面向智能诊断的大跨斜拉桥状态评估方法研究

In view of the lack of fusion problems on manual inspection information and health monitoring data in the assessment data sources of cable-stayed bridges, the accurate defects diagnosis and condition assessment method of long-span cable-stayed bridge is studied in order to achieve more intelligent and comprehensive maintenance. First of all, two types of evaluation indexes are performed for multidimensional processing. For artificial inspection information, fuzzy theory is used to quantify the qualitative index, and data mining of disease distribution and characteristics is conducted for the multidimensional display; for monitoring data, the multiscale degradation law is explored from the perspective of time and space based on the normalization of environment. Secondly, multiple views are formed from the evaluation information sources, and a multi views based hybrid attribute selection algorithm is proposed. The technology complementation is formed by fusion strategy for combining the correlation degrees, and the key attribute set is obtained. Thirdly, large span cable-stayed bridge evaluation index system and weights are established through literature research and field investigation. Assessment standards of various indicators are determined combined with the dimensionless model and the envelope diagram obtained from multi scale finite element model. On this basis, the establishment of condition evaluation theory of long-span cable-stayed bridge is performed based on Fuzzy Evidence Reasoning, and the corresponding software is developed. Finally, the above method is applied to the condition diagnosis and evaluation of several test bridges, and the correctness and applicability of the research theory of the project are verified.

针对目前斜拉桥评估数据源中人工检查信息和健康监测数据融合不足的问题，开展大跨斜拉桥的精确病害诊断和状态评估方法研究，以实现更智能全面的养护维修。首先对两类评估指标进行多维化处理：对于人工检查检测信息，采用模糊理论实现定性指标定量化，通过对病害分布和特征的数据挖掘实现病害多维化展示；对于监测数据，在环境归一化的基础上从时间、空间角度探索其多尺度退化规律。其次，从评估信息源中形成多个视图，提出基于多视图的混合属性选择算法，通过结合相关度的融合策略，形成技术互补，得到关键属性集合。再次，通过文献调研和实地考察，建立大跨斜拉桥的评估指标体系和权重，结合无量纲化模型以及多尺度有限元模型包络图确定各类指标的评定标准，在此基础上，建立基于模糊证据推理的大跨斜拉桥状态评估理论，并开发相应软件。最后，将上述的方法应用到若干试验桥的状态诊断和评估中，验证本项目研究理论的正确性和适用性。

斜拉桥中多源检查信息各有优劣，如何融合得到更可靠的斜拉桥状态评估结果具有重要意义。本项目主要研究内容包括：斜拉桥综合评估指标体系研究；评估指标的属性多维化处理方法；病害关键属性选择方法；基于证据理论的斜拉桥状态评估模型；斜拉桥智能评估软件开发。主要研究成果如下：（1）通过文献总结建立包含人工检查和健康监测的评估指标体系并统计得到指标权重；（2）分析了人工检查信息和健康监测数据的空间分布特征，这有利于进一步指导巡检和养护；（3）建立了索梁塔三者力学简化模型，并依此建立索力变化率的评定标准和索力缺失值填补验证；（4）提出基于多视图的斜拉索病害混合属性选择算法框架，根据斜拉索的检测数据属性，将斜拉索系统分成两个视图，分别定义为外观检查检测视图和健康监测视图。在每个视图里采用混合智能算法进行属性选择，然后合成为关键属性集合，这样可以发挥各个视图的优势，使得最终的属性集合包含的信息多样，能更全面的反映出拉索的损伤。（5）建立考虑部件相关性的斜拉桥状态评估方法和基于证据推理框架的斜拉桥状态评估模型，并以实例验证了两种评估方法的适用性。（6）主梁外观检查和支座是故障表象部件，预防性养护时要着重关注；排水系统和机电设施是故障源部件，在日常检查中需保持这两个部件的良好工作状态。本项目所研究的成果和开发的软件可为养护管理部门提供养护决策指导。

内容获取失败，请点击重试