Data-driven remaining useful life estimation of subsea pipelines under effect of interacting corrosion defects

This research presents a method for analyzing the Remaining Useful Life (RUL) of pipelines impacted by corrosion defects through the integration of Latin Hypercube Sampling (LHS), Finite Element Analysis (FEA), and Machine Learning (ML). A dataset consisting of 200 samples and 8 random variables is...

Celý popis

Uložené v:

Podrobná bibliografia
Vydané v:	Applied ocean research Ročník 155; s. 104438
Hlavní autori:	Hosseinzadeh, Soheyl, Bahaari, Mohammadreza, Abyani, Mohsen, Taheri, Milad
Médium:	Journal Article
Jazyk:	English
Vydavateľské údaje:	Elsevier Ltd 01.02.2025
Predmet:	Latin hypercube sampling Machine learning Offshore pipeline engineering Pipeline integrity management Random sampling Remaining useful life Structural reliability ECDF ANN DTR Random sampling CDF LR Pipeline integrity management ILI Structural reliability GPR KNN SVM Remaining useful life K-S MLP Latin hypercube sampling PDF LHS MAOP ML FEA MIC POF IP AI RUL MFL-TFI Offshore pipeline engineering MVMS PSO FEM PCA SGD SCC Machine learning PIMS TBRA MTTF
ISSN:	0141-1187
On-line prístup:	Získať plný text
Tagy:	Pridať tag Žiadne tagy, Buďte prvý, kto otaguje tento záznam!

Popis
Shrnutí:	This research presents a method for analyzing the Remaining Useful Life (RUL) of pipelines impacted by corrosion defects through the integration of Latin Hypercube Sampling (LHS), Finite Element Analysis (FEA), and Machine Learning (ML). A dataset consisting of 200 samples and 8 random variables is generated, representing various pipeline and corrosion defect specifications. Finite element modeling is performed using ABAQUS software and Python scripting to calculate the Failure Pressure and failure Maximum Von-Mises Stress (MVMS) under varying conditions of longitudinal spacing (Sl) and Internal Pressure (IP). This model generates a dataset that includes internal pressure, longitudinal spacing, and other relevant variables for the training and evaluation of ML models. Model performance is assessed through grid search and overfitting checks. A corrosion growth algorithm is incorporated to update input data dynamically, allowing for the prediction of future MVMS values and associated failure probabilities. The Probability of Failure (POF) is calculated, and Probability Density Functions (PDFs) for failure pressure are analyzed using standard distributions and Kolmogorov-Smirnov tests to identify the most accurate model. This approach provides a robust framework for predicting RUL by evaluating pipeline failures and probabilistic failure pressure over time, contributing valuable insights into the reliability and safety of pipeline systems under various conditions and time intervals.
ISSN:	0141-1187
DOI:	10.1016/j.apor.2025.104438