Scholarlink

<p>The present article investigates methods for calculating CO<sub>2</sub> and CH<sub>4</sub> adsorption isotherms on coconut-shell-derived activated carbon (AC) at pressures up to 50 bar and temperatures ranging from 253 to 343 K. The research aims to improve the accuracy of existing isotherm models by using empirical models (DA, modified DA and Toth) and a new machine learning model developed with a multilayer feed-forward neural network (MLFNN), benefiting gas storage and separation applications. Before performing empirical/artificial neural network (ANN) modelling, the equilibrium adsorption concentration as a function of pressure and temperature must be determined. Hence, an experimental facility is designed to investigate the equilibrium adsorption of CH<sub>4</sub> and CO<sub>2</sub> onto AC. To ensure the adsorbent’s suitability, its structural, textural and elemental characteristics are also evaluated. The AC exhibited a high Brunauer-Emmett-Teller surface area (1254&#xa0;m<sup>2</sup>/g), micropore volume (0.4&#xa0;cm<sup>3</sup>/g), and dominant microporosity with pore widths primarily between 4 and 20 Å, confirming its effectiveness for gas adsorption. Elemental analysis revealed a high carbon content (89.5 wt.%) with low ash (2.1 wt.%) and moisture (3 wt.%), indicating favourable surface chemistry and high purity. Following this characterization, adsorption experiments are conducted using the developed test rig. A maximum CO<sub>2</sub> concentration of 0.70 kg/kg is observed at 278 K and 32 bar, while CH<sub>4</sub> adsorption reached 0.140&#xa0;kg/kg at 253 K and 40 bar. These experimental results are then used to train and evaluate different models. The results show that, the ANN model outperforms the empirical models, with AARD values of 3.12% for CH<sub>4</sub> and 4.39% for CO<sub>2</sub>. The study also shows that ANN models are more effective at predicting isotherm data at higher temperatures (i.e. 283–343 K) than empirical isotherm models.</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

Experimental and neural network-based modelling of CO2 and CH4 adsorption

  • 期刊论文

  • 发表日期：

    2026年2月13日

查看全文
  • Anupam Chaudhary,
  • Vidushi Omar,
  • Biranchi Prasad Mishra,
  • Satyabrata Sahoo

摘要

The present article investigates methods for calculating CO2 and CH4 adsorption isotherms on coconut-shell-derived activated carbon (AC) at pressures up to 50 bar and temperatures ranging from 253 to 343 K. The research aims to improve the accuracy of existing isotherm models by using empirical models (DA, modified DA and Toth) and a new machine learning model developed with a multilayer feed-forward neural network (MLFNN), benefiting gas storage and separation applications. Before performing empirical/artificial neural network (ANN) modelling, the equilibrium adsorption concentration as a function of pressure and temperature must be determined. Hence, an experimental facility is designed to investigate the equilibrium adsorption of CH4 and CO2 onto AC. To ensure the adsorbent’s suitability, its structural, textural and elemental characteristics are also evaluated. The AC exhibited a high Brunauer-Emmett-Teller surface area (1254 m2/g), micropore volume (0.4 cm3/g), and dominant microporosity with pore widths primarily between 4 and 20 Å, confirming its effectiveness for gas adsorption. Elemental analysis revealed a high carbon content (89.5 wt.%) with low ash (2.1 wt.%) and moisture (3 wt.%), indicating favourable surface chemistry and high purity. Following this characterization, adsorption experiments are conducted using the developed test rig. A maximum CO2 concentration of 0.70 kg/kg is observed at 278 K and 32 bar, while CH4 adsorption reached 0.140 kg/kg at 253 K and 40 bar. These experimental results are then used to train and evaluate different models. The results show that, the ANN model outperforms the empirical models, with AARD values of 3.12% for CH4 and 4.39% for CO2. The study also shows that ANN models are more effective at predicting isotherm data at higher temperatures (i.e. 283–343 K) than empirical isotherm models.