|
Abstract
|
A lab-scale vacuum pressure swing adsorption process for oxygen production was investigated both experimentally and theoretically. The experiments were conducted with up to 91% purity and 17% recovery. A complete set of governing equations were solved and compared using the finite difference method (FDM) and differential quadrature method (DQM). Based on the theoretical achievements, a new comprehensive algorithm is proposed, which is compatible with various numerical methods. The DQM method with 12 points combined with the FDM for time integration was determined to be accurate enough for predicting system behaviour. The artificial neural network (ANN) with two hidden layers and up to eight neurons was used to predict the process behaviour at more complex conditions. The agreement between the simulation results and experimental data shows that the algorithm accurately simulates the cyclic adsorption process, and the ANN is reliable for prediction of system behaviour considering variations in all parameters.
|