An efficient method for determining the constitutive relationships of pipeline girth welded joints based on micro-area material testing and machine learning

Accurately obtaining the constitutive relationships of materials is the fundamental requirement for structural design and safety assessment. The small punch test (SPT), as a testing method for acquiring the mechanical properties of materials in narrow areas, possesses unique advantages and great potential in characterizing the mechanical properties of pipeline girth weld joints. Taking the materials in various regions of X80 pipeline girth weld joints as the research objects, this paper combines in-situ tensile tests, SPT, finite element (FE) simulation, and machine learning (ML) methods to achieve the goal of efficiently determining the constitutive relationship of materials in various regions of pipeline girth weld joints. Specifically, the true stress-strain curves and SPT load-displacement curves of materials in various regions of X80 pipeline girth weld joints are obtained through in-situ tensile tests and SPT. And the reliability of the SPT FE model is verified with the help of experimental data. Many numerical simulation works are carried out to establish a ML database. Finally, a deep learning model (DLM) with a complex structure is established, which can realize the function of determining the true stress-strain relationships of materials from the SPT load-displacement data. This DLM performs well on both FE simulation data and experimental data, with a prediction error lower than 4.3 %. Moreover, it has a relatively fast prediction speed, enabling it to efficiently predict the constitutive relationships of pipeline steel materials and providing the possibility for practical applications in on-site engineering.