Prediction of H2S Solubility in Liquid Electrolytes by Multilayer Perceptron and Radial Basis Function Neural Networks. Issue 2 (21st December 2016)

• Record Type: Journal Article
• Authors: Barati‐Harooni, Ali; Nasery, Saeid; Tatar, Afshin; Najafi‐Marghmaleki, Adel; Isafiade, Adeniyi Jide; Bahadori, Alireza
• Other Names: Thrän Daniela sponsoringEditor.; Pfeiffer Diana sponsoringEditor.
• Is Part Of: Chemical engineering & technology. Volume 40:Issue 2(2017)
• Page Start: 367
• Page End: 375
• Publication Date: 2016-12-21
• Subjects: Genetic algorithm -- H2S solubility -- Ionic liquids -- Radial basis function -- Simple multilayer perceptron; Chemical engineering -- Periodicals; Chemical processes -- Periodicals; 660.05
• DOI: 10.1002/ceat.201600110
• Languages: English
• ISSNs: 0930-7516

Abstract

Abstract: Industrial natural gas treating plants commonly employ amine‐based treatments for hydrogen sulfide elimination from crude oil and gas. Some deficiencies boost the motivation to find an appropriate alternative. Due to their advantageous properties, liquid electrolytes are considered as possible substitutes for classical alkanolamine solvents in such processes. The solubility of gases in ionic solutions at different temperatures and pressures is a crucial factor in the examination of ionic liquids as a potential alternative. Two intelligent methods, namely, simple multilayer perceptron (MLP) and radial basis function neural networks, are proposed to accurately predict the solubility of H2 S in various ionic liquids. The predicted values agree well with the experimental data. A comparison to other intelligent models, which were recently suggested, reveals the superiority of the proposed simple MLP model. Abstract : Two intelligent models named multilayer perceptron and radial basis function optimized by a genetic algorithm were developed to accurately predict the solubility of H2 S in various ionic liquids based on experimental data from literature. A comparison to previously presented intelligent models demonstrates the superiority of the proposed models without performing lengthy experiments.