Numerical simulation of large-scale structures may incur tremendous computational cost due to huge number of degree of freedom. To improve the efficiency of finite element analysis without accompanying loss of accuracy, a super element method is proposed in this paper. Based on this method, the entire structure can be divided into several zones, viz. primary and secondary zones. It is assumed that members within the primary zones exhibit nonlinear behavior due to localized extreme loading such as blast or fire events, but members in the secondary zones show linear behaviour since they are located further away from the primary zones and their responses are not critical to the survival of an entire structure. Using the proposed approach, all the secondary zones can be merged into just one super element, regardless whether they are inter-connected or not. This super element consists of all the boundary nodes between the primary and secondary zones, and an additional node which only has one degree of freedom. Thus, the total number of degrees of freedoms of the original finite element model can be reduced significantly.