Hi, i'm doing some DNS/LES computations of a jet flow. The code used a fourth order dinite difference scheme on a cartesian grid. Artificial non reflecting boundary conditions are used at all boundary. A velocity profile is prescribed at the inflow and the computations are started with the same condition in the whole domain. Density profile are also prescribed. Pressure is uniform and the flow is isothermal.

The non dimensionalization of the equations are made and several paremeters are to be specified, namely the Re number which depends on the nozzle exit diameter.

My question is, how to increase the re number of the computed flow? do i need simply to change the numerical value of the Re number i specify for the initialisations (keeping constant all the other values) or should i change also the diameter of the jet to increase the reynolds number? Thanks

Re is defined as rho*v_infty*D/viscosity

So Re is defined by the condition you are prescribing!!

If you retain constant all the parameters that are currently accepted by the codes (rho, V_infty and D) a change in Re will cause a change in the viscosity.

Pay attention to the condition that you have to simulate and calculate the Reynolds you have to impose.

In general, Re = density * velocity scale * length scale / viscosity, so for a set geometry of characteristic length L, the best way to increase Re is to increase the characteristic velocity scale U, or descrease the viscosity.

In general the Reynold number is given as an input parameter. For a given geometry let say length scale L is known (i.e jet diameter), velocity scale is know (centerline jet velocity), etc... So prescribing a Re number let say Re=1000 is the same as to prescribe a given viscosity. Rigth? Then if a just increase the initial Reynold parameter from 1000 to 10 000 i'm decreasing the initial viscosity. How can i be sure that i'm simulating a jet at that given Re number? How high can i increase such a value? is it depending on the numerical stability of the scheme, resolution, domain length?

Thanks for our answers.

Well if you have a reference length and velocity, then prescribing an Re as in initial condition either sets the viscosity or the density of the flow. What commercial program are you using to do the simulation, or are you using a custom-written code?

Do you have a reference experiment to compare your data with? Does it have its initial length, velocity, density and viscosity specified? It's unwise to arbitrarily increase the Reynolds number of a simulation until you are sure that the code produces reliable results.

I'm using a custom-written code. I'm comparing the results with incompressible experiment results and with low reynold compressible DNS. It compares well at low reynold cases less than 50 000. Now my question before computing higher Reynold case is how far can i arbitrarily grow the Re number. Shall i just refine the mesh (how fine for a given Re number?) and increase the Re number or they are other feature to be taken in account? I also read in the litterature that using LES does not compute the real reynold number of the flow since you are adding an extra viscosity by LES filtering. Some people rather use MILES approach to avoid this. Can someone explain this? Thanks

Arbitrarily increasing the Re by, say, increasing the input velocity of the jet will cause some issues. The momentum thickness of the boundary layer inside the channel from which the jet issues will descrease, requiring a significant increase in the mesh resolution around this area. Adequate simulation of any kind of free shear flow requires this quantity to be particularly well-resolved, as the roll-up frequency of the shear layer, and its subsequent development is very sensitive to this parameter.

LES is actually very good at simulating this type of flow. With a sufficiently fine mesh, the subgrid-scale model tends to zero in the laminar regions, hence the effect it has on the viscosity of the flow is quite low. The advantage that LES has over URANS and other such methods is its ability to resolve the large-scale structures that are present in the shear layer, leading to much better predictions for transition and mixture fraction in the jet. MILES is a very contentious simulation paradigm, as it uses numerical dissipation to account for the subgrid fluctuations. It has its limitations, as a finer grid may result in poorer simulation results than on a coarser grid for a given numerical scheme. MILES has also only really been tested on decaying isotropic turbulence, which has limited applicability to flows of practical interest. It also has some severe deficiencies for wall-bounded flows.