The paper reports the development of a Risk-Based Inspection (RBI) - Machine Learning perspective. The Optical Emission Spectrometry (OES), Tensile and Hardness Test, Scanning Electron Microscope (SEM), Energy Dispersive X-Ray Spectroscopy (EDS), Sulfate Reducing Bacteria Check, and X-Ray Diffraction (XRD) was used to analyze the root cause of the pipeline's failure. Corrosion attack shows at the cross-section microstructure based on SEM results. Carbon, Manganese, Phosphorous, and sulfur's chemical composition is dramatically lower than the standard API 5L Grade X42. Siderite and hematite dominate the composition of the corroded area as a result of CO2 dissolving in water. In contrast, hematite is generated due to the pipe and outdoor atmosphere reaction. Severe local wall thinning of the sand abrasion causes the degradation of the material's mechanical properties and increases the corrosion rate. This result amplifies by the development of Machine Learning (ML) of Pearson Multicollinear Matrix and Supervised ML (Random Forest, Support Vector Machine, and Linear Regression) to estimate the corrosion degradation of the material. The source of datasets provided by ILI inspection includes the calculated PoF Remaining Useful Life (RuL) as input data, while Probability of Failure (PoF) prediction serves as output data. The Random Forest shows superior predictions of 92,18 %, with the lowest validation loss of 0,0316. The modeling result confirms the experimental outcome. This work demonstrates the implementation strategy to reduce the analysis time, minimize human bias, and serve as a reliable reference tool and guideline to maintain the integrity of the subsea pipelines.

 Наукова періодика України 

---

  Якщо, ви не знайшли інформацію про автора(ів) публікації, маєте бажання виправити або відобразити більш докладну інформацію про науковців України запрошуємо заповнити "Анкету науковця"