Abstract:This paper considers the problem of parameter fitting for polynomial models. Based on the theory of optimization method, we transform the problem into a constrained nonlinear programming problem, using the fastest descent method to iterate in the calculation to solve the nonlinear problems without linearization. Firstly, the rationality of the improvement of the external point method is illustrated by analysis of the penalty factor, the objective function, and the initial value of the external point method. Secondly, using an example we explain the feasibility of the method; compared with the traditional method, this method possesses high precision and multi-solution, and provides a new idea for studying the spatial data processing theory of nonlinear models.