A data-driven model reduction of cyclic adsorption in coated tubes

In this study, cyclic adsorption inside coated tubes was investigated using computational fluid dynamics (CFD) and reduced-order modeling (ROM). An axisymmetric 2D CFD model was developed to simulate the problem. In this model, the coated layer was treated as a wall boundary condition, assuming a negligible radial variation in the layer. The coated layer and the air field were coupled through the transmission of heat and mass diffusion fluxes at the interface. The collected CFD data were used to develop ROMs. Dynamic mode decomposition (DMD) and proper orthogonal decomposition with long short-term memory networks (POD-LSTM) were introduced as spatiotemporal ROMs. The outlet condition of the tube was also reconstructed using discrete Fourier transform (DFT) as a temporal ROM. The proposed ROMs demonstrated their advantages as fast, accurate, and stable surrogate models for predicting cyclic adsorption under laminar conditions at Reynolds numbers between 30 and 600. Using 720 time samples, DMD with 30 modes, DFT with 20 modes, and POD-LSTM with 10 modes showed suitable performance in terms of the reconstruction accuracy and computational cost.