Abstract: A mass of aerodynamic data obtained with experiments, computations and theory methods are required in an aerodynamic database construction or aerodynamic sharp optimization for hypersonic vehicle. The multi-fidelity data fusion method, i.e., variable complexity model (VCM) is proposed to obtain high quality aerodynamic data with low price. In this paper, it is found that small deviations can led to the fusion data oscillation, the loss of the fidelity, and deterioration of the data quality when original data is near to zero. The adaptive preconditioning method for VCM is proposed to solve this problem. The preconditioning parameter is adaptively computed to shift the original data to the values away from zero. The new method is validated by applying to the construction of 6 degrees of freedom aerodynamic design data base for a hypersonic vehicle. The wind tunnel experiment data, CFD computation data, and theory computation data are fused in this new method, which leads to the consistently high fidelity aerodynamic design data base.