基于贝叶斯推断和动力测试的既有砌体结构性能分析

发布时间：2018-01-15 21:34

本文关键词：基于贝叶斯推断和动力测试的既有砌体结构性能分析　出处：《上海理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：由于砌体材料具有良好的物理性能,且施工方法简便,造价低廉,所以至今仍是我国主导的建筑材料。砌体结构在我国有着悠久的历史,岁月的沉淀给这些已有的建筑结构带来了潜在的威胁,所以对于这些既有砌体结构的性能检测也变得尤为重要。虽然原位测试及钻芯法等现场测试方法是当前主要的检测方法,然而,这些方法会对结构造成新的损伤,所以这些方法仍然存在着一定的缺陷,并不完善。因此,基于无损的动力测试方法的结构性能检测方法受到了越来越多的关注。频率是既有砌体结构安全性的重要指标,由于损伤情况、结构形式以及测试方法等不确定因素的影响,导致其现场实测结果数据量有限且非常离散,难以确定合理的估计值。传统的回归分析方法完全依赖于实测样本,如果实测样本离散,也会造成结果判断的严重偏差。为解决这一问题,尝试采用贝叶斯推断方法,通过马尔科夫蒙特卡洛(Markov Chain Monte Carlo,MCMC)抽样,结合先验信息,将结构频率作为未确知量,形式上借助数理统计的方法来处理。首先通过三维数值模拟确定结构动力测试的采样频率以及加速度传感器的布置方案,然后分别采用环境激励与人工激励方法,获得结构的加速度时程曲线,通过快速傅里叶变换(Fast Fourier Transform,FFT)获取结构基本频率的样本数据。运用贝叶斯推断方法,通过Metropolis-Hasting(MH)抽样,结合对基本频率的先验判断,构建基本频率的后验概率密度模型,从而确定结构基本频率的估计值。通过有限元模型,拟合结构频率与弹性模量之间的关系式,反算出结构等效弹性模量的估计值。结果表明,以上方法在实测样本量少、噪音影响较大的情况下具有很好的鲁棒性,能够较快收敛于稳定、合理的后验概率密度模型,为推断结构弹性模量提供可靠的基础。针对某既有砌体结构,对其进行动力测试,结合先验判断,获取其频率的后验概率密度模型,根据有限元模型反算结构等效弹性模量估计值,验证本文所提方法的工程应用。说明该方法可为砌体结构性能的长期监测提供一种经济、有效的手段。
[Abstract]:Masonry materials have good physical properties, simple construction methods and low cost, so they are still the leading building materials in China. Masonry structure has a long history in China. The sedimentation of years has brought a potential threat to these existing building structures. Therefore, for these existing masonry structure performance testing has become particularly important. Although in-situ testing and core drilling methods are the main testing methods, however. These methods will cause new damage to the structure, so these methods still have some defects and are not perfect. More and more attention has been paid to the structural performance testing method based on the non-destructive dynamic testing method. Frequency is an important index of the safety of existing masonry structure due to damage. Due to the influence of uncertain factors, such as structure form and test method, the field measured results are limited and discrete, so it is difficult to determine the reasonable estimation value. The traditional regression analysis method is completely dependent on the measured samples. In order to solve this problem, Bayesian inference method is used to solve this problem. The structural frequency is taken as an unascertained quantity through Markov Chain Monte Monte MCMC sampling and a priori information. Firstly, the sampling frequency of structural dynamic test and the layout of acceleration sensor are determined by three-dimensional numerical simulation. Then the acceleration time history curve of the structure is obtained by using the environmental excitation method and the artificial excitation method, and the fast Fourier Transform is obtained by the fast Fourier transform. The sample data of the basic frequency of the structure are obtained by FFT.The Bayesian inference method is used to sample the basic frequency through Metropolis-HastingMH sampling and the prior judgment of the basic frequency is combined. The posteriori probability density model of the basic frequency is constructed to determine the estimated value of the basic frequency of the structure. The finite element model is used to fit the relationship between the structural frequency and the elastic modulus. The results show that the proposed method is robust in the case of small sample size and large noise effect, and can converge to stability quickly. A reasonable posteriori probability density model provides a reliable basis for inferring the elastic modulus of a structure. For an existing masonry structure, the dynamic test is carried out and the posterior probability density model of its frequency is obtained by combining the prior judgment. The engineering application of the method proposed in this paper is verified by the inverse calculation of the equivalent elastic modulus of the structure based on the finite element model. It is shown that the method can provide an economic and effective means for the long-term monitoring of masonry structure performance.
【学位授予单位】：上海理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TU364

【参考文献】