CFD Simulations of Structured Packing

Sponsor: Lummus Technology

Principal Investigator: Bruce Eldridge

Graduate Research Asst: Michael Basden

Status: In Progress

Introduction

High performance computing (HPC) resources have been used for decades in the aerospace, automotive, and naval industries to aid in the development and evaluation of new designs. Using computational fluid dynamics (CFD), engineers and researchers evaluate modifications and improvements before building any prototypes. In this manner, only the most promising designs are actually built and tested. Furthermore, a CFD model provides much more information than a single experiment can provide. For example, stresses and structure response can be predicted on naval vessels, which allows for the design of an optimum shape and ideal placement of reinforcing material.

CFD offers similar benefits in the design of distillation equipment. State-of-the-art distillation design methodologies rely on past experience and empirical correlations derived from experimental observations. These data typically covers only a limited range of operation with limited theoretical basis. CFD analysis is firmly rooted in theory and promises to be more accurate and robust than current empirical methods. A much larger volume of information can also be gathered from a CFD simulation; both macroscopic variables (e.g. overall pressure drop) and local variables (e.g. velocity vectors) can be calculated.

Geometry Generation and Simulation Details

A digital representation of a Mellapak N250Y half-element with a diameter of 146mm was generated through the use of x-ray computed tomography (CT) scans. The packing element was then imported into the CD-adapco’s CFD software, STAR-CCM+, where it was manipulated to yield a stack of packing consisting of three half-elements as seen Figure 1. A solid tube was then placed around the packing elements. A 190mm length of empty tube was added below the stack of packing to allow the flow to develop naturally from a flat velocity profile at the inlet. A 100mm section of empty tubing was added above the packing stack to allow for any vortices to dissipate before reaching the pressure outlet. No slip boundary conditions were assumed for the packing surface and tube walls, and incompressible, turbulent flow was assumed for nitrogen flows with F-factors in the range of 0.610-3.05Pa^0.5.

Figure 1: Packing stack utilized for simulations.

Experimental System

A schematic and photograph of the experimental system can be seen in Figure 2. The experimental system was designed to match the simulation geometry as nearly as possible. In particular, the experiments and simulations shared the column diameter, pressure measurement elevations, and bed depth consisting of three half-elements of Mellapak N250Y.

 Figure 2: Experimental system used to measure pressure drop in the bed of strucutred packing.

Results

Joint Rotation Study

The joint, the region where two layers of structured packing meet, is generally rotated 90° to ensure redistribution of the liquid and gas; an example of this rotation is provided in Figure 3.

 Figure 3: Example of typical stuctured packing layer orientation.

In an effort to understand the impact of joint rotation on dry pressured drop, four different joint rotation angles were examined via CFD: 0° (no rotation between successive packing layers), 30°, 60°, and 90°. For these simulations, the SST-kω turbulence model was utilized. The results are presented in Figure 4.

Figure 4: Presure drop as a function of joint rotation angle. 

The CFD simulations predict the pressure drops for the 30°, 60°, and 90° cases to be within 3% of each other. The 30° case has a predicted pressure drop slightly lower than the 60°, which is again only slightly lower than the 90° case. The 0° case has a pressure drop prediction roughly 10% lower than other cases examined. In the 0° case, the triangular packing channels were aligned from one layer to the next. Thus, the nitrogen could transition between layers of packing without significant flow separation and turbulence caused by an abrupt change in flow direction. 

Figure 5 shows the comparison between experimental data and CFD predictions for the 0° and 90° joint rotation cases. In all cases, the CFD under-predicted the experimental pressure drop; predictions were within 20% in all cases. Of note, the CFD accurately predicts the difference in pressure drop between the 0° and 90° cases, indicating differences in geometry are being accurately captured by the simulations.

Figure 5: Comparison between experimental data and CFD prediction.

Evaluation of Turbulence Models

Figure 6 shows the pressure drop predicted by different turbulence models (Realizable k-ε and SST k-ω) and a semi-empirical design (Stichlmair equation). In all cases, the CFD simulations outperform the Stichlmair model when predicting experimental pressure drop. The SST k-ω simulations predicted pressure drops no more than 20% lower than experimental data. The Realizable k-ε turbulence model performed best of the models examined, predicting a pressure drop no more than 15% lower than experimental data. The CFD simulations also outperformed the Stichlmair model when predicting the power law dependence of pressure drop on the f-factor with the Realizable k-ε model once again performing best.

Figure 6: Comparison between turbulance models, experimental data, and semi-empirical model.

Future Work

Multiphase simulations to predict irrigated pressure drop, wetted area, and liquid holdup will be simulated using STAR-CCM+. To accurately capture the interface between phases, the volume-of-fluid (VoF) model will be used. The impact of physical properties and flow conditions on wetted area and liquid holdup will be examined. Single phase simulation data will be examined to determine appropriate pressure drop mechanisms.