Affiliation:
1. Chemical Engineering Department University of Waterloo Waterloo Ontario Canada
2. Natural Resources Canada (NRCan) CanmetENERGY‐Ottawa (CE‐O) Ottawa Ontario Canada
3. Department of Chemical Engineering Khalifa University Abu Dhabi United Arab Emirates
Abstract
AbstractTo address the deficiency and predict the adsorption performance in different adsorbents, this study proposes a new optimizer linked to the machine learning (ML) model considering the performance of the adsorption process. The main goal is to predict adsorption under different process conditions with different adsorbents and provide a unified framework, leading to the prediction of adsorption phenomena instead of traditional isotherm models. This research focuses on predicting the adsorbed amount of O2 and N2 on several carbon‐based adsorbents using the ML approach linked to the grey wolf optimizer algorithm (GWO). Experimental isotherm data (dataset 1344) on adsorbent type, temperature, pressure, gas type, and adsorption capacity of the process adsorption were used as input and output datasets. The best algorithm was Broyden–Fletcher–Goldfarb–Shanno (BFGS), a two‐layer network from a multi‐layer perceptron (MLP) method applying 28 neurons. The new MLP‐GWO network would have the best mean square error (MSE) efficiencies of 0.00037, while the R2 (r‐squared) was 0.9934. The new ML‐generated model can accurately predict the adsorption process behaviour of different carbon‐based adsorbents under various process conditions. The results of this research have the potential to assist a wide range of gas separation industries.
Funder
Natural Sciences and Engineering Research Council of Canada
Natural Resources Canada
Subject
General Chemical Engineering