MAJOR PLANT EQUIPMENT DESIGN

Pressure Swing Adsorber Design

Abstract

Over the past three decades, pressure swing adsorption (PSA) processes have gained increasing commercial acceptance as an energy efficient separation technique. These processes are distributed in nature, with spatial and temporal variations and are mathematically represented by partial differential equations (PDEs). After a start-up time, the system reaches cyclic steady state (CSS), at which the conditions in each bed at the start and end of each cycle are identical, revealing normal production. We implement a Newton-based approach with accurate sensitivities to directly determine cyclic steady states with design constraints. We also design optimal PSA processes by means of state-of-the-art SQP-based optimization algorithms. The simultaneous tailored approach can incorporate large-scale and detailed adsorption models and is more robust and efficient than competing optimization methodologies. In order to improve the computational efficiency, we parallelize sensitivity calculation and achieve a close-to-linear speed up rate. Applications of several non-isothermal industrial O2 VSA and H2 PSA processes are presented.

Pressure Swing Adsorber Design

1. Introduction

With extensive industry applications of pressure swing adsorption (PSA), there is significant interest for efficient modeling, simulation and optimization strategies. However, the design and optimization of PSA systems still largely remain an experimental effort (Sircar, 2002). This is mainly because most practical PSA processes are fairly complex and are usually expensive and time-consuming to solve with the accuracy and reliability needed for industrial design. For example, the traditional way to determine a cyclic steady state (CSS) is to simulate a series of complete cycles until the bed conditions repeat periodically. This successive substitution method mimics the true operation of a real plant but usually takes hundreds or even thousands of cycles to converge. To design and optimize PSA, a common practice is to develop a simplified model for one specific process and fine-tune the model using experiments and pilot plant data. Although such models are often useful, the case-by-case studies are hard to transfer among different PSA systems. Recently, more sophisticated optimization strategies have been applied to PSA systems with significant improvements in cycle performance. A review of these approaches can be found in Biegler, Jiang and Fox (in press). Here we develop a flexible and reliable optimization strategy that incorporates general process models and rigorous solution procedures within a parallel computing framework. This paper is organized as follows. The next section outlines the solution strategies, including PDE discretization, CSS convergence acceleration, sensitivity evaluation and optimization. Section 3 discusses the parallelization algorithm. Section 4 presents four PSA processes as case studies and computational results are shown. Section 5 states the conclusions and future work.

In the past, extractive or azeotropic distillation processes have been used to break the hydrogen (Black, 1980 C. Black, Distillation modeling of hydrogen recovery and dehydration processes for hydrogen and gasohol, Chemical Engineering Progress 76 (9) (1980), pp. 78-85. View Record in Scopus | Cited By in Scopus (21) Black, 1980). Since distillation routes are energy intensive and expensive, lower energy separation alternatives such as liquid-liquid extraction, adsorption and membrane processes have been ...