【作者】 姜谙男；

【导师】 冯夏庭；

【作者基本信息】 东北大学 ， 工程力学， 2005， 博士

【摘要】 大型洞室群规模已超出现行规范,本身结构比较复杂,又处于复杂的地质环境中,因而影响围岩稳定性的因素错综复杂。如何准确地根据监测信息快速动态调整施工方案,是有待解决的重要问题。本论文以清江水布垭电站地下厂房作为工程应用背景,综合三维数值计算、智能方法和计算机决策技术,提出了大型洞室群施工期集成智能动态反馈分析方法,并用于水布垭地下厂房施工期的方案反馈分析与优化。具体来说,主要进行了如下的工作: (1) 针对经验公式方法的不足,提出基于工程实例的隧洞最大允许变形的支持向量机(SVM)自动获取方法。 (2) 针对以往反馈优化方法容易陷于局部最优及耗时的问题,提出了大型洞室群反馈分析的3D并行进化SVM-数值模拟优化方法,并且基于MPI开发了相应并行计算平台。该方法能够全局快速搜索最优解,与并行神经网络进化有限元方法相比,计算速度提高近十倍。 (3) 地下大型洞室群锚固参数优化具有以下特点:稳定性目标和经济性目标是矛盾的;洞室的稳定性分析非常复杂;评价指标要求全面反映锚固效果;锚固参数组合的方案数量多,优化计算量大。针对这些特点,建立了地下大型洞室施工锚固参数优化模型,确定了约束条件及优化的指标,给出了优化的方法和步骤。采用主成分分析法处理多指标评价问题,既可以在保持评价系统信息、达到降维,简化评价系统的目的,又可以避免权重确定的人为因素。 (4) 建立了地质描述模糊评判和变形速率比值判别知识库。在课题组原有推理机和知识库基础上,提出了基于IDSS的大型洞室群施工期反馈分析的集成智能分析系统的模型。进行了软件系统的部分开发工作,并将该系统初步应用于水布垭电站地下厂房施工期反馈分析。 (5) 以三维弹塑性数值模拟开挖支护为基础,将多种反馈分析的方法,包括从工程实例获取隧洞最大允许变形的SVM方法、围岩变形SVM时间序列分析、地质描述模糊评判法、变形速率比值判别法、基于3D并行进化SVM-数值模拟方法等进行集成。建立了清江水布垭电站地下厂房水施工期动态集成智能反馈分析流程,提高分析结果的可靠性。 (6) 在水布垭电站地下厂房施工期间,利用开挖监测信息,基于3D并行进化SVM-数值模拟的方法进行该地下厂房围岩参数的反演。利用反演得到的岩更多还原

【Abstract】 The scale of large cavern group is beyond current regulation. The structure of cavern group and the geologic environment where is located in are complicated, therefore, the factors affecting the surrounding rock stability are complicated too. How to dynamically and rapidly redesign the support system based on monitoring data? That is a very important problem need to be solved. In the dissertation, aiming at the underground powerhouse engineer of Qingjiang Shuibuya hinge, the method of integrated dynamic feedback analysis in construction for large cavern group is proposed, which has been used in the feedback analysis and optimization of support schemes of the Shuibuya underground powerhouse. Concretely, such works carried through as below.(1) Aiming at the disadvantages of experience formula, the case-based SVM method for maximal deformation forecasting of surrounding rocks of tunnels is proposed.(2) Because of the local optimization and time wasting of conventional feedback optimizing method, a new 3D parallel evolution SVM-numerical simulation optimizing method is proposed to optimize large cavern group anchor parameters. Based on MPI, its computing platform is developed. Using this method, the global optimal solution could be gotten rapidly, and the computing velocity improves 10 times more than the ANN-finite element method.(3) The optimization of large underground cavern anchor parameters has some characters such as : the object of stability is incompatible with the object of economics,the analysis of large underground cavern’s stability is very complicated, appraising guide lines are required to denote the anchor effect roundly, a lot of schemes of different anchor parameters will occur and a great deal of calculate work is needed. Aiming at the characters, the anchors optimizing model is constructed to optimize large cavern group anchor parameters, the restriction condition, optimization indexes and optimizing steps are decided. The main component analysis method is used to deal with multiple indexes appraisement, not only the system information can be kept but also the system can be predigested, avoiding the artificially deciding the weights of indexes.更多还原