Learning Forum

[Scott Ormiston]

My research group is starting to use the EWF model for a steady-state falling film
of pure water adjacent to a steam-air mixture. We are finding that the convergence rate is poor. Are there parameter settings that can we can use to improve the performance?

[Rob]

I need to tread carefully as I'm unable to use "detailed engineering knowledge" on the Forum. I've done a fair bit of work with EWF with partners. 

First pointer is to fully understand the condensation rates and likely film flow regime. Make sure the facet growth rate is low, and avoid splits in the surface if you have poly mesh: note my comment on cell size. From there you may find transient film with steady flow to be appropriate, read the solver theory to see how energy is handled for phase change. 

[Scott Ormiston]

Thank you for replying. While there are some pointers in your response about concepts, it doesn't tell me how to check facet growth and how to avoid splits in the surface.  I was looking for settings in the parameters or guidelines for time step size (or CFL number) or relaxation or model setup parameters. We are very familiar with the governing equations for falling film condensation heat and mass transfer and see that the EWF model is a simplified version. Is there something about the phase change equations that we need to specify?   If you are constrained from giving detailed responses, how can I find out what I need to know?

[Rob]

Facet growth would come from the meshing tools. Splits in poly face meshes occur where you have a surface boundary split, possibly from a surface label change or sharp angle. 

Re the film heat transfer, the problem tends to arrise from the amount of energy being added to the (almost nonexistant) film or near wall cell. If the near wall region is over refined then the source terms can become high. 

I am constrained to using public knowledge on here. So, as I come across questions like this I'm writing up content to make public. 

[Scott Ormiston]

Can you please provide more details on the parameters to choose when setting up the EWF model that will have the greatest impact on convergence behaviour. You can assume that we will want to solve momentum and energy for the film.

[Rob]

Now I've posted a couple of public solutions (to be signed off) I have a little more leeway.  Alter the film timestep to maintain a Courant Number below about 0.1 drop further if you're looking at condensation. 

[Scott Ormiston]

Here are some details and questions from the researcher in my group who is struggling with the EWF model. Can you please provide suggestions on how to move forward to get a physically realistic converged solution?

=========================================================

EWF Model Notes

Main problem: Non-physical temperatures (above the inlet temperature) appear in the domain, near the inlet, near the cooled wall on which the film is located. The solution converges with these non-physical temperatures. If the temperature is limited to just above the inlet temperature, the solution does not converge.

Model Options and Setup:

·        Solve Momentum, Energy, Phase Coupling

·        Solving Momentum Equation (not analytical solution)

·        Gravity Force, Surface Shear Force enabled

·        Phase change enabled, film vapor material is water-vapor, film material is water-liquid

Solution Method and Control:

·        Time: Second Order Implicit

·        Continuity, Momentum, Energy: Second Order Upwind

o   Note: First Order Upwind did not improve results

·        Adaptive Time Stepping:

o   Max CFL: 0.05

§  Anything above this is unstable

o   Initial Time Step: 0.0005 [s]

o   Increase Factor: 1.1

o   Decrease Factor: 1.2

o   Subiteration Stop: 1e-08

o   Sub-Iterations: 5-10

·        Maximum Thickness: 0.02 [m]

·        Continuity and Momentum Coupling: Coupled Solution

Wall Film Options:

·        Boundary Type: initial condition (Film Thickness: 0 [m], Film Velocity (X, Y, Z): <0, 0, 0> [m/s], Film Temperature: same is inlet temperature.

·        Flow Momentum coupling engaged

·        Phase Change Enabled, using wall-boundary-layer model

Domain:

Domain is initialized without EWF model engaged, with CFL=200, default under-relaxation then EWF model is turned on and CFL and under-relaxation factors are lowered.

