基于GPU平台的水泥水化微结构实体纹理合成

发布时间：2018-05-13 09:07

本文选题：水泥微观结构 + 水泥水化　；参考：《济南大学》2017年硕士论文

【摘要】：随着社会经济的发展,水泥在世界范围内被广泛应用,已经成为基础建设中必不可少的建筑材料,并且需求量越来越大,同时水泥的高质量和高性能也越来越受关注。从微观上来看,水泥的微观纹理结构的变化会直接影响水泥的各项特性,所以通过水泥实体微观结构来分析、预测水泥水化过程,进而提高水泥的性能具有非常重要的现实意义。水泥水化的过程中不仅伴随着复杂的化学变化,同时水泥的微观纹理结构也发生着微妙的变化。传统的方法是采用物理化学的方法来测定水泥样本指标,并且仅能局限于细微颗粒规模的实验,不能进行庞大规模的实验,例如高楼建筑、高架桥等建筑物。随着近年来的计算材料学的发展,利用计算机对材料进行仿真模拟实验成为了当下材料研究的热点。所以,针对水泥的各项特性,本文通过对水泥实体微观结构进行建模,并利用GPU平台对水泥微观建模过程进行加速,通过建立准确高效的水泥实体纹理模型来实现大规模水泥纹理合成,为预测大规模建筑的水泥不同时期水化过程提供初始状态。本文主要从以下几个方面进行水泥实体纹理建模研究:(1)基于二维真实水泥图像的水泥图像纹理结构建模(2D-2D)本文提出了基于真实二维水泥图像的纹理随机合成算法。水泥图像数据运用的是真实的实体水泥二维切片,以马尔科夫随机场模型为算法的基础思想,以及采用多层高斯模糊的方式,对二维水泥纹理进行合成。通过多组对比实验,对影响二维水泥纹理合成效果的因素进行详细分析。(2)基于三维真实水泥实体的水泥实体微观结构建模(3D-3D)将二维纹理合成算法融入到三维水泥实体纹理合成的方法中,首先针对二维参数的三维化,将算法重新定义。通过多组对比实验,分析影响三维水泥实体纹理合成的相关因素,同时采用三维显示软件对实验结果做可视化处理,便于实验结果观察评估。(3)基于二维真实水泥图像的水泥实体微观结构建模(2D-3D)由于水泥具有各向同性的特点,因此可以从二维水泥纹理图像获得水泥实体的三维化数据信息。通过,,ZYX三个方向来获取水泥纹理信息,通过邻域和灰度进行匹配,完成二维到三维实体纹理合成算法,然后详细分析实验结果。(4)三维水泥实体微观结构建模并行化对三维水泥实体纹理合成的串行算法进行分析,通过统计串行算法每个步骤的时间消耗,总结出消耗时间巨大并且独立性强的算法步骤,提出水泥实体微观结构合成的并行算法,其中包括基于三维真实水泥实体的水泥实体纹理合成算法和基于二维真实水泥图像的水泥实体合成算法,通过实验结果分析,并行实验的效率比较高,证明本文的并行算法具有高效性。
[Abstract]:With the development of social economy, cement has been widely used in the world, and has become an indispensable building material in infrastructure construction, and the demand is increasing. At the same time, the high quality and high performance of cement are paid more and more attention. From the microscopic point of view, the change of cement microstructure will directly affect the properties of cement, so the cement hydration process can be predicted by analyzing the microstructure of cement entity. And then improve the performance of cement has very important practical significance. The process of cement hydration is accompanied not only by complex chemical changes, but also by subtle changes in the microstructure of cement. The traditional method is to use physical and chemical methods to measure the cement sample index, and can only be limited to the fine particle size of the experiment, can not carry out a large scale of experiments, such as high-rise buildings, viaducts and other buildings. With the development of computational materials in recent years, computer simulation of materials has become a hot topic. Therefore, according to the characteristics of cement, this paper models the microstructure of cement entity, and uses GPU platform to accelerate the process of cement micro-modeling. An accurate and efficient cement solid texture model is established to realize large-scale cement texture synthesis, which provides the initial state for predicting the hydration process of cement in different periods of large-scale construction. In this paper, the following aspects of cement solid texture modeling research: 1) based on two-dimensional real cement image texture structure modeling cement image 2D-2D) this paper proposes a texture random synthesis algorithm based on real two-dimensional cement image. The cement image data is based on the real two-dimensional slice of solid cement. The basic idea of Markov random field model and the multi-layer Gao Si fuzzy method are used to synthesize the two-dimensional cement texture. Through a number of comparative experiments, The factors that affect the effect of 2D cement texture synthesis are analyzed in detail. (2) 3D cement entity microstructure modeling based on 3D real cement entity 3D-3D) integrates 2D texture synthesis algorithm into 3D cement entity texture synthesis method. Firstly, the algorithm is redefined for the 3D of two-dimensional parameters. Through multi-group contrast experiments, the related factors affecting the texture synthesis of 3D cement entity are analyzed, and the experimental results are visualized by using 3D display software. It is convenient to observe and evaluate the experimental results. (3) the microstructure modeling of cement entity based on 2D real cement image (2D-3D) can obtain 3D data information of cement entity from two-dimensional cement texture image because cement has isotropic characteristics. The cement texture information is obtained from three directions of ZYX, and the 2D to 3D solid texture synthesis algorithm is completed by matching the neighborhood and gray level. Then we analyze the experimental results in detail. 4) the parallel modeling of 3D cement entity microstructure is used to analyze the serial algorithm of 3D cement entity texture synthesis, and the time consumption of each step of the serial algorithm is calculated. Summing up the time-consuming and independent steps of the algorithm, a parallel algorithm for the microstructure synthesis of cement entities is proposed. It includes the texture synthesis algorithm based on 3D real cement entity and the cement entity synthesis algorithm based on two-dimensional real cement image. Through the analysis of experimental results, the efficiency of parallel experiment is relatively high. It is proved that the parallel algorithm in this paper is efficient.
【学位授予单位】：济南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TQ172.1;TP391.41

【参考文献】