14 Replies Latest reply on Aug 15, 2018 3:37 PM by Stephen Endersby

flow simulation on blade design is different from real life result

Kai Chan Aug 3, 2018 7:32 AM

I am trying to create flow simulation of existing fan design. I have real-life lab test that I like to compare the accuracy of this simulation, but the result is far different from real-life test. Below is my summery of test:

Objective:

Visialize air flow in 1.8m in front of fan blades, width 1.5m
Find target wind speed at 75cm in front of fan blade, (done same things in real lab test, so I can compare them)

Problem on simulation result:

The wind flow pic (trajectories) look different from different refinement level. I think my cell on solid-fluid, small gap, and the blade itselt is fine enough. But, I cant tell if this is refect to reality, I cant tell which/where the refinement is need in order to optimize the result that is close to real situation.
Whatever resolution (or refinement level or cell) that I have set, the wind speed at 75cm location is far different from real lab test. (the lab result is around 1.3-1.6m/s, but simulation is only around 0.1-0.5m/s) At this point, I don’t quite believe on the trajectories result becauase the wind speed is dfifneret from real-life lab test

Setting highlight:

I have different set of refinement setting, and only keep all atmosphere pressure and density the same, rotation at fan blade is 800rpm, and all the goals is about velocity and z velocity in average + bulk. But all goals I set not to ‘use for convergence control’ because I don’t the calculation stop at any point so that I can visialize the whole air flow)

My question:

Why with differnet refinement level setting, the wind speed result is still far different from real-lifel ab test?
Could anyone suggest a right strategy on setting refinement level, so I know everytime when I update the cell, I can tell the new one is always better then old one, and my experiement is always a step closer to the reality.
Is anything other than cell/refinement level I set it wrong so that my wind speed result is different? Such as atmosphere pressure or density or anything?

attached pic are all my setting and result of final-try:

cell3.jpg 243.2 KB
cell2.jpg 175.7 KB
cell1.jpg 163.0 KB
goal result.jpg 254.3 KB
mesh2.jpg 73.8 KB
mesh1.jpg 34.0 KB
domain.jpg 123.3 KB
setting 4.jpg 83.6 KB
setting 3.jpg 55.8 KB
setting 2.jpg 76.7 KB
setting 1.jpg 75.9 KB
trajectories.jpg 97.2 KB

Re: flow simulation on blade design is different from real life result

Ryan Navarro Aug 3, 2018 8:07 PM (in response to Kai Chan)

Hi Kai,

My biggest concern with your study setup is that you turned off convergence control for your goals. Your flow trajectories to me do not look fully developed so it is possible the solution just needs to run longer.

What was your stopping criteria for the study? How many travels did it solve for?

I would recommend re-running or resuming the study with the goals set for convergence control. You could also add additional goals such as Force on the turbine blades to make sure they are converged as well.

If you don't want to re-run you can show a Goal Plot or export goals to Excel to see the progress of the goals over time and that may give you some idea if they are converged.

It looks like you have a dummy body setup in front of the fan. This is useful so you can track surface goals/parameters, but did you remember to disable that body through component control? That would be another thing I would check. It's hard to tell from the screenshot but it almost looks like the flow may be diverting around it.

I would check into these things first before worrying more about the cell refinement. At first glance your refinement looks adequate to me. I might recommend expanding the computational domain slightly behind the fan. Can you also show how you have your rotating region defined?
Like Show 1 Likes(1)

Actions
- Re: flow simulation on blade design is different from real life result
  
  Kai Chan Aug 6, 2018 8:09 AM (in response to Ryan Navarro)
  
  My dummy body for goal is off for component control.
  
  I have few pic showing solid body of rotation region in transparent color. What I did is to cover all body volume of blade and not touch with other body while leave gap between(some of the gap is small since the blade is close to nearby bodys, but the rotation region have no touch with nearby body at all).
  
  For travels and stopping criteria, please see pic (stop criteria.jepg). I do not know much area on travel time so I set it 'auto'. and I haven't set up anything for stopping criteria because I have a concept that let the calculate run it in full mode so I can obtain complete accurate view on flow trajectories. Am I wrong? could you suggest what travel time and stopper criteria is suitable for my purpose of study.
  
  I have a pic showing goals result in excel. The 'process' indicates 0% and 'use for convergence' is 'no'. does it mean the result is not usable? I have to converge it in calculation in order to get far more accurate result?
  
  I may have to ask fundamential setting about the simulation.
  One of my purpose of doing simulation is to visualize whole air flow in trajectories view within my computational domain, I want to know how wide the wind can reach after 180cm in front, as some of blade design, wind can go straight, or wide spread. The reason I turn all the goals off is because I think it will will stop in middle of calculation process when it satisfy result of the goals. For example, I have set a goal to check wind speed in front of blade 75cm ahead. When the calculate process complete those goals, obtain the wind speed result straight in front of blade 75cm location, it will ignore the rest of calculate, such as wind condition on left/right side of my wind-speed checking location. I could eventually get goals result but missing detail or obtain less accuracy view on other area in flow trajectories. Am I correct?
  
