11.9.1. General guidelines on boundary conditions
¾ Ensure that appropriate boundary conditions are available for the case being considered. For swirling flows at an outlet consult the manual to check that appropriate boundary conditions have been implemented (for example, radial equilibrium of pressure field instead of constant static pressure). Special non-reflecting boundary conditions are sometimes required for outflow and in-flow boundaries where there are strong pressure gradients (Giles [1990]).
¾ Check whether the CFD code allows inflow at open boundary conditions. If inflow cannot be avoided at an open boundary then ensure that the transported properties of the incoming fluid in-cluding turbulence properties are properly modelled.
¾ Examine the possibilities of moving the domain boundaries to a position where the boundary con-ditions are more readily identified, are well-posed and can be precisely specified.
¾ Check whether upstream or downstream obstacles (such as bends, contractions, diffusers, blade rows etc.) outside of the flow domain may be present which might have significant effects on the flow distribution. Often information on the components upstream of the inlet boundary or down-stream of the outlet boundary of the domain are lacking or not available at the beginning of a pro-ject.
¾ For each class of problem that is of interest, carry out a sensitivity analysis in which the boundary conditions are systematically changed within certain limits to see the variation in results. Should any of these variations prove to have a sensitive effect on the simulated results and lead to large
changes in the simulation, then it is clearly necessary to obtain more accurate data on the bound-ary conditions that are specified.
¾ Be aware that the code may have default boundary conditions for regions of the domain bounda-ries where the user has specified nothing.
11.9.2. Guidelines on inlet conditions
¾ Carry out a sensitivity analysis in which the key inlet boundary conditions are systematically changed within certain limits. Depending on the problem, the key parameters that might be exam-ined are:
• Inlet flow direction and magnitude.
• Uniform distribution of a parameter or a profile specification, for example a uniform inlet veloc-ity (slug flow) or an inlet velocveloc-ity profile.
• Physical parameters.
• Turbulence properties at inlet.
¾ Particular care is needed with the specification of inlet mass flow as a boundary condition for compressible flow-fields with supersonic regions, as the flow may be choked and unable to pass the mass flow specified.
11.9.3. Guidelines on specification of turbulence quantities at an inlet
¾ Use verified quantities from experiments as inlet boundary conditions for turbulent kinetic energy k and dissipation ε, if these are available, as the magnitude can significantly influence the results. If there are no experimental data available, then the values need to be specified using sensible en-gineering assumptions (see below), and the influence of the choice should be examined by sensi-tivity tests with different simulations.
¾ Specify values of the turbulent kinetic energy k which are appropriate to the application. These values are often specified through a turbulence intensity level Tu, which is defined by the ratio of the fluctuating component of the velocity to the mean velocity. In external aerodynamic flows over airfoils the turbulence level is typically Tu = 0.003 (0.3%). In atmospheric boundary layer flows the level can be two orders of magnitude higher (Tu = 0.30 (30%)) and details of the actual boundary layer profiles are needed. In internal flows the turbulence level of Tu = 0.05 to 0.10 (5 to 10%) is usually appropriate.
¾ Specify values of the turbulent length scale, as an equivalent parameter for the dissipation ε, that are appropriate to the application. For external flows remote from boundary layers a value deter-mined from the assumption that the ratio of turbulent and molecular viscosity µT/µ is between 1 and 10 is a reasonable guess. For internal flows a constant value of length scale derived from a characteristic geometrical feature can be used (e.g. 1 to 10% of the hydraulic diameter for internal flows). For simulations in which the near-wall region is modelled with a low Reynolds number model the length scale should be based on the distance to the wall and be consistent with the in-ternal modelling in the code.
¾ If more sophisticated distributions of k and ε are used, check that these are consistent with the ve-locity profile. An inconsistent formulation may lead to an immediate unrealistic reduction of the tur-bulence quantities after the inlet.
¾ For RSTM models the stresses themselves need to be specified. If these are not available, as is often the case, make an assumption of isotropic flow conditions (that is with the normal stresses given by 2/3 k and with zero shear stresses).
¾ For inlets that are representative of fully developed pipe or channel flows, algebraic profiles might be used.
¾ Check the consistency of the definitions of k and ε by making a plot of the ratio of turbulent to mo-lecular viscosity µT /µ. Note that the ratio of turbulent to laminar viscosity depends on Tu, L, k and ε as µT/µ ~ k2·ε or µT/µ ~ Tu·L.
