PARALLEL PROCESSING FOR SOLUTION-ADAPTIVE COMPUTATION OF MOVING-FRONT PROBLEMS

THE OBJECTIVE OF THE WORK IS TO DEVELOP EFFICIENT AND ACCURATE COMPUTATIONAL TECHNIQUES FOR THE SOLUTION OF MOVING FRONT PROBLEMS SUCH AS THOSE OCCURRING IN PHASE-CHANGE, POLLUTANT DISPERSAL, GROUNDWATER TRANSPORT, AND OTHERS. MIMD PARALLEL PROCESSING IS TO BE USED TO OBTAIN THE COMPUTATIONAL SPEED NECESSARY TO SOLVE THESE COMPUTER-INTENSIVE PROBLEMS. A FIXED GRID METHOD AND A NEW NUMERICAL FORMULATION USING TRIANGULAR/TETRAHEDRAL UNSTRUCTURED MESHES ARE TO BE EMPLOYED. UNSTRUCTURED MESHES ALLOW THE RESOLUTION OF THE MOVING FRONT BY MESH ADAPTION AND TO AVOID THE COMPLICATIONS OF FRONT-TRACKING TECHNIQUES. FOR PARALLEL PROCESSING, A PARALLEL PARTITIONING ALGORITHM WHICH WILL REDISTRIBUTE LOAD DYNAMICALLY AS THE MOVING FRONT PASSES THROUGH THE DOMAIN IS TO BE USED. NOT ONLY DEDICATED PARALLEL ARCHITECTURES, BUT ALSO HETEROGENEOUS WORKSTATION CLUSTERS ARE TO BE ADDRESSED.