Computational Fluid Dynamics (CFD): Questions Answered by Skill-Lync Experts

Steady state. We want the infinite-time solution (thermal equilibrium). There is no time dependency. Only the boundary conditions are known and all inner node values are unknown. The system is implicit, so we cannot solve each node individually and instead iterate in a while loop. We start from initial guess values and reach converged values after several iterations. T_old holds the T values from the previous convergence iteration and must be updated on every iteration of the convergence loop.

Transient, explicit. We start from the initial condition and compute each new time step from the previous one. T_old holds the previous time step values and is updated on every iteration of the time loop.

Transient, implicit. We time-march, but at each time step we solve a steady-state-like problem, because unknown terms on the right-hand side make the equations implicit. For every time step we start with guess values, iterate to convergence, and then move to the next time step. So two stored variables are needed:

• T_previous: values from the previous time step. Update it on every iteration of the time loop.
• T_old: values from the previous convergence-loop iteration. Update it on every iteration of the convergence loop.

icoFoam is a transient solver for incompressible, laminar flow of Newtonian fluids. You choose a solver based on the physics of the problem and what you want to study. The OpenFOAM user guide describes all the standard solvers and their capabilities.

Why velocity at the inlet and pressure at the outlet

• If you solve only the continuity and momentum equations, each boundary needs two boundary conditions, one for each unknown (pressure and velocity).
• For subsonic flow there is a two-way exchange of information between boundaries and inner nodes.
• Information travelling from the boundary to the inner nodes needs a fixed value (Dirichlet) at the boundary. Information travelling from the inner nodes to the boundary needs a floating condition (Neumann).
• Fixing inlet velocity and outlet pressure (so inlet pressure and outlet velocity float) is the most convenient and commonly used combination, especially for simple flows.
• Walls follow the same logic: fix velocity (no slip) and let pressure float.

Why a small wedge angle is better

• A wedge is a 2D axisymmetric boundary condition. For axisymmetric flow it reduces the domain from a whole pipe to a single 2D plane along the flow axis.
• The closer your domain is to that plane, the more accurate the representation. So a smaller wedge angle gives better results.
• A larger wedge angle also reduces the vertical height of the single cell used in the neglected third direction, which is not desirable.
• The 5 degree threshold is not absolute. Accuracy gradually decreases as the angle increases.

Wedge vs symmetry Wherever a wedge BC can be used, a symmetry BC can also be used, but not the other way round. Flow through a circular pipe is axisymmetric. A square duct has no axisymmetric plane but has two symmetry planes, so there you must use symmetry and not wedge.

A facet is a face of a cell. A facet average of a quantity, say temperature on a pipe outlet patch, adds the values at all cell faces on the patch and divides by the number of faces:

Sum of temperature at all faces / number of faces

It ignores how much of the patch each face covers.

An area-weighted average weights each face value by its area:

Sum of (temperature x face area) / sum of all face areas

It is generally preferable to use the area-weighted average for surface averages, because it accounts for face size and makes more physical sense. The definitions of these terms are in the ANSYS Fluent documentation.

For non-linear problems, an adjustable time step scheme lets the solver pick a time step that keeps the solution stable.

How to start

• Start with a low time step, for example 1e-7 or 1e-8.
• The time step typically grows each time the maximum CFL criterion is satisfied. A multiplier of 1.10 is typical, growing it by 10%.
• When the maximum CFL criterion is hit, the time step is cut by a factor of 1.50.

For linear problems, a stability criterion can be derived analytically, so you can calculate the maximum time step quite accurately. Dynamic time stepping can be used for both linear and non-linear problems.

There is no fixed procedure for choosing the multiplier, because it depends on the problem. The strategy above makes conservative time-step jumps and aggressive cuts, and works for most problems as long as you start small enough.

1. After the design phase, go to the Flow Simulation tab and click the Wizard option to create the project.
2. The wizard asks for inputs such as initial conditions, simulation type, type of fluid flow, the time-dependent option and the rotation option.
3. After you specify these, the computational domain is created automatically.
4. Define the boundary conditions.
5. Add goals if needed.
6. Mesh the part. Meshing can be manual or automatic.
7. If you need more results, such as a cut plot or flow trajectories of velocity, select them in the results options.
8. Click Run to start the simulation.