抄録
Adsorption-based CO2 capture is widely recognized as an effective strategy for mitigating CO2 emissions from industrial processes. The integration of deep learning has accelerated materials science by enabling the analysis of large datasets and accurate property prediction, thereby facilitating the rational design of next-generation carbon capture materials. In this study, a fully connected neural network (FCNN) was developed to predict the CO2 adsorption isotherms of KOH-activated AC (activated carbon), based on 19,540 data points spanning a wide variety of precursors, including biomass-derived and high-carbon-content compounds. All the ACs were synthesized using the same activation route, with nitrogen either inherently present or introduced via doping. KOH was treated as a controlled descriptor because the influence of different activating agents may not be fully captured by the existing physicochemical features. The results identified the volume of narrow micropores (Vn, pore diameter < 1 nm) as the most critical factor affecting CO2 uptake, while the nitrogen content (N%) exhibited a nonlinear contribution, enhancing adsorption at low levels but becoming detrimental when exceeding 13 %. To validate both the predictive power of the model and the reliability of its variable interpretation, external validation was performed using nine unseen AC/CO2 isotherms. The model achieved an R2 of 0.9606 and a root mean squared error of 0.0106 kg/kg. These results demonstrate that the proposed FCNN framework is capable of accurate performance prediction and serves as a valuable tool for guiding the design and optimization of CO2 adsorbents based on their physicochemical characteristics.
| 本文言語 | 英語 |
|---|---|
| 論文番号 | 104188 |
| ジャーナル | Thermal Science and Engineering Progress |
| 巻 | 67 |
| DOI | |
| 出版ステータス | 出版済み - 11月 2025 |
!!!All Science Journal Classification (ASJC) codes
- 流体および伝熱
フィンガープリント
「Adsorption isotherm prediction of KOH activated AC/CO2 pairs employing a fully connected neural network」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。引用スタイル
- APA
- Standard
- Harvard
- Vancouver
- Author
- BIBTEX
- RIS