考虑损失位移及支护设计耦合优化的大型洞室群动态反馈分析

Large cavern group is huge and multi-cavities effect that makes the construction of cavern group is much more complicated than single cavern. In recent years, dynamic feedback analysis method is put forward to effectively guide the construction of large cavern group. However, lose of the monitoring displacement and support time-parameters coupling optimization are not considered. In this project, the whole displacement back analysis model is established based on intelligent optimization back analysis theory using swarm intelligence-machine learning collaborative optimization idea, which considers the lose displacement that is difficult to monitor as back analysis object. Thus the rock mass mechanics parameters can be obtained accurately and efficiently. On this basis, the support time-parameters coupling optimization model is established through three-dimensional numerical simulation of cavern group excavation and using staging dynamic feedback principle, in which the influence of lose displacement to the stability of surrounding rock and support design is considered. Then the new method for rapidly dynamic feedback analysis of large cavern group during construction period is proposed combining the early warning standard of high side wall based on risk assessment. Finally, the proposed theory and model are improved by application study on typical engineering case in western region of China. The information software system is established that can powerfully predict the next construction and reinforce the current local support. Besides that, it can help the field engineer to make a concise and rapid decision. The study results will provide scientific references for fast and accurate decision to the security construction of large cavern group and formulating construction specifications for large cavern group.

大型洞室群规模巨大、多个洞室贯通影响，使得洞室群施工远比单一洞室复杂得多。近年提出的动态反馈分析方法能够有效指导大型洞室群施工，然而对于监测位移的损失及支护时机-支护参数耦合作用未做考虑。本项目基于智能优化反演理论，采用群智能与机器学习协同优化思想，将难以监测到的损失位移纳入反演对象，建立考虑损失位移的全位移反分析模型，从而准确、高效获得岩体力学参数；在此基础上，考虑损失位移对洞室群稳定性及支护设计的影响，运用分期动态反馈原理，建立支护时机-支护参数耦合优化模型，结合基于风险评估的高边墙预警标准，提出大型洞室群施工期快速动态反馈分析新方法；最后通过西部地区典型工程应用研究，验证完善提出的方法与模型，建立能够对后期施工提供有力预测，对当前支护进行局部补强，并让现场工程师对围岩的稳定性给予简洁、快速判定的信息化软件系统。为大型洞室群安全施工的快速、精准决策以及相关规范的制订提供科学依据。

项目针对大型洞室群施工期动态反馈问题，采用数值模拟、现场监测与软件开发相结合的思路，将智能科学与工程科学交叉融合，对大型地下洞室群施工期反馈进行系统研究。主要研究内容及成果为：（1）在动态反馈分析过程中，为了准确获得岩体力学参数，基于围岩位移量测信息的参数优化反分析理论和技术成为当前借助的主要手段，反分析方法的效率、精度直接影响动态反馈分析的耗时性、准确性。基于群智能与机器学习协同优化思想，通过动态更新学习样本的策略，实现寻优过程中机器学习模型预测能力不断提升的目的，提出基于粒子群-高斯过程的全局智能优化方法。（2）洞室开挖后监测仪器到位前已经发生的位移称为损失位移，以监测位移作为优化反分析的依据是损失位移求取的有效途径。在数值模拟的基础上，将损失位移的求取问题转化为一种优化反分析问题，借助群智能与机器学习协同优化方法，将损失位移纳入反演对象，建立高效率、高精度的全位移反分析模型。（3）针对当前隧道支护结构多为钢框架及钢筋混凝土结构，开展钢框架、钢筋混凝土结构在动力扰动情况下的性能分析，提出在不同动荷载作用下合理的结构型式，为支护结构的有效优化提供参考依据。（4）大型地下洞室群施工期反馈分析的研究对象是复杂的多空间、多维度耦合的动态系统，这种系统存在着数值模型优化、施工方案优化等多种优化问题，这些问题多是全局优化问题。针对现行动态反馈分析方法的不足，提出一种基于优化前、后信息，高边墙预警标准和开挖、前后支护方案联合优化的施工期动态智能反馈分析新方法。（5）利用工程现场监测数据，完善提出的反馈分析新方法，集成各个模块，建立信息化软件平台，为大型地下洞室群的合理设计与安全施工提供科学依据。