AnsweredAssumed Answered

Convergence Problems

Question asked by Dario Denzler on Mar 29, 2019
Latest reply on Apr 2, 2019 by Peter Francis

Like • Show 0 Likes0
Comment • 9

Hello everyone,

I am facing some convergence issues in a flow simulation. Basically, I do have a housing with air with the following boundaries:

Inlet: 2 faces with ambient pressure

Outlet: 3 faces with ambient pressure

Fans: Internal fan simplified with fan curve

Inside the housing, there are some channels which should guide the airflow accordingly. I created goals for volume/mass flow rate for all inlet/outlet. Additionally, I checked the net volume/mass flow (sum of all flows), which should be zero in case of incompressibility from my understanding of fluid dynamics. Well, this is not the case here as I have air, but as pressure differences in this simulation are quite small, I assume this can be neglected.

Attached you find plots during the solve for the mentioned goals. After arround 50 iterations, I run into oscillation on the flow rate's @ the inlet's. Now the question is, how to overcome this phenomena. Is the problem not well defined? Too coarse mesh? Maybe some issue with the internal fan (fan curve not suitable)?

Peter Francis Apr 1, 2019 2:03 AM

Correct Answer

Hi Dario,

If your housing has both inlets and outlets set as ambient pressure there will be no fluid flow as there is no pressure difference.

You need to set up either the inlets or outlets as a volume/mass flow then the internal fan will function according to its fan curve.

In order to use an internal fan with a fan curve you need to have a pressure differential across it that can either be caused by a volume/mass flow or a pressure difference.

Hope that helps

See the reply in context

Attachments

No one else had this question

Outcomes

9 Replies

Peter Francis Apr 1, 2019 2:03 AM
Unmark Correct
Correct Answer
Hi Dario,
If your housing has both inlets and outlets set as ambient pressure there will be no fluid flow as there is no pressure difference.
You need to set up either the inlets or outlets as a volume/mass flow then the internal fan will function according to its fan curve.
In order to use an internal fan with a fan curve you need to have a pressure differential across it that can either be caused by a volume/mass flow or a pressure difference.

Hope that helps
Like • Show 0 Likes0
Actions
- Dario Denzler @ Peter Francis on Apr 1, 2019 3:17 AM
  Mark Correct
  Correct Answer
  Dear Peter,
  
  Thanks for your reply. I checked your suggestion and it works pretty well. However, for me it is surprising, that the oscillation in my approach starts at around iteration 55. Until then, it looks like everything is going well. I also had some simulation where I used an outlet fan instead of ambient pressure at the outlet and this works fine.
  
  For some general understanding, a fan should create this pressure difference as it will have some negative pressure @ inlet and positiv pressure @ outlet so there is some driving force across a fan in my opinion. Is this some physics, that Flow Simulation cannot reproduce due to the solver or is this a general issue that needs to be "modelled away" in order to be solved in CFD?
  
  I am asking this because the next question already arises....how do I define the volume flow @ inlet? An idea that comes to my mind would be to take the necessary volume flow calculated from the thermal input into my system as this will also dictate, together with the system resistance, the selection of the fan
  Like • Show 0 Likes0
  Actions
  - Peter Francis @ Dario Denzler on Apr 1, 2019 4:26 AM
    Mark Correct
    Correct Answer
    Hi Dario,
    When using a fan curve in FloEFD you have to provide a pressure differential other wise it will not give you any airflow. The reason this doesn't show initially is that the software has to perform the necessary calcs to determine the pressure differential to apply to the fan (over simplified but generally true).
    Generally you can apply a volume/mass flow boundary condition at the inlet and a ambient pressure boundary condition at the outlet or the otherway around depending on your setup.
    
    I tend to think of the fan curve as a lookup table rather than a true fan.
    
    Of course you could model a true fan with all the necessary geometry , rotation etc but that would be too complex for most applications. I used to work for a major UK fan manufacturer and I didn't do that when I employed our fans in an application !
    
    In order to apply the correct volume/mass flow you can perform a simple calc to determine the volume flow you require to cool your application, you would need to do this anyway to size/select your fan(s).
    Apply that volume flow to your inlet, add goals to check for mass balance (very important - what goes in MUST come out) and pressure at your inlets and outlets.
    This should give you a converged result which you can check againt your fan selection (you need to find your fan operating point on the system curve)
    
    I cover this in a little more detail on my website https://thermal-design.net/airflow-and-thermal-design-procedure/
    
    Hope that helps a little , if you have further queries let me know
    
    Regards
    
    Peter
    Like • Show 0 Likes0
    Actions
    - Dario Denzler @ Peter Francis on Apr 2, 2019 1:41 AM
      Mark Correct
      Correct Answer
      Well explained and now it makes sense. Of course, I always try to simplify my models as much as possible, especially on system scale. That is why I was not aiming for a rotational region or things like that.
      
