Hints & Tips: Choosing Boundary Conditions
I am confused. I want to have fluid move smoothly out of a computational boundary, but I don't know what boundary condition to choose that will do this best.
There are many choices for a boundary condition where fluid is expected to pass out of the computation. Included in these choices are Continuative, Periodic, Specified Pressure and Outflow. Which one is best depends on the physical conditions of the overall problem. One thing to ask yourself first is whether or not it's possible to realize a particular boundary condition in a laboratory experiment. If the answer is "no, the condition cannot be physically realized," then that is a good indication that it is also unlikely to work in a computation.
At a Continuative boundary all flow quantities are assigned a zero normal
derivative. This is not a proper physical condition and should only be used when
there is no other option. Obviously, a Periodic boundary condition makes sense
when the assumption of periodicity in the direction of flow can be made. A Specified
Pressure condition is a well-defined physical state that can often be closely
approximated in an experiment. This condition, however, does not always work
in the presence of free surfaces, because one does not always know the shape
and/or location of the surface at the boundary, or because surface tension forces
are not consistent with a uniform pressure at the boundary.
Based on its name, the Outflow boundary condition might seem to be the best choice under any circumstance. Unfortunately, the term "outflow" was intended primarily for problems having outgoing waves. The Outflow boundary condition was designed to pass waves (surface or internal) through a boundary with a minimum of wave reflection back into the computational region. Some changes have been made to this condition to also permit a new outflow of material in addition to waves, so this may be a good choice for many problems.
In most cases, when the flow across a boundary is truly going out, and the boundary is well removed from regions of most interest, there is a good chance that it is performing properly. The greatest difficulty associated with outflow boundaries occurs when they have regions of incoming flow. When this happens it is very important to assess whether or not the inflow could be adversely affecting the results. Because an outflow boundary condition does not specify conditions for inflow. What comes in is anybody's guess.