Title: Computational Science Research and Education: Perspectives Evolving from Modeling Porous Medium Dynamics
1 Computational Science Research and Education
Perspectives Evolving from Modeling Porous Medium
Dynamics
- C.T. Miller
- Department of Environmental Sciences
- and Engineering
- University of North Carolina-Chapel Hill
2Overview
- Modeling porous medium systems
- Motivation
- Example scientific issues
- Current challenges
- Computational science research and education
- Experiences
- Desirable skill set
- Useful educational components
- Funding considerations
3Some Motivating Problems
- Water supply
- Irrigation
- Drainage
- Infiltration
- Subsurface systems coupled to surface and
atmospheric systems - Domestic and industrial waste disposal
4Some Motivating Problems
- Nuclear waste disposal
- Contaminant remediation
- Risk assessment
5Environmental Modeling Framework
6Mechanistic Modeling Framework
7Standard Formulation Approach
- Conservation of mass for species
- Conservation of mass for a phase
- Conservation of energy
- Closure relations Darcys law, equations of
state, capillary pressure-saturation-permeability
relations, interphase exchange, and reactions
8Conservation Equations
9Identities
10Closure Relations
11Closure Relations
12Darcys Law
Henry Darcy (1803-1858)
13Darcys Law
Figure 1.
Darcy's data plotted as q versus H/e (Glenn
Brown).
14Multiscale Nature of Porous Medium Systems
- Range of scales is immense, lt10-6 to gt103 m
- Morphology and topology of pore structure is
highly complex - Impossible to represent detail at finest scale
- Must therefore represent smaller scale phenomena
through appropriate closure schemes
15Natural Systems are Complex
- Heterogeneity exists across all observed scales
- Flow and transport process strongly influenced by
heterogeneity - Complete characterization is impossible
- Systems and models thus stochastic in nature
- Direct measurement of properties only one source
of data
16Borden Site, Sudicky (1986, WRR)
17Subsurface Remediation Example
18Pump and Treat Modeling
19Microscale Heterogeneity
20Macroscale Heterogeneity
- Two-dimensional cell
- Heterogeneous porous medium
- TCE dyed with oil red
- Vertical gravity displacement
- Heterogeneous media effects dominant because of
variation in capillary forces - Brine layer establishment
- Vertical mobilization under gravity forces
- Surfactant added to reduce IFT
- TCE mobilized and collected from brine layer
21Immiscibile Instabilities
- From Imhoff et al.
- NAPL dissolution fingering
- Results from high-resolution finite volume
simulator - Length scale of features can be cm to meters
- At field scale very likely to be sub-grid scale
- Standard closure schemes based on laboratory
experiments grossly in error
22Pore-Scale Modeling
- Need information at pore scale---either simulate
or measure - Solve equations with various levels of
approximation at the pore scale - Detailed information on flow and transport can be
obtained
23Pore-Scale Modeling
24Standard LB Approach
- Memory waste
- For a typical porous medium, up to 70 of the
memory is wasted - Load imbalance due to
- Heterogeneity of porous media
- Use of large number of processors for homogeneous
media
25Domain Decomposition Methods
- Regular domain decompositions 1D, 2D, 3D
- Rectilinear partitioning
- Orthogonal recursive bisection (ORB)
26Speedup as a Function of Number of Processors
27Current Challenges
- Existing multiphase porous medium modeling
approaches suffer from many serious limitations - No specific account for interfaces
- Lack of a direct relation to microscale
properties - No rigorous thermodynamic constraints applied
- Ad hoc and inconsistent closure relations for
pressure-saturation-permeability relations
28Current Challenges
- Intent of evolving approach
- First principles set of balance equations
- Explicit link between microscale and macroscale
- All quantities rigorously defined
- Thermodynamic constraints imposed
- All assumptions explicit and changeable if model
found to be inadequate
29Current Challenges
- Thermodynamically constrained model formulation,
closure, and analysis - Wettability, viscous coupling, transient psk, and
interphase exchange in multiphase systems - Multiscale closures
- Numerics, solvers, and optimization
- Spatial and spatial/temporal estimation
- Efficient modeling environments
30Computational Science Research and Education
Experiences
- Student backgrounds engineering, mathematics,
computer science, physics, chemistry, and geology - Student training porous medium physics,
stochastic hydrology, environmental processes,
numerical methods, applied mathematics, computer
science, high-performance computing - Graduates academic engineering, science, and
mathematics departments, research, consulting - Challenges rigor, cross-disciplinary
requirements, disciplinary acceptance
31Computational Science Research and Education
Desirable Skill Set
- Science and engineering (10)
- Porous medium physics
- Stochastic hydrology
- Process dynamics and fluid dynamics
- Environmental sciences
- Thermodynamics
- Mathematics and statistics (15)
- Numerical methods including solvers, ODEs, PDEs
- Advanced and evolving methods for PDEs
- Analysis through functional analysis
- Probability
32Computational Science Research and Education
Desirable Skill Set
- Mathematics and statistics (continued)
- Mathematical statistics
- Random fields
- Spatial/temporal estimation
- Stochastic differential equations
- Computer science (5)
- Data structures
- Algorithms
- Parallel and distributed computing
- Compilers
33Computational Science Research and Education
Useful Educational Components
- Interdisciplinary curriculum
- Other identifiable skills (computer languages
etc) - Strong committee presence in science, math, and
computer science - At least an appreciation for all aspects
experimental, theory, and computation - Interdisciplinary research skills---demonstrated
through publishing and presentations, and
internship
34Computational Science Research and Education
Funding Considerations
- Fellowship programs
- Need for critical mass
- Focus
- Baseline support of sufficient duration is
optimal