Scholarlink

<p>In this paper, a novel computational scheme based on the integration of Molecular Dynamics (MD) simulation and Artificial Neural Networks (ANNs) modeling was developed to analyze and predict the selectivity separation of noble gases (Kr, Xe, Ne, and He) with activated carbon nanochannels. In the ANN model presented herein, a 20-20-20 hidden layer configuration was adopted to establish the non-linear relationship between physical characteristics and adsorption capacity. Results show that the ANN prediction model achieves satisfactory predictive accuracy (R² ≈0.93, rMAE = 13.13%, rRMSE = 21.59%). The model can therefore serve as an efficient interpolation tool for estimating adsorption behavior within the range of conditions represented in the MD simulation dataset. Moreover, the results obtained from the variable importance analysis indicate that the two main features that influence the process are pore size and gas type. Finally, the hybrid scheme indicated that the P14 configuration, where pore size is 8.3&#xa0;A, is the best configuration for the separation of heavy noble gases.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adsorption and permeation behavior analysis of activated carbon nanochannels using an MD–ANN approach: a case study of the Kr–Xe–He–Ne mixture

  • 期刊论文

  • 发表日期：

    2026年8月4日

查看全文
  • Hamed Rouhi,
  • Iraj Jabbari,
  • Mahdi Tourang

摘要

In this paper, a novel computational scheme based on the integration of Molecular Dynamics (MD) simulation and Artificial Neural Networks (ANNs) modeling was developed to analyze and predict the selectivity separation of noble gases (Kr, Xe, Ne, and He) with activated carbon nanochannels. In the ANN model presented herein, a 20-20-20 hidden layer configuration was adopted to establish the non-linear relationship between physical characteristics and adsorption capacity. Results show that the ANN prediction model achieves satisfactory predictive accuracy (R² ≈0.93, rMAE = 13.13%, rRMSE = 21.59%). The model can therefore serve as an efficient interpolation tool for estimating adsorption behavior within the range of conditions represented in the MD simulation dataset. Moreover, the results obtained from the variable importance analysis indicate that the two main features that influence the process are pore size and gas type. Finally, the hybrid scheme indicated that the P14 configuration, where pore size is 8.3 A, is the best configuration for the separation of heavy noble gases.