Decision-Making Support for Reeled Pipeline Installation Using a Machine Learning Based Method

Abstract

Installation of subsea pipelines used for transportation of hydrocarbons, water, or CO2, is carried out by ship-type installation vessels, which are highly sensitive to wave conditions. The prediction of installation loads in the pipeline is an essential input to the decision-making process for safe operation during offshore execution. Predictions may be required up to five days into the future. They can be produced from a physics-based simulation model with nonlinear calculations in the time domain and probabilistic representations of the response parameters based on multiple simulations for the forecasted wave spectra. Such calculations are computationally costly and, therefore, normally produced in advance by considering a set of generic parameter-based wave spectra. This paper describes how a machine learning model can be established, verified, and used to support decision-makers during a reeled pipeline installation operation. Compared to a physics-based simulation model, this method enables computationally efficient calculation of pipeline responses from forecasted wave spectra during offshore execution to provide more accurate input to decision-makers.