Pipeline Stress Test Simulation Under Freeze-Thaw Cycling via the XGBoost-Based Prediction Model

This study conducted ten freeze-thaw cyclic tests to clarify the effect of cycles on forces acting buried oil pipeline. The stress evolution in Q345 steel pipeline versus number was obtained. test results were consistent with COMSOL simulation different moisture contents bottom stress. Besides proposed XGBoost model, eleven machine-learning prediction models also applied 10–20 cycling tests. showed that during process, compressive at did not exceed −69.785 MPa. After eight cycles, extreme value principal -252.437MPa, i.e., 73.17% yield stress, reached. When initial content exceeded 20%, eighth cycle’s decreased remarkably. model effectively predicted pipeline’s each cycle 10 tests, R 2 = 0.978, MSE 0.021, and MAE 0.102. above fluctuated from −131.226 −224.105 values well matched experimental ones, being concert “ratcheting effect” by theory. obtained provide references for design, operation, maintenance pipelines.