ABSTRACT In this paper, multi-objective crashworthiness optimisation of the new multi-corner cross-section tubes has been studied. The experimental validated Abaqus/ExplicitTM finite element simulations have been utilised to analyse the force and energy absorption–displacement of the different cross-section specimens. The results of the crashworthiness performances of the tubes for the metaheuristic optimisation algorithm of non-dominated sorting genetic algorithm II (NSGA-II) have been extracted from the machine learning process of the neural network-type group method of data handling (GMDH). The results show that the optimisation of these sections may lead to higher crashworthiness abilities without any increase in cost and difficulties in production. The comparison of the crashworthiness results between the 12 and 20-cornered cross-sections have shown that the higher the design parameters, the more the capability of the energy absorption by multi-objective optimisation. Finally, the crashworthiness performances of the optimised and the best SEA of the 12, 20-cornered specimens with the square section have been compared. It has been demonstrated that in the same IPF, the specific energy absorption in the optimised and the best SEA of the 12, 20 multi-cornered sections are about 27%, 40% and 67%, 78% higher than the square one, respectively.