Experimental and Neural Network Modeling of Partial Uptake for a Carbon Dioxide/Methane/Water Ternary Mixture on 13X Zeolite

Abstract

In this work, GERG2008 EoS embedded in a volumetric–gravimetric technique was utilized to measure multicomponent partial uptakes into the mixture. The sophisticated combination may overlap recent theoretical measurements and replace it with real‐time and experimental selective adsorption analysis. 13X zeolite was utilized as a solid adsorbent for the adsorption of binary and ternary CO2/CH4/H2O mixtures. Premixed and preloaded water vapor was studied at 323 K temperature and up to 10 bar pressure. The isotherms of individual components within the mixture were identified and compared to the adsorption data of the pure components for assured benchmarking and validation. Artificial neural network (ANN) modeling was used to predict ternary mixtures. The ANN results showed a good agreement with the experimental data. Moreover, simulated configurations by utilizing an ANN model reflected the high consistency. We identified the behavior of the single components in ternary and higher multicomponent mixtures.

Experimental investigation, binary modelling and artificial neural network prediction of surfactant adsorption for enhanced oil recovery application

“…In binary mixtures, competitive adsorption is a controlling factor that usually takes place for between the adsorbate species on the solid surface. This phenomena shows that the adsorption of both adsorbates on the solid surface occurs with certain fractional loadings [47] . A revisited mathematical equation was proposed by Kurniawan et al [51] for and for a binary adsorption predictions as shown in the following equations (Eqs.…”

“…Where and are the amounts of APG and AEC adsorbed on the rock surface, respectively, that are calculated directly from pure experimental data [47] .…”

“…Typically, connecting nodes have varying values, making certain nodes more important than the others. Therefore, if a certain node is connected to different inputs, the system is capable of deciding the node that is more important than the other according to the connection weights [47] .…”

“…Over the years, several theoretical models have been developed to provide the behaviour of predicted mixtures by the exploitation of the individual components’ adsorption data. Among these models are the Langmuir model, the ideal adsorbed solution theory (IAST), molecular simulations two-dimensional equations of state (2D EOS) and Discrete Fourier transform (DFT) [47] . The Langmuir adsorption isotherm has been utilised extensively to model the adsorption of individual surfactants.…”

High‐temperature CO2 removal from CH4 using silica membrane: experimental and neural network modeling

“…A feedforward back‐propagation neural network (FFBP) is considered one of the most applied learning algorithms as it moves forward with the data sets in hidden layers, composed of artificial functional neurons. The algorithm predicts outcomes according to the provided experimental data, checks for errors, and repeats the process with a different number of neurons and network weights, to minimize mean square errors (MSEs) …”