1. INTERACTIVE CONTROL OF LARGE-SCALE SIMULATIONS Richard E. Ewing Institute for Scientific Computation Texas A&M University Robert C. Sharpley Industrial Mathematics Institute University of South Carolina

2. LEAKY UNDERGROUND STORAGE TANKS N EED TO D EVELOP M ONITORING AND C LEAN U P M ETHODS UNSATURATED ZONE SATURATED ZONE AQUIFER

3. BIOREMEDIATION STRATEGIES M ACROSCALE M ESOSCALE I NJECTION R ECOVERY F LOW M ICROSCALE GROWTH MECHANISMS Attachment Detachment Reproduction Adsorption Desorption Filtration Interaction INPUT Substrate Suspended Cells Oxygen

4. PETROLEUM APPLICATIONS G AS O IL W ATER F AULT S ALT D OME

9. MODELING PROCESS PHYSICALPROCESS PHYSICALMODEL MATHEMATICALMODEL OUTPUTVISUALIZATION DISCRETEMODEL NUMERICALMODEL

10. NEED FOR SIMULATION  D EVELOP B ETTER U NDERSTANDING OF N ONLINEAR B EHAVIOR –C OMPUTATIONAL L ABORATORY E XPERIMENTS –U NDERSTAND S ENSITIVITIES OF P ARAMETERS –I SOLATE P HENOMENA THEN C OMBINE  S CALE  U P I NFORMATION AND U NDERSTANDING –P ORE L ABORATORY F IELD  O BTAIN B OUNDING C ALCULATIONS  D EVELOP P REDICTIVE C APABILITIES –O PTIMIZATION AND C ONTROL

11. TRANSPORT IN POROUS MEDIA  P HYSICAL R ELATIONS: C ONSTITUTIVE L AW ( D ’ ARCY) C ONSERVATION OF M ASS C OMPONENT B ALANCES E QUATIONS OF S TATE T HERMAL B ALANCE

12. TRANSPORT IN POROUS MEDIA (CONTINUED)  A RE U SUALLY U NKNOWN –P ARAMETER E STIMATION / I NVERSE P ROBLEMS  P ARAMETERS A RE H ETEROGENEOUS AT V ARIOUS L ENGTHS –N EED E FFECTIVE P ARAMETERS / E QUATIONS  N EED A CCURATE F LUID V ELOCITIES –M IXED F INITE E LEMENT M ETHODS  I MPORTANCE OF T RANSPORT –E ULERIAN  L AGRANGIAN M ETHODS  C OUPLED N ONLINEAR S YSTEMS –E FFECTIVE L INEARIZATION / S OLUTION M ETHODS

13. TWO-PHASE FLOW D ’ARCY’S L AW: M ASS B ALANCE: w = Wetting Phase n = Nonwetting Phase

14. P HYSICAL P ROCESS IDENTIFICATION (INVERSE) PROBLEM  D ETERMINE S UITABLE M ATHEMATICAL M ODEL  E STIMATE P ARAMETERS W ITHIN M ATHEMATICAL M ODEL I NPUTS O UTPUTS I NPUTS M EASUREMENTS M ATHEMATICAL M ODEL

15. P HYSICAL P ROCESS IDENTIFICATION (INVERSE) PROBLEM  D ETERMINE S UITABLE M ATHEMATICAL M ODEL  E STIMATE P ARAMETERS W ITHIN M ATHEMATICAL M ODEL I NPUTS O UTPUTS I NPUTS M EASUREMENTS M ATHEMATICAL M ODEL

16. LARGE SCALE INTERACTIVE APPLICATIONS ON REMOTE SUPERCOMPUTERS  Model Development and Formulation  Coupled Flow and Transport Codes with Complex Boundary Conditions  Numerical Discretization and Parallel Algorithm Development  MPP Code Development  Field Testing and Production Runs  User Environments and Visualization Tools Need for Interactive tracking and steering

17. PRIMARY RESEARCH TEAM  Texas A&M –Richard Ewing –Steve Johnson –Hao Lu –Joseph Pasciak  South Carolina –Robert Sharpley –Ronald DeVore –L. Scott Johnson

18. SAVANNAH RIVER SITE  Difficult topography  Highly Heterogeneous Soils  Saturated and Unsaturated Flows  Reactions with disparate time scale  Transient/Mixed Boundary Conditions

28.  Software Engineering –Motif with OpenGL / MESA rendering areas –Self-Interrupting Work Stack  3D Pre-processing –Data set construction –GIS Import of Field Data –Data Verification  Post-processing –Data Verification –Model and Algorithm Verification  Controller library coupling G3D - GRAPHICS FRONT-END

29. GRAPHICS PRE-PROCESSING  3D grid creation and editing  Material properties  Initial conditions  Time dependent boundary conditions  Multiple views

30. GRAPHICS POST-PROCESSING  Multiple vector/scalar fields  Time animation  Multiple slices/Iso-surfaces  Stereo rendering, lighting models  Overlay images for orientation  Volume rendering Hierarchical Representations

31. CONTROLLER LIBRARY  User callable  Client - Server – server MPP - client workstations  Non-interrupt approach  Controller - “Graphics” Node – processes client requests and performs scale analysis/compression  MPI Message passing  Remote procedure calls

33. COMPRESSION MODULE FOR ENCODED TRANSMISSION OVER CONGESTED NETWORKS  Wavelet based processing – Fast decomposition/reconstruct – Nonlinear approximation in various metrics – Compression – Feature identification at various scales – Scale up/scale down  Progressive transmission – Lossy to lossless – Adaptive local spatial resolution  Scattered data assimilation and denoising

34. INTERACTIVE CONTROL OF LARGE-SCALE SIMULATIONS Richard E. Ewing Institute for Scientific Computation Texas A&M University Robert C. Sharpley Industrial Mathematics Institute University of South Carolina