[f0208secretx]

I am using the pressure-based segregated solver with transient formulation. Pressure-velocity coupling is done by PISO. If I understand the algorithm correctly, it solves u,v,w momentum equation based on the states of the fluids (p, U) updated in the last time step. After this, SIMPLE family of algorithms will perform a pressure projection in an attempt to satisfy the continuity equation. In this process, the pressure Poisson equation is solved, using the boundary condition $grad(p) dot n = rho * surfaceAcceleration$. If the surface we are solving is static, then the right hand side of this equation will just be zero. PISO simply iterates a couple of times to make the solution closer to both continuity and momentum equation.

Upon agreement of that solution process, I have a question regarding the "transient export" option in Fluent, using which I exported the pressure gradient and normal vector on a patch that I specified as wall (static) in ASCII format. However, no matter I used "cell center" option or "node" option, the computed $grad(p) dot n$ is always non-zero. I attached one time step of the exported solution for your reference at the last time step of my simulation. In my processing, I simply do $dp-dx*x-face-area + dp-dy*y-face-area + dp-dz*z-face-area$, which should approximate the boundary condition, which should be zero exactly on the boundary. I have also computed the normal vector myself based on the geometry of the surface, using face-weighted technique and dot it with grad(p), still far from zero. The largest number I get from doing this can reach values on the order of 0.01~0.02, which is far from being simply a machine precision error. The residual is around 1E-5~1E-7, which seems to suggest that convergence is reached. Also, I have done a temporal resolution test, and it seems that cutting the time step by half does not alter the solution by much, this also indicates a convergence.

I understand this error could potentially be a result of the least-square based gradient computation, or the Rhie-Chow interpolation, but the magnitude of the offset is too great to be considered as a minor truncation error or anything like that.

I am puzzled. Can someone provide an explanation? Thanks

[f0208secretx]

hello,

does anyone know, or care about how fluent setup the pressure boundary conditions...?