1 and S2 were ratios of the inner tube side length to the outer tube side length at the ends of the structures. Values of both S1 and S2 were assumed as 0, 0.25, 0.5, 0.75 and 1. Longitudinal geometry of the outer tube was straight; while, both the straight or tapered geometry could be generated for the inner tube depending on the values of S1 and S2. An experimentally validated model generated in finite element code LS-DYNA was utilized to study crushing behavior of these structures under axial impact. Geometrical dimensions of these structures were optimized using ANNs (artificial neural networks) and GA (genetic algorithm) by considering three different scenarios. The optimal structures were compared together from the crashworthiness point of view by considering two conflicting crashworthiness indicators namely SEA (specific energy absorption) and PCF (peak crush force) using a decision making method called TOPSIS (technique for ordering preferences by similarity to ideal solution). Ranking of the studied cross sections was obtained as Octagonal-Circular-Decagonal-Hexagonal-Square for all the mentioned scenarios. Consequently, octagonal multi-cell structure was selected as the best cross sectional configuration among the studied cross sections. In addition, the proposed octagonal multi-cell structure was found to have higher crashworthiness capacity than conventional single-cell and simple multi-cell ones.