With EWF on:

·        Scheme: SIMPLEC

o   Coupled scheme produces instabilities in film thickness

·        Pressure: Second Order

·        Momentum, TKE, SDR, Species, Energy: Second Order Upwind

·        Pseudo-Time CFL: 1-5 (increasing past 5 produces instabilities in film thickness)

·        Under-relaxation: h20 = 0.99

·        Under-relaxation: energy = 0.2-0.5

o   Higher under-relaxation produces instabilities when the film is forming

o   Under-relaxation of energy can be increased (up to 0.99) after the film thickness has progressed down the entire domain

o   Low under-relaxation (e.g. 0.001-0.1) does not improve convergence of results.

Mesh:

·        Current mesh has wall-adjacent cell size of 1e-04 [m]. K-omega SST model is used.

 

Theory:

In the theory manual, Chapter 18.2.1.6 ‘Coupling of Wall Film with Mixture Species Transport’ indicates the various available transport models. Currently, we are using the wall boundary layer model as it best matches the approach we wish to take.

Chapter 18.2.1.6 redirects us to Chapter 12.9.5.2 ‘Mass Transfer from the Film’ for the wall boundary layer model, which is in the Lagrangian Wall-Film Model Theory chapter. However, the exact subsection dealing with the wall boundary layer model is not indicated.

We believe that Chapter 12.9.5.2.2 Film Condensation is the subsection that corresponds to the wall boundary layer model based on its content. Is this correct?

Chapter 12.9.5.2.2 begins with Fick’s law for determining the condensation rate, however, it then presents a different method where the mass fraction is integrated and mass transfer coefficients are used. For turbulent flow, it is indicated that the mass transfer coefficient is determined via wall functions from equation 4-341 (an equation that gives the law of the wall for species transport).

-        What is the rationale for moving from Fick’s Law to a mass-transfer coefficient approach?

-        How does equation 4-341 determine the film mass-transfer coefficient for turbulent flow? No mass transfer coefficient is evident in the equation.

-        When the domain is turbulent, does the model change its approach when the viscous sublayer is resolved (i.e. wall y+ <=1)?

=========================================================

 

[Rob]

The model uses the wall shear to calculate film motion. But it's also a thin film model. This can mean trying to get a y+ =1 mesh means the cell is thinner than the film. That then causes problems with phase change as the source terms are added to the near wall cell. 

In regions where the film is thin-nonexistant there's no liquid to absorb heat which may explain the temperature issues. If you tighten the limits too much any excess heat (or cold) is thrown away, that tends to give unrealistic results and may prevent the solver reaching a stable solution. 

There is some of the maths for EWF in the LWF section of the manual - that has been raised before to the DOC team. 

The wall function won't alter the wall film mass transfer. However, it may effect the rate a species reaches the wall which may then effect how the film forms. 

[Scott Ormiston]

Thank you for these comments. Does this mean that the EWF is not a subgrid model?  I thought that it solved simplified fluid film equations and applied the film interface velocity to the gas phase wall-adjacent cells and a wall face mass flux instead of volumetric sinks in the wall-adjacent cells.

It is crucial that we get a clear picture of what the EWF model is doing and how to get it to converge as soon as possible. We are presently working on a journal article which assesses CFD code condensation modelling for a vertical channel with conjugate heat transfer to a coolant channel.  We are planning to compare Fluent and STAR-CCM+ for both in-house volumetric sink models (via UDFs and field functions) and built-in fluid film models.  We have everything running except the EWF model. I must say that it will be disappointing if we are forced to omit the EWF from the article.

[Rob]

It's a wall facet model, so everything takes the near wall cell and wall facet values as needed. 

Please note, I am NOT permitted to go beyond what's in the manual to comply with Legal's rules for me answering anything on the Forum. 

[Ashish Khemka]

Hi Scott,

As Rob pointed out it looks that this question is the beyond the scope for Ansys employees to answer on a public forum, so hopefully other forum users can chime in. In case you have access to Ansys Customer Support Space (ACSS) then you can open a new support case.

Customer Support and Help Desk: Submit a Request | Ansys

Regards,

Ashish Khemka