  I had tried with another blade design different time with goals on and off (pic ex_blade goal-off.jepg, ex_blade goal-on.jepg). The goal's surface/body is still velocity check at location 75cm in front, and off for component control. The result seems to me that when goal involve, the trajectories view is driven to be different. how can I tell which flow trajectories is truly reflect in real-life condition or which one is closer to real-life? When this is important to converge goals, my question in general is, what goal should I apply in order to obtain accurate view on flow trajectories that I want to visualize real air flow direction and strength.
  Although alot of discussion and question above, you agree my refinement/cell is ok. So I guess my problem could be setup about the goal, travel time, and stop criteria? Could you suggest how I can adjust those parameters and get accurate result both on wind speed(which should be close to my real life lab test) and accurate view on flow trajectories? Many thank Ryan.
  
  Like Show 0 Likes(0)
  
  Actions
  - Re: flow simulation on blade design is different from real life result
    
    Ryan Navarro Aug 6, 2018 11:43 AM (in response to Kai Chan)
    
    Kai Chan wrote:
    ...I guess my problem could be setup about the goal, travel time, and stop criteria? Could you suggest how I can adjust those parameters and get accurate result both on wind speed(which should be close to my real life lab test) and accurate view on flow trajectories?
    This is still what I believe could be the problem
    
    If you aren't using goals with convergence criteria, then the software will stop after achieving predefined number of travels (currently set to 4) You can think of a travel as approximately the amount of time it takes for a particle to make its way from one end of the computational domain to the other.
    
    If you use this approach it's a manual guess and check game to figure out how many travels you need to solve for to get your fully developed flow field. It could be 1-2 travels or 16 or more travels depending on the conditions.
    
    This is the reason for using goal convergence control. Rather than manually increasing number of travels, the recommended technique would simply be set goals for your parameters of interest and use those for convergence control. Then the solution won't stop until all of these goals are satisfied.
    
    I would re-run the study with your goals checked as "use for convergence control" to rule out that as a problem. Alternatively you could re-run the project with a manually increased number of travels (8 travels for instance) to see if results improve.
    
    Like Show 0 Likes(0)
    
    Actions
    - Re: flow simulation on blade design is different from real life result
      
      Kai Chan Aug 6, 2018 12:03 PM (in response to Ryan Navarro)
      
      Thank Ryan.
      I have already tried to run calculation with the criteria, and include your suggestion that add force on blade, but still get wind speed around 10-20% of real lab test.
      Then I have tried another simulation for control experiment: I calculate the model only with blade, and without the surrounding frame (or any body in my model). The wind speed result is 90% close to my real lab test.
      So I assume that are some missing setup I may have to do with any element other than the blade, maybe I will try to simulate the model with same size of blade, but resize the rest of element, and compare the wind speed test to see how air flow change. And this would be a topic more about flow dynamic rather than simulation operation.
      And also I will increase travel # and run again.
      Thank helping.
      
      Like Show 0 Likes(0)
      
      Actions
      - Re: flow simulation on blade design is different from real life result
        
        Ryan Navarro Aug 6, 2018 5:10 PM (in response to Kai Chan)
        
        If you are seeing that big of a discrepancy with and without the surrounding frame, maybe your rotating region is not setup ideal.
        
        Reference the built in tutorial on the CPU Cooler... you may want to try expanding the rotating region to include a portion of the surround frame, and then manually define those as stationary wall.
        
        Like Show 0 Likes(0)
        
        Actions
        
        Re: flow simulation on blade design is different from real life result
        
        Kai Chan Aug 8, 2018 11:18 AM (in response to Ryan Navarro)
        
        I will try the cpu example, set up the rotation region and include some area of other bodies and set stationary boundary.
        let see what will happen.
        
        And I have tried 10 travel, usually 4, and the result is similar so see new how's the new rotation region goes.
        Thank suggestion.
        
        Like Show 0 Likes(0)
        
        Actions
Re: flow simulation on blade design is different from real life result

Igor Fomenko Aug 4, 2018 4:07 AM (in response to Kai Chan)

First of all you deal with a MODEL of real situation, MODEL - is only model, so you never get real behaviour till including of all properties of real situation.
And second - I don`t know about features of SW flow simulation engine but I know that Ansys has several math flow models and a lot of settings inside each one. So, as a result, you have a huge amount of settings which control result you obtain.
So, before calculations you have to be familiar with a not simple theory (phisics) and math of all available models to make right decision.
Like Show 2 Likes(2)

Actions
- Re: flow simulation on blade design is different from real life result
  
  Kai Chan Aug 6, 2018 8:09 AM (in response to Igor Fomenko)
  
  hi Igore:
  I know it will never get 100% compare to real life.
  Let say, if I have wind speed on lab result is 1.5m/s, my simulation is 1.4, or 1.3....it seems to me the simulation is still in on target or close to real situation.
  But if I get 0.5, or even 0.15, I know I need some improvement on it.
  The reason using simulation is because I can testify the design and see if the result is acceptable (acceptable mean this is close to target)
  so save time and cost on prototype. I could make 20 prototype without using this tools, or now I only need to make 5 for another accurate control in real life after simulation done.
  
