Learning Forum

[MarkA]

Hello everyone,

I am trying to simulate a quarter of a 3 dimensional diffuser in steady state in Fluent. To this end, I have split the diffuser in 3 different bodies belonging to the same part as per the image below, and I applied symmetry planes to the faces on which symmetry is required (faces with normals along the z and -y direction). The flow is in the +x direction, and the inlet area is 0.5 x 0.5 mm2.

[YasserSelima]

Excellent work!nThe mesh seems to be great. The oscillation of the residuals could mean the problem is actually transient because of the vortex happening. But as they converge and reach steady value, the solution is converged. nFrom rough manual calculations, what is the expected throat pressure? Are the fine mesh getting you closer to the theory?n

[Rob]

I'd avoid the transitional models unless you've carefully read up on the boundary conditions. nHave a careful look at pressure and velocity around the expansion. As notes you've probably picked up flow separation. n

[MarkA]

Firstly, it's very comforting that my mesh looks great.nSecondly, the velocity vector profile depends heavily on the number of iterations. For example, the picture below is from the same simulation but with 5 more iterations. Clearly the velocity around the walls has changed in some spots.nThirdly, since I am using a small Reynolds number (Re = 100-ish), the flow is laminar from the inlet to the throat. In the diffuser, the geometry expands and a seperation point is expected, as both of you point out. Since I am expecting some turbulence in the boundary layer after the seperation point, a transitional model seemed fit to me, and was recommended by my uni professor as well.nFourthly, regarding the boundary conditions: at the inlet, the turbulence is defined by the intermittency, turbulent intensity and hydraulic diameter. The intermittency is set to 0 since I am expecting laminar flow at the inlet, TI = 1% due to the low turbulence involved here and the hydraulic diameter is based on the geometry. At the outlet, the turbulence is defined by the same three parameters. Since some backflow is expected at the outlet, which is both laminar and turbulent, the intermittency is set to 0.5, while the TI is set to 5%. The hydraulic diameter is once again calculated from the geometry at hand. do these BCs seem reasonable to you?nFifthly, the intermittency is rather low, as can be seen in the below picture, as well as the turbulent viscosity ratio, one picture further below. Does that imply that a laminar model would be sufficient in this case? This seems unlikely to me due to the seperation layer which is present in the system.nnLastly,, the only analysis I could think of to get a feeling of the pressures involved around the throat and exit is by means of an energy approach, (1 = throat, 2 = exit)np1 + 1/2*rho*v1^2 = p2 + 1/2*rho*v2^2 + K_LnnIn the above formula, v1 and v2 are known, and hence p1 can be expressed in terms of the unkowns p2 and K_L and the known velocity terms. However, since K_L is not known beforehand and often calculated by means of simulations (which is also the purpose of my simulations by the way), it is hard to find exact values for p1 and p2. Is this what you had in mind as well, or could you otherwise provide some material that I can read up on?n

[Rob]

TI of 5% for potentially transitional flow is high, 2-3% is generally used for most applications. n

[YasserSelima]

nYes this what I meant but overall the whole tunnel. K_L could be estimated from theory using Re. nYour finding of the different velocity profile after few iteration confirms the reason why you are not getting convergence. This the same kind of residuals expected when solving unsteady problem using steady solver.nTry to do the following, save case and data away from the working directory. Then, change the case to Unsteady. Compare the time averaged pressure at the throat between different mesh sizes. n