A simplified digital twin of a pressure swing adsorption plant for air separation

TL;DR

This paper introduces a simplified yet accurate digital twin CFD model of a pressure swing adsorption (PSA) plant for air separation, aiding design and optimization through simulation of key processes and components.

Contribution

The work presents a robust, generalizable digital twin model of a PSA system that accurately replicates plant performance and cyclic operation, facilitating design and industrial application.

Findings

Model closely matches experimental oxygen purity and pressure transience.

Optimal process parameters identified: pressurization 26s, purge 2s, equalization 4s.

Versatile model applicable to hydrogen purification and carbon capture.

Abstract

The pressure swing adsorption (PSA) process is one of the widely utilized techniques for air separation. Operating on the Skarstrom cycle, the porous adsorbent columns of a PSA system alternate between adsorption and desorption phases to selectively enrich the desired component in a gas mixture. The current work presents a robust and generalizable digital twin CFD model of a PSA system that can significantly help in design and device characterization. Using an axisymmetric representation, the digital twin accurately mimics all the key components of an air separation plant, including the air reservoir, adsorbent columns, product buffer tank, pressure regulator, solenoidal valves, and mesh filters. The model simulates the flow and adsorption processes in the system by solving the conservation equations for mass, momentum, energy, and species, along with the equation for adsorption kinetics. The cyclic operation of the PSA plants, typically controlled by solenoid valves, is emulated by dynamically modifying the boundary conditions of different subdomains. Such an integrated approach is shown here to closely replicate the performance of an in-house PSA pilot setup producing oxygen in terms of purity and pressure transience. Also, both the numerical and the experimental results yield an optimum performance for the same process parameters, such as pressurization time (26 s), purge time (2 s), and equalization time (4 s). The proposed numerical model is versatile and can be adapted to various industrial applications of PSA technology, such as hydrogen purification and carbon capture. Thus, it offers a cost-effective tool for designing and optimizing PSA systems.