Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

3D microstructural generation from 2D images of cement paste using generative adversarial networks

Xin Zhao, Lin Wang, Qinfei Li, Heng Chen, Shuangrong Liu, Pengkun Hou, Jiayuan Ye, Yan Pei, Xu Wu, Jianfeng Yuan, Haozhong Gao, Bo Yang

Abstract

Establishing a realistic three-dimensional (3D) microstructure is a crucial step for studying microstructure development of hardened cement pastes. However, acquiring 3D microstructural images for cement often involves high costs and quality compromises. This paper proposes a generative adversarial networks-based method for generating 3D microstructures from a single two-dimensional (2D) image, capable of producing high-quality and realistic 3D images at low cost. In the method, a framework (CEM3DMG) is designed to synthesize 3D images by learning microstructural information from a 2D cross-sectional image. Experimental results show that CEM3DMG can generate realistic 3D images of large size. Visual observation confirms that the generated 3D images exhibit similar microstructural features to the 2D images, including similar pore distribution and particle morphology. Furthermore, quantitative analysis reveals that reconstructed 3D microstructures closely match the real 2D microstructure in terms of gray level histogram, phase proportions, and pore size distribution. The source code for CEM3DMG is available in the GitHub repository at: https://github.com/NBICLAB/CEM3DMG.

3D microstructural generation from 2D images of cement paste using generative adversarial networks

Abstract

Establishing a realistic three-dimensional (3D) microstructure is a crucial step for studying microstructure development of hardened cement pastes. However, acquiring 3D microstructural images for cement often involves high costs and quality compromises. This paper proposes a generative adversarial networks-based method for generating 3D microstructures from a single two-dimensional (2D) image, capable of producing high-quality and realistic 3D images at low cost. In the method, a framework (CEM3DMG) is designed to synthesize 3D images by learning microstructural information from a 2D cross-sectional image. Experimental results show that CEM3DMG can generate realistic 3D images of large size. Visual observation confirms that the generated 3D images exhibit similar microstructural features to the 2D images, including similar pore distribution and particle morphology. Furthermore, quantitative analysis reveals that reconstructed 3D microstructures closely match the real 2D microstructure in terms of gray level histogram, phase proportions, and pore size distribution. The source code for CEM3DMG is available in the GitHub repository at: https://github.com/NBICLAB/CEM3DMG.
Paper Structure (26 sections, 5 equations, 19 figures, 2 tables)

This paper contains 26 sections, 5 equations, 19 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Experimental Method
  3. Specimen Preparation
  4. BSE Image Acquisition
  5. 3D Microstructural Image Generation
  6. CEM3DMG
  7. Multi-scale Learning Strategy
  8. Block-wise Synthesis Strategy
  9. Adversarial Training
  10. Experimental Configuration
  11. Analytical Evaluation Criteria
  12. Gray Level Histogram
  13. Various Phases Proportions
  14. Pore Size Distribution
  15. Results and Discussion
  16. ...and 11 more sections

Figures (19)

  • Figure 1: Various 3D microstructures. (a) represents the microstructure generated by our method, (b) denotes the real microstructure, and (c) indicates the spherical-based microstructure.
  • Figure 2: 3D microstructural image acquisition process. (a) the processed of hardened cement paste specimen preparation, (b) the processed of specimen is scanned to obtain BSE images, (c) the 3D microstructural images are generated from a given single 2D BSE image.
  • Figure 3: 3D microstructural image generation framework (CEM3DMG). (a) the generator and discriminators in GAN-based adversarial training. (b) the multi-scale learning strategy.
  • Figure 4: Conceptual diagram of the block-wise synthesis strategy. (a) Input data, (b) and (c) show (a) divided into two smaller input data blocks. The red numbers indicate the overlapping parts between (b) and (c). (d) represents the convolutional operator in the generator (no padding). After processing by the same convolutional operator (d), (a), (b), and (c) produce the output (e), (f), and (g), respectively.
  • Figure 5: Phases segmentation. Pores, hydration products and unreacted clinkers are segmented based on histogram. In (b), the vertical axis represents the proportion of the number of pixels with the same gray level value to the total number of pixels.
  • ...and 14 more figures