Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Help Needed - Lift and Drag for a Flat Plate, Normal to Flow

Please login with a confirmed email address before reporting spam

I am trying to determine the lift and drag coefficients for a 2-D flat plat ( 0.1 m chord x 0.0166667 m height ) oriented perpendicular to an air flow with Re of 10^4 using the Spallart-Almaras Turbulent Flow module. My primary concern is in matching the "textbook value" for C_d -- My COMSOL run returns 2.00615 whereas I'm expecting the value to be around 0.9.

I am following the method described in the Flow Past a Cylinder Tutorial. I am evaluating the drag force over the boundary integral of my flat plate. I then compute C_d = - 2*reacf(u)/(spf.rho*V^2*height).

Can someone tell me where I might be going wrong? I have attached my COMSOL model for your reference.

v/r
Connor


8 Replies Last Post 9 lug 2014, 15:29 GMT-4

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 apr 2014, 21:14 GMT-4
I am not an expert and defer to other input when available. However, I noticed a few things about your model. You didn’t have an initial condition set for the velocity, which I am sure would increase your time to convergence. Also, you didn’t set the upper and lower walls to moving, to simulate an infinite domain. In order to evaluate the Cd, I created a line integral evaluation around the plate which is the following:

-2*reacf(u) / (spf.rho* (Re*spf.nu/chord)^2 *height) / (2*height+2*chord)

That is equivalent to the following:

(Reaction Force in U) / ( 1/2 Density * U0^2 * Height ) / (Perimeter)

The perimeter term comes from the line integral. You can see that in my second line integral which integrates the #1... It gives the perimeter back as an answer. The units now come back as unitless, as they should (I think). I also made the height a parameter and included it everywhere, instead of carrying the 0.1. the domain is now 10 chords wide and 10 heights tall.

The new value I get is 0.86, which is getting closer to where you want to be, I think. Anyone else have some thoughts?

-TCL

P.S. The updated model file is attached. But I think I am 4.4... hmmm, didn't think that one through.
I am not an expert and defer to other input when available. However, I noticed a few things about your model. You didn’t have an initial condition set for the velocity, which I am sure would increase your time to convergence. Also, you didn’t set the upper and lower walls to moving, to simulate an infinite domain. In order to evaluate the Cd, I created a line integral evaluation around the plate which is the following: -2*reacf(u) / (spf.rho* (Re*spf.nu/chord)^2 *height) / (2*height+2*chord) That is equivalent to the following: (Reaction Force in U) / ( 1/2 Density * U0^2 * Height ) / (Perimeter) The perimeter term comes from the line integral. You can see that in my second line integral which integrates the #1... It gives the perimeter back as an answer. The units now come back as unitless, as they should (I think). I also made the height a parameter and included it everywhere, instead of carrying the 0.1. the domain is now 10 chords wide and 10 heights tall. The new value I get is 0.86, which is getting closer to where you want to be, I think. Anyone else have some thoughts? -TCL P.S. The updated model file is attached. But I think I am 4.4... hmmm, didn't think that one through.


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 apr 2014, 21:54 GMT-4
Excellent, thank you!

Does the moving wall have the same velocity component as the free stream velocity: Re*spf.nu/ chord in the x-hat direction?
Excellent, thank you! Does the moving wall have the same velocity component as the free stream velocity: Re*spf.nu/ chord in the x-hat direction?

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 apr 2014, 22:02 GMT-4
Yes. They are set at (Re*spf.nu/chord). -TCL
Yes. They are set at (Re*spf.nu/chord). -TCL

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 apr 2014, 22:35 GMT-4
Okay, I have attached what I have. As you mentioned, your model file was for version 4.4 so I had to base my model off the explanation and run it in 4.3a.

I set the wall condition for the top and bottom of the domain to 'moving' at a velocity of ( Re*spf.nu/chord ) and
added a separate no-slip condition for the flat plate's boundary.

