Computational Modeling of Industrial Scale Reactor System Components Using Parallel Processing on a Networked Cluster of Computers

Abstract

Industrial scale reactor system components are frequently characterized by a complex geometry that requires a large number of nodes in a computational grid to resolve significant flow features and by reaction and additional complex physics, such as spray vaporization, that greatly increase the number of partial differential equations (PDEs) to solve. On a fast serial workstation, simulations of these systems can take several days, making use of the model for extensive parametric and optimization studies impractical. A parallel multiphase reacting flow code was developed from a family of related serial codes using the industry standard Message-Passing Interface (MPI) and domain decomposition in the primary flow direction. The parallel computational fluid dynamics (CFD) code can now be run on a cluster of networked computers. This capability makes detailed simulation of many industrial reactors feasible without incurring the large increase in cost of moving to simulation on massively parallel computing platforms. The required computers and network are often already present in organizations and idle at night. The parallel CFD code has been applied to studies of transient heat transfer to monolith catalyst substrates and to the design of inlet chambers that evenly distribute a high speed hot air inlet stream over the monolith inlet plane. These studies are part of a program to investigate the feasibility of rapid start fuel reformers. The use of fuel reformers in light duty fuel cell powered vehicles requires a reformer start up time of less than 60 seconds. Parallelization issues and parallel performance are presented with examples of reformer component flow and comparison with experimental data from monolith heat up studies.