[joel.lehikoinen]

Hi,

I'm having trouble getting my solution to converge. The computation becomes unstable already at the first time step; the final residual of pressure (p_rgh) is always huge. I'm using buoyantPimpleFoam, because it is the solver we usually need to use. However, this specific case is isothermal and buoyancy effects are negligible.

I have tried all the usual tricks: playing around with the relaxation factors, increasing viscosity and using first order discretization schemes. I have also tried simpler meshes, but to no avail. My initial conditions are the interesting thing: the simulation runs well with pisoFoam, with temperature and buoyancy omitted. I've taken the initial conditions for p_rgh and U from the pisoFoam simulation. I have only changed the boundary conditions for p_rgh to buoyantPressure. For U, I've remembered to change the volumetric flow to mass flow.

I don't really understand why buoyantPimpleFoam fails where pisoFoam works. As I mentioned, the buoyancy effects should not affect the case that much. The computation fails also if I set g to zero. I set the pressure reference level in the pisoFoam simulation to something I expect is reasonable also when compressibility is accounted for. Also, I've tried using incompressible thermophysical model, so I don't think the problem is in the thermophysics.

I am at a loss what to try next. What could be the difference that causes buoyantPimpleFoam fail when pisoFoam succeeds? I could run this case using pisoFoam, but for the sake of generality and future use, I'd prefer to see buoyantPimpleFoam at work.

[aujamal20]

Originally Posted by joel.lehikoinen

Hi,

I'm having trouble getting my solution to converge. The computation becomes unstable already at the first time step; the final residual of pressure (p_rgh) is always huge. I'm using buoyantPimpleFoam, because it is the solver we usually need to use. However, this specific case is isothermal and buoyancy effects are negligible.

I have tried all the usual tricks: playing around with the relaxation factors, increasing viscosity and using first order discretization schemes. I have also tried simpler meshes, but to no avail. My initial conditions are the interesting thing: the simulation runs well with pisoFoam, with temperature and buoyancy omitted. I've taken the initial conditions for p_rgh and U from the pisoFoam simulation. I have only changed the boundary conditions for p_rgh to buoyantPressure. For U, I've remembered to change the volumetric flow to mass flow.

I don't really understand why buoyantPimpleFoam fails where pisoFoam works. As I mentioned, the buoyancy effects should not affect the case that much. The computation fails also if I set g to zero. I set the pressure reference level in the pisoFoam simulation to something I expect is reasonable also when compressibility is accounted for. Also, I've tried using incompressible thermophysical model, so I don't think the problem is in the thermophysics.

I am at a loss what to try next. What could be the difference that causes buoyantPimpleFoam fail when pisoFoam succeeds? I could run this case using pisoFoam, but for the sake of generality and future use, I'd prefer to see buoyantPimpleFoam at work.

Hi joil,
Have u got the solution to overcome the convergence problems in buoyantPimpleFoam. Right now I am trying to simulate a simple cube filled with fluid and fluid properties are defined by icoPolynomials. The vertical walls of the cube (2D case) have a fixed temperature gradient. I can run the case if set the case according to buoyantBoussinesqPimpleFoam but when I try to use the buoyantPimpleFoam is shows the convergence problem... I have tried whatever I can do but in vain.

Any suggestion will be welcomed.

Thanks.