Thermal conductivity analysis of polymer-derived nano-composite via image-base structure reconstruction, computational homogenization and machine learning

This dataset includes supplementary data and utilities for validating simulation results and training machine learning models as outlined in the publication titled "Thermal Conductivity Analysis of Polymer-Derived Nanocomposite via Image-Based Structure Reconstruction, Computational Homogenization, and Machine Learning" (Fathidoost, 2024). This dataset containes the microstructure images (identified by particle diameters size \(D_1\) and \(D_2\) volume fraction \(V_\mathrm{f}\) and aspect ratio \(A_\mathrm{r}\)) (see Table 1) and their corresponding homogenized thermal conductivity. these images resemble the microstructure of the monolithic \(\mathrm{(Hf,Ta)C/SiC}\) ceramic following FAST sintering, the material system of this work (Fathidoost, 2024). White and black colors within the images represent distinct regions of the material system, respectively referring to former powder particles (FPPs) and sinter necks (SNs), which is explained in this work. Table 1. Parameterized descriptors extracted from the mesoscale SEM image analysis Param. Mean [unit] Std. \(D_{1}\) 40, 50, 60 [μm] 20% \(D_{2}\) 20, 25, 26, 30, 33, 40 [μm] 30% \(V_\mathrm{f}\) 1.5, 2.0 - \(A_\mathrm{r}\) 35, 40, 45, 55, 60 [%] - This dataset contains: dataset.csv: containing a summary of data including the names of microstructure images, their corresponding geometric details, as well as the first and third principal components of two-point statistics for all images, along with the effective thermal conductivity of the corresponding microstructures. Further details can be found in the associated publication. microstructures_images.zip: containing binary cross-section images of the RVEs from synthetic microstructures。 results.zip: contains all the simulation results based on digitized diffuse-interface microstructures, which can be opened by the post-processing software, such as ParaView.本数据集包含补充数据与实用工具，用于验证论文《基于图像结构重建、计算均匀化与机器学习的聚合物基纳米复合材料导热性能分析》（Fathidoost, 2024）中所述的仿真结果验证与机器学习模型训练工作。 本数据集包含由颗粒直径(D_1)、(D_2)、体积分数(V_mathrm{f})与长径比(A_mathrm{r})表征的微观结构图像（详见表1）及其对应的均匀化导热系数。这些图像模拟了本研究采用的材料体系——经FAST烧结制备的块体(mathrm{(Hf,Ta)C/SiC})陶瓷的微观结构（Fathidoost, 2024）。图像中的白色与黑色分别代表该材料体系的两个不同区域，即前驱体粉末颗粒（Former Powder Particles, FPPs）与烧结颈（Sinter Necks, SNs），相关细节已在本论文中阐述。 表1 介观尺度扫描电子显微镜（Scanning Electron Microscope, SEM）图像分析提取的参数化描述符 | 参数 | 均值[单位] | 标准差 | | :--- | :-------- | :----- | | (D_1) | 40、50、60 [μm] | 20% | | (D_2) | 20、25、26、30、33、40 [μm] | 30% | | (V_mathrm{f}) | 1.5、2.0 | - | | (A_mathrm{r}) | 35、40、45、55、60 [%] | - | 本数据集包含以下内容： - `dataset.csv`：包含数据汇总文件，内容涵盖微观结构图像名称、对应几何参数、所有图像的两点统计量第一与第三主成分，以及对应微观结构的有效导热系数。更多细节可参见相关论文。 - `microstructures_images.zip`：包含合成微观结构的代表性体积单元（Representative Volume Element, RVEs）二值截面图像。 - `results.zip`：包含基于数字化扩散界面微观结构的全部仿真结果，可通过ParaView等后处理软件打开。