      I guess I will also verify the implementation of the fan curve in a simplified model that I also use for my MSS. Currently, it looks like with the defined volume flow rate, I do not get the corresponding pressure difference given by the curve. Anyway, as I see how everything works, I should be able to keep going. Thank you so far!
      Like • Show 1 Like1
      Actions
      - Tg R2 @ Dario Denzler on Apr 1, 2019 12:33 PM
        Mark Correct
        Correct Answer
        When you measure the pressure difference across the fan you should measure the (averaged) pressure at both faces of the fan. In my experience however the corresponng pressure/flow point is often not following fan's curve, and this is a frequently asked question about SWFS. The support people from SWFS advise to exetnd the airdomain in the direction of the fan and have a fan curve with more points, but this is more like "you are on your own here pal" recommendation.
        Like • Show 0 Likes0
        Actions
    - Tg R2 @ Peter Francis on Apr 1, 2019 12:08 PM
      Mark Correct
      Correct Answer
      The selection of volume flow@inlet is quite artificial in my opinion and will affect results accuracy. What about making the analysis external and extending the air-domain in the direction of the Inlet and outlet. In this case you could assign ambient pressure without worrying. You could also assign some initial velocity to speed up the convergence
      Like • Show 0 Likes0
      Actions
      - Peter Francis @ Tg R2 on Apr 2, 2019 2:13 AM
        Mark Correct
        Correct Answer
        As with most things in Flow Simulation there are many ways to achieve the end result.
        I regard the setup I suggested as a starting point.
        You would need to look at the pressure drop of the entire system to correctly judge the operating point and if you are speed controlling the fan you need a specific fan curve for that speed or range of speeds, you can usually get these from the manufacturers or measure them yourselves.
        How and where you instal the fan will affect how it operates.
        Goal selection is also important and monitoring mass flow in and out is crucial to ensuring you have a valid simulation.
        A Mesh Sensitivity Study is necessary if you want results you can rely on.
        
        An understanding of Fan performance is also crucial , for instance if you are using a centrifugal fan pressure recovery can be beneficial to the design.
        
        Fans are part of the entire system and they both react to changes in the system and affect the system. It is vitally important to understand how the fan fits into the overall system.
        Like • Show 0 Likes0
        Actions
    - Dario Denzler @ Peter Francis on Apr 2, 2019 7:44 AM
      Mark Correct
      Correct Answer
      Hello Peter
      
      So far, I estimated my cell size in a MSS, managed to get my solution to converge and started to put everything together in my block model for a first evaluation.
      
      Peter Francis wrote:
      
      ...This should give you a converged result which you can check againt your fan selection (you need to find your fan operating point on the system curve)...
      
      For the fan selection, I am curious on how you would determine your system resistance. From the thermal input of my components, I can estimate the necessary flow volume rate which is also used as inlet boundary. Now for the fan selection, I want my fan curve intersecting my system resistance curve at this specific flow volume rate.
      The following article describes the system resistance curve in equation 4 All You Need to Know About Fans | Electronics Cooling . My approach would be to determine the pressure loss of the system at a given flow rate without a fan to determine K. For n, I would probably go for a worst-case scenario. Would you confirm this approach?
      Like • Show 0 Likes0
      Actions
      - Peter Francis @ Dario Denzler on Apr 2, 2019 9:08 AM
        Mark Correct
        Correct Answer
        Hi Dario,
        Glad to hear its all moving in the right direction !
        
        Interesting article on Electronics Cooling website . Formula (3) is the most useful one
        If it is possible to measure the system pressure drop on in the actual equipment I would do it as it will provide a check for any CFD analysis.
        I would then run the analysis as follows.
        
        Run the model with the estimated volume flow rate you've calculated.
        Making sure you use the air density & specific heat capacity at the temperature of the application otherwise you will have the wrong volume flow - remember it is the MASS of air that transfers the heat not the volume as this changes with temperature.
        For interest check the volume difference between NTP conditions and at your working temperature and pressure.
        
        Incidentally I use SMath Studio (MathCAD clone - but much better) for all my calcs as it provides a great record for use in reports etc and you can use CoolProp for fluid properties.
        
        Setup pressure goals to calculate the pressure drop between your inlet and outlet boundaries., subtract one from the other and you will have the system pressure drop. This value is an approximation of the system resistance.
        
        Next step is to select a fan or fans.
        You need to look at the fan curve and select one where the volume flow is close to your calculated value and at the pressure drop from your analysis. Avoid using a fan where these values are at one end or other on the curve.
        
        Setup the internal fan with the fan curve from the manufacturers datasheet.
        Make sure you model the physical geometry of the fan inlet / outlet as accuratly as possible.
        You don't need to model the blades but you will need to model the hub for an centrifugal / axial fan.
        If you have guards fitted to the final application make sure you add them in as they have a significant effect.
        
        Hope that helps
        Like • Show 0 Likes0
        Actions

Incoming Links

Re: Flow results do not correspond to input fan curve

Convergence Problems

Attachments

Outcomes

Related Content

Recommended Content

Incoming Links