Study on the collapse behavior of multi-cell conical structures and their optimization using artificial neural networks

In the present research, the collapse behavior of multi-cell conical structures has been studied under axial dynamic loading. These conical structures consisted of two inner and outer walls which have been connected together by several plates as stiffeners. These structures were assumed to have five different cross-sections of square, hexagonal, octagonal, decagonal and circular. Before performing the numerical simulations using LS-DYNA, the numerical results were validated by experimental results. After ensuring correctness of the created finite element models, indicators of SEA and F_max were then computed for all the structures to find the best structure from the crashworthiness point of view. The artificial neural networks and genetic algorithm methods were used to obtain the optimized dimentions of the mentioned structures including θ (cone angle) and S (ratio of the inner wall size to the outer wall one). Among the optimized structures, the best structure was selected using the decision making method called TOPSIS. The multi-cell conical structure with circular cross-section having dimensions of S=0.578 and θ=3.94°, was found to perform the best in terms of crashworthiness capability. Effect of triggers (inner wall and stiffeners) was finally studied, and the results revealed that the triggering by inner wall had a suitable result.