A Novel Strain Capacity Prediction Model of Girth Welded Joints of High-Grade Steel Pipeline Via Machine Learning Techniques and Refined FE Model

Abstract In recent years, incidents of fracture failure in girth welds of high-grade steel pipelines have been increasingly frequent, posing a serious threat to the safe operation of these pipelines. Therefore, it is urgent to conduct a comprehensive study on the ultimate strain capacity of pipeline girth welds. The study establishes a three-dimensional finite element simulation model for the crack driving force of the girth weld using ABAQUS, employing parametric programming to quantitatively assess the influence of material properties, geometric structure, and load parameters on the crack driving force. Via parameterized calculation and processing of Python programs and parallel computing, a database of girth weld crack driving forces with 21870 samples is constructed. Additionally, a regression prediction model for the crack propagation driving force of full-size pipeline girth welds considering the influence of multiple factors is established by using neural network technology. The results of Charpy impact tests (CVN value) serve as the fracture toughness, enabling the determination of the ultimate strain capacity of the pipeline girth weld. The model was verified by the verification data of full-scale test and wide plate test. The relative error between the experimental value and the predicted value is basically below 15 %. Comparative analysis with the evaluation results of internationally recognized strain capacity prediction models (PRCI-CRES & Exxon Mobil) reveals that the prediction model proposed in this study achieves a minimal average relative error. These research findings provide valuable insights for the design and safety assessment of girth welds of high-grade steel pipelines, and further ensure the safe operation of pipelines.