In order to address the problems of high modeling difficulty, poor accuracy and low computational efficiency in the process of reliability analysis of complex aerospace agencies, a method combining data augmentation Latin hypercube sampling (DALHS), adaptive partitioned threshold rejection sampling (APTRS) and active learning Kriging is proposed for agency reliability analysis. First, the data enhancement technique is used to improve Latin hypercube sampling to obtain initial sample points and improve the diversity and representativeness of the initial sample points; second, an adaptive partitioning strategy is used to divide the design space and perform rejection weight sampling within the subspace to improve the local and global search capability of the samples; third, the active learning NU (Normalize U) function is proposed to screen high-quality samples, combined with the quasi-random fractal algorithm (QRFA) to dynamically optimize the Kriging model, and construct the DALHS – APTRS – Kriging model; finally, we use the convergence criterion of the coefficient of variation to realize the efficient calculation of the reliability of the aviation mechanism. The results show that the mechanism reliability of the general aviation piston engine is 0.987, with only 72 model calls, and a calculation error of only 5.7% compared to traditional methods. This indicates that the proposed method can not only obtain high-quality Kriging models with a small number of samples but also improve the efficiency and accuracy of reliability calculation by combining local and global search.