Dynamic Modeling, Predictive Control and Optimization of a Rapid Pressure Swing Adsorption Syste

Rapid Pressure Swing Adsorption (RPSA) is a gas separation technology with an important commercial application for Medical Oxygen Concentrators (MOCs). MOCs use RPSA technology to produce high purity oxygen (O2) from ambient air, and provide medical oxygen therapy to Chronic Obstructive Pulmonary Disease (COPD) patients. COPD is a lung disease which prevents O2 from entering a patient's blood, and reduces the blood oxygen level. The standard therapy for COPD is to provide the patient with high purity (~90%) O2. MOCs have become more popular than traditional O 2 gas cylinders due to their improved safety, and smaller device size and weight. The MOC market is growing rapidly and was expected to grow from $358 million in 2011 to $1.8 billion in 2017. Recently, a novel, single-bed MOC design was developed and tested to further reduce the size and weight of the device, and provide a continuous supply of O2 to the patient. This single-bed design uses a complex RPSA cyclic process with many nonlinear effects. Flow reversals, discrete valve switching, nonlinear adsorption effects, and complex fluid dynamics all make operating the RPSA system very challenging. Feedback control is necessary in a final commercial product to ensure the device operates reliably, but feedback control of PSA systems is not well studied in the current literature.;In this work, a study of dynamic modeling, predictive control and optimization of this single-bed RPSA device is presented. A detailed, nonlinear plant model of the RPSA device is used to study the dynamics of the system as well as design a Model Predictive Controller (MPC) for the RPSA system. The plant model is a fully coupled, nonlinear set of Partial and Ordinary Differential Equations (PDEs and ODEs) which act as a representation of reality when design and evaluating the MPC. A sub-space model identification technique using Pseudo-Random Binary Sequence (PRBS) input signals generate a linear model which reduces the computational cost of MPC, and allows the algorithm to be implemented as an embedded controller for the RPSA device. The multivariable MPC independently manipulates the RPSA cycle step durations to control both the product composition and pressure. This MPC strategy was designed and tested in simulation before being implemented on a lab-scale device.;The MPC is implemented onto a lab-scale MOC prototype using Raspberry Pi hardware, and evaluated using several MOC-relevant disturbance scenarios. The MPC is also expanded using piece-wise linear modeling to improve the performance of an RPSA device for other concentrated O2 applications. The embedded MPC features a convex quadratic optimization problem which is solved in real time using online output measurements. Additional hardware in the embedded controller operates the RPSA cycle and implements control actions supplied by the MPC.;Design and optimization of RPSA systems remains an active area of research, and many PSA models have been used to optimize RPSA cycles in simulation. In this work, a model-free steady state optimization approach using the embedded hardware is presented which does not require a detailed process model, and uses experimental data and a nonlinear solver to optimize the RPSA operation given various objectives.