基于移动窗口最小拟合法的空间网格结构损伤识别和状态评估方法研究

Substantial economic losses and fatalities during recent collapses of space latticed structures in the last ten years have highlighted the particular need to proper structural damage identification and their condition assessment. Determination and selection of appropriate methods and instrumentation layout for condition assessment of space latticed structures, as well as the corresponding damage identification, are still not clear. In particular, due to collected data on site usually contaminated by noise effect and other types of interventions, damage location detection using these data is the biggest challenge. To bridge this gap, this project will develop reliability-based analysis with random imperfections and the corresponding sensitivity for the benchmark of determination of instrumentation layout, which responds for the reliability of space latticed structures due to impacts of environmental effects, fatigue and fracture, and corrosion, and their sensitivity to various physical loadings, including self weight, earthquake, wind and snow loads. With the aid of a probability density function in terms of random imperfections, this project aims to develop a new moving-window least squares fitting method for damage detection and localization with applications to condition assessment of space latticed structures. The new fitting method will provide a capability of mutual adjustment for damage identification through its moving window, in accordance with data collected on site against predicted ones. In addition, the proposed fitting method will be experimentally validated with testing of scaled single-layer reticulated domes under a concentrated impact load and shake table, respectively. New in-house software consisting of those aforementioned functions will be developed in the end. This understanding will provide a robust structural condition assessment for space latticed structures.

近十年来，空间网格结构倒塌事故引起了工程界对其结构损伤识别以及状态评估科学性的广泛关注。空间网格结构状态评估方法的选择和测点布置，以及监测识别是研究难点，尤其是由于现场噪音以及其他因素的干扰，结构监测识别困难很大。本项目针对空间网格结构在环境荷载、疲劳效应、腐蚀等作用下的可靠度降低这一问题，结合结构在自重、地震、风、雪荷载等作用下的结构响应敏感性程度，建立考虑缺陷随机变量概率分布的可靠度分析方法和以敏感性为基础的测点布置方法；发展新的移动窗口最小拟合方法，以缺陷随机变量的概率分布函数为基础，对空间网格结构进行损伤识别和状态评估，并通过现场实测数据和理论计算数据的对比调整最小拟合法的移动窗口，实现交互式损伤识别；通过空间网格结构模型的冲击激励和振动台试验研究，对上述结构损伤识别和状态评估的新方法进行验证，同时开发相关的状态评估软件，为空间网格结构的健康评估提供更为高效准确的方法。

大跨度空间网格结构损伤识别和状态评估研究是结构健康监测中的热点课题。基于移动窗口最小拟合（MWLS）法的损伤识别方法具有高适应性和敏感性的优点。本项目较完整的完成了MWLS方法从理论研究到数值模拟研究再到实验研究的整个过程，得出了MWLS方法在空间网格结构损伤识别和状态评估中的有效性和先进性。具体的研究包括：（1）基于狄里克莱函数推导和建立了MWLS方法的损伤定位和定量指标，利用数值计算验证了MWLS方法在损伤定位和定量计算中的适用性，得出MWLS方法在损伤定位和定量识别中都具有比较高的准确率，并且对上述算法进行编程，建立了MWLS损伤识别方法小软件。（2）对大跨度网格结构进行足尺模型试验，验证基于MWLS方法在网格结构损伤识别中的敏感性和有效性，得出MWLS方法在损伤定位中在具有足够数量的传感器条件下具有较高精度。（3）以MWLS方法为基础，推导和验证了基于广义应变差方法在空间网格结构损伤识别中的应用，得出广义应变差方法具有较高的损伤定位和抗噪声性能。（4）利用深度贝叶斯神经网络建立在不同损伤程度下的结构状态评估方法，并利用实验数据验证，得出该方法具有比常规方法更高的准确率和抗噪声能力。（5）将MWLS方法构造的损伤识别指标推广到基于统计模型的损伤识别研究领域，对大跨度结构体系动力特性进行分析，并结合支持向量机，较为系统的研究MWLS方法的损伤敏感性和大跨度结构动力特性的损伤敏感性，得出支持向量机方法可以充分考虑结构在不同荷载作用下的随机性和不确定性的优点，并具有较强的损伤识别敏感性。