用于泡沫金属材料内部变形场分析的高精度数字体图像相关法研究

Metal foams are a type of important light-weight and energy-absorption engineering material. The deformation of metal foams is highly inhomogeneous when they are subjected to considerable external loading. It is essential for the study on the relation between the mechanical behaviors of metal foams and the evolution of their meso-structure to precisely measure the deformation distribution within the metal foams. Currently, the internal deformation field of metal foams is measured using the digital volume correlation (DVC) based on the digital volume image obtained by computed tomography (CT) technique. However, the conventional DVC based on the assumption of small deformation satisfying the affine transformation is unsuitable to the highly inhomogeneous deformation of cells in metal foams. Furthermore, DVC computation with sufficiently high resolution and accuracy is extremely time-consuming. There is little investigation on the experimental protocol for the real application of DVC in the deformation measurement of metal foams heretofore. These issues have become a serious barrier to the wide application of DVC in the quantitative research of metal foams. The proposed project is focused on the methodological study of DVC for the internal deformation measurement of metal foams. A high accuracy DVC algorithm will be developed based on the feature identification and registration of digital volume images, which considering the characteristic cellular structure of metal foams and their deformation mechanisms. Synergistic parallel computing based on GPU and multi-core CPU will be introduced to fulfil super high-speed DVC computation. To meet the need for the experimental study of metal foams at relatively high strain rate, the method combining ex-situ CT scanning and DVC will be investigated. The exploration proposed in this project will lead to significant advances in the principle and implement of DVC, which will promote the applications of DVC in the research on metal foams and other solid cellular materials, as a practical and reliable measurement technique.

泡沫金属做为一类重要的轻质吸能工程材料，变形往往呈现高度非均匀性。准确测量泡沫金属的变形场，是研究其细观结构演化与力学行为关系的基础。泡沫金属内部变形场的测量，主要利用数字体图像相关法（DVC）处理计算层析成像（CT）技术获取的数字体图像。然而，传统的基于仿射变换和小变形假设的DVC并不适合分析泡沫金属中高度非均匀的胞孔变形；高分辨率和高精度DVC计算极为耗时；相关实验方法的探索稀少。这些问题严重阻碍了DVC在泡沫金属内部变形场定量分析中的应用。本项目针对泡沫金属的结构特点和变形机制，研究其数字体图像的特征识别和匹配方法，开发适用于泡沫金属的高精度DVC算法。引入GPU和多核CPU协同并行计算机制，实现超高速DVC运算。发展离位CT扫描结合DVC测量方法，适应动态加载实验的需求。本项目的开展，能够拓展充实DVC的原理和方法，促进其成为泡沫金属及其它固体泡沫材料研究中可靠实用的测量技术。

数字体图像相关法（Digital volume correlation，DVC）是目前少数测量非均匀材料内部变形场的有效方法。在处理MicroCT扫描重构的泡沫金属三维体图像时，如何在图像信息量较少，变形幅度较大且不均匀的情况下进行准确测量；如何在保持高精度的情况下提升DVC算法的计算效率；对于无法进行原位MicroCT扫描的情况，如何应用DVC方法处理离位扫描获得的体图像；如何评估DVC算法测量泡沫金属内部变形场的准确性？本研究就上述DVC方法的关键基础问题开展了系统的探讨，提出了完整可行的解决方案。利用尺度不变特征变换算法提取变形前后体图像中的特征，开发了图像特征辅助的DVC算法。实验表明，该算法对于泡沫金属经历大角度旋转、高度非均匀变形、并且变形场存在显著间断性的各种复杂情况，均能够准确而自动的测量，无需任何人工干预。应用基于GPU和多核CPU的并行计算技术对DVC算法进行加速，在个人电脑上实现了超高速的高精度DVC运算，将DVC处理的速度从每秒数十或数百考察点大幅度提升至每秒约10000考察点。对于常规分辨率的体图像，处理时间从过去的数小时缩短至数十秒。发展了一套离位MicroCT扫描结合DVC算法测量泡沫金属各阶段内部变形场的实验方法，设计了粘接在试样边缘的刚性标记块，通过图像轮廓分析，可准确估测离位扫描获得的多帧体图像之间的相对刚体平移和旋转，消除因此带来的DVC测量偏差。开发了一套基于三维体图像构造有限元模型和从有限元模型导出体图像的图像处理方法，提出了有限元单元变形映射DVC考察点变形的转换方法。在此基础上根据真实泡沫铝试样的体图像建立高保真度有限元模型，利用有限元仿真实验获得的变形场对DVC测量结果的准确性进行了可靠的评估。本研究对DVC方法的原理、实现和实验方法进行了创新性的探索，取得了一系列重要成果，有力地推动了DVC这一定量分析材料内部变形场演化的关键实验技术的发展和应用。