Abstract: This paper presents a temporal-spatial global surrogate model that can be used for reverse analysis as a substitute for the complicated FEM. It extracts the representative base vectors of space and time for a response field through two-layer proper orthogonal decomposition, and expresses this field as a linear combination of the base states. Then, theoretical equations are derived to obtain the basis state combination coefficients of sample response fields. And for any parameters, the corresponding combination coefficients can be calculated via a nonlinear parameter-coefficient relationship that is established forehand using a radial basis function interpolation model. This method can be used to model the temporal-spatial response fields of any degree of freedom and achieve real time outputs, keeping the same number of outputs as that of a FEM model.