Hybrid-RC-Model-of-Dynamic-Adsorption-Cycle
Description
A new hybrid mesoscopic RC model has been developed to describe dynamic adsorption cycles on isothermal surfaces. The model has been rigorously validated through three independent campaigns: one inter‑model comparison and two validations based on published scientific studies. This file repository consolidates the reference datasets employed in these validation exercises and provides the computational codes necessary to extract and calibrate the parameters of the proposed model. The results confirm the robustness and general applicability of the hybrid approach, establishing a reliable framework for the simulation and optimization of adsorption/desorption processes across diverse adsorbent systems.
Files
Steps to reproduce
Inspired by the Python program created by N. Fette available at this link: https://www.mycompiler.io/view/DTj7zH0xIB6 , a new program has been written to simulate the amount of adsorbate X (in kg adsorbate/kg adsorbent) and the temperature T (K) of an adsorber over several complete cycles of adsorption-desorption. This code allows the simulation of these X and T outputs according to seven different dynamic models: Freundlich associated with an LDF dynamics (F+LDF), Dubinin-Radushkevich (DR+LDF), Dubinin-Astakhov (DA+LDF), Saha-Boelman-Kashiwagi (SBK+LDF), DA with a Logarithmic Mean Temperature Difference dynamics (DA+LMTD), DA with the Finite Volume Method (DA+FVM), and the new hybrid model. The whole of this program is in file sorption-mass-temperature.py and requires the CoolProp module (library of thermophysical properties of fluids) to work. At the end of the code execution, the graphs representing the quantity of adsorbate X and the temperature T are drawn. The table models.csv exports the data for each of the models as well as the average of the six reference models (F+LDF, DR+LDF, DA+LDF, SBK+LDF, DA+LMTD, DA+FVM). Please, note that in this case, the new model settings are determined manually. Then, the other programs (comparison-models.py, Amin.py and Brancato.py) are built on the same structure. It is to determine the optimal parameters of the new model in order to minimize the deviation from the reference data, respectively models.csv, Amin.csv and Brancato.csv. This optimization is achieved by the Broyden-Fletcher-Goldfarb-Shanno method with the Large-scale and Bound-constrained options (L-BFGS-B) and requires the scikit-learn module for least squares calculations. This method was chosen according to the article by Bemporad [4]. To obtain the reference data corresponding to the studies by Amin and Brancato et al., the web tool WebPlotDigitizer (see https://automeris.io) was used to digitize the Figures. The extracted data were then discretized to match a chosen time step (10 s for Amin et al., 60 s for Brancato et al.). Inflection points in the curves were used to determine the duration of each phase: adsorption, preheating, desorption, and precooling. For Amin’s study, the phase durations are identical for each cycle, whereas in the Brancato experiments, the durations vary between cycles. The complete data processing workflow, from the original graphs to the reference datasets, is documented in the Excel workbooks Amin.xlsx and Brancato.xlsx.