Surrogate model based design optimization has been widely adopted in automotive industry. For non-linearity crashworthiness response, hybrid surrogate model which combines different surrogates is considered to be a better choice comparing with a single surrogate. In the existing methods of constructing of a hybrid surrogate model, the number of component surrogates has to be determined before-hand. In fact, it is unknown that how many surrogates are fit for specify training set. This paper systematically compare three popular hybrid modeling method – heuristic computation strategy and two kinds of optimal weighted surrogates. The objective of this study is to investigate the effect of samples size on the number of individual surrogates included into the final hybrid surrogate models for crashworthiness response, under different hybrid modeling techniques and multiple criteria, and to sum up some meaningful conclusions about the selection of component surrogates in hybrid surrogate modeling for crashworthiness optimization.