地下工程地震灾害风险的贝叶斯网络评价方法及其可靠性研究

In recent years, that some serious or catastrophic accidents occurred in underground engineering resulted by earthquake risk, has drawn extensive concerns from all the circles. How to grasp the complexity of the earthquake environment as well as the complex risk factors, was such a prerequisite of the earthquake disposition, so as to control the risk within the acceptable range. Therefore, it was urgent to carry out a study on earthquake risk evaluation in underground engineering. This study mainly included: ①. After standardizing, specifying and improving the related concepts of earthquake risk system in underground engineering, earthquake risk ontology was erectted for underground engineering. ②. Through analysizing or demonstrating earthquake risk event types, risk symptom factors and its mechanism, risk identification methods was studied in earthquake-caused underground engineering risk system; ③. With the simplification of risk variables and data structure, and the constraint of minimal loss about the original data, earthquake risk evaluation model was established on the basis of Principal Component Analysis and Bayesian Network; ④. In the restriction of keeping the capability of earthquake risk judgments, earthquake risk evaluation model was established on Rough Set Theory and Bayesian Network; ⑤. Along with the design about the confidence levels and object functions in earthquake risk evaluation in the least risk control, the reliability of Bayesian Network evaluation method was contrasted under the respective condition of Principal Component Analysis and Rough Set Theory for earthquake-based underground engineering risk system.

近年来，地震致使地下工程出现的严重事故或灾难性事故，越来越引起社会各界的关注，如何把握复杂的震灾环境以及复杂多变的风险因素，从而将风险控制在可接受的范围之内，是震灾处置的前提条件，因此亟需开展地下工程地震灾害风险评估方面的研究。研究内容主要包括：①.规范并完善地下工程地震灾害风险体系的相关概念，建立地下工程地震灾害风险本体；②.分析地下工程地震灾害风险事件类型及其征候指标和致险机理，研究地下工程地震灾害风险辨识方法；③.在简化震害风险变量和数据结构且保证原始数据损失最小的约束下，研究基于主成分分析与贝叶斯网络的地下工程震害风险评估模型；④.在保持地下工程地震灾害风险判决能力不变的约束条件下，研究基于粗糙集约简和贝叶斯网络的地下工程地震灾害风险评估模型；⑤.设计支持最小风险的震害风险评估置信水平和目标函数，比对主成分分析和粗糙集约简条件下地下工程震害风险贝叶斯网络评价方法的可靠性。

按照复杂系统演变规律，集成解释结构模型、相似性及关联度计算、熵、变异系数、可靠性等方法，提出、试验、评估、改进地下工程震害风险的贝叶斯网络模型与推理方法。①应用系统论方法进行地下工程地震灾害综合风险体系的结构化分解和元素集成，建立任务、事件、要素的地下工程地震灾害综合风险三元支持结构形式化描述，提出了基于集成化管理和全寿命管理的地下工程地震灾害的动态集成化风险管理模式;②给出工程震害风险解释结构模型组件方案，探究地下工程震害致灾元素和承灾元素之间的逻辑关联和作用机理，构造统一的地下结构震害状态空间和风险概率表达方式，建立地下结构震害风险融合证据框架，挖掘地下工程地震灾害的口部区域、围岩、孕险环境的风险及其演化规律和致灾机理，得出地震灾害风险的关联因素和辨识方法；③全面考量多重震害风险要素的主成分因子规律，简化震害风险变量和数据结构且保证原始数据损失最小的约束下，构建地下工程震害风险的主成分贝叶斯网络模型；④设计保持地下工程地震灾害风险判决能力不变的约束条件，构建基于粗糙集约简的工程地震灾害风险评估贝叶斯网络模型；⑤集成主成分分析和粗糙集等方法进行地下工程地震灾害风险评估的模型试验，应用改进蚁群算法进行贝叶斯网络结构的学习，应用改进遗传算法提高粗集约简的速度和可信度，以改进地下工程震害风险的贝叶斯网络模型可靠性，为地下工程地震灾害风险及其演化规律和致灾机理进一步深化研究奠定基础。