fluent.com home page

   
 

Meshing Strategies & Solvers Compared

 

Motor Industry Research Association (MIRA)

Engineers at MIRA used the flow field in the vicinity of an inlet valve port to make a comparison of meshing strategies and commercial CFD solvers. Simulations were carried out for three valve lift positions. A final analysis included considerations of meshing time and solver time. Solver accuracy was not evaluated, since no field data were available.

Starting with a CAD geometry, three mesh types were built: a pure tetrahedral mesh of 162,000 cells, a pure hexahedral mesh of 100,000 cells, and a hybrid mesh of 182,000 cells with prisms in the near-wall regions and tetrahedral elements elsewhere. The tet mesh was the quickest to build and the hex mesh the slowest. The hybrid mesh took less than twice the time as the tet mesh, but a third of the time of the hex mesh.

FLUENT 5 was compared to two other commercial solvers: an alternative RANS solver (referred to as the RANS solver or code) and an extended lattice gas dynamics code (the LGDS code). In all simulations, pressures were set at the inflow and outflow boundaries, and the mass flow rate through the engine port was computed. Results were compared among the solver/mesh configurations for each of three valve lift positions.

Contours of velocity magnitude on a slice through the valve port and particle traces are used to illustrate the velocity field.

Predictions of mass flow rate all fell within a +10% span of the average value for each position, with the RANS code con-sistently returning the maximum value and the LGDS code returning the minimum value.

Solver times were compared between the RANS code and FLUENT for each of the three grid types. When using the pure tetrahedral mesh, the RANS solver was a factor of 6 times slower than FLUENT. When using the pure hexahedral mesh, the RANS solver was 80% slower than FLUENT. When using the hybrid mesh, the RANS solver was 3 times slower than FLUENT. When comparing the FLUENT solver working on the three different grid topologies, the pure tet mesh was the quickest to converge and the hex mesh the slowest. Based on these simulations, engineers at MIRA concluded that the hybrid mesh and the FLUENT 5 solver offer the best combination for rapid meshing, accurate boundary treatment, and increased solver speed.


Previous Article FluentNEWS Next Article