Parallel Processing in Lagrangian Treatment of Particulate phase in a Power Utility Boiler

摘要

The complexity of the flow generated in power utility boilers, (3D, turbulent and two-phase flow), has pushed designers to use empirical information to investigate the problems associated with erosion reduction, heat transfer enhancement and more efficient boiler heat exchangers. The annual cost of erosion is as high as several million dollars; this is too expensive procedure (Tu et al., 1997). Hence, it is desirable to utilise CFD codes to predict the very detailed flow within power utility boilers under different operating conditions. The present paper describes a general three-dimensional calculation procedure based on the Lagrangian approach to predict complex fly-ash flow in a power utility boiler. The RNG k-ε turbulence model (Orszag et al., 1993) is used to characterise the time and length scales of the continuous phase turbulence. Models investigated are used to predict turbulent, fully developed gas-solid boiler configuration. The architecture considered is a HP, 7200, or 4 processors system. In single-phase modelling the grid has been partitioned into multiple sub-domains such that the number of partitions is an integral multiple of the number of compute nodes available. Each partition is resident on a different computer node. This paper outlines the effect of partitioning on parallel processing and illustrates the rate of process and real time savings in going to more processors. The present study will show that as the number of computer nodes increases, the turnaround time for the solution will decrease. However, the parallel efficiency decreases as the ratio of communication to computation increases. The paper will address the important issues of achieving high parallel efficiency for interacting physical processes in complex geometric domains.