en Offshore Structure مقاله پژوهشي Research Paper <span style="font-size:8pt"><span new="" roman="" style="font-family:" times=""><span lang="EN-GB" style="font-size:10.0pt">Internal corrosion poses a significant risk to offshore pipeline operations. This study aims to utilize a combination of the Finite Element Method (FEM) and Latin Hypercube Sampling (LHS) to create a database of structural response data for corroded pipelines experiencing longitudinally interacting internal corrosion defects under internal and external pressure loading. The database includes input data such as pipeline geometry parameters, pipeline material data, corrosion defect data and loading data. This generated database will be utilized to train various advanced machine learning (ML) models to develop a predictive model capable of estimating the Maximum von Mises Stress occurring in the outermost mesh layer of a mesh ligament within the thickness of the corroded pipeline at the defected area. Such predictive capabilities of the ML model will enhance the ability to forecast leakage based on pipeline and defect specifications, thereby saving costs and time. To achieve the optimal model, various ML algorithms have been compared. Finally, to assess the prediction accuracy of the models, results of models were compared and evaluated.</span></span></span><br> &nbsp; Offshore Pipeline Engineering,Pipeline Integrity Management,Structural Reliability,Random Sampling,Latin Hypercube Sampling,Machine Learning 14 31 http://ijmt.ir/browse.php?a_code=A-10-8417-1&slc_lang=en&sid=1 Soheyl Hosseinzadeh s_hosseinzadeh@ut.ac.ir 10031947532846004430 10031947532846004430 Yes University of Tehran Mohammad Reza Bahaari mbahari@ut.ac.ir 10031947532846004431 10031947532846004431 No University of Tehran Mohsen Abayni mohsen.abyani@ut.ac.ir 10031947532846004432 10031947532846004432 No University of Tehran