Session: 12-04-01 Blue Economy IV

Submission Number: 157462

Machine Learning-Augmented Multi-Criteria Decision Analysis for Optimizing The
Decommissioning of Subsea Assets 

The decommissioning of subsea assets, including pipelines, equipment, and infrastructure, presents a complex and multifaceted challenge for the offshore energy industry. It demands careful consideration of environmental, social, technical, safety, and economic factors, making decision-making inherently intricate. Previously, a Multi-Criteria Decision Analysis (MCDA) methodology was developed to support holistic decision-making for subsea asset decommissioning. However, this process is often time-consuming, requiring extensive resources for data acquisition and specialized analyses. To address this limitation, this study aims to automate the decommissioning decision-making process using machine learning techniques. A machine learning algorithm was trained using data derived from multiple case studies, incorporating key variables such as field characteristics (e.g., water depth, distance to the coastline), asset features (e.g., pipe length, equipment types), and available resources (e.g., ships, port facilities). By integrating machine learning with MCDA, the proposed model seeks to rank potential decommissioning alternatives automatically, thus streamlining the decision process. This enables human experts to focus their analysis on the alternatives with the highest probability of good performance, avoiding unnecessary expenditure of time and resources on low-ranked options. Results demonstrate high accuracy in identifying optimal decommissioning options, with the machine learning model effectively replicating expert decisions. This approach significantly reduces costs by minimizing the need for exhaustive studies and reducing dependence on costly data collection resources such as ROVs. Furthermore, it allows human experts to focus their efforts on high-potential alternatives, maximizing efficiency and decision quality. The original contribution of this research lies in leveraging machine learning to enhance traditional decision-making frameworks, providing a scalable, cost-effective solution for the offshore industry's decommissioning challenges.

Presenting Author: Jean-David Caprace Federal University of Rio de Janeiro (UFRJ)

Presenting Author Biography: Jean-David Caprace has been a professor at the Alberto Luiz Coimbra Institute for Graduate Studies and Research in Engineering (COPPE) at the Federal University of Rio de Janeiro (UFRJ) since 2013. He earned his PhD in Applied Sciences from the University of Liège, Belgium, in 2010. Currently, he serves as the Director of Academic Affairs at COPPE and actively contributes to the Graduate Program in Ocean Engineering (PEnO). His research focuses on naval and ocean engineering, where he develops innovative solutions for offshore operations, port infrastructure, and international shipping logistics.