Probabilistic fatigue failure assessment of free spanning subsea pipeline using dynamic Bayesian network

Xinhong Li , Yi Zhang, Rouzbeh Abbassi, Faisal Khan, Guoming Chen

Research output: Contribution to journalArticlepeer-review

44 Citations (Scopus)


The pipeline spanning is triggered by the unevenness of the seabed. The cyclic fatigue loadings from the subsequent scouring make the spanning pipeline be prone to be fatigue failure which can cause the loss of pipeline integrity and even catastrophic accident. However, probabilistic modeling of a free spanning subsea pipeline fatigue failure is a challenging task due to its dynamic characteristics and uncertain information. This paper proposes a dynamic probabilistic methodology that could capture the uncertainty and time dependence of the fatigue failure scenario of free spanning subsea pipelines. Dynamic Bayesian Network (DBN) is adopted to develop the accident scenario by finding the fatigue failure causations and derived events. The probabilities of basic causations are estimated by fuzzy set and evidence theory considering their uncertain characteristics. The established model can capture the dynamic fatigue accumulation of pipelines over their entire service life and assess the dynamic probability of fatigue failure at different time slices. Besides, the most credible causations of fatigue failure can be figured out due to the diagnostic ability of the model. A real-field case study is used to demonstrate the application of different steps of the developed methodology. It is observed that the methodology can be a useful tool to analyze dynamic fatigue failure risk of spanning subsea pipelines.
Original languageEnglish
Article number109323
Pages (from-to)1-9
Number of pages9
JournalOcean Engineering
Publication statusPublished - 15 Aug 2021


  • Subsea pipelines
  • Fatigue failure
  • Dynamic Bayesian network
  • Probabilistic assessment


Dive into the research topics of 'Probabilistic fatigue failure assessment of free spanning subsea pipeline using dynamic Bayesian network'. Together they form a unique fingerprint.

Cite this