I am still getting a different value for C_d: 6.19649
Okay, I have attached what I have. As you mentioned, your model file was for version 4.4 so I had to base my model off the explanation and run it in 4.3a. I set the wall condition for the top and bottom of the domain to 'moving' at a velocity of ( Re*spf.nu/chord ) and added a separate no-slip condition for the flat plate's boundary. I am still getting a different value for C_d: 6.19649


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 apr 2014, 22:53 GMT-4
Well... odd. The file you posted earlier had a chord of 0.6 and a height of 0.1. That gives a difference answer than the one you have here with 1/6th the scale. I have both versions posted here.
Well... odd. The file you posted earlier had a chord of 0.6 and a height of 0.1. That gives a difference answer than the one you have here with 1/6th the scale. I have both versions posted here.


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 8 apr 2014, 12:34 GMT-4
I see what you're saying. I went ahead and ran a study for other aspect ratios and got the same result.

It seems that it isn't necessary to divide by the perimeter within the integral evaluation which is strange as the line integral of '1' over the surface yields the perimeter. I have attached an excel spreadsheet that shows my data thus far. The values labeled "COMSOL - C_d" were returned by the equation for C_d divided by the perimeter whereas the "COMSOL w/ Perim" values are the aforementioned values multiplied by the perimeter.
I see what you're saying. I went ahead and ran a study for other aspect ratios and got the same result. It seems that it isn't necessary to divide by the perimeter within the integral evaluation which is strange as the line integral of '1' over the surface yields the perimeter. I have attached an excel spreadsheet that shows my data thus far. The values labeled "COMSOL - C_d" were returned by the equation for C_d divided by the perimeter whereas the "COMSOL w/ Perim" values are the aforementioned values multiplied by the perimeter.


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 apr 2014, 20:05 GMT-4
That seems to work out then. It is getting closer for sure. It looks like the answer is to leave the perimeter (integral) alone and just run with it. I have some similar runs for a sphere in 2 and 3D that I will follow this up with as soon as I can package them appropriately. Attached is a 2D sphere (axi) which gets near (but not on) the ~0.5 answer for 100,000 Re. This seems to be VERY mesh dependent due to the line integral (I would assume). I must do more work on investigating that. I will run a parameter sweep over Reynolds number and report back to see if it captures the dip at approximately 3E5.

-reacf(w) [N] / ( 1/2 * spf.rho * (Re*spf.nu/Diam)^2 * (pi*Diam^2/4) ) (1)
0.37155
That seems to work out then. It is getting closer for sure. It looks like the answer is to leave the perimeter (integral) alone and just run with it. I have some similar runs for a sphere in 2 and 3D that I will follow this up with as soon as I can package them appropriately. Attached is a 2D sphere (axi) which gets near (but not on) the ~0.5 answer for 100,000 Re. This seems to be VERY mesh dependent due to the line integral (I would assume). I must do more work on investigating that. I will run a parameter sweep over Reynolds number and report back to see if it captures the dip at approximately 3E5. -reacf(w) [N] / ( 1/2 * spf.rho * (Re*spf.nu/Diam)^2 * (pi*Diam^2/4) ) (1) 0.37155


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 9 lug 2014, 15:29 GMT-4
Hi all,
I'm trying to plot the lift and drag force on a rotating cylinder in a uniform stream (2D) for low Re=100. I've been successfull with calculating/plot the coefficient for the non-rotating cylinder which is well in order with literature.

For the non-rotating cylinder I've used the non-slip boundary condition on the cylinder and the expression: -reacf(v)*2/(spf.rho*U_mean^2*d) for the lift coefficient. Where d is the cylinder diameter and U_mean is the free stream velocity.

Now, when I set the boundary to "sliding wall" and set a rotation velocity the calculated lift force C_L=-reacf(v)*2/(spf.rho*U_mean^2*d)=0. Why is that? please help me asap

thanking you
susobhan
Hi all, I'm trying to plot the lift and drag force on a rotating cylinder in a uniform stream (2D) for low Re=100. I've been successfull with calculating/plot the coefficient for the non-rotating cylinder which is well in order with literature. For the non-rotating cylinder I've used the non-slip boundary condition on the cylinder and the expression: -reacf(v)*2/(spf.rho*U_mean^2*d) for the lift coefficient. Where d is the cylinder diameter and U_mean is the free stream velocity. Now, when I set the boundary to "sliding wall" and set a rotation velocity the calculated lift force C_L=-reacf(v)*2/(spf.rho*U_mean^2*d)=0. Why is that? please help me asap thanking you susobhan

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.