灾变环境下土石坝多源异构信息融合方法研究

For the complex and uncertain influence of cataclysmic environment, the dam safety monitoring information are obviously multi-source and heterogeneous, which increase the difficulty of information processing and decision making. Therefore, typical cataclysmic environment would be chosen for investigation, and the evolvement influence on structural condition of earth-rock dam would be described with the K-L information distance. Then the time-series model would be established to realize the real-time simulation of cataclysmic environmental influence. Subsequently, based on blind number theory, the multi-source monitoring project information would be quantitatively analyzed according to their characteristics of uncertainty, such as information obtained from field monitoring, inspection, detection and hydrologic meteorological forecast, project operation scheduling and etc. Then combined with the Petri network, the layered mixed fusion analysis model of dam multi-source monitoring blind information would be made up for the qualitative and quantitative data fusion.On the basis of above, the heterogeneous characteristics of project hazard-formation information and dam monitoring blind information would be analyzed. Then with combination of fuzzy reasoning and Bayes network technology, the fusion analytical model of multi-source heterogeneous information for earth-rock dam would be established, which would realize the information integration and reconstruction of multi-source information containing the cataclysmic environment data for earth-rock.The research has important theoretical and practical significance for risk analysis and decision making of earth-rock dam project under cataclysmic environment.

灾变环境致灾影响的复杂性及不确定性，使得土石坝安全监控信息具有明显的多源异构特征，增加了信息处理与决策的难度。为此，首先基于典型灾变环境及其对工程结构致灾影响的调研分析，采用K-L信息距离等描述灾变环境对土石坝结构性态演化的影响，并构建时序模型，实现灾变环境影响的实时模拟；继而对灾变环境下由现场监测、检测、巡查及水文气象预测预报、工程运行调度等多源渠道所获取的土石坝安全监控信息，根据其不确定性特征基于盲数理论进行量化分析，并结合Petri网络等构建土石坝多源监控盲信息的分层混融分析模型，实现工程定性与定量信息的融合；在此基础上，辨析灾变环境工程致灾信息及土石坝监控盲信息的异构特征，结合模糊推理与Bayes网络等技术，构建土石坝多源异构信息融合分析模型，实现融入灾变环境影响的土石坝多源信息的集成与重构。该项目对灾变环境下进行土石坝工程的风险分析及决策，具有重要的理论和现实意义。

灾变环境对土石坝的致灾影响具有极强的复杂性及不确定性，选取台风和旱涝急转作为典型灾变环境进行分析。采用不完备信息系统的约简方法挖掘了台风行成库区降雨的影响因子，并采用非线性分析方法构建了台风作用下的库区降雨量的时序模型；基于非饱和非稳定渗流理论，研究了典型粘土心墙坝水位骤升所引起的非稳定渗流场的变化规律，建立了渗透坡降与库水位上升历时的经验关系，对旱涝急转全过程的坝坡瞬态抗滑稳定进行了分析。针对灾变环境下土石坝安全监控信息的多源异构特征，基于均值漂移模型研究了误差数据和强影响数据的统计诊断方法，包括以模型扰动值大小为依据的误差数据t检验法，和以模型扰动对拟合参数的影响为依据的强影响数据Cook距离检验法；基于盲数理论研究了土石坝工程监控信息的盲表征方法和盲融合流程，包括因子状态盲数描述、AHP法指标因子权重确定、盲信息合并及结构归一化处理等；采用多重蕴含命题的特征展开近似推理的相关理论，研究了多源异构信息融合的多重多维近似模糊推理方法，首先对因子模糊划分和语言变量定义，构建推理大前提，然后以因子表征状态的模糊推理小前提，特征展开近似模糊推理获取融合向量，较好的避免了信息异构造成的分析困难。研究成果对提高土石坝工程的风险分析能力及土石坝应对灾变环境的决策管理水平，具有重要的理论和现实意义。依托本项目公开发表论文19篇（SCI/EI检索5篇），获得发明专利2项，获省部级科技进步二等奖1项，培养研究生3名。