多目标堆石坝参数反演方法研究

发布时间：2018-07-04 23:40

  本文选题：高面板堆石坝 + 参数反演　； 参考：《武汉大学》2017年硕士论文

【摘要】：随着国家经济发展,国内水电开发力度进一步加强,其主战场逐步向西和西南方向转移,随之带来的是面板堆石坝的规模从200m级向300m级跨越,300m级面板堆石坝的堆石体颗粒所处的应力环境相对于200m级面板堆石坝发生了较大的变化,高围压下所表现出来的变形特性更加复杂。由于尺缩效应的存在,目前常用的堆石体本构模型以及室内试验的参数测定方式无法对大坝应力变形进行有效模拟,对大坝修建过程中或后期运行阶段的应力变形特点和趋势无法有效预测,使得面板堆石坝建设的安全性、经济性得不到有效保障。基于实测资料对堆石体本构模型参数进行修正是指导大坝安全施工和变形预测的有效手段。已有的堆石体参数反演方法主要为单目标方法,对整个坝体或某个断面仅采用一个目标函数,其未考虑堆石材料对变形的区域性影响,将所有材料整合到一起,映射关系复杂,反演效果不佳,结果不稳定。本文引入多目标的思想,将坝体划分多个子区间,在每个子区间内建立反演子目标函数,从而充分考虑材料的区域性影响,并基于非支配排序的多目标遗传算法(NSGA-Ⅱ),从而兼顾了材料的全局性影响,提出了基于NSGA-Ⅱ和RBF神经网络的堆石坝多目标参数反演方法。本文的主要研究内容如下:1)介绍了非支配排序的多目标遗传算法(NSGA-Ⅱ)和RBF神经网络的基本原理,以及与堆石坝参数反演工程应用的搭接技术,搭建了基于NSGA-Ⅱ算法和RBF神经网络的高堆石坝多目标参数反演平台,并系统阐述了该反演平台的各个环节。2)总结了堆石坝的分区原则和典型形式,并以此提出了子区间划分的基本原理、方法和优势。以一种典型材料分区模式为例,建立标准坝模型,进行材料区域敏感性分析研究,总结规律,验证了堆石材料对变形的区域性影响以及提出了针对该分区模式的反演子区间划分方案。3)阐述了堆石体E-B本构模型和流变本构模型的基本原理,以及瞬变-流变的有限元联合计算法方法。并通过对瞬变和流变本构模型的力学参数敏感性研究,选出了最合适的反演参数。通过整理和分析水布垭面板堆石坝的变形监测资料,选出了最合适的反演断面和测点。4)运用上述方法对水布垭主堆石和次堆石的瞬变-流变参数进行了反演,并与单目标反演方法进行了对比分析。计算结果表明,多目标参数反演方法不仅在变形量和变形趋势上与实测值基本一致,且所有监测点的反演计算值与实测值均吻合较好,计算结果的稳定性较好,整体变形和应力分布规律均符合实际,而单目标参数反演的计算结果只在局部区域与实测值较为接近,计算结果不稳定,多目标参数反演计算结果明显优于单目标参数反演,说明了本文提出的多目标参数反演方法具有合理性和优越性。
[Abstract]:With the development of the national economy, domestic hydropower development has been further strengthened, and its main battlefield has gradually shifted to the west and southwest. As a result, the stress environment of the rockfill grain in the face rockfill dam from 200m to 300m is changed compared with that of 200m CFRD. The deformation characteristics under high confining pressure are more complicated. Due to the existence of scale shrinkage effect, the commonly used constitutive model of rockfill and the measurement of parameters in laboratory tests can not effectively simulate the stress and deformation of the dam. The characteristics and trends of stress and deformation in the process of dam construction or in the later stage of operation can not be effectively predicted, which makes the construction of face rockfill dam safe and economic can not be effectively guaranteed. It is an effective method to guide dam safety construction and deformation prediction by modifying the parameters of constitutive model of rockfill based on the measured data. The existing inversion methods of rockfill parameters are mainly single-objective method. Only one objective function is used for the whole dam body or a section. The method does not take into account the regional influence of rockfill material on deformation, so all materials are integrated together and the mapping relationship is complex. The inversion effect is not good and the result is unstable. In this paper, the idea of multi-objective is introduced, the dam body is divided into several sub-regions, and the inversion subobjective function is established in each sub-interval, so as to fully consider the regional influence of the material. A multi-objective genetic algorithm (NSGA- 鈪，

本文编号：2097959