simtk.org

Hi Developers,

I am running the Simvasuclar to do a 3D pusatile blood flow simulation on a benchmark of bifurcation case, see geomtry.png. The inflow is a function 100*sin(x) with x in [0,pi], and I use RCR boundary condition. Details of the Solver Parameters are attached in 1, 2, 3 .png.

My questions are:
1. Why the inflow at the inlet boundary is forced to do a Fourier approximate (Fourier models), even though my inflow is a smooth function? Is there any way to cancel this approximation?

2. I now know that the outflow rate will fluctuate if the Time Integration Rho Infinity is set as 0.5 (see the inflow outlow.jpg, where red is the theoretical inflow, black is the total outlfow, blues are the two outlet flow). When I change the value to 0, the flunctuation disappears (The value 0 corresponds to maximal numerical dissipation, where the under-resolved frequencies are annihilated within one time step), but there is a shift in the profile and the magnitude is decreased a little bit. It is obvious when the use a large mesh. Since I use an incompressible flow, why the shift emerges. The mass quality should be convervative.

Thank you so much.

Best regards,
Shanlin

I guess the reason for using Fourier series to approximate inflow data is to create a periodic inflow condition. You can modify the bct.dat to impose the nodal velocity.
My understanding is that discretized numerical schemes usually will not give you a perfect conservation and unlike finite volume schemes for conservation laws, where the quantities are conserved in each control volume, the weak form used in the finite element method tries to find solutions to satisfy the equation in an average sense. Contintuity may not be accurately satisfied in each element. You can consider reducing the mesh size and residual crieria to improve the difference. If the inflow rate is 100*sin(t), t=[0, >pi], I am not sure if the backflow stabilization has any influences. You can use periodic inflow without regurgitation to see if it is significant. \rho_{inf} has an impact on dispersion and dissipation. I suggest you adjust the value between 0-1 instead of using 0 or 1 for most problems based on the properties you want to preserve. Reducing the time step may help reduce dispersion errors.