基于强震观测的混凝土重力坝动力有限元模型修正方法研究

发布时间：2018-05-21 00:42

本文选题：强震观测 + 混凝土重力坝　；参考：《西安理工大学》2017年硕士论文

【摘要】：我国处于环太平洋和地中海—喜马拉雅两大地震带之间,是地震灾害最严重的国家之一。同时,我国也是世界上已建和在建混凝土重力坝最多的国家之一。对于服役中的混凝土重力坝进行抗震分析时,尤其已经经受过地震考验的混凝土重力坝,其材料的动力参数的选取是一个值得深入研究的问题。基于混凝土重力坝的强震观测,进行坝体结构的模态参数识别,并据此进行混凝土重力坝动力有限元模型修正是解决该问题的方法之一。混凝土重力坝属多自由度复杂空间结构,采用传统的矩阵模型修正和参数模型修正法进行混凝土重力坝动力有限元模型修正需要进行大量的有限元计算,计算效率低,对混凝土重力坝这种大型工程结构的适用性不强。本文考虑混凝土重力坝结构本身和强震观测数据的特点,利用混凝土重力坝强震观测资料结合模态分析理论以及智能优化算法对其动力有限元计算模型进行修正,主要研究内容和成果如下:(1)对常见的几种强震记录去噪方法进行了介绍,同时,提出了基于主成分分析法(Principal Component Analysis, PCA)结合小波滤波的混凝土重力坝强震观测数据去噪方法,并通过仿真试验将该方法与小波软、硬阈值去噪法的去噪效果进行了对比;以去噪后的强震观测数据为基础,提出了基于自动频域分解(Auto Frequency Domain Decomp--osition, AFDD)法的混凝土重力坝模态参数识别方法,并采用该方法对某混凝土重力坝实测得到强震观测记录进行了模态参数识别。(2)对“库水-大坝-基础”三维动力有限元分析所涉及到的库水-大坝和基础-大坝动力相互作用以及粘弹性人工边界的基本理论进行了介绍,并在此基础上,研究了待修正混凝土重力坝动力有限元模型参数选取和参数的敏感性分析。采用实际工程算例实现了上述分析过程。(3)首先,将混凝土重力坝的动力有限元模型修正问题就转化为一个强震识别模态参数与计算有限元模型模态参数之间的差异最小化问题。考虑到混凝土重力坝结构本身和强震观测数据的特点,利用拉丁立方抽样的试验设计能力,多输出支持向量机(Multi-output Support Vector Machine, M-SVM)算法对多变量之间非线性隐式关系的拟合能力以及遗传算法(GeneticAlgorithm,GA)神经网络的搜索寻优能力提出了一种基于强震观测和M-SVM的混凝土重力坝动力有限元模型修正方法。并采用该方法对某混凝土重力坝的有限元模型进行了修正,在保证计算精度的前提下有效的提高了计算效率。
[Abstract]:China is one of the two largest earthquake belts in the Pacific rim and the Mediterranean - Himalaya. It is one of the most serious earthquake disaster countries. At the same time, China is one of the countries that have built and built the most concrete gravity dams in the world. The selection of dynamic parameters of the gravity dam is a problem worth studying deeply. Based on the strong earthquake observation of the concrete gravity dam, it is one of the methods to solve this problem by identifying the modal parameters of the dam structure and modifying the dynamic finite element model of the concrete gravity dam. The concrete gravity dam is a complex space of multi freedom. Structure, using the traditional matrix model correction and parameter model correction method to modify the dynamic finite element model of concrete gravity dam needs a lot of finite element calculation, the calculation efficiency is low and the applicability of the concrete gravity dam is not strong. This paper considers the structure of the concrete gravity dam and the strong earthquake observation data. Characteristics, using the strong earthquake observation data of concrete gravity dam combined with modal analysis theory and intelligent optimization algorithm to modify the dynamic finite element calculation model, the main research contents and results are as follows: (1) several common strong earthquake recording de-noising methods are introduced. In the same time, a principal component analysis (Principal Component) method is proposed. Analysis, PCA) combined with wavelet filtering for strong earthquake observation data de-noising method of strong earthquake of concrete gravity dam, and comparing this method with the denoising effect of wavelet soft and hard threshold denoising method by simulation test. Based on the strong earthquake observation data after de-noising, the method is based on the automatic frequency domain decomposition (Auto Frequency Domain Decomp--osition, AFDD). The method of identification of modal parameters of concrete gravity dam, and using this method to identify the modal parameters of a concrete gravity dam measured by strong earthquake records. (2) the interaction between reservoir water dam and foundation dam and the viscoelastic artificial boundary involved in the three-dimensional dynamic finite element analysis of "reservoir dam foundation" The basic theory is introduced. On this basis, the parameters selection and the sensitivity analysis of the dynamic finite element model of the concrete gravity dam are studied. The above analysis process is realized by practical engineering examples. (3) first, the dynamic finite element model correction problem of the concrete gravity dam is transformed into a strong earthquake recognition model. The discrepancy between the state parameters and the modal parameters of the finite element model is minimized. Taking into account the characteristics of the structure of the concrete gravity dam and the strong earthquake observation data, the Multi-output Support Vector Machine (M-SVM) algorithm is used for the nonlinear implicit expression between the multivariable variables by using the experimental design ability of the Latin cubic sampling. The fitting ability of the relationship and the search and optimization of the GeneticAlgorithm (GA) neural network proposed a modified method for the dynamic finite element model of the concrete gravity dam based on strong earthquake observation and M-SVM, and modified the finite element model of a concrete gravity dam with this method, which is on the premise of guaranteeing the calculation precision. The efficiency of calculation is improved effectively.
【学位授予单位】：西安理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TV642.3;TV312

【参考文献】