Abstract: As a key weak link of oil and gas pipelines， the girth weld of high-grade steel pipeline was always concerned by the engineering and scientific research communities.  As a typical welding structure， the welds had obvious heterogeneity， which led to the inability to accurately test the axial mechanics properties of girth weld materials， which seriously affected the accuracy of safety evaluation of pipeline girth welds.  Based on MATLAB-PYTHON-ABAQUS co-simulation， an optimization inversion method of the material stress-strain constitutive relationship was proposed in the weld zones for the high-grade steel pipelines.  Four groups of uniaxial tensile tests with different notch sizes were carried out， and the load-displacement curves of each sample were obtained.  The true stress-strain constitutive relationship of the weld zone materials was obtained by BRBP neural network and GWO， and the accuracy of the constitutive relationship was fully verified by the testing data， the results show that the relative error is less than 1%.  The inversion method proposed is also suitable for the determination of stress-strain curves of homogeneous metal materials in large strain ranges.  The inversion method may provide accurate stress-strain constitutive relationship and strength matching relationship for the safety evaluation of girth welds of high-grade steel pipelines， and further ensure the safe operation of oil and gas pipelines. 

Key words: notched bar tensile test, stress-strain constitutive relationship, Bayesian regularization back propagation（BRBP） neural network, grey wolf optimizer（GWO）, optimization inversion

摘要： 高钢级管道环焊缝作为油气管道关键薄弱环节一直受到工程界与科研界的关注，它作为一种典型的焊接结构具有明显的非均质性，这会导致环焊缝材料轴向力学性能无法准确测试，严重影响管道环焊缝安全评价的准确性。基于MATLAB-PYTHON-ABAQUS联合仿真提出了一种高钢级管道焊缝区材料应力应变本构关系优化反演方法。开展了4组不同缺口尺寸的单轴拉伸试验，得到了各试样的载荷位移曲线；利用贝叶斯正则化反向传播（BRBP）神经网络与灰狼优化算法（GWO）得到了焊缝区材料真实应力应变本构关系，并通过试验数据充分验证了本构关系的准确性，结果表明相对误差小于1%。所提出的反演思路同样适用于均质金属材料大应变范围应力应变曲线的测定。该反演方法的提出可为高钢级管道环焊缝安全评价提供准确的应力应变本构关系及强度匹配关系，进一步保障了油气管道的安全运行。

关键词: 缺口圆棒拉伸试验, 应力应变本构关系, 贝叶斯正则化反向传播神经网络, 灰狼优化算法, 优化反演