Finding optimal maintenance policy for pipeline corrosion using data fusion

The overarching goal of this work is to estimate remaining useful life (RUL) for a segment of an oil and gas pipeline under internal pitting corrosion as part of the Pipeline System Integrity Management Project (PSIM). We are creating a hybrid physics-based and data-based prognostic model for pipeline corrosion. Once the model is developed, it will be used to estimate the RUL of pipeline segments and subsequently used to propose an optimal maintenance policy, considering both cost and reliability. The available pit depth prediction models in the literature are based on the assumption that operational conditions remain the same during the life of the pipeline. In this paper, we address an actual case where operational conditions change over time. In this way, we fuse together data from infrequent, full pipeline inspection with in-line inspection (ILI) tools with data from frequent, high-accuracy sensors in localized sections of the pipeline. Specifically, we define a similarity index which allows us to fuse inspection data of two different pits at two different locations. This index is defined as the average of the ratio of the estimated depth of one pit and another pit. We use augmented particle filtering and hierarchical Bayesian method to fuse available inspection data for those pits. The results will be used to inform the selection of maintenance actions and also the optimal next inspection interval.