Register Blogs Members List Search Today's Posts Mark Forums Read

April 24, 2014, 06:10
#1
Member

Join Date: Dec 2012
Location: South Africa
Posts: 40
Rep Power: 12
Hello all

I've recently run a simulation for flow over a 3-D vertical flat plate using FLUENT 14.5. I've attached the geometry for a clearer understanding. I have run the simulation using the RNGk-epsilon model using second-order upwind differencing for all variables. I found that under steady state conditions I get a fluctuating, somewhat periodic, velocity measured in the wake of the plate. When I then switch to an unsteady simulation the fluctuating velocity appears to be damped out to a steady value. Can anyone assist me in understanding if there is an issue with the simulation or if this is a valid result? I have attached the monitor for the y-velocity to give a clear picture.

Attached Images
 vel-timestep.png (8.2 KB, 20 views) vel-iteration.png (16.4 KB, 18 views) geom.png (10.1 KB, 15 views)

 April 24, 2014, 15:12 #2 Senior Member   lore Join Date: Mar 2010 Location: Italy Posts: 460 Rep Power: 16 Hi, what is your free stream velocity? You mention y-velocity, is that your side velocity (x being along the freestream direction and z being vertical and normal to the ground)? Where are you measuring that velocity exactly? What is the frequency of your measurement?

 April 24, 2014, 15:32 #3 Senior Member   Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,266 Rep Power: 67 Try to compute the kinetic energy integrated in all the domain and see the evolution in time... just the velocity component at a point is not meaningful...

 April 24, 2014, 15:45 #4 Senior Member   lore Join Date: Mar 2010 Location: Italy Posts: 460 Rep Power: 16 Filippo, could you please explain why you think that the velocity signal at a point is not meaningful? If you want to characterize the vortex shedding downstream a bluff body I don't see any other way apart from looking at the velocity or pressure fluctuations in the wake…

 April 24, 2014, 16:41 #5 Senior Member   Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,266 Rep Power: 67 In order to characterize the shedding a better way is to check the integrated pressure coefficients over the body. A single velocity point is not meaningful in statistical sense. The total kinetic energy allows to assess the fully development and the energy equilibrium

 April 25, 2014, 04:45 #6 Member   Adhikar Hariram Join Date: Dec 2012 Location: South Africa Posts: 40 Rep Power: 12 Hello In response to Lore: my velocity was approximately 24m/s as I am trying to replicate experimental work: VORTEX SHEDDING FROM SQUARE PLATES NEAR A GROUND PLANE: AN EXPERIMENTAL STUDY, Masters thesis by Rosemary Ricohermoso Matty. And by y velocity I meant perpendicular to the ground where x is along the freestream and z being to the side. The velocity is being measured 1.5 plate lengths down stream, halfway up the plate and it is measured every timestep, at 0.00022sec. In response to Prof. Denaro: You mention that in order to characterise vortex shedding is to check integrated pressure coefficients over the body, will this also relate to drag and lift coefficients on the body as I have recorded data over the simulation time for those quantities and they display the same form of curves. You also mention that I should check kinetic energy integrated in the entire domain, if I were to re-run the simulations what must I look for in the integrate kinetic energy in order to assess the frequency of the shedding, as this is all I am interested in, not so much the strength of the vortices?

 April 25, 2014, 08:46 #7 Senior Member     Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 25 Coming back to the original question: You were wondering, why you get unsteady results in the steady simulation (by means of some velocity value over the interation counter) and steady results in the unsteady compution, if I get you correctly. Now, having these "unsteady" results over the iterations just means, that the solver does not converge. This is not a usefull result at all, except that this might indicate that your actual simulation is just not steady. But the results themselfs are not physical. For some reason, the unsteady solver seems to actually converge to steady state, which (if done correctly) is a physical meaningful result. It seems like your flow is "slow enough" that the turbulence model will damp out all fluctuations. It does what it is supposed to do - calculate the average profiles. Only if you get strong / slow / large laminar vorticies the RANS model should switch to URANS mode. FMDenaro likes this. __________________ The skeleton ran out of shampoo in the shower.

 April 25, 2014, 08:49 #8 Senior Member   Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,266 Rep Power: 67 I agree, again, I suggest to plot the total kinetic energy versus time for URANS and versus iterations for RANS

 April 25, 2014, 09:07 #9 Member   Adhikar Hariram Join Date: Dec 2012 Location: South Africa Posts: 40 Rep Power: 12 Thank you Philipp and Filippo Looking at the non-transient solution there is a large recirculation zone being the plate as expected. I suppose this may be a large vortex creating the issue when run as steady state? I will re-run the simulations and keep track of the total kinetic energy as well as the other variables. In the case of vortex shedding being present, would the kinetic energy plot fluctuate at the frequency of the shedding? I will report back on my findings in a few days if you are willing to assist further at that stage. Many thanks again Adhikar

April 28, 2014, 08:07
#10
Member

Join Date: Dec 2012
Location: South Africa
Posts: 40
Rep Power: 12
Hello all

I have re-run the simulation and recorded the total TKE integrated in the domain and plotted it against iterations and flow time. Please see the attached images for these plots. As far as I can tell they display the same behaviour as with the velocity plots, so can I conclude that the simulation results in a steady solution, however this steady solution cannot be reached without running a transient simulation?

Attached Images
 int_tke.jpg (19.2 KB, 16 views) int_tke_trans.jpg (21.1 KB, 13 views)

 April 28, 2014, 08:32 #11 Senior Member   Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,266 Rep Power: 67 Both Solutions reached the same equilibrium level of energy, the Rans has a sufficiently low residuals?

 April 28, 2014, 10:50 #12 Member   Adhikar Hariram Join Date: Dec 2012 Location: South Africa Posts: 40 Rep Power: 12 The RANS simulation reaches residuals of around 0.0015 for continuity, 0.000003 for x,y,z velocity and around 0.0002 for k and epsilon. Could you let me know if the RANS simulation is not in equilibrium in the regions where the TKE is fluctuating, such as between 4000 and 5000 iterations? Thank you for the help thus far

 Tags damping, oscialltion, rng k-epsilon, unsteady, urans