This paper presents the implementation of the mixed numerical experimental technique for the finite element (FE) model calibration of a sandwich structure by using modal data. Model calibration is conducted by minimizing the difference between the numerical and experimental dynamic parameters. In this study, instead of the direct minimisation of the response discrepancy, the experiment design and the response surface method is employed to solve the inverse (calibration) problem. Numerical eigenfrequencies are obtained by performing FE calculations in sample points derived by Latin Hypercube experiment design where each sample point represents a unique engineering constant configuration in the FE model of the sandwich panel. Using the information on the dynamic responses of the panel in the sample points, response functions describing the relationship between the engineering constants and the calculated eigenfrequencies are obtained by means of response surface method. Genetic algorithm is employed to solve the minimisation of the response discrepancy where the response functions instead of FE calculations are used to obtain the numerical modal frequencies. The verification results show that proposed method is capable to calibrate a numerical model with a good prediction accuracy and small uncertainties.