Learning Forum

[Anthony Bowers]

Hi all, 

I am in need of some advice or direction regarding my simulation. To begin, the image below is a layout of the simulation.

The governing equations are as followed: 

With the source term's closed as such: 

(For now the F_disp is ignored)

The counterflow phenomenon is implemented by setting the backflow volume fraction , of the gas phase , as 1 and having a negative velocity at the gas inlet boundary. The porous zone is modeled using either ANSYS built-in porous media model or a momentum source term. Either method is closed by the following relations: 

Substituting results in the porous term as :

 

 

Then the momentum exchange of the two phases is closed by : 

 

 

Now, I have verified my insertion of the terms by setting constant values and hand-calculated the terms. After I let the simulation run until steady-state is achieved my volume fraction results look like this: 

 

However, I need my results to look like this: 

 

What does the community recommend I do/try? 

[Anthony Bowers]

Should I make the densities of the two phases near each other for a better estimation of volume fraction? 

[Anthony Bowers]

I am thinking I need to enforce a pressure gradient for the gas phase

[Anthony Bowers]

 

Also this is the pressure drop I am in need of 

 

 

This is what the authors have said about their source terms (porous force term) 

[Rob]

Which multiphase model did you use? I'm fairly sure someone has asked about a column fairly recently, see if the search option returns anything: you may need to try a few terms. 

[Anthony Bowers]

I have been asking, haha. All of the posts were mine, I am still severely struggling and worried about meeting my dissertation deadline now. I am using a eulerian-eulerian model. 

[Anthony Bowers]

I am looking into the 2013 UDF manual because I find it strange that the authors implemented the porous term with define properties 

[Rob]

There isn't a 2013 version, unless you mean the version that was available in 2013? The current release is 2023R2.

Porous zone values are probably DEFINE_PROFILE (I've not checked), DEFINE_PROPERTIES tends to be for density, viscosity and the like. Assuming you've got an axi-symmetric model your results look reasonable. What does the pressure field look like? 

Does the paper list any of the porous media coefficients? Or give any pressure against flow data to let you work it out? 

[Anthony Bowers]

I do mean the version that was available in 2013. The paper does not list explicit values for the porous media coefficients. The paper only list the force terms and porous coefficient equations as such : 

The results look reasonable but the values are not correct. My fe which is the wettability is higher than the maximum value of 1. Moreover, the volume fraction should not be 1 in the packed bed region. 

I can not have my current volume fraction results because I want to enable mass transfer later. If mass transfer is enabled with the current results than species transport will only occur at the gas-liquid interface. 

[Anthony Bowers]

The only pressure against flow data provided is as seen below 

[Rob]

OK, so you're trying to replicate a result from a paper but don't have the resistance coefficients that they used? From the last image, have you got any data for Mellapak 250.X? That looks to be a brand name for a column packing material, so data may exist in the public domain. 

As an aside, putting all of the liquid down the axis is an awful way to design a scrubber or gas-liquid contactor. If my old Unit Ops lecturer saw that he'd have several things to say about the design! 

[Anthony Bowers]

I will look for data. I could not agree more with your latter statement. I am only mimicking this simulation to ensure the validity of my source terms. After I ensure that the source terms are working correctly I will begin my design 

[Anthony Bowers]

Do you have or know where I can locate the equations on how Ansys treats volume fraction estimation in an eulerian-eulerian simulation? 

[Rob]

Of the Eulerian model or porous media? Both are in the Documentation. 

For Eulerian in porous media you can work out two sets of coefficients and apply to each phase. 

[Anthony Bowers]

I found the documentation on the volume fraction. See the issue is the over estimation of volume fraction, that is it. I just don't know what to do to correct the issue. 

[Rob]

Overestimating compared to the experiment? Remember that porous media doesn't occupy volume, it's just a resistance. 

[Anthony Bowers]

Yes over estimate.

I did read that the porous media doesn't occupy a volume. So to account for the solid volume I added a ternary phase with a constant velocity of 0 m/s in all zones. I then patched the volume fraction of the solid phase equal to (1-porosity) = 0.03. This enforces the volume fraction limitation of 0.970  (V_L+V_S+V_G=1). 

 

What I am regarding is that in their results their average volume fraction of the liquid phase in the porous zone is ~0.40 (see image way up above).

 

 

 

[Rob]

That's going to do all sorts of interesting things to the flow, so getting different results isn't surprising. 

[Anthony Bowers]

I also implemented the porous media model through a source term rather than using the build in porous media model. Many authors went this route. I only used the built-in model for stability and simplicity of insertion ( as well as told to by my advisor). 

 

What would you recommend?

[Anthony Bowers]

 A paper called "Calibration of drag models for mesoscale simulation of gasliquid flow through packed beds" 

Was my inspiration for addition of the ternary phase

[Rob]

I would suggest keeping things very simple to start with. With any lumped parameter model you need some caution, not least as it'll not (easily) pick up on bed flooding and the like. For that you'd need to model the porous material itself, and that's usually not practical. 

 

[Anthony Bowers]

Yeah, I figured that would be ideal. However, my home machine is not capable of simulating such computational expense as I only have 32gb of ram and 16 cores. 

[Rob]

Which is where the engineering comes in. And where I need to be very careful to just use public domain knowledge. 

32Gb RAM will get you 20M cells or so, plenty to model a small section of the packed bed. However, you're looking to solve a problem. In trickle bed reactors/scrubbers etc we generally want to make them work better. So, we then look at why they fail, or work at a lower efficiency than the books tells us. That's where CFD comes in, and where I get involved. 

In your case, I think you need to carefully define the objective of the work, with help from your supervisor. Then decide what CFD can do to answer part/all of that question.