Radial Compressors

Recommended Process

Qualify Using Steady State

Before running an unsteady analysis, it is vital to first qualify the model by ensuring that the fan performance curve is predicted well from steady state simulations. Use the Realizable K-Epsilon or the K-Omega SST turbulence models for consistency with the DES model in the unsteady analysis.

To avoid pressure reflections, set a free-stream boundary at either the inlet or the outlet, depending on which side is most important. Having a free-stream boundary on both inlet and outlet is unstable in a steady analysis. If it is necessary to control the mass flow, then set the mass flow on the downstream boundary, since acoustics on the inlet side are often more important.

Use Compressible Flow

It is essential to use compressible flow modeling because of the operating conditions of these machines. The peak Mach number is generally in the supersonic range, and a 'choke' regime is present due to a passage shock. Choke occurs when a shock wave spans the full space between the blade pressure and suction sides.

Recommendations for Obtaining Performance Curves in Steady State

1. Initialize the flow field using GSI (Grid-Sequencing Initialization).
2. For high speed radial compressors, validations show that the mixing plane interface is superior to the frozen rotor.
3. The segregated solver is preferred to the coupled solver due to the follow-on step to transient where it is definitely faster. Prediction accuracy between the two is marginal.
4. Marginal differences exist between high and low Reynolds number resolution on the blade. A high Reynolds number wall treatment is preferred as the mesh can be coarser and the grid sequencing initialization is more effective. Practice shows that low Reynolds number meshes produce more accurate predictions. You can refine the mesh and interpolate between high and low Reynolds number predictions.
5. The preferred boundary condition combination is to use stagnation inlet and static pressure outlet. Start with a low pressure ratio between the inlet and outlet. To get points at higher pressure ratios, increase the back pressure. As the pressure ratio reaches the maximum for the device, activate the target mass flow setting in the pressure boundary properties. This setting allows you to fine-control the mass flow while Simcenter STAR-CCM+ still applies an outlet pressure through successive corrections. (Alternative boundary combinations are also possible - you can choose either an inlet mass flow boundary or an outlet mass flow boundary in steady state. An outlet mass flow boundary is preferred when moving to an unsteady analysis.)
6. The SST K-Omega Turbulence model seems to be marginally more accurate than Spalart Allmaras turbulence model as conditions approach choked flow. Otherwise both perform with a similar level of accuracy.

A performance curve of 10 points can be generated in approximately a day when running a 4–6 million-cell mesh on 48 processors.

Eliminating Reflective Boundary Effects

1. Extract reports of the mass-flow averaged Mach number, pressure, and temperature from the steady-state run at inlet and outlet boundaries.
2. Change the steady state case inlet boundary type to Freestream and set the Mach number, pressure, and temperature from the steady state reports.
3. Run the simulation to convergence with these new boundary settings.
4. Ensure that the pressure ratio and mass flow remain unchanged. Alternatively, wait until the mass flow balances across the inlet and outlet.

Recommendations for the Unsteady Simulation

1. Start the unsteady simulation from a converged steady solution, as described in the previous section.
2. Convert to transient.
3. Change the turbulence model to Detached Eddy Simulation (DES) using the SST k-omega model.
4. Apply the recommended Segregated Flow settings.
5. Use the Segregated Flow Model setting Hybrid-Bounded Central Differencing (BCD) or full BCD, if cell quality allows with blend=0.3. Use second-order time.
6. Freeze the Wall Distance Solver.
7. Run the case.
8. Run for between 100 and 200 full rotations to get a good statistical sample for digital signal processing.

Using these settings approximately 15 full rotations a day are possible when running a 4–6 million-cell mesh on 48 processors.

Choosing Transient or Harmonic Balance

Radial compressors are usually close to a downstream scroll volute, diffuser vanes, or other geometries which are not rotating. The correct way to simulate this system is to account for the periodic interaction of the rotating and stationary parts, namely downstream wake, and upstream pressure gusts of turning vanes, blades, and supports.

There are two ways to do model radial compressors in Simcenter STAR-CCM+: using transient Rigid Body Motion (sliding mesh) or the Harmonic Balance method. Both are more time-consuming, but more accurate, than steady state approximations that use frozen-rotor or mixing-plane stator-rotor interfaces. The harmonic balance method runs faster than transient. However, the harmonic balance method only resolves tones that are based on the modes solved (blade passing frequency, harmonics, and Strouhal shedding). It does not resolve broadband noise which is associated with resolved turbulence.

Recommended Time-Step

Recommended Solver Settings