Abstract: To address the problem that traditional indoor and outdoor rock mechanics testing methods were difficult to accurately obtain macroscopic creep parameters of surrounding rock in tunnel-type anchorages within mudstone formations, which in turn affected the long-term stability assessment of engineering, the parameter inversion method based on the modified Burgers nonlinear viscoplastic model (Bugers-NVPB) was proposed. A backpropagation neural network (BPNN) was constructed to establish mathematical mapping between mudstone creep parameters and anchor plug body displacements. Combining with the non-dominated sorting genetic algorithm Ⅱ (NSGA-Ⅱ), the multi-objective optimization inversion was carried out, and the effectiveness of the proposed method was verified through practical engineering. Research results demonstrate that when using the Bugers-NVPB parameters obtained from inversion for numerical simulation, the displacement curves of the front anchor surface, mid-section top, and rear anchor surface of the anchor plug are very close to those from the 1∶10 in-situ test. The proposed approach not only confirms the effectiveness of the Bugers-NVPB model in simulating long-term stability of tunnel-type anchorages in mudstone formations but also validates the engineering applicability of NSGA-Ⅱ-based parameter inversion.

Key words: bridge engineering; geotechnical engineering; anchor plug; mud rocks parameters; NSGA-Ⅱ; multi-objective optimization

摘要： 针对传统室内外岩石力学试验方法难以准确获取泥岩地层隧道式锚碇围岩宏观蠕变参数，进而影响工程长期稳定性评估的问题，提出了基于改进伯格斯非线性黏塑性模型(Bugers-NVPB)的参数反演方法。通过构建反向传播神经网络(BPNN)建立了泥岩蠕变参数与锚塞体位移的数学映射，结合非支配排序遗传算法(NSGA-Ⅱ)进行多目标优化反演，并依托实际工程验证了该方法的有效性。研究结果表明：采用反演所得Bugers-NVPB参数进行数值模拟时，锚塞体前锚面、中段顶部和后锚面位移曲线与1∶10原位试验十分接近。该方法不仅证实了Bugers-NVPB模型对泥岩地层隧道式锚碇长期稳定性模拟的有效性，同时验证了NSGA-Ⅱ算法反演参数的工程适用性。

关键词: 桥梁工程；岩土工程；锚塞体；泥岩参数；NSGA-Ⅱ；多目标优化