Research On The Theory And Approach Of Multi-parameter Optimization For Occupant Restraint System

This subject is carried out with the support of the National Natural Science Foundation (51275164), named robust optimization for human injury based on global sensitivity analysis and metamodel of fitting basic function in vehicle collision.The design of occupant restraint system is an important content of safety design for vehicles. In the research and development of occupant restraint system, it is an effective method to shorten the product R&D cycle and improve the product performance by combining the computer simulation with optimization design. Vehicle collisions are complicated response period of the system with multi-parameters, in which the injury of the occupant is not only directly related to the performances of some devices such as seat-belts and air bags, but also related to boundary conditions such as car body acceleration and occupant seating position, etc. With the number of system parameters increasing, the cost for the calculation of optimization design rises sharply. So how to identify important parameters in the highly nonlinear system and achieve the optimization quickly in concept phase makes a big sense for improving the vehicle’s safety performance and shorten the development cycle.In view of the above problems, a strategy for the optimization design of the complex nonlinear system with multi-parameter is proposed, which is applied to the optimization design of the occupant restraint system in100%frontal crash.20parameters from the four assemblies including curve, seatbelt, airbag and interior are studied as design variables. In view of the characteristic of multi-parameter, a method of global sensitivity analysis based on divided variables in groups is adopted. By making relevant variables in one group, global sensitivity analysis is conducted in grouped variables and the importance of parameters is evaluated by calculating the contribution value of each parameter to the total variance of system response. The described Monte Carlo simulation is applied to sampling in the entire design region during the analysis period and the sensitivity of variables was analyzed by using the metamodel instead of using the simulation one. Secondly the results of the analysis were applied to HAM(Hybrid and adaptive metamodeling Method). This method integrates three different metamodels, including quadratic function (QF) model, Kriging model and radial basis function (RBF) model, among which the appropriate one can be selected automatically according to the specific problem. A certain number of the samples is selected to update and reconstruct the model in the optimization process regularly and the value of the samples is set to the optimal solution. Meanwhile, the design space is partitioned according to the rank of the value and priority region is constructed.The accuracy of the region is improved by sampling to find the global optimal solution quickly. Finally, considering the possible fitting error of the metamodel, three samples whose value is the smaller in all samples are applied to construct the key region. The metamodel is applied to search in the key region to complete the optimization design.The results show that the strategy based on the global sensitivity analysis and hybrid metalmodels in this paper is efficient for the optimization of occupant restraint system with multi-parameter. The important parameter can be identified quickly by the global sensitivity analysis based on variance. Meanwhile the hybrid metamodels techniques can rapidly, economically and accurately solve the problems of occupant restraint system optimization by breaking the limit of the single metamodel, which is referred by the optimization design of complex nonlinear system.