Based on the AMSAA model, which is centered on the failure data of the same or similar products collected during the R&D stage, the principle of minimization of discrete coefficients is innovatively introduced as an optimization criterion, which is used to guide the selection of goodness-of-fit test statistics. After integrating multivariate information such as mean and variance, a new algorithm for solving the time-environmental folding coefficients is constructed, aiming at effectively folding the original failure data and then accurately estimating various parameters of the model. By solving the resulting folding coefficients, the one-sided lower confidence limits of MTBF (Mean Time Between Failure) are calculated one by one for each of the 10 different confidence settings when the product design is being molded. Example studies show that when the confidence interval is between 0.9 and 0.99, the method consistently produces better results than the existing literature regardless of the same confidence level, i.e., the accuracy of the method is significantly improved for predicting the reliability growth of aerospace products. In addition, the values of the time-environment folding coefficients solved by the improved method do not change at different confidence levels, i. e., the quantitative relationship between the environmental stresses and the real environmental stresses for each test item does not change due to the increase in confidence level, which also proves that the improved method is closer to the engineering practice from another perspective.