Efficient Surrogates Construction of Chemical Processes: Case studies on Pressure Swing Adsorption and Gas-to-Liquids

07 October 2021, Version 1
This content is a preprint and has not undergone peer review at the time of posting.


We propose a workflow for reduction in the time required for data generation during generation of statistical digital twins. This methodology is particularly relevant for real-world engineering problems when data generation is expensive. A prerequisite for building surrogates is sufficient input/output data, whereas over-sampling can hardly improve the regression accuracy. The time for data generation can be reduced via (1) reduction of the average time spent on generating individual data points and (2) reduction in the total number of data points, by reducing the sampling rate with the improvement of surrogate quality. Examples of a dynamic process and a steady-state process from the field of carbon capture and utilization are used as two case studies: pressure swing adsorption (PSA) and Gas-to-Liquids (GTL). With the proposed methodology, the time for surrogate generation can be reduced by 88% for PSA and 60% for GTL, respectively.


machine learning
pressure swing adsorption
digital twin


Comments are not moderated before they are posted, but they can be removed by the site moderators if they are found to be in contravention of our Commenting Policy [opens in a new tab] - please read this policy before you post. Comments should be used for scholarly discussion of the content in question. You can find more information about how to use the commenting feature here [opens in a new tab] .
This site is protected by reCAPTCHA and the Google Privacy Policy [opens in a new tab] and Terms of Service [opens in a new tab] apply.