A probabilistic approach is proposed for reliability analysis of composite honeycomb sandwich panel under air-blast loading in the paper. Firstly, an adaptive method based on universal Kriging model is developed to estimate the probability of failure. The U learning function is used to adaptively update finite element model (FEM). The weighed K-means clustering strategy is added to accelerate the convergence speed. The updated universal Kriging model is combined with Monte Carlo simulation (MCS) for reliability analysis. Secondly, the FEM is established and the results of deformation and energy absorption are discussed. Thirdly, a three-layer sandwich shell is analyzed to verify the accuracy and efficiency of the proposed approach. Finally, based on the results of MCS, compared with adaptive support vector regression (SVR), polynomial chaos expansion (PCE) and ordinary Kriging approach, it is found that the accuracy and efficiency of the proposed adaptive universal Kriging approach are the best for reliability analysis of composite honeycomb sandwich panel under air-blast loading.