Marginal Iinearization method in modeling on fuzzy control systems is proposed in this paper, in order to deal with nonlinear model with variable coefficients. The method can turn a nonlinear model with variable coefficients into a linear model with variable coefficients by means of the way that the membership functions of the fuzzy sets in fuzzy partitions of the universes are changed from triangle waves into rectangle waves. However, the Iinearization models are incomplete in their forms because of their lacking some items. For Solving this problem, joint approximation by using linear models is introduced. The simulation results in this paper show that marginal linearization models are of higher approximation precision to their original nonlinear models.