Development of a Framework for Restraint System Optimization, and Illustration for an Obese Anthropometry

Restraint system optimization, which is performed through computational simulations and is a state-of-the-art approach to increase the safety of motor vehicle occupants, requires a substantial computational cost, which is a disadvantage. The primary objective of this dissertation was to develop a framework for restraint system optimization that incorporates metamodeling using machine learning to decrease the number of required simulations for the optimization. The optimization framework was used to investigate strategies for increased safety of occupants with obesity, who are shown to be at a higher risk of injury in motor vehicle collisions than occupants with normal Body Mass Index (BMI). Thus, the secondary objective of this study was to optimize the restraint system for two occupants, one with obese anthropometry (BMI=35) and one with a normal BMI (BMI=25), and compare the two designs.
This study consisted of five tasks. The objective of the first task was to statistically compare the injuries of occupants with obesity and normal BMI in frontal impact cases of a field crash database. The results showed that the occupants with obesity have a higher risk of injury to the extremities and spine compared to the occupants with normal BMI. The objective of the second task was to evaluate the performance of an obese (BMI=35) human body model (HBM) in frontal impact sled tests. The obese HBM was capable of representing biomechanical characteristics of occupants with obesity, which were reported to be potentially challenging for designing an effective restraint for obese. In task three, 450 frontal impact parametric simulations with 14 different restraint parameters and two HBM types (obese, BMI=35, and non-obese, BMI=25) were performed. Then, statistical and biomechanical analyses were carried out on the simulation results to study the effects of restraint parameters on the HBMs’ responses and to compare the responses of the two occupants.
In task four, machine learning was leveraged to develop metamodels of occupants’ responses as a function of different restraint parameters in simulations. In task five, a genetic algorithm was applied to the metamodels to optimize the restraint system for the obese and non-obese HBMs. The results revealed that while most of the restraint parameters between the optimized design for obese and non-obese HBMs were similar, the main difference was that the restraint for the obese HBM included an under-the-seat air bag, which improved the occupant’s kinematics and decreased its lower extremity and lumbar spine injury risks. Several design recommendations were suggested, which should be considered to improve the safety of occupants with obesity. Also, the framework developed in this study can be used to optimize the restraint system for a variety of occupants and crash characteristics.