超高层建筑动态位移全天候摄像测量及数据挖掘研究

Wind effects are one of the utmost concerns for the structural design of super-tall buildings in China southeastern coastal cities. Nevertheless, in-situ monitoring of wind-induced responses have rarely been conducted for super-tall buildings. Especially, the long-term continuous monitoring of dynamic displacements is still quite rare. Therefore, this study carries out an investigation into the data mining of displacements obtained by long-term viedogrammetry-based monitoring of super-tall buildings. First, in-situ monitoring of dynamic displacements of two buildings (one 220 m high hotel and office building, and one 189 m high residential building) in Wenzhou will be conducted. The in-plane rigid-body displacements of the buildings are measured by making use of viedogrammetry technique. Then, methods like empirical mode composition decomposition (EMD), wavelet shrinkage noise reduction, and Kalman filter, will be combined to preprocess the displacement measurement data. By so doing, the systematic errors are suppressed, the measurement noises are reduced, and the high-frequency components are integrated. Finally, data mining of long-term displacement measurement data will be performed to extract the behavior and ambient excitation mechanism of displacement. Cluster analysis and HHT will be used to extract the statistical characteristics, time-frequency characteristics of displacements, as well as the environment-displacement relationships. Furthermore, the quasi-static and dynamic displacement components will be separated and extracted using EMD, in order to separately investigate their ambient excitation mechanism. Methods like time-sequence analysis, linear regression analysis, BP neural network will be employed to reveal the ambient excitation mechanism and formulate the environment-displacement correlation model.

风致响应是超高层建筑结构设计需要关注的共性问题。然而，超高层建筑风致响应的现场实测仍较缺乏，特别是动态位移的长期连续监测。鉴于此，本项目开展超高层建筑动态位移全天候摄像测量及数据挖掘研究。首先，开展超高层建筑动态位移全天候摄像测量试验研究。以温州置信广场（一座220 m主塔楼和一座189 m住宅楼）为对象，利用摄像测量方法进行平面刚体位移现场实测试验。然后，开展超高层建筑实测动态位移数据处理方法研究。采用EMD、小波阈值降噪、卡尔曼滤波等方法处理实测动态位移数据，抑制系统误差、降低环境噪声、弥补高频成分。最后，开展超高层建筑动态位移的特征规律及环境激励机制研究。采用聚类分析、HHT变换等方法分析动态位移的统计特征、相关性、及时频域特征，提炼超高层建筑动态位移的特征与规律。采用EMD方法分离与提取准静态位移与动态位移，采用时间序列相似性分析、线性拟合、BP神经网络等方法研究其环境激励机制。

本项目开展了超高层建筑动态位移全天候摄像测量及数据挖掘研究。首先，开展了超高层建筑动态位移摄像测量方法研究。利用单个摄像机，获取结构三维位移，形成了一种结构三维位移单目摄像测量方法。针对以剪切变形为主的高层结构，提出了一种能够精确测量层间位移、克服光路不通视局限的高层结构剪切变形Z型摄像测量方法。然后，开展了超高层建筑实测动态位移数据处理方法研究。针对摄像测量系统热效应引起的图像漂移，提出了一种准参照点图像漂移消除方法。针对多种因素耦合作用引起的摄像测量图像漂移，提出了一种基于弥散圆光斑的图像漂移消除方法。采用EMD与小波阈值降噪结合的方法对动态位移数据进行降噪处理，通过EMD得到从高频至低频分布的各IMF分量，对各个高频IMF分量分别设定阈值进行小波阈值降噪处理。采用卡尔曼滤波方法融合台风条件下的位移和加速度数据，提高动态位移数据的精度。最后，开展了超高层建筑动态位移的特征规律及环境激励机制研究。采用谱系聚类方法对风、温度、动态位移数据进行聚类分析，对每个聚类分别进行特征统计、相关性分析等，提炼超高层建筑动态位移的特征与规律。采用盲源分离方法分离与提取不同环境激励因素引起的位移分量，采用线性拟合、神经网络等方法建立位移分量与风、温度的相关模型。对比分析风洞试验结果、规范提供的计算方法等，对抗风设计的有效性和准确性进行了初步验证。比较分析不同结构的监测数据，初步探讨了结构型式、高度、高宽比等因素对超高层建筑动态位移的影响。本项目研究结果为结构动态位移摄像测量的发展和应用、为认知超高层建筑在环境激励下的变形行为、为完善超高层建筑的抗风设计理论等提供数据及理论上的支持。