Dear all,

I would like to calculate a turbulent jet using LES. I have problems at the outflow. The BCD are as follow:

ENTRAINMENT |---------------------------------| | | | I |O ||N |U ||L |T ||E |L | T |E | |T |---------------------------------|

ENTRAINMENT

at the Inlet (nozzle): tanh-velocity profile Entrainment region : v- or w velocity

(for entrainment mass flow) outlet : convective bcd or

extrapolated bcd Model : Smagor.

The domain has a size of 18D x 10D x 10D where D is the size of the nozzle.

The problem: At the outlet the pressure decreases and subsequently the velocity increases and finaly the solution "runs away".

Has anybody calculated such a problem? Where are my mistakes. Any suggestions/remarks are very appreciated.

Matthias

Matthias,

How do you specify v- and/or w-velocities at the entrainment boundaries. What you really need are some sort of unsteady entrainment BCs which are hard to prescribe. The scaling laws for turbulent jets would only provide steady state estimates of entrainment.

Some have suggested the computation of entrainment velocities by extrapolating from the inside of the computational domain. Such an approach would not be mathematically sound.

Could you settle for a confined jet simulation instead of a free jet simulation. That is perhaps the easiest way of simulating jets. Also the computational domain needs to be extended far more than 18D. Since no convective outflow condition is perfectly non-reflecting (unless you use parabolized NS solvers or fully upwinded schemes), the domain of interest needs to be as far from the outflow condition as possible. It would not hurt to have a sponge region that would hinder (if not fully prevent) the noise created at the outflow boundary from propagating upstream.

Here are a few more observations from my brief experience with LES of jets. I had noticed that the effects of outflow boundary are minimal if I used a 5th order upwind-biased scheme instead of a second order scheme. Though, a recent AIAA paper by Mittal and Moin suggests the use of 2nd order central scheme over a 5th order upwind biased scheme, it was Rai and Moin who suggested earlier that the 5th order upwind-biased scheme was more preferable. Also, much is being made of the so-called kinetic energy preserving schemes in LES which are necessarily central. I have not seen a single one of these extended for variable density flows or for use on general grids. So the answer to the question : which scheme is better depends on the problem and for jets upwind-biasing helps. Since the truncation error is 5th order, I would suspect that the numerical dissipation would be much smaller than the dissipation resulting from the viscous and subgrid stresses (which usually tend to be second order in grid spacing). Hope this helps.

Dear Kalyan

thanks for the answer, currently I am trying to use convectiveinflow BCD.

Of course a confined jet simulation is much more easier to model. But I'd like to simulate this problem.

You mentioned a sponge layer at the outflow, I have read about this in a paper of Freud, but without details. Can you please refer to a paper where it is described in more detail or can you say sth. about the implementation.

Thanks a lot

Matthias