Robust Design of An Inflatable Knee Bolster

Yan Fu (Ford Motor Company) - UM04

Computational analysis of occupant restraint system has become a powerful and efficient tool to reduce the cost and development time for a new product. Multi-body computer models (e.g. Madymo models) that simulate vehicle interior, restraint system and occupants in various crash modes have been widely used. To ensure public safety, many important injury numbers, such as head G, chest G, chest deflection,shoulder belt load, femur loads, neck load, and neck moment, are monitored.
In the past, deterministic optimization methods have been employed to meet various safety design requirements. With the further emphasis on product quality and the consistency of product performance, uncertainties in modeling, simulation, and manufacturing, need to be considered. 
There are many difficulties involved in the optimization under uncertainty for occupant restraint systems, 
such as 

  1. highly nonlinear and noisy nature of occupant injury numbers;
  2. large number of constraints; 
  3. computational intensity to obtain the statistic information of injury numbers by the traditional Monte Carlo method.

This paper investigates an integrated robust design approach for occupant restraint system by taking advantages of design of experiments, variable screening, stochastic meta-modeling, and genetic algorithm. An inflatable knee bolster design is used as an example to demonstrate the methodology, however, the proposed method is applicable for all occupant restraint system design problems.