Artificial neural network metamodeling-based design optimization of a continuous motorcyclists protection barrier system

Abstract

Longitudinal barriers are considered as passive safety systems designed to shield hazards located at roadsides against motor vehicle impacts. Since these barriers are manmade obstacles, they also pose a threat to drivers using the road. Recent motorcyclist accidents with longitudinal barriers have proven that a particular barrier successfully protecting vehicle occupants may wound or kill motorcyclists due to its components. For this reason, sharp and blunt edges in steel longitudinal barrier parts, such as posts are usually shielded against contact from unprotected motorcyclists during a high-speed impact event. In recent years, more longitudinal barriers have been designed with motorcyclists in mind and these motorcycle protection barriers have become wide spread especially on urban high-speed roads. However, since the development of these barriers are fairly new compared to conventional longitudinal barriers, there is limited guidance on their design criteria, such as thickness, geometry, connections. For this purpose, this paper intends to provide an artificial neural network metamodeling-based design optimization methodology to an existing continuous motorcycle protection barrier design to make it more competitive in terms of weight and thus, cost. As a result of this study, the optimized barrier has become 34% more economical compared to its original design while its protection level remained intact.