Corrosion poses a significant and persistent challenge in the oil and gas industry. Pigging is the most widely used technology used by operators to sample corrosion progress at different times during pipeline operation. Despite improvements in recent years, measuring anomalies with inline inspection (ILI) tools is subject to significant uncertainties. One or several ILI measurements of depths along with their uncertainties and times are required to estimate the future growth of a given anomaly.

In this work, we develop a probabilistic growth model under an assumption that individual depth errors are normally distributed. Its advantage is that it is derived analytically as a natural extension of simpler linear models. As a result, it does not require costly computations to run as opposed to models based on Monte Carlo (MC). The model provides direct predictions of depths at future times without relying on growth. More generally, it outputs probability distributions of both true depth and growth based on historical ILI depths and their tolerances.

We utilize an extensive dataset to validate and analyze the model. 165K miles of ILI data and 50K of repaired metal loss anomalies from 1294 pipelines of 4 large pipeline operators were used. We compare the probabilistic model with other models used in the industry. Various aspects and advantages of its application in pipeline integrity are considered. For example, a depth probability distribution predicted by the model at a specific future time could be used when calculating probability of exceedance for pipeline burst pressures defined with the modified B31G model.