OCR Text |
Show SlR[EAW~ [FUNCT~ON Fine grtd 66x66 nodes Figure 8: Reactor geometry and flow pattern for the selected multigrid comparison. Figure 8 illustrates an example geometry that was used for multigrid evaluation. The chamber diameter and length are 0.2 m and 0.8 m, respectively. One inlet stream was included, with a diamet.er of 0.1 m and maximum velocities greater than 30.0 m/s. It is thought that this example, due to its' inlet conditions and large recirculation zone, would provide for a satisfactory evaluation of multigrid methods in combustion applica t ions. Figure 9 compares the computational requirements for for two predictions (axial and radial velocity components and pressure) of the,above conditions. In both predictions~ turbulence was modeled using a globally constant eddy diffusiyity (0 .02 cp) . One prediction was n1ade using distributive relaxation and the multigrid algorithm. The other prediction was obt ained utilizing the SI~1PLER pressure correction algorithm on a single grid, which is identical to the finest grid of the previous prediction (66x66 nodes). For the SI~1PLER calculation, under-relaxation factors of 0.7 were used for the two momentum equations. A work unit is defined to be the an10unt of computational time required to make one complete calculation (both momentUJ11 con1- ponents and continuity) on the finest grid using distributive relaxation. It is clear that a converged solution is obtained using the multigrid code at lea st two orders of n1agnitude faster than for the non-multigrid case. \\lith multigrid processes, con\ ergence is fast and constant. It is thought that this demonstrates conclusively the ad\-antages of muhigrid processes for flow conditions of interest to confined combustion. Current efforts are being focused on 3-dimensional combustion applications . It is thought that this improved performance can be maintained for these n10re con1plex flows and geolnetries. Indeed , there is the potential for even greater impro\·ements in a full -con1bustion code. One of the primary difficulties in PCGC-2 is in resoh ing 16 |