【作者】 郝书亮；

【导师】 党发宁；

【作者基本信息】 西安理工大学 ， 岩土工程， 2009， 硕士

【摘要】 目前,混凝土材料力学行为的研究大都是建立在试验的基础上,需要花费大量的人力、物力,所得到的试验成果又往往受到试验条件、环境以及材料本身复杂性的影响,试验成果的离散性较大。在这种情况下,结合理论和试验的细观数值模拟是一种行之有效的方法。而在细观尺度上有限元模型的建立是数值模拟首要的一环。随着混凝土力学性能研究的深入,学者们越来越重视其细观力学性能的研究,随之产生了各种各样的细观结构有限元模型,例如格构模型,梁—颗粒模型,随机骨料模型,随机力学特性模型。以上模型大多是基于随机抽样方法和统计学的知识,所建立的模型虽然在骨料的分布和形状上已越来越接近混凝土的真实结构,但不可避免的与混凝土的真实细观结构存在着差异。CT技术的出现,实现了在无损状态下将混凝土内部细观结构用数字化方法进行呈现。如果能利用CT平面图像的信息,建立真实或近似真实的细观结构有限元模型,将对混凝土数值模拟研究产生一定的作用。混凝土包括了骨料、水泥砂浆和孔隙三部分,从图像特征来看,骨料接近于白色,孔隙趋于黑色,水泥砂浆则介于两者之间。由于骨料、砂浆与孔隙具有相对明显的密度差异,通过CT扫描并转换成图像后,混凝土各个组分之间具有较好的对比度,能比较容易的观察和提取出骨料的轮廓。本文选用平滑滤波、阀值分割、形态学开闭运算等比较成熟的数字图像处理技术对混凝土圆柱试件的截面图像进行预处理,以获取准确的骨料几何与位置信息。利用这些信息,本文提出了两种三维重建的思路。第一种思路也就是本文第三章,基本步骤是利用处理好的图像,提取骨料的轮廓,基于矩理论对骨料的轮廓进行识别,将同一骨料在不同的层的轮廓提取出来,进行拓扑重构。对重构好的骨料数据进行处理,提取骨料的几何信息如长短轴长,长轴的倾角,圆度,矩形度等,并提取有限数量且能准确描述骨料结构的特征点,将这些特征点的位置信息导入ANSYS,连点成线,由线生成面,并利用ANSYS中的蒙皮技术利用同一骨料的不同面生成体,最后剖分网格,生成有限元模型。第二种方法也就是本文的第四章,基于混凝土的CT平面图像信息,用APDL语言生成的程序对混凝土的三维细观结构在大型商业有限元软件ANSYS中进行三维重建。重建过程没有生成点、线、面、体等几何元素,而是利用CT平面图像信息直接生成节点,连接节点生成单元,避免了网格无法剖分现象的出现。生成的单元形状规则,长宽比合理。第一种方法生成的模型忽略了小体积组分结构,由于特征点选取的有限,生成的较大体积的骨料与试件中骨料的圆形状也存在着差异,并且剖分后的一小部分网格质量不高。第二种方法直接对图像信息数据压缩,没有对不同的骨料进行区分,并且由于界面层的引入,使生成模型与原混凝土试件相比骨料率明显降低。但基于CT图像三维重建后的模型是以混凝土的真实细观结构信息为依据的,从而使生成的模型与真实模型差异更小,并且彻底解决了多级配骨料无法投放的难题。更多还原

【Abstract】 Nowadays, researching on mechanical behavior of concrete is based on experiment, which takes plenty of manpower and material resources, and test results obtained from the experiments is largely discrete because of experimental conditions, circumstance an complexity of material of itself etx. In this case, numerical simulation in mesolevel, which combine experiment and theory together, is an effcctive way of researching concrete.With the researching on concrete mechanical property, the scholar put more and mor focus on the micromechanical behaciors of the concrete. So many kings of meso-structure finite element model was been made, such as lattice model, beam—granular model, random aggregate model and random mechanics model. These methods were all based on the random sampling methods and the statistics knowledge. So the FEM models made by these methods were differert from the real concrete structure inevitablly. With the CT apperence, the meso-structure of concrete was seen non-destructivly by digital method. If we can use the message given by the CT images to constructed the more lifelike FEM model, it will be very meanful for the concrete digital simulation. The concrete was consist of aggregate, mortar and pore. In the images, the aggregates were white, the pores were black and the mortar was between of them. Because the density of the three materials were different apparenttly, the contrast between of the three was very well in the images transferred from the CT and we can get the contours easilly. In the article some digital image processing technics such as smooth filter, threshold dividing and the open and close on morphology were used to process the images of the concrete beam. The shape and location messages of the aggregates was got. In the article, two 3-D reconstruction methods were given. The first method was described by the third section in the article. Using the images processed, the contours were extracted, then these contours were identied based on the moment theory to reconstruction topologically. Processing the aggregate data to get the messages of the aggregate shapes such as length of the long and short axis, the pitch angle of the long axis, circularity and rectangular degree etc. a few number of characteristic points were extracted to describe the shape of the aggregate. the location of these characteristic points were loaded into ANSYS, then connecting these points to generate lines and generate areas by these lines, finally the volume was generated using envelope technology by areas. The FEM model was generated after the mesh divided. The second method described by the forth section in the article. Based on the CT images, the 3-D meso-structure of concrete was reconstructed using APDL language by ANSYS. In the reconstruction procession, the geometic constituents such as points, lines, areas and volumes had not been generated. Using the messages from the CT images, the nodes were generated directly, then the elements were generated by jointing the nodes. The phenomenons that the elements could not been divided was avoided. The shapes of elements were regular, and the aspect ratio was rationally. In The first method the small volume compoment constructions were ignored. Because the characteristic points chosen were finitly, the big aggregates generated in the FEM model was also a little different from the real aggregate shapes and the qualitity of some mesh was not very well. In the second method the images data was compressed directly, the specification between aggregates were not very well.The boundary layer was generated in this model, so the ratio of the aggregate was cuted down apperently contrast to the real concrete model. But the 3-D reconstruction FEM model generated from the real construction of the concrete beam. So the model was more likely with the real model and the problem that the different aggregate graded concrete can hardly be generated was ravel out rippingly.更多还原