Parallel processing of three-dimensional field-scale ...

Object Details

View

XVI International Conference on Computational Methods in Water Resources (CMWR-XVI) Ingeniørhuset

Parallel processing of three-dimensional field-scale reactive transport applications.
Author:Adil Sbai <a.sbai@brgm.fr> (Water Department, BRGM (French Geological Survey))
Presenter:Adil Sbai <a.sbai@brgm.fr> (Water Department, BRGM (French Geological Survey))
Date: 2006-06-18     Track: Special Sessions     Session: Multi-Disciplinary Approaches To Reactive Transport Simulation In Aquifer Systems
DOI:10.4122/1.1000000285

In order to achieve field-scale reactive transport simulations in a reasonable CPU time, a parallel computational model is developed for RTAFF a fully three dimensional reactive transport simulator. The MPICH2 portable implementation of the message passing interface standard is used for communication among nodes in a Beowulf cluster of PC's. Parallel computations are carried out in different subdomains mapped to each processor in the Local Area Network during the advection and dispersion steps of the time-split procedure. The domain decomposition method is implemented using row-wise and column-wise overlapping subdomains simultaneously for the mixed hybrid finite element kernel. Computations of the reaction step are performed using a newly developed load balancing algorithm. This algorithm maps cells into processors based on global convergence history analysis from the previous time step. The efficiency of this approach is showed on a demonstrative field scale example showing that the load balancing algorithm efficiently localizes and redistributes reaction fronts cells among the processors. Furthermore, this application shows the scalability of the overall parallel algorithm and the achievement of several-orders-of-magnitude speedup among a range of processors. Modeling results provide new insights into detailed subsurface reactive-transport processes which could be taken simultaneously by scientists and engineers in either research or field-scale applications.