ABSTRACTPost‐maintenance crack size distribution is assumed to be the same as the initial crack size distribution for conventional structural reliability analysis (SRA). This may be true for replacement; however, other maintenance methods, including crack blending and repair welding, will likely have significantly different crack size distribution. Bayesian search theory was applied to estimate the initial crack size distribution, which is critical for the SRA of aircraft structures due to its high uncertainty. The basic idea is to formulate a likelihood function that filters unlikely long or too‐short crack sizes. Both objective and subjective information were used to construct the likelihood function. The result shows that posterior crack size distribution after maintenance using the information of eddy current inspection likelihood function was comparable to field measurement data. Using the Bayesian search theory reduces the uncertainty in the crack size distribution after maintenance; thus, more accurate SRA can be conducted.