machine learning image

Machine Learning for Riser Engineering Projects

Real-time operational and installation support for riser systems through leading-edge machine learning tools and industry-recognised structural dynamics expertise.

Machine learning is a key enabling technology for riser structural design, integrity management and life extension which is helping mitigate risk and improve performance and delivery time while keeping costs down. 2H Offshore has developed several machine learning tools which are helping optimise the riser engineering process during design, installation and operations.

Optimised Riser Design Process for Flexibles

Flexible riser analysis typically requires a two-step approach; non-linear global riser analysis to characterise global structural response, and local time domain finite element cross-sectional analysis using software such as BFLEX to capture pressure and tensile armour layer stresses. We have developed a more efficient single-step approach that can be deployed quickly offshore. The novel coupled FEA/machine learning model can predict riser local stress response using vessel motion and/or riser hang-off load response data.

Riser System Digital Twins for Operational Integrity and Life Cycle Management

Our experience using field data to benchmark riser design and estimate “best fit” modelling parameters has enabled the development of augmented learning from simulated and measured field/lab-scale data. The digital twin approach facilitates stress and fatigue damage tracking for long-term integrity management and life extension for all dynamic equipment on an asset such as risers, subsea wellhead, conductors, tendons, mooring lines, and subsea jumpers.

Installation and Operational Advisory Tools

We have developed tools to conduct real-time dynamic analysis and determine allowable installation windows for production and water injection risers. These windows are based on pre-trained algorithms, which rely on field measurements from the vessel and environment and are not constrained by conservative or simplified approaches used in design. This can also be used for drilling riser operational guidance for operating limits, vessel position optimisation, drift-off, and emergency disconnect advice.

Our Approach

  • Data from FEA simulations and field/lab measurements
  • Algorithm and feature selection based on data complexity
  • Unsupervised machine learning techniques including spectral, hierarchical, K-means clustering and Gaussian mixture models
  • Predictive analytics and supervised learning approaches including random forest, boosted decision trees, support vector machines and artificial neural networks
  • Automated data analytics tools for field measurement and real-time condition feedback using MATLAB and Python/TensorFlow
  • Edge and/or cloud-based deployment of GUIs for on-site visualisation


  • Backed by proven engineering and system dynamics domain expertise
  • Integrated data science expertise in-house and with strategic partnerships
  • Input data stream quality review and automated pre-processing providing improved accuracy of predictions
  • Experienced team with practical understanding of industry requirements
  • Outcome/decision-driven providing strong business value
  • Independent and data platform agnostic