Optimization problems of chemical process can generally be formulated as nonlinear programming problems, but the objective functions and the constraints of problems are often linear for most variables. In this paper, we propose a new decomposition method for nonlinear programming problems with two types of variables; linear and nonlinear variables. The proposed method is more effective than a nonlinear optimization method without decomposition to solve problems containing a relatively small number of nonlinear variables, because the dimension of the decomposed problem is much smaller than that of the original problem. Some computational results for the Rosen-Suzuki test problem indicate that the proposed method is useful for reducing the computational time.