A Comparison of Machine Learning Algorithms for Predicting Fatigue Life of Steel Catenary Risers
A Comparison of Machine Learning Algorithms for Predicting Fatigue Life of Steel Catenary Risers
Steel Catenary Risers (SCR) are essential components in offshore oil and gas development, yet their fatigue life assessment remains challenging due to the combined effect of vessel motion, hydrodynamic loads, and seabed soil conditions. Current dynamic methods for flexible structures, such as the lumped mass method in OrcaFlex, struggle with large deformation analysis. This study employs the Absolute Nodal Coordinate Formulation to model SCR, integrating environmental loads, vessel motion, and nonlinear pipe-soil interactions. Based on the numerical model, a database of 10,000 fatigue scenarios were constructed to train and test machine learning models for quick prediction of SCR fatigue life. In this study, traditional machine learning models, including Linear Regression, Polynomial Regression, and Ridge Regression, were compared with deep learning models, including Multilayer Perceptron, Deep Neural Networks and Transformer for predicting SCR fatigue life. The results demonstrate that the deep learning models outperform traditional machine learning methods, offering an effective tool for fatigue life prediction in offshore engineering.
Lusheng Jia; W. Wu; YI Liu; C. Zhang; Zhichao Shen
5th International Symposium on Frontiers in Offshore Geotechnics (ISFOG2025)
15 - Mooring lines, Cables, Pipelines, Immersed tunnels and Risers
https://doi.org/10.53243/ISFOG2025-328
Contact Information
- Secretary General - Dr A M McNamara
- Email: [email protected]
- Phone: +44 20 7040 8154
- City, University of London
- Northampton Square
- London EC1V 0HB. United Kingdom