Fast and accurate prediction of failure pressure of oil and gas defective pipelines using the deep learning model

•A deep neural network (DNN) is developed for predicting failure pressure of defective pipelines.•The sample data are obtained from FEM simulations and burst experiments.•Prediction accuracy of DNN is far better than that of empirical formulae.•Computational efficiency of DNN is at least two orders...

Full description

Saved in:
Bibliographic Details
Published inReliability engineering & system safety Vol. 216; p. 108016
Main Authors Su, Yue, Li, Jingfa, Yu, Bo, Zhao, Yanlin, Yao, Jun
Format Journal Article
LanguageEnglish
Published Barking Elsevier Ltd 01.12.2021
Elsevier BV
Subjects
Online AccessGet full text
ISSN0951-8320
1879-0836
DOI10.1016/j.ress.2021.108016

Cover

More Information
Summary:•A deep neural network (DNN) is developed for predicting failure pressure of defective pipelines.•The sample data are obtained from FEM simulations and burst experiments.•Prediction accuracy of DNN is far better than that of empirical formulae.•Computational efficiency of DNN is at least two orders of magnitude faster than FEM simulations. Determining the failure pressure of defective pipelines is an important part in pipeline reliability engineering, which affects the assessment of pipelines residual service life. In this work, a fast and accurate method for predicting the failure pressure of defective pipelines using the deep learning model is developed. The calculation results of ASME-B31GM, DNV, PCORRC codes and finite element method (FEM) are compared and analyzed in detail to obtain high-quality sample data. 150 groups of validated FEM simulation data and 142 groups of burst pressure test data are selected for the training and validation of deep learning model. In the training process, influences of key model parameters of deep neural network (DNN) on the prediction accuracy are investigated. Prediction results indicate that the used deep learning model can offer high prediction accuracy. In addition, the calculation of deep learning model is accelerated by at least two orders of magnitude compared with that of FEM simulations under same calculation conditions. Finally, the influence of defect sizes on pipeline failure pressure is analyzed by the DNN. This work is expected to shed a light on efficient and accurate predictions of failure pressure of oil and gas defective pipelines.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0951-8320
1879-0836
DOI:10.1016/j.ress.2021.108016