Developing a Digital Twin of the Subsea System

A digital twin of a physical system models its behaviour in its intended environment. It has found application in testing, monitoring and maintaining crucial systems large and small while avoiding the risk of any actual damage to the physical systems. Subsea laboratories offer a safe, reliable and easy way to study, visualise, monitor and test various components of an offshore oil rig without needing to be physically present or use real equipment. This paper proposes a digital twin for a double stage reciprocating air compressor at the subsea laboratory at MIT World Peace University, Pune - the Centre for Subsea Engineering Research (CSER). In this work, we propose a framework for end-to-end functioning of the digital twin, including the usage of Machine Learning tools for modules such as testing and predictive maintenance. The User Interface, 3D Simulation and ML models used are all detailed in the paper, with several machine learning and deep learning models being implemented. The best accuracy was provided by Polynomial Regression for Outlet Temperature and Pressure with a Root Mean Square Error (RMSE) of $\mathbf{0 . 0 9}$ and $\mathbf{0 . 0 3 6}$ respectively, while AdaBoost gave the best accuracy for Air Flow with an RMSE of 0.019. This system will allow virtual access to the physical system for various purposes, with work set to continue on more components from the laboratory, making simple, safe and costefficient testing and monitoring in subsea environments a widely accessible reality.