DeePore: a deep learning workflow for rapid and comprehensive characterization of porous materials (2005.03759v2)

Abstract: DeePore is a deep learning workflow for rapid estimation of a wide range of porous material properties based on the binarized micro-tomography images. By combining naturally occurring porous textures we generated 17700 semi-real 3-D micro-structures of porous geo-materials with size of 256^3 voxels and 30 physical properties of each sample are calculated using physical simulations on the corresponding pore network models. Next, a designed feed-forward convolutional neural network (CNN) is trained based on the dataset to estimate several morphological, hydraulic, electrical, and mechanical characteristics of the porous material in a fraction of a second. In order to fine-tune the CNN design, we tested 9 different training scenarios and selected the one with the highest average coefficient of determination (R^2) equal to 0.885 for 1418 testing samples. Additionally, 3 independent synthetic images as well as 3 realistic tomography images have been tested using the proposed method and results are compared with pore network modelling and experimental data, respectively. Tested absolute permeabilities had around 13 % relative error compared to the experimental data which is noticeable considering the accuracy of the direct numerical simulation methods such as Lattice Boltzmann and Finite Volume. The workflow is compatible with any physical size of the images due to its dimensionless approach and can be used to characterize large-scale 3-D images by averaging the model outputs for a sliding window that scans the whole geometry.

• The paper introduces DeePore, a deep learning workflow using a CNN trained on 17,700 3-D microstructures to rapidly estimate 30 physical properties of porous materials from binarized images.
• Numerical results show DeePore achieves high prediction accuracy with an average R^2 of 0.885 and a 13% permeability error relative to direct numerical simulations, providing results in milliseconds.
• DeePore's rapid characterization capabilities have significant practical implications for industries like petrochemicals and renewable energy by accelerating workflows and saving computational resources.

Overview of "DeePore: A Deep Learning Workflow for Rapid and Comprehensive Characterization of Porous Materials"

The paper presents a robust deep learning methodology designed to rapidly characterize porous materials using binarized micro-tomography images. This workflow, named DeePore, leverages a convolutional neural network (CNN) model to estimate 30 diverse physical properties of porous geo-materials, demonstrating an innovative intersection between machine learning and material science.

Methodological Innovation

The authors have generated a dataset comprising 17,700 semi-realistic 3-D micro-structures of porous geo-materials, sourced from combining naturally occurring porous textures. Each sample within the dataset is evaluated for 30 physical properties through simulations using pore network models. The CNN is trained to predict morphological, hydraulic, electrical, and mechanical characteristics with an average $R^2$ of 0.885 on 1418 testing samples, indicating a high-level predictive capability. Distinctively, the CNN model offers dimensional compatibility, allowing accurate assessments of varying image sizes through a dimensionless approach—a notable advantage for scaling applications.

Numerical Findings and Implications

The relative error in permeability estimates offered by DeePore compared to direct numerical simulation methods such as the Lattice Boltzmann and Finite Volume methods stands at 13%, underscoring the competitive accuracy of this deep learning approach. DeePore facilitates rapid characterization, achieving predictions in milliseconds, contrasting drastically with the computational demands of classic numerical methods.

Practical and Theoretical Implications

The implications of this research span both practical applications and theoretical advancements. Practically, the rapid estimation capabilities of DeePore can revolutionize workflows in industries reliant on porous material characterization, such as petrochemicals and renewable energy systems. Its application extends to diverse fields requiring quick assessments of materials' properties, potentially conserving both time and computational resources.

Theoretically, this research enriches the field of material characterization by illustrating deep learning’s capability to encode complex, nonlinear interactions within material micro-structures. DeePore’s success underscores the value in training models on large, diversified datasets, thereby enhancing predictive accuracy across varied morphological contexts.

Speculation on Future Developments in AI

Looking forward, the approach embodied by DeePore paves the way for further integration of machine learning techniques with material science, harnessing AI’s potential to address increasingly complex characterization challenges. There is scope for extending similar methodologies to characterize other material properties and behaviors, driving innovations across computational material science.

Conclusion

This paper effectively combines deep learning techniques with established methods in material science to offer a comprehensive workflow for porous material characterization. The results provide convincing evidence of the power and efficiency of AI-driven approaches, setting a significant precedent for future research and application in the domain of material characterization.