【作者】 向衍；

【导师】 吴中如； 顾冲时；

【作者基本信息】 河海大学 ， 水工结构工程， 2004， 博士

【摘要】 本文针对混凝土坝坝体与坝基之间的互馈问题，利用神经网络、数据挖掘技术、遗传算法、分形理论等前沿数学方法以及接触力学、工程力学等知识，并结合大坝原位监测资料，研究了高坝坝体与复杂坝基互馈的力学行为及其分析理论。在此基础上，研究了互馈的演变规律和转异挖掘方法。为使上述理论应用于实际工程，本文建立了坝体与坝基互馈的预警系统。主要研究内容如下： (1)针对混凝土坝坝基中软弱结构面多的特点，本文利用人工神经网络和决策树技术，以众多工程实例为依据，建立了基于坝基稳定的八个参数的神经网络和决策树分类器，从而确定复杂坝基中的互馈优势面。对于多个互馈优势面，采用极限平衡理论计算各个面的抗滑稳定安全系数，定义安全系数最小的面为互馈面。 (2)针对坝体与坝基之间的相互作用通常假定为平面型接触问题的不足，考虑互馈面实际接触面积的影响以及接触界面上的力学参数(特别是凝聚力)随接触面积的改变而改变，建立了基于坝体与坝基实际接触面积变化的互馈模型。 (3)由于分形能定量刻画接触面的粗糙程度，利用分形维数得到混凝土坝坝体与坝基互馈面的应力应变本构关系，在此基础上，考虑分形维数的变化较真实的模拟了混凝土坝坝体与坝基的互馈过程。 (4)在弄清混凝土坝坝体与坝基互馈的力学行为及其分析理论和方法的基础上，研究了混凝土坝坝体与坝基互馈的演变规律和转异特征。并基于突变性判据，利用互馈面实际接触面积或分形维数曲线随名义应变变化的特征，得到了坝体与坝基互馈的转异判据。 (5)本文研制了混凝土坝坝体与坝基互馈的预警系统，并将其概化为数据融合、警源分析、警兆辨识、警情分析和警情发布等五个子系统，给出了各系统的物理模型以及警兆、警情指标的确定方法。该系统可以更好地预警混凝土坝的安全。更多还原

【Abstract】 Aiming at the crossfeed problem between high concrete dam body and complex foundation, some new models and methods are presented based on the advanced mathematics methods (such as artificial neural network, data mining technology, genetic algorithm, fractal theory) and contact mechanics and engineer mechanics, etc. With the help of the prototype observation data, the mechanical behavior and analysis theories of crossfeed between high dam body and complex foundation are studied. Based on the above, the evolvement rules and the anomalism mining methods of crossfeed between dam body and foundation are presented. The main contents are as follows,Firstly, although there are many weak structure surfaces in dam foundation, a classify mode based on artificial neural network and decision tree technology is developed, which has eight input parameters for the dam foundation stabilization with the help of many engineering examples. With the classify mode, the crossfeed dominant surface can be confirmed. For the more than one crossfeed dominant surfaces, the one whose safety index is least is defined as the crossfeed surface based on limit equilibrium theory.Secondly, for the deficiency of the tradition hypothesis, which the interaction surface between dam body and foundation is considered as a plane, a new kind of crossfeed model is built with the variational contact area, including the variation mechanical parameters (especial the cohesion) change with the contact area.Thirdly, fractal geometry can describe the coarseness of the contact surface, so with the help of fractal dimension the constitutive relation is acquired based on which the crossfeed process of concrete dam body and foundation is truly simulated.Fourthly, based on the above research of the mechanical behavior and analysis theories & methods of crossfeed between high dam body and complex foundation, the evolvement rules and the anomalism mining methods of crossfeed are studied. With the help of catastrophe criterion, the anomalism criterions of the crossfeed between dam body and foundation are acquired from the real contact area or fractal dimension curve vs. representative strain of the crossfeed surface.Finally, the early-warning system of crossfeed between dam body and foundation is developed, which consists of five sub-systems, such as data fusing, warning root analysis, warning sign analysis, warning instance analysis and promulgating. The physical models of the five sub-systems and the index of warning sign and warning instance are presented.更多还原