Minimization of Accident Severity Index in concrete barrier designs using an ensemble of radial basis function metamodel-based optimization

Özcanan S. , Atahan A. O.

OPTIMIZATION AND ENGINEERING, 2020 (SCI İndekslerine Giren Dergi) identifier identifier


Along with the advantages provided by the material and ease of assembly/disassembly, the ease of repair provided by minimum deformation after a collision and its sustainability highlight the preference of concrete barriers for roadside safety. However, concrete barriers, as rigid systems, are highly risky in case of a collision. Because the top priority of the application purpose is safety, it is desirable to have designs that provide the necessary safety criteria in the relevant standards and which are highly safe in terms of environment, especially for drivers. In this study, the optimum safety design of the New Jersey (NJ) concrete barrier which reduces injury levels up to the acceleration severity index (ASI) and meets safety criteria for the EN1317/2 standard, is achieved by simulation-based design optimization. For this purpose, the critical design points of the NJ type barrier have been determined. Then the critical design points were taken as variables and the safety criteria in EN1317 were taken as the objective function for multi-objective optimization (MOO). Once the variables and objective functions were determined, data was prepared with finite elements (FE) and the surrogate model was constructed using an optimal weighted pointwise of radial basis function (OWPE-RBF). Afterward, the multi-objective genetic algorithm was employed to solve the MOO problem. The optimum safety design that was obtained was compared with the validated original design, i.e. full-size modeling in the finite element (FE) environment. As a result, both the critical design points of NJ type barriers and important determinations regarding the OWPE-RBF model were obtained. Above all, a design has been achieved with an ASI/injury level which is 22-23% safer by way of the proposed analytical model and the monitored path.