(531g) Surrogate-Model-Based Wetted Wall Column Data Interpretation for Amine-Based Carbon Capture Kinetics Identification

The successful deployment of amine-based post-combustion carbon capture technologies is greatly facilitated by the development of predictive process models that adequately describe all the phenomena occurring within an absorber or desorber column. Among these phenomena is the reaction kinetics, which contributes largely to the differentiation between one solvent and another. Therefore, assessing the reaction kinetics for a given solvent or solvent system is the subject of extensive research.

Wetted-wall columns (WWCs) have become the standard apparatus for measuring these kinetics due to their simplicity and reliability. WWCs also provide data at flow and contact conditions relevant for large-scale gas absorption systems. A reliable framework has been developed by many researchers to quantify the overall mass transfer coefficient (commonly denoted by K_G), and to distinguish the effect of gas-phase (k_g) and liquid-phase (k_g')coefficients in WWCs. The liquid-phase coefficient is of particular interest due to its dependence on the solvent. Using this framework of data interpretation, one can readily make meaningful comparisons between the liquid phase resistance to mass transfer exhibited by different solvents. In addition, heuristic estimates of key process parameters such as the packing area needed for achieving a specified capture rate can be made using this framework for WWC data interpretation.

Nevertheless, a method for the derivation of reaction rate expressions directly from WWC data remains a challenging task. A notable contribution is that of Sherman [1], who implemented an Aspen Plus RateSep model for the WWC, and explored several approaches to obtain reaction rate expression parameters from the WWC data. These included (1) a manual fitting method, which involves adjusting the parameters manually to minimize the difference between the model and the experimental data, (2) the Aspen Plus Data Fit tool, which uses the maximum-likelihood estimation method to automate the search, and (3) the Response Surface Methodology (RSM). Sherman noted that the manual approach, although laborious, outperformed the Aspen Plus Data Fit tool which was computationally expensive and did not yield better parameters than the manual approach. The RSM approach combined the strengths of the two other approaches giving more reliable results, but was highly sensitive to the initial guess and was difficult to implement.

In this work, we present a surrogate-modeling based, easy-to-implement methodology for fitting the parameters of the kinetic rate expressions using WWC data. We first develop an Aspen Plus^® simulation of a WWC with monoethanolamine (MEA) as the solvent [2] incorporating appropriate models for mass transfer, interfacial area, and holdup within the column. The model is validated against literature data. We then employ Latin Hypercube sampling to generate WWC performance predictions over a broad range of kinetic parameter values. Surrogate models are subsequently trained to relate the predicted WWC performance to the kinetic parameters. Finally, an optimization is conducted to minimize the difference between the predicted and experimental fluxes by varying the kinetic parameters. The final parameter estimates are validated using the full-fidelity model, and the surrogate-modeling approaches are compared in terms of accuracy and computational efficiency.

References

[1] Sherman, B.J. (2016). Thermodynamic and Mass Transfer Modeling of Aqueous Hindered Amines for Carbon Dioxide Capture [Doctoral Dissertation, The University of Texas at Austin].

[2] Chinen, A. S., Morgan, J. C., Omell, B., et al. (2018). Industrial & Engineering Chemistry Research, 57(31), 10448–10463.