Concurrent 18. Presentation for: Minimising pipeline leaks and maximising operational life by application of machine learning at Cooper Basin

Abstract

Presented on Wednesday 18 May: Session 18 The development of technologies in the last few decades has enabled operators to collect significantly more data than previously possible. Despite availability, making data-driven decisions on asset health, and developing efficient asset management strategies, is not common. This is mainly due to challenges with compilation, and alignment of all the data into a comprehensive picture of pipeline integrity, as it consumes significant resources deploying conventional methods. A critical advantage of modern data storage, analysis and visualisation techniques is the relative ease of performing statistical assessments of integrity data. Analysis of correlated data can be equally challenging as algorithms used can be overly simplistic and inaccurate. Machine learning algorithms parse, analyse and learn from data, enabling the operators to make an educated decision. This has been extensively deployed in other industries such as finance, healthcare and supply chain management but has never been fully developed and enhanced in pipeline integrity industry until very recently. This paper provides an overview of the development in machine learning tools in pipeline integrity, allowing enhancement of asset performance, through the application of machine learning and automation, to predict integrity threats, and prevent leaks and failures. It provides a case study where a tool was developed, and this technique was successfully implemented across a significant number of upstream pipelines in the Cooper Basin, enabling the Santos integrity engineering team to make the most effective decisions on asset condition and to develop a data-driven asset management plan. To access the presentation click the link on the right. To read the full paper click here