Speaker
Description
Understanding the saturated and unsaturated flow in porous media by producing ceramic porous model samples with controlled morphology. By controlling the morphology over a large range of microstructure, the study aims to isolate the parameters influencing resin impregnation and permanent flow in porous media. This subject has been treated by the community with many different approaches [1]. Unfortunately, existing models often fail to predict flow behavior correctly in cases where the porous medium is unsaturated, particularly during infusion. Compared to deformable fibrous media, porous ceramic model samples allow limiting and controlling the geometric variability of the porous network. Aiding in isolating the parameters influencing resin impregnation regimes in the material. This study has applications in the medical field (ceramics/polymers).
The medium-term objective is to develop models to better understand fluid flow in complex and controlled porous media [2]. To support this goal, a comprehensive experimental database is currently being built based on the study of porous ceramics manufactured with the sacrificial template method. First, an active-learning algorithm based on Gaussian Process Classification (GPC) has been developed to efficiently identify the parameters and boundaries of the chosen porous ceramic manufacturing process, with a minimal number of trial iterations. This approach is particularly advantageous for processes involving multiple parameters, where classical experimental designs would require extensive testing. We demonstrate the predictive capability of the algorithm for a test case involving two varying parameters: porogen volume and size.
Second, instrumented infusion tests are performed with an in-house set-up able to measure samples permeability from 10^(-16) to 10^(-12) m². Based on these measurements, a regression model is developed to predict permeability from the porogen characteristics (volume fraction of 2 classes of porogen). In parallel, the samples are characterized to quantify their internal structure (e.g., pore-size distribution) [3], enabling the quantitative assessment of how these parameters influence the fluid flow behavior.
Finally, dedicated descriptors are used to represent the 2D pore morphological features extracted from image-based characterization. These features are projected into a latent space using dimensionality-reduction techniques to obtain a compact representation of the pore morphology. Thus, regression is performed between reduced descriptors and permeability to establish a quantitative pore structure–property relationship. The study could bring insight into the relevant features of porous geometry that affect the permeability.
References
[1] D. Lee, M. Ruf, N. Karadimitriou, H. Steeb, M. Manousidaki, E.A. Varouchakis, S. Tzortzakis, A. Yiotis, Development of stochastically reconstructed 3D porous media micromodels using additive manufacturing: numerical and experimental validation, Sci. Rep. 14 (2024) 9375. https://doi.org/10.1038/s41598-024-60075-w.
[2] L. Xie, Q. You, E. Wang, T. Li, Y. Song, Quantitative characterization of pore size and structural features in ultra-low permeability reservoirs based on X-ray computed tomography, J. Pet. Sci. Eng. 208 (2022) 109733. https://doi.org/10.1016/j.petrol.2021.109733.
[3] S. Nickerson, Y. Shu, D. Zhong, C. Könke, A. Tandia, Permeability of porous ceramics by X-ray CT image analysis, Acta Mater. 172 (2019) 121–130. https://doi.org/10.1016/j.actamat.2019.04.053.
| Country | France |
|---|---|
| Acceptance of the Terms & Conditions | Click here to agree |
