中文摘要：

动态载荷作用时,具有应变率效应的材料在碰撞仿真中会展现出不同于静态载荷时的性能,准确的材料参数是获得可靠仿真结果的前提。主流的近似模型优化方法忽略了对变量间耦合关系的判定,造成近似模型中存在不必要的耦合项,增大误差项所占比重,降低模型的效率和泛化能力。为此,提出了解耦型数据分组处理方法（GMDH）,在建模初期判断变量之间的耦合关系,进而确定模型的耦合项。在高维非线性函数测试中,该方法表现出优良的建模性能;将该方法与台车试验结合,反求出两种材料构成的拼焊板参数,与试验结果的对比表明,该反求方法具有较高的精度。

英文摘要：

Since materials with strain rate effect showed different performances in crash simulation compared with they were under static conditions,accurate constitutive model parameters were the premise of reliable simulation results.The determination for the coupling relationship among the input variables was ignored during the construction program in traditional metamodel method,there would be coupled terms which were not necessary,on the contrary the proportion of the error term could be increased and efficiency and generalization ability of the metamodel were reduced.To deal with this problem,a decoupled GMDH algorithm was proposed：there would be a judgment process for the cou-pling relationship of all the input variables before the construction process,and the results are taken to determine whether choose the coupled term.In the test program of high dimension functions,the algo-rithm could provide accurate metamodel.Based on the trolley test,the method was implemented into match of the material parameters for two kinds advanced high-strength steel,and the results show that this algorithm owns good modeling performance.