[dhruv]

Hello all,

I am trying to simulate a fan in an open channel with a radiator in front of the fan, modeled as a porous zone. The flow in the channel is entirely driven by the fan. I am using the fan boundary condition. My mesh is made of almost 3 million cells.

Some details are as follows:

p (inlet and outlet)

type totalPressure;

U (inlet)
type pressureInletOutletVelocity;
value uniform (0 0 0);

U (outlet)
type inletOutlet;
inletValue uniform (0 0 0);
value uniform (0 0 0);

solver: porousSimpleFoam

The pressure curve of fan is taken as linear for simplicity, with 2 coefficients in the f List<scalar>. the equation is of the form delP = a - bx, where x is the velocity.

My observations are that as I change the coefficients of the fan in the same geometry, my calculations take longer and longer to converge. So, if I take the first coefficient (value of "a" in the fan equation above) as (say) 500, my solution converges in 6000 steps, but if I increase the value of the coefficient to 800, the solution takes almost 9000 steps to converge. So, it takes very long time to reach the duty point of the fan.

I do not understand why this happens . Can anyone help me with this?

Regards,

Dhruv.

[eugene]

Dhruv,

I might be wrong here, but the probable reason this thing takes so long to converge is because your fan has to accelerate the flow from a stationary starting point. The faster the fan turns, the more the flow has to be accelerated before it reaches equilibrium, leading to longer integration times.

Try starting with a initial velocity field closer to the final one and see if this improves matters.

Eugene

[dhruv]

Hi Eugene,

Thanks for the reply. I did get a faster convergence to one of my problems, but not in some cases. But I would like to point out that the problem was the explicit and implicit solution method applied to porousZones. When I changed my fvSolution file to implicit from explicit, my problem reached convergence in 2500 steps (rather that 9000-10000 steps in explicit), and also was much more stable and accurate.

Do you know, why is there so much of a difference between these two solution methods?

Regards,

Dhruv.