Harbin Institute of Technology (Zhang Youyuan, Yu Yuan/Report, School of Environment/Photos) Recently, Professor You Shijie’s team from the School of Environment at Harbin Institute of Technology (HIT) has made significant breakthroughs in deep learning computer vision and transfer learning for visualizing catalytic reaction transport. Their research findings, titled Visualizing nexus of porous architecture and reactive transport in heterogeneous catalysis by deep learning computer vision and transfer learning were published in Nature Communications.
In fields such as chemistry, energy, and environmental science, heterogeneous catalysis is a core driver of technological innovation. Its reaction efficiency is highly dependent on the composition of the porous catalyst and the complex internal microstructure. However, precisely analyzing the structure–performance relationship, especially quantifying the local reaction transport processes in three-dimensional space, has long been a major challenge for the scientific community. Traditional experimental and simulation methods are not only expensive and time-consuming but also struggle to capture the dynamic details of reaction processes, thus limiting the design and development of high-performance catalysts.
cGAN-TL Model Framework and Workflow
To address the above-mentioned challenges, Professor You Shijie’s team proposed a novel deep learning computer vision (DLCV) method based on the integration of Conditional Generative Adversarial Networks (cGAN) and Feature Representation Transfer Learning (FRT). This method utilizes a three-dimensional reconstruction network to restore the internal 3D structure of the catalyst from two-dimensional side-view images. By combining reaction conditions, it predicts the distribution of local reaction rates, achieving the mapping from 2D images to 3D local reaction information. DLCV, through its attention mechanism, identifies key areas such as pore throats, curved flow channels, fractal pore structures, and their combinations as critical regions for controlling reaction transport performance. The method, in conjunction with the theory of physical field synergy, reveals the intrinsic correlation mechanism between mass transfer enhancement and reaction kinetics, making the visualization of catalytic reaction kinetics possible. This highlights the vast application potential of artificial intelligence in solving complex scientific problems and signals a shift in catalyst research from the traditional trial-and-error approach to the efficient AI-driven design paradigm. In the future, this method is expected to be applied to broader scenarios, such as fuel cells, carbon dioxide electro-reduction, and environmental pollution control, significantly accelerating the research and development process of new materials and structures, thus providing strong theoretical and technical support for addressing global energy and environmental challenges.
The State Key Laboratory of Urban-rural Water Resource and Environment is the first affiliation for this paper, with Assistant Researcher Yu Yuan from the School of Environment as the first author, and Professor You Shijie as the corresponding author. The research was supported by the National Natural Science Foundation of China, the China Postdoctoral Science Foundation, and the Heilongjiang Provincial Postdoctoral Foundation.
