[nicholas.jones]

Hey All,

I have been trying to simulate the flow of a liquid through a packed bed reactor. To keep things simple to start, I am using 10 particles in a tube. I simulate the location of the particles using Blender, export the .stl into Salome, label my boundary patches and mesh using netgen 1D-2D-3D. I then export the .unv and use ideasUnvToFoam to get my mesh in OpenFOAM.

My checkMesh test passes with 2% severely non-orthogonal faces, so I increase the correctors to compensate.

Unfortunately, when i run with icoFoam using basic parameters, my Courant number creeps up past 1 after 3-6 iterations. I have tried reducing my time step, or dropping my initial velocity to compensate, but it has little impact.

The same workflow is successful with more simple systems, so I suspect it has something to do with my mesh.

Any insight, advice, or a point in the right direction would be greatly appreciated! My .unv mesh is available at the following link.

[chegdan]

Nicholas,

I did my PhD work on interstitial flow in packed bed reactors, so this is something that i have toiled over for years. If you are generating a 3D tetrahedral mesh, run away as fast as you can as you're gonna have a bad time with tet meshes in OpenFOAM. If you need to use a tet mesh, i suggest generating one with nice boundary layers cells and convert the bulk cells to arbitrary polyhedral cells. That has its own challenge. As far as solver, what is your goal? What are your Re you are modeling and is this steady or transient? icoFoam is an incompressible laminar flow solver for transient flows. let me know....i bet i can offer some advice and get you moving.

[nicholas.jones]

Hey Dan,

Thanks for the reply.

I am trying to simulate the flow of a liquid (starting with water, eventually a high viscosity liquid) through the packed bed. Re should be very low, less than 10 at all times.

In a few months, I would like to be able to simulate the heat transfer in the system (assuming heated walls). A few months past that, I would like to look into two phase systems (Gas/liquid moving through my reactor, and interacting with internal components such as static mixers or thermocouple wells).

Im only interested in steady state solution, so I will need to change solvers. As well, your comments about tet meshes gives me something new to try.

Would you recommend SnappyHexMesh?

I have played around with viscous layers in Salome, but im running into issues with overlapping layers. Because im using blender for creating my packed column, there may be a bit of overlap in my spheres in some cases, and spacing between particles is not consistent.

I have had some success using Netgen to make a very fine tet mesh (15 million cells), but I run into computational limits as I am using a i3 processor for my calculations.

I am limited to opensource or free software, or low cost commercial solutions. I may be able to justify some expenses/training with my employer, as long as the price is reasonable.

Any tips or documentation is appreciated. Thanks for linking your thesis, I will review your work and your references for further information.

Best,

Nicholas

[chegdan]

snappyHexmesh should do what you need since you are working in a low Re regime. You may have some issues with boundary layer cells being added, but it should be fine.

I suggest doing your best to get more computational power as these can be rather computationally intensive (especially multiphase). If that is not possible, I would focus on a single topic rather than trying to hit everything since these low Re flows will be a much different beast than a multiphase simulation. What really lacks in the field is what to do with the data you get from these systems. Many Authors in the past have been able to achieve single phase simulations in both spherical and non-spherical particles...but the true task is what to do with all the data. There have been recent efforts in the last few years to "reflect" the knowledge from these higher order fluid mechanical methods back to lower-order chemical engineering models (i.e. 1D or 2D reactor models). This can then be leveraged on a large scale simulation that will use orders of magnitudes less compute power than CFD. Plus those models still hold water.

In my experience, packed bed simulations go two main directions.

1. Larger scale realistic packing models with a high mesh count in the tens or hundred of millions of cells.
2. Take what we learn on a representative groups of particles and apply that to lower-order models

Both have a real necessity to address the "what to do with the data?".

[nicholas.jones]

Hey Dan,

Ill start working with SnappyHexMesh. I have been meaning to get around to it, but until now, I have stuck with Salome for most of my meshing. With my current computational limits, I am very interested in getting away from a GUI.

Im looking towards greater computational power (im thinking i7-5820), but it may be a while before that happens.

As for what to do with the data, we currently have a pilot scale catalytic packed bed reactor, and we are looking to improve on our design, specifically when it comes to liquid/gas interaction and thermal profiles. As the ultimate goal is scale up, I try to argue that a geometric scale up would be a bad idea in our case. Having some supporting data would be very useful in the decision making process.

Due to the hazardous nature of our reaction, we have little information beyond what comes in/goes out, and photos taken during maintenance.

On a different note, as someone who is relatively new to CFD, could you recommend any tools or documentation which you rely on or have found useful in the past?

I have been enjoying this process, so any tips/tricks/experiences/comments are always greatly appreciated.