Dynamics And Convergence Accelerations of Rapid Pressure Swing Adsorption

• Main Author: Soo, Ching Yee
• Format: Thesis
• Language: English
• Published: 2005
• Online Access: http://psasir.upm.edu.my/id/eprint/6066/1/FK_2005_51.pdf; http://psasir.upm.edu.my/id/eprint/6066/

Rapid Pressure Swing Adsorption (RPSA), a type of Pressure Swing Adsorption (PSA) process, offers promising solution to size and portability issues for the development of fuel cell powered vehicles and medical-grade oxygen concentrator. In this dissertation, numerical simulations of RPSA models are carried out using the Instantaneous Local Equilibrium (ILE) and Linear Driving Force (LDF) as the mass transfer models. A physically consistent RPSA model identified by Choong et al. (1998) is adopted for the numerical simulation of RPSA process. The verification of the applied numerical method and computer programs has been carried out successfully. The numerical simulation of RPSA models requires a sufficiently large number of cycles to reach cyclic steady state (CSS), resulting in large computational time. Hence, convergence acceleration methods are examined. The methods are extended from the accelerated successive substitution method of Choong (2000). The problem where the Aitken-like extrapolator (first extrapolator) and the MSS approach the CSS fi-om the same side, with condition kc"' >k'"-", which was not solved by Choong (2000), is investigated in this work. The Secant-like extrapolator (second extrapolator) provides a satisfactory extrapolator for reaching CSS from the opposite side of the first extrapolator. Further, the Muller method is found to reach CSS faster than the Secant-like extrapolator by approximately 250%. The numerical simulation of URPSA models is challenging because small time steps are required to capture the process dynamics of cycle time within fractions of seconds. Studies are carried out to assess the suitability of LDF model to describe the mass transfer mechanism for URPSA process by: (1) considering the effect of external fluid film resistance provided by Choong and Scott (1998); (2) comparing the LDF model with a full diffusion model provided by Todd and Webley (2002). The LDF model is considered sufficient to describe the mass transfer mechanism in the particle for the numerical simulation of URPSA considered in this work. The effects of axial dispersion and feed pressure boundary conditions on the performance of URPSA are studied. Increasing the value of effective axial dispersion reduces substantially the oxygen product purity. However, the axial dispersion has no effect on the cycle-averaged feed gas rate. The simulated cycle-averaged feed gas rate using the step function as the feed pressure boundary condition overestimates the experimental cycle-averaged feed gas rate by 100%. However, this is a substantial improvement over the 600% overestimation by Murray (1996). Using the exponential function as the feed pressure boundary condition provides a better prediction of the experimental cycle-averaged feed gas rate than that of the step function. Nevertheless, the forms of feed pressure boundary condition have no effect on the oxygen product purity.