Research On The Theory And Approach Of Robustness Design For Occupant Restraint System

With the steady increasing of automobile amount, traffic accidents occurfrequently. The automobile occupant restraint system can significantly reduce theharm to the body in the crash. The conventional design for occupant restraint systemis mainly focused on deterministic optimization. Since it neglects the fluctuation ofuncertainties including the design variables and the boundary conditions, the methodmay always lead to an optimal object beyond constraints. In order to get a strongrobustness of the protection, we have to make a robust design for occupant restraintsystem, which has a greater resistance to the uncertainties.On the basis of the study made by predecessors, this paper does a furtherresearch on the application of the robust theory and method for occupant restraintsystem. The main research contents and achievements are listed as follows:Firstly，the paper summarizes the current method of robust design and suggeststwo different optimization approaches for the front impact and the side impact basedon their own characteristics. Meanwhile, a six sigma mathematical optimal model isoffered combined with the quality engineering.Secondly，the front impact model and side impact PSM model are created withthe MADYMO software, and then the models are validated by crash tests. Some keyissues for the models are discussed, such as the pre-simulation of seatbelt, airbaginflation mode and the loading boundary conditions for PSM model.Thirdly，based on optimization approach nested by Monte Carlo, the variableselection method, the Kriging surrogate model and the Sequential QuadraticProgramming Algorithm are applied to the six sigma robust design for the frontalrestraint system. The optimal design is gained and compared with the baseline design.The results indicate this method can largely decline the injury level of occupants andenhance the resistance to the fluctuation.Fourthly, an Optimal Latin hypercube experiment design is performed to build asurrogate model for objective and constraint functions. Then，the Second-OrderSecond-Moment Method is employed to estimate the means and standard deviationsof the responses when taking the uncertainties of the design variables intoconsideration. Next, the Mutli-island Genetic Algorithm is applied to search theglobal optimal design and moreover the optimal result is verified by the Monte Carlosimulation technique. Engineering example indicates the SOSM method can reach ahigh accuracy and is much more effective than Monte Carlo method.Finally, the paper makes a deeper summarization for these two different robustdesign methods. Moreover, some more complex questions and deficiencies in thisstudy are introduced for further research.