  Like Show 0 Likes(0)
  
  Actions
Re: flow simulation on blade design is different from real life result

Mark Keown Aug 5, 2018 2:32 AM (in response to Kai Chan)

Here is my two cents:
Increase the domain at the inlet - about as big as the outlet.
Are you sure the lab fan is rotating at 800 rpm?
At the time of the lab test measure the air pressure and temperature - use same values in simulation.
Dose the Flow Simulation match the lab setup? Do you want to include bench top, walls, structure other obstructions (anything with in ~10 fan diameters)?
I think the mesh drops off to quickly. Suggest using local mesh, cylinder to make the fine mesh volume much larger thus including in & out air flow details. If you want detailed output you need detailed input.
How are you measuring the air velocity in the lab? If using a hot-wire may want move this around
Mesh - how do you know you have enough? When you add more mesh the results do not change (see link below).
https://docs.wixstatic.com/ugd/ea3635_835c0ae5a7654ff28fe783889f6f6542.pdf
Like Show 1 Likes(1)

Actions
Re: flow simulation on blade design is different from real life result

Peter Francis Aug 7, 2018 2:20 PM (in response to Kai Chan)

Hi,
Sorry I've come to this a little late.
I've been using FloEFD in the fan industry for a number of years and my initial thoughts on your problem are as follows;

1. You need a lot more mesh around your fan blades , I'd also create some local ititial mesh both upstream and down stream of your fan.
2. Your rotation body around your blades looks a little big , I usually keep the body as close to the blade/hub as possible
3. make sure your boundaries are at least 10D (more is better) from any part of your model
4. Simulating fan blades in FloEFD is difficult, its going to be a long run
5. Definatly use convergence on your values of interest , and a few others as a cross check
6. I agree with Marks comments re modelling the real test scenario including desk , wall , temperature and pressure

I hope thats some help

Peter
Like Show 3 Likes(3)

Actions
- Re: flow simulation on blade design is different from real life result
  
  Kai Chan Aug 8, 2018 11:26 AM (in response to Peter Francis)
  
  Peter:
  
  I can also include lab setting, especially the wall behind fan could be a key.
  and I will next try rotation region conclude part of surrounding bodies, but set those body as stationary boundary. Like the tutorial setting on CPU cooler.
  I have a question:
  you said I should have more local mesh up and downstream of fan. If I am looking for downstream flow condition 1800mm away, how far does the local detail mesh should I include for 300mm diameter of blade. (if I increase the area too large, the calculation take few days!!! I accept max 9 hrs for calculation, unless this is the true that it have to take for few days)
  
  Thanks,
  
  Kai
  
  Like Show 1 Likes(1)
  
  Actions
  - Re: flow simulation on blade design is different from real life result
    
    Peter Francis Aug 8, 2018 3:30 PM (in response to Kai Chan)
    
    Hello Kai,
    Now there is a question !
    I would start with at least 1D (300mm) up & down stream of the fan and make sure it is at least 1.5D(450mm) wide . Air does not enter or leave fans parallel to the axis of rotation.
    Don't forget the flow into the fan has an influence on the downstream flow so if its not got enough detail it won't give valid results.
    
    To get accurate results from a fan model will need a lot of cells and extended run time.
    
    The way I'd approach this is to build the analysis up piece by piece as follows
    I'd start with just the fan and the Local Initial Mesh (upstream and downstream) , place some measurement bodies in the relevant areas
    Add goals for velocity , static pressure , and any other relevant goals and use these for convergence
    Make sure your temperature & pressure are the same as during the testing.
    
    I'd run this to see how close to your real life values you can get, If necessary refine the model , more cells especially in the LIMs
    
    Repeat until you can't improve your results.
    
    Good Luck
    Peter
    
    Like Show 1 Likes(1)
    
    Actions
Re: flow simulation on blade design is different from real life result

Jon Dahlgren Aug 9, 2018 6:32 AM (in response to Kai Chan)

You have a few good answers above.

I'd emphasize on the mesh resolution tho which needs to be refined with the gradients of the flow, not only with the solid walls. Domain also looks small. To get an idea of the mesh strategy I'd suggest you to use the adaptive mesh refinement control.

Good luck.
Like Show 1 Likes(1)

Actions
Re: flow simulation on blade design is different from real life result

Stephen Endersby Aug 15, 2018 3:37 PM (in response to Kai Chan)

Hi Kai,

take a look at this presentation which discusses external aerodynamics. Especially the tips for the mesh. https://www.mentor.com/products/mechanical/multimedia/player/aerodynamicanalysis-of-a-red-bull-air-race-world-championship-raceplane-55d763df-5a2a-4302-9cae-9f49ded1020d
You need a good mesh at the leading and trailing edges of the fan blade to get a decent solution
Like Show 0 Likes(0)

Actions

Go to original post