Conflict of input parameters u1 usage: Two different meanings in the same case

[crismze]

Dear Xcompact Team:

I'm having trouble understanding why, for a cylinder wake (itype=5), the parameter u1 is set as an inflow velocity and then, apparently, becomes both a flag option and an outflow velocity.

subroutine inflow:
Here, I don't have any trouble understanding that u1 is the inflow velocity. Also, I interpret that this is an adimensional velocity since the equations are also adimensionalized (with the exception of the ABL case, I think).

...
do k=1,xsize(3)
       do j=1,xsize(2)
          bxx1(j,k)=u1+bxo(j,k)*inflow_noise    ! ******Here it is used as inflow velocity
          bxy1(j,k)=zero+byo(j,k)*inflow_noise
          bxz1(j,k)=zero+bzo(j,k)*inflow_noise
       enddo
enddo
...

subroutine outflow:
However, I found myself confused in this case. For u1=0, I get that the advection factor uses an Uc=1/2*(ux_max + ux_min). But what's the meaning of u1=one and u1=two in this case, beside being a numerical flag? If this parameter is being repurpose, shouldn't be better to create a new parameter for the selection of the characteristic or advection velocity Uc?

...
! *****Here is used as a flag option to compute the advection factor cx based on velocity Uc
if (u1 == zero) then
       cx=(half*(uxmax+uxmin))*gdt(itr)*udx
    elseif (u1 == one) then
       cx=uxmax*gdt(itr)*udx
    elseif (u1 == two) then
       cx=u2*gdt(itr)*udx    !works better
    else
       cx=(half*(u1+u2))*gdt(itr)*udx
endif
...

And then, u2 is repurpose again...

...
 if (iscalar==1) then
       if (u2==zero) then
          cx=(half*(uxmax+uxmin))*gdt(itr)*udx
       elseif (u2==one) then
          cx=uxmax*gdt(itr)*udx
       elseif (u2==two) then
          cx=u2*gdt(itr)*udx    !works better
       else
          stop
endif
...

Thanks in advance to their guidance in this matter!

[slaizet]

Hi,

The 1D advection equation used in the outflow can be unstable from time to time especially when large vortical structures are advected outside of the computational domain. As a results it is possible to use different convection velocity (the natural one being u1, same as in the inlet), but sometimes it crashed. We worked out that using a convective velocity of two instead of one, could be more stable. My advice is to try different values and keep the one that will be stable for your flow configuration.

Best,

Sylvain