Data-driven collapse strength modelling for the screen pipes with internal corrosion defect based on finite element analysis and tree-based machine learning

Abstract

Aiming to address the difficulties in predicting the collapse strength of sand control screen pipes with internal corrosion defects under the external pressure in offshore unconsolidated sandstone oil and gas reservoirs development, a parametric analysis model of a screen pipe with corrosion defects was established based on secondary development of ABAQUS by Python script, and was verified by the published experimental data. Then, the effect of corrosion defect and hole parameters on the collapse strength of the screen pipes was analyzed. Furthermore, a novel data-driven model with the gradient boosting regression tree (GBRT) method was used to explore and characterize the laws contained in the numerical simulation data based on a large volume of numerical simulation data. The results demonstrated that the proposed data-driven prediction model for the collapse strength of screen pipes exhibited a stable prediction average difference of 3.97% in the test set. And a large number of generalization practices based on data-driven prediction model indicated that corresponding corrosion parameter warning interval existed when the screen pipe design structure was determined, and when the corrosion defect size was greater than the warning interval, the collapse capacity of the screen pipe was considerably weakened. Moreover, the collapse strength was the lowest when the corrosion defect was located at the center of the screen pipe. And the collapse strength of the screen pipe gradually decreased with an increase of corrosion defects parameter and hole arrangement parameter. This study can provide a reference for evaluating the strength and establishing a safety-related warning mechanism of sand control screen pipes with corrosion defects.