In order to fulfill the potential of resistance wall in filling capacity and extending die life，it’s very important to find the optimal combination of structural parameters. The multi-objective optimization was studied using combinatorial optimization strategy which was radial basis function approximation model methods (RBF) combined with genetic algorithm (GA).Latin hypercube sampling design was applied to select the appropriate design parameters for the FEM experiments. The RBF meta-models were obtained after the comprehensive comparison of the fitting accuracy of five different kernel functions. Then the GA was introduced into meta-models to carry out multi-objective optimization. The optimal parameters of the resistance wall were determined，such as height h=48 mm，inclination a=0，clearance n=2.5 mm，flash thickness t=5 mm and bridge widthb=10 mm， which were acquired by weighting function optimization.
