Application of Machine Learning for Fatigue Prediction of Flexible Risers - Digital Twin Approach
- Nitin Repalle (2H Offshore Engineering Ltd) | Ricky Thethi (2H Offshore Engineering Ltd) | Pedro Viana (2H Offshore Engineering Ltd) | Elizabeth Tellier (2H Offshore Engineering Ltd)
- Document ID
- Society of Petroleum Engineers
- SPE Asia Pacific Oil & Gas Conference and Exhibition, 17-19 November, Virtual
- Publication Date
- Document Type
- Conference Paper
- 2020. Society of Petroleum Engineers
- 7.6.7 Neural Networks, 6.1.5 Human Resources, Competence and Training, 6 Health, Safety, Security, Environment and Social Responsibility, 7 Management and Information, 6.1 HSSE & Social Responsibility Management, 7.6 Information Management and Systems, 4.2.4 Risers, 4 Facilities Design, Construction and Operation, 4.2 Pipelines, Flowlines and Risers, 7.6.6 Artificial Intelligence
- Fatigue damage, Flexible Riser, Digital twin, Machine learning, Asset life extension
- 16 in the last 30 days
- 19 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 5.00|
|SPE Non-Member Price:||USD 28.00|
Flexible pipes have a range of potential failure modes, however fatigue damage of the tensile, and eventually, the pressure armour, is one of the most common problems affecting the longevity of service life and the OPEX due to the common need for flexible riser replacement. With increasing utilisation of flexible pipe for current and future field developments, compounded by the recurrent need for field life extension, it is essential to monitor the riser fatigue regularly to maintain integrity, maximise asset life and to allow for informed appraisal before extending its operational life.
This paper presents a novel method of using the refined finite element analysis (FEA) in combination with Artificial Neural Network (ANN) to develop a riser digital twin that can be utilised as an operational decision-making tool for integrity management and life extension. A digital twin model is trained on a subset of available metocean and vessel motion data utilising advanced neural networks which can then be utilised to predict fatigue under the full spectrum of metocean and internal pressure conditions. This approach allows for a significant reduction in the estimation time of the fatigue damage compared to conventional FEA as well as improved accuracy of prediction.
The methodology presented in the paper has been primarily developed with the view of deepwater riser applications but is easily adaptable to shallow water application in combination with various floating vessels. A case study is presented to demonstrate how this technology is being deployed offshore. A comparison of FEA and digital twin approach is also presented to highlight the speed and efficiency of digital twin model whereby real-time insights on fatigue life can be evaluated for informed operational decisions.
|File Size||1 MB||Number of Pages||14|
H. L. Fawaz, G. Forestier, F. Weber, L. Idoumghar and P. Muller: "Deep Learning for Time Series Classification: a review", http://dx.doi.org/10.1007/s10618-019-00619-1. 2019.