基于分布式监测数据的信息延拓与融合决策方法

The structural health monitoring system is an effective way to solve the selection on the design parameters of the large complicated structures and structures with novel architecture, the control in stresses and deformations of the structure in its construction period, as well as the tracking on structural responses to the structure in its construction and working periods and so on. However, because of the mass data from measurements, the uncertainties and wide sources in measurements, and the diversities in types of sensors and the limitation in the number of the placed sensors, how to use the measurements effectively becomes one of the most important problems for development in structural health monitoring system theories and the applications. The information prolongation and fusion decision methods based on distributed measurements are proposed and researched in this proposal, which is aimed to solve the scientific problems such as stratified multi-level management mode of the structural health monitoring system, reliability and integrity of the measurements, synthesized evaluation methods with fusion to various aspects. The major research schemes are firstly to build and analyze the distributed measurements subsystems, propose the associate model for checking the reliability of the measurements and give the data diagnosis method. Second research scheme is to research the structural response information prolongation based on the limited measurements. Third research scheme is to propose the stratified multi-level fusion decision method on synthesized structural performances based on the measurements from various types of the sensors and multiple distributed measurements subsystems. This proposal can give the theoretical and technical basis for the structural health monitoring system with cost-effective characteristics, stable operations, and reliable measurements.

结构健康监测系统的研究为大型复杂结构与新型结构的设计参数验证、结构施工过程中的应力与变形控制，以及结构在施工与使用过程中的结构响应跟踪等提供了有效的技术途径。然而，结构健康监测的海量数据具有不确定性、类型多、测点少、来源广等特点，如何充分运用监测数据，成为结构健康监测系统理论发展与推广应用的主要问题之一。本项目以结构健康监测数据为基础，研究基于分布式监测数据的信息延拓与融合决策方法，旨在解决结构健康监测系统的分层多级管理模式、监测数据可靠完备、多角度融合的综合评价方法等科学问题。主要研究内容包括：建立分布式监测数据子系统理论模型，提出监测数据可靠性关联模型，给出筛选有效数据的方法；研究基于有限监测数据的结构响应信息扩展与延拓方法；通过多类型传感器与多个分布式监测数据子系统，研究结构综合评定的分层多级融合决策方法。为发展性价比高、运行稳定、数据可靠的结构健康监测系统奠定理论和技术基础。

本课题以结构健康监测数据为基础，研究了基于分布式监测数据的信息延拓和融合决策方法，提出了结构健康监测系统的分层多级管理模式，解决了监测数据的可靠性判定，给出了多角度融合的综合评价方法，并实现了理论方法在实际结构健康监测工程项目中的应用与验证。主要取得以下成果：.(1) 建立了基于多智能体的结构健康监测系统搭建策略。提出了多智能体法方法在结构健康监测系统中的运用，通过层级控制策略与功能模块的集成，分散了系统运行风险，降低了由于子系统故障率而引起的监测系统瘫痪。.(2) 建立了传感器冗余优化的布设准则。提出基于键能算法的传感器优化布设准则，考虑了监测数据的独立性及其有效监测范围，降低测点之间的冗余性，提高了监测数据的完备性。.(3) 提出了监测数据失真的辨识方法。基于结构整体受力协同工作关联特性，建立了结构响应失真识别的关联数学模型，提出了关联度偏离率，建立故障诊断函数。.(4) 提出了局部响应参数延拓获取方法。通过改进模式匹配准则，选定多组模式对应力进行融合预测，提高了应力预测的精度，同时提出了融合应力响应数据和位移响应数据对未布置应变传感器构件位置的应力进行延拓获取。.(5) 建立了混合尺度参数的损伤识别与诊断方法。提出了混合运用反映结构整体和局部响应参数的结构损伤识别方法，建立了以应力和加速度监测数据为基础的损伤参数，确定了损伤参数的灵敏度，得到了最优的反向传播神经网络，实现了局部损伤的识别。