结构监测系统"多维传感网络"的节点优化及其监管方法研究

In this project, the systematically theoretical researches on the key scientific problems in the "Node Placement Optimization and its Management of Multidimensional Sensor Networks (MSN) in Structural Health Monitoring System" are developed. Firstly, the principle of selecting the number of modes in sensor optimal placement is formulated and the multidimensional modal assurance criterion of node placement optimization in MSN is established based on the mutual information of load and measuring points. Then the coordinated constraint conditions on balancing the node coverage and information redundancy is proposed and the node placement optimization in MSN is given according to the improved artificial fish swarm algorithm. Furthermore, the online outlier detection of nodes in MSN is presented based on the grey relational analysis, the abnormal data classification mechanism is set up by the regional correlation analysis and the abnormal range identification algorithm based on the adaptive negative selection algorithm is given. Moreover, the global performance management model of MSN via the collective intelligence is built and the test method of its robustness and stability is provided. Finally, the above results are integrated on the MATLAB platform and the validity & effectiveness of the proposed theories and methods are verified by both numerical simulation and model experiments. The completion of the project will push the traditional node placement optimization of one-dimensional sensor networks expanding to multidimensional space, which can lay a sound theoretical foundation for satisfying the need of the large-scale distributed sensor network placement and thus have a great value in theoretical development and practical significance.

本项目针对"结构监测系统多维传感网络的节点优化及其监管"中的关键科学问题开展系统性的理论和试验研究。研究传感网络优化布置时模态数目的选取标准，建立基于荷载-测点互信息的"多维传感网络"节点优化多维模态保证准则；研究能够均衡节点覆盖率及冗余性的协调约束机制，给出基于智能算法的"多维传感网络"节点优化布设方法；提出基于关联分析的"多维传感网络"节点异常在线检验方法，建立基于区域关联的异常成因分类机制，给出基于自适应负选择的异常范围界定策略；建立基于群集智能的"多维传感网络"整体性能监管模型，给出性能检验方法。将以上成果在MATLAB平台上进行集成，通过数值仿真和模型试验共同验证所提出理论和方法的正确性与有效性。项目的完成将使传统的"一维"传感网络节点优化方法拓展到"多维"空间，为满足结构大规模、分布式传感网络布设的需求奠定坚实的理论基础，具有重要的科学意义和工程实用价值。

考虑到实际结构中激振出的振型与荷载有关，建立了考虑参数因子的模态分量加和指标，进而通过信息熵来处理考虑参数因子的模态分量加和指标的不确定性，建立了基于信息熵传感器一维优化布置准则，采用将节点的三个自由度作为一个整体，推导得到基于信息熵传感器三维优化布置准则；提出了基于条件信息熵的传感器布置方法以解决针对模态识别的传感器布置中的不确定性问题，在量化模态识别结果的不确定性时，考虑模型误差带来的Fisher阵的变化，提出了一个条件信息熵指标；提出了一种基于结构模态估计的应变计与加速度计联合布设方法，利用模态置信准则和模态信息冗余度进行初始加速度计位置选择，定义需要进行位移模态估计的位置，利用应变模态估计待估计位置的位移模态，同时利用模态估计效果好坏选择应变计的位置；提出了一种针对结构局部变形和整体模态信息同时获取的应变计和加速度计联合布设方法：将应变计布置在结构的大变形处用于监测结构局部变形，并调整应变计所在位置使其包含尽可能多的重要位移模态信息，使用应变计位置的应变模态估计结构的位移模态，并增加加速度计以改善估计所得位移模态的可区分性，同时减少获得的位移模态间的信息冗余；提出一种基于帕累托的多目标传感器优化布置策略，以模态参数识别误差最小和振动信号能量最大的两个目标建立双准则传感器优化模型，采用非支配排序遗传算法得到一组帕累托最优前沿集，可根据偏好从帕累托前沿中选出满意的传感器布置方案；针对结构监测系统传感器优化问题，在基本萤火虫算法的基础上引入等级划分策略和免疫机制，提出了一种分级免疫萤火虫算法；提出一种时-空相关性建模方法，定义一种典型相关生成算子，并通过典型相关系数是否为零，将算子划分为系统相关与系统无关两部分；对应于算子的两部分，分别定义两类统计量以实现桥梁测点监测数据的连续异常预警，并结合贡献分析定义异常变量隔离指标以识别发生异常的传感器。