Summary: Due to some uncertainty factors, including noises, interferences and the coupling among variables, the similar linear correlation among columns may exist in fuzzy relation matrix. To reconstruct the input space of the fuzzy model, an algorithm based on the objective function is proposed to confirm the structure and parameters of fuzzy model for nonlinear systems. It simplifies the structure of model and deletes redundant rules. The UD matrix decomposition is used to greatly reduce the computation cost in the fuzzy modeling. The convergence proof of the presented algorithm is shown. The simulation results demonstrate the effectiveness of the proposed method.