Application of machine learning algorithms on VIV fatigue life assessment of multi-spanning subsea pipelines

Abdolalipour Miandoab, Mobin (2023) Application of machine learning algorithms on VIV fatigue life assessment of multi-spanning subsea pipelines. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (12MB)

Abstract

Subsea pipelines are the most efficient and reliable method of transportation of liquid, gas, and multiphase products through an aquatic medium. Pipelines laid on the uneven seabed lose contact with the seabed, leaving it suspended in some areas. The length of the pipeline hung over the seabed is called freespan. The free-spanning length of the pipeline is subjected to different static and hydrodynamic loads where the Vortex-Induced Vibrations (VIV) are among the most severe factors threatening pipeline integrity. Repetitive deflection on the free-spanning length of the pipeline imposes fatigue damage on the pipeline structure. Consequently, it results in the early failure of the structure before the expected operational life. Regarding the critical role of subsea pipelines and the uncertainty of pipeline behavior against VIV damages, high safety factors are usually applied in designing subsea free-spanning pipelines, bringing unnecessary financial burdens on offshore pipeline projects. Designing subsea pipeline projects is primarily conducted based on standard codes (e.g., DNVGL RP-F105, DNVGL RP-C203, and DNVGL RP-F114). However, accurate assessment of the free-spanning pipeline integrity against the VIV needs advanced experimental and numerical modeling with extensive cost and time impacts. In this study, machine-learning algorithms are adopted to develop a cost-effective and easy-to-implement solution for VIV-induced fatigue analysis of free-spanning and multi-spanning pipelines. The numerical model was developed based on recommendations provided by DNVGL RP-F105 and verified by numerical results from FATFREE software, which is developed by DNV for analysis of subsea free-spanning pipelines (DET NORSKE VERITAS, 2021). The comparison conducted using the data presented by Pereira et al. (Pereira et al., 2008). The employment of machine learning methods requires a high-quality dataset covering all of the possible cases. In this regard, a Python script was developed to control ABAQUS for creating different case studies and interacting with output results to perform data cleaning and preparation. The procedure of creating case studies, reading and analyzing outcomes, and cleaning and preparing results for machine learning was performed by a fully automated cycle managed by Python scripts. More than 200000 configurations of single free-span and multi-spanning conditions were analyzed. In order to have a more realistic study, pipeline characteristics, and soil properties were selected from available industrial products and trusted industrial data such as standard codes of DNVGL. The study showed that machine learning algorithms can effectively predict the fatigue life of the free-spanning pipelines subjected to VIV oscillations. Particularly, this can be of significant importance during the initial design projects for a fast and fairly accurate assessment of the fatigue lives.

Actions (login required)

View Item