Abstract: Based on the function approximation theory, the formulas for calculating surface roughness parameters are obtained by optimizing the coefficients of the function template. At first, a global approximate solution set is obtained by optimizing the initial function template using genetic algorithm. Then, the set is taken as the initial value of Levenberg–Marquardt（LM） algorithm to obtain the better local optimal solution set, and the two algorithms are used in turn until the solutions are convergent or the largest switching times are reached. At last, according to the convergence precision of the solutions and the value of each monomial, the growth or trim operation is conducted on the function template, and the two optimization algorithms are executed again; until the termination conditions are satisfied. The numerical simulation examples show the algorithm has the better ability to find the optimal solution and the good robustness. Moreover, the method is easy to carry out, and can be used in engineering practice.