TL;DR
This paper investigates three methods (BSE image analysis, Katz-Thompson, and Winland models) for predicting the gas permeability of mortar samples, a crucial indicator of durability. The Winland model, using pore size data from mercury intrusion porosimetry, proved to be the most accurate for predicting gas permeability, outperforming BSE-based methods and the Katz-Thompson model.
What Problem Does This Paper Address?
The durability of cementitious materials, like concrete and mortar, is paramount for the longevity of infrastructure. Gas permeability is a key metric for assessing this durability, as it dictates the ingress of harmful gases like CO2 and O2, which can lead to material degradation (e.g., carbonation, steel corrosion). However, directly measuring gas permeability can be complex, requiring specialized equipment and precise sealing, making it time-consuming and costly. Therefore, developing reliable methods to predict gas permeability based on a material's pore structure is highly desirable. This paper addresses the challenge of accurately predicting mortar's gas permeability by comparing the effectiveness of different predictive models that utilize pore characteristics derived from two common characterization techniques: backscattered electron (BSE) imaging and mercury intrusion porosimetry (MIP).
Methods
The researchers employed a multi-pronged approach to predict gas permeability:
- Laboratory Measurement: Intrinsic gas permeability was directly measured using a CEMBUREAU method apparatus, employing helium gas and a confining pressure of 4 MPa. The Klinkenberg equation was used to account for slippage effects at low pressures.
- BSE Image Analysis: Mortar samples were analyzed using scanning electron microscopy (SEM) with backscattered electrons (BSE). Image processing techniques, including Yen's algorithm for thresholding and a watershed algorithm, were used to extract pore structure information and generate pore size distributions. The Hagen-Poiseuille equation, a fundamental fluid flow equation for porous media, was then applied using these extracted pore parameters.
- Mercury Intrusion Porosimetry (MIP): MIP was used to characterize the pore structure, providing data on porosity, pore size distributions, and critical pore diameters.
- Predictive Models:
- Hagen-Poiseuille Equation (with BSE data): This model, based on fluid flow through capillary tubes, was adapted using pore parameters derived from the processed BSE images.
- Katz-Thompson Equation: This model, rooted in percolation theory, uses the critical pore diameter obtained from MIP to predict permeability.
- Winland Model: An empirical model that uses the pore diameter corresponding to a specific cumulative mercury saturation (35% in this case) from MIP data.
Datasets
The study utilized mortar samples with varying silica fume (SF) content (0%, 5%, and 10%). These samples were prepared according to standard procedures and characterized using the methods described above. No new benchmark datasets were introduced; rather, existing characterization techniques and predictive models were applied to the prepared mortar samples.
Key Insights and Ideas
- BSE Limitations: While BSE imaging provides visual insights into pore structure, its ability to accurately predict gas permeability is limited. The relatively small observation range and the difficulty in capturing the complete, interconnected pore network mean that predictions can be significantly off.
- Katz-Thompson Underestimation: The Katz-Thompson model, while theoretically grounded, tends to underestimate gas permeability. This is attributed to its strong reliance on the critical pore diameter and its failure to adequately account for the contribution of larger, coarse capillary pores and pore connectivity.
- Winland Model's Empirical Strength: The Winland model, despite being empirical, demonstrated a strong correlation with measured gas permeability. This suggests that empirical relationships, when calibrated with relevant data, can be highly effective for predicting transport properties like permeability.
- Silica Fume's Impact: The incorporation of silica fume consistently reduced the gas permeability of the mortar samples. This is attributed to its pozzolanic reaction and pore-filling effects, which refine the pore structure and reduce overall porosity.
Key Results
- Measured Gas Permeability: The intrinsic gas permeability of the mortar samples decreased with increasing silica fume content, with values ranging from 5.49 × 10⁻¹⁷ m² (SF0) to 7.61 × 10⁻¹⁸ m² (SF10).
- BSE Prediction Accuracy: Predictions from BSE images showed significant deviations. Using 1000x magnification, the predicted permeability was 1.5-3 times higher than measured values, while 4000x magnification resulted in predictions approximately 5 times lower.
- Katz-Thompson Prediction Accuracy: The Katz-Thompson model consistently underestimated gas permeability by about two orders of magnitude compared to laboratory measurements.
- Winland Model Prediction Accuracy: The Winland model provided the closest predictions to the measured gas permeability values, with deviations ranging from -62.94% to +24.77% (though the paper states "lower or larger," the table suggests it's generally closer).
Limitations and Future Work
- BSE Limitations: The authors acknowledge that BSE image processing has limitations in fully capturing the actual pore structure due to the limited observation scope and magnification. They suggest that 3D imaging techniques (e.g., X-ray micro-CT, FIB/SEM) could offer a more comprehensive understanding of pore tortuosity and connectivity. Additionally, the difference in confining pressure between the permeability measurement (4 MPa) and BSE observation (zero) could influence the results.
- Katz-Thompson Limitations: This model's theoretical basis, while valuable, overlooks pore connectivity and the contribution of coarser pores, leading to underestimation.
- Future Work: The paper suggests that future research could explore 3D imaging techniques for a more complete pore structure analysis and investigate the influence of confining pressure on permeability prediction. Incorporating additional parameters into BSE-based predictions, such as confining pressure, could also improve accuracy.
Summary updated