OPEN ACCESS
This research reports a vehicle occupant restraint system design by using evolutionary multi-objective optimization with response surface model. The vehicle occupant restraint systems are composed of restraint equipment, such as an airbag, a seat belt and a knee bolster. The optimization aims to improve the safety of the system by evaluating some indexes based on some safety regulations. Estimation mod- els of the safety indexes are introduced for accelerating the optimization. The estimation models, which are called the response surface models, are constructed by using Gaussian Process, which is a kind of machine learning method. The Gaussian Process constructs the estimation model from sampling results, which are calculated by using multi-body dynamics simulation. Some helpful information for designing the restraint systems, such as trade-off information of safety performance and contribution of design variables for the safety performance, is obtained by analysing the Pareto optimal solutions.
evolutionary algorithm, machine learning, multi-objective optimization, occupant safety
[1] Fu, Y., Yang, R.J. & Yeh, I., A genetic algorithm for optimal design of an inflatable knee bolster. Structural and Multidisciplinary Optimization, 26(34), pp. 264–271, 2004. http://dx.doi.org/10.1007/s00158-003-0344-1
[2] Fu, Y. & Abaramoski, E., Robust design for occupant restraint system, reliability and robust design in automotive engineering 2005, SAE Technical Papers 2005-01-0814, 2005.
[3] Horii, H., Estimate modelling for assessing the safety performance of occupant restraint systems. WIT Transactions on the Built Environment, 134, pp. 627–635, 2013. http://dx.doi.org/10.2495/SAFE130561
[4] Rasmussen, C.E. & Williams, C.K.I., Gaussian Processes for Machine Learning, MIT Press, 2006.
[5] Sasaki, D. & Obayashi, S., Efficient search for trade-offs by adaptive range multi-objective genetic algorithms. AIAA Journal of Aerospace Computing, Information and Communication, 2, pp. 44–64, 2005.