The most important aspects of modern vehicle design are crashworthiness and strength requirements. Their structural analysis involves a large number of structural and safety components. The durability and crash analysis of this problem is performed with the help of finite element analysis (FEA). These FEA based simulations requires high computational power. The optimization of these vehicle structures requires many such costly simulations. The optimization of vehicle design involves more than one objective functions such a mass of the car body, occupant acceleration and toe- board intrusion. For a good vehicle design, all these parameters should be minimized simultaneously. So Multi- objective optimization is required in such cases. For reducing the overall computational cost of this multi- objective optimization, surrogate models are found to be very effective. The surrogate models approximate the expensive computer simulations and reduces the overall computational cost required for the optimization. In the literature, surrogate assisted multi- objective optimization (SAMO) algorithm are found to be very effective in handling these complex engineering problems. Some of these algorithms are based on the evolutionary algorithms (EA). In these cases it is very difficult to guarantee the convergence. In the current research an advanced surrogate assisted multi- objective optimization (ASAMO) algorithm is developed for vehicle crashworthiness applications. ASAMO algorithm ensures that most effective surrogate models are used for the objective and constraint functions. This is achieved by the MATLAB based tool box- MATSuMoTo. Most effective single and their mixture surrogates models are created by Dempster- Shafer theory (DST). This theory has a capability to handle more than one surrogate model error metrics. The effect of multiple single and mixture surrogate models on the quality of non- dominated solutions and performance metrics are studied for multi-objective optimization of vehicle crashworthiness.
Volume 12 | Issue 6
Pages: 2278-2288
DOI: 10.5373/JARDCS/V12I6/S20201187