Abstract: The distribution regularity of pipeline corrosion defect size not only has an important effect on the calculation accuracy of pipeline reliability, but also can predict the remaining life of corroded pipelines more accurately. In this paper, in-line inspection data of 25 pipelines of one domestic pipeline company was taken as the research objects. Based on the probability theory and the principle of mathematical statistics, their corrosion defect sizes were fitted by means of the commonly used distribution functions, such as normal distribution, lognormal distribution, Gumbel distribution, Weibull distribution and exponential distribution. Then, the corresponding distribution parameters were calculated using the maximum likelihood estimation method. Finally, the priority of different distribution types was compared on the basis of K-S test and P-P scatterplot, and their applicability to the corrosion defect depth, length and width was discussed. It is indicated that the lognormal distribution can better describe the distribution regularities of depth, length and width of corrosion defects of pipelines.