Matching pipeline In-line inspection data for corrosion characterization

Abstract In-line inspection (ILI) is critical to pipeline integrity management program in oil and gas industry. The corrosion leading to potential pipe leakage and failure needs to be detected through periodical ILI and trigger the alarm for further field investigation. The proactive management strategy relies on the capability of predicting the corrosion growth rate over time. To determine the growth of corrosion, the corroded areas on the pipeline need to be matched with the extracted features from the data acquired through the ILI performed at different times. However, manual matching is time consumption, labor intensive, and prone to errors. This paper proposes the automated methods to match multiple ILI data. Informative features are extracted from ILI data and input to the machine learning models. Both individual- and ensemble-based machine learning models were investigated in this study. The experimental results demonstrate an accurate matching can be achieved with the ensemble learning method. Thus, the characterization of matched corrosion will further contribute to pipeline integrity management and risk analysis.