¾ In cases where problems arise, move the inflow boundary sufficiently far from the region of inter-est so that a natural inlet boundary layer can develop.
11.9.4. Guidelines on outlet conditions
¾ Place open boundary conditions (outflow, pressure) as far away from the region of interest as possible and avoid open boundaries in regions of strong geometrical changes or in regions of re-circulation. If the latter point cannot be achieved, consult the code manual to see what boundary condition is considered best.
¾ Take special care with the orientation of outlet planes with regard to the mean flow, especially when the boundary condition consists of a constant static pressure profile
¾ Select the boundary conditions imposed at the outlet to have only a weak influence on the up-stream flow. Extreme care is needed when specifying flow velocities and directions on the outlet plane. The most suitable outflow conditions are weak formulations involving specification of static pressure at the outlet plane.
¾ Particular care should be taken in strongly swirling flows where the pressure distribution on the outlet boundary is strongly influenced by the swirl. It is therefore not acceptable to specify constant static pressure across the outlet.
¾ Be aware of the possibility of inlet flow inadvertently occurring at the outflow boundary, which may lead to difficulties in obtaining a stable solution or even to an incorrect solution. If it is not possible to avoid this by relocating the position of the outlet boundary in the domain, then one possibility to avoid this problem is to restrict the flow area at the outlet, provided that the outflow boundary is not near the region of interest. If the outflow boundary condition allows flow to re-enter the domain, appropriate Dirichlet conditions should be imposed for all transported variables.
¾ If there are multiple outlets, impose either pressure boundary conditions or mass flow specifica-tions depending on the known quantities.
11.9.5. Guidelines on solid walls
¾ Care should be taken that the boundary conditions imposed on solid walls are consistent with both the physical and numerical models used (for example, adiabatic walls, local heat fluxes, free-slip boundaries, roughness, moving or rotating walls). In cases with moving or rotating walls the boundary conditions need to be specified consistent with the motion of the walls.
¾ Great care is needed if roughness on the wall is not negligible. If no detailed information is avail-able, significant levels of uncertainty can arise through incorrect specification of equivalent rough-ness on solid walls.
11.9.6. Guidelines on symmetry and periodicity planes
¾ If symmetry or periodicity planes cross the inlet or outlet boundaries then care should be taken to specify inlet or outlet variables that are consistent with these. Symmetry and periodicity planes as-sume that the gradients perpendicular to the plane are either zero (for symmetry) or determined from the flow field (periodicity).
11.9.7. Guidelines on initial guess or initial condition
¾ Be aware that in special cases the initial guess can influence the converged flow field e.g. buoyant flows, flows with bifurcations or flows with a hysteresis depending on the direction of the changes.
¾ Care should be taken to impose realistic conditions that consistent with the flow equations as the initial condition of an unsteady-state flow calculation (for example, use a steady-state flow simula-tion as the initial condisimula-tion for a gas release simulasimula-tion).
¾ Special attention should be directed to the initial guess of k and ε, especially with low Reynolds number models.
11.9.8. Guidelines on uncertainties with steady flow, symmetry and periodicity
¾ Check carefully whether the geometry is symmetric or whether a geometrical distortion or distur-bance in the inlet conditions is present which can trigger asymmetric solutions.
¾ Estimate the Reynolds-number of the inflow and check whether the flow could be asymmetric, tur-bulent and/or unsteady (e.g. by sources or literature). Coupling between competitive physical phe-nomena might also be considered as a potential source of unsteadiness.
¾ After obtaining a steady solution, switch to the transient mode and check whether the solution re-mains stable.
¾ Switch to the transient mode if it is difficult to get a converged steady solution, especially if there is an oscillation of the residuals.
¾ In case of doubt, the simulation should be unsteady and without symmetry assumptions as boundary conditions.
11.9.9. Guidelines on physical properties
¾ Ensure that the correct physical properties are specified (e.g. water, air, ideal gas). If constant property assumptions are being used make sure that they are valid.
¾ Ensure that the specified value is the correct one (e.g. dynamic and not kinematic viscosity).
¾ Ensure that the values are specified in the correct units.
¾ Check whether user coding is being used to specify non-constant properties via tabulated values, as this can be an additional source of uncertainty (e.g. interpolation, extrapolation, user coding er-rors, inconsistency with base code).