Shuyu Sun Earth Science and Engineering program KAUST Presented at the 2009 annual UTAM meeting, 2:05-2:40pm January 7, 2010 at the Sutton Building, University.

Slides:



Advertisements
Similar presentations
School of something FACULTY OF OTHER School of Computing An Adaptive Numerical Method for Multi- Scale Problems Arising in Phase-field Modelling Peter.
Advertisements

A parallel scientific software for heterogeneous hydrogeoloy
1 Numerical Simulation for Flow in 3D Highly Heterogeneous Fractured Media H. Mustapha J. Erhel J.R. De Dreuzy H. Mustapha INRIA, SIAM Juin 2005.
1 A new iterative technique for solving nonlinear coupled equations arising from nuclear waste transport processes H. HOTEIT 1,2, Ph. ACKERER 2, R. MOSE.
Conductivity Testing of Unsaturated Soils A Presentation to the Case Western Reserve University May 6, 2004 By Andrew G. Heydinger Department of Civil.
Dongxiao Zhang Mewbourne School of Petroleum and Geological Engineering The University of Oklahoma “Probability and Materials: from Nano- to Macro-Scale”
A modified Lagrangian-volumes method to simulate nonlinearly and kinetically adsorbing solute transport in heterogeneous media J.-R. de Dreuzy, Ph. Davy,
Features of POLLUSOL Flow model Flow model Homogeneous, Isotropic, Heterogeneous and Anisotropic medium Homogeneous, Isotropic, Heterogeneous and Anisotropic.
High performance flow simulation in discrete fracture networks and heterogeneous porous media Jocelyne Erhel INRIA Rennes Jean-Raynald de Dreuzy Geosciences.
The Finite Element Method Defined
Dual Mesh Method in Upscaling Pascal Audigane and Martin Blunt Imperial College London SPE Reservoir Simulation Symposium, Houston, 3-5 February 2003.
Ground-Water Flow and Solute Transport for the PHAST Simulator Ken Kipp and David Parkhurst.
An Introduction to Multiscale Modeling Scientific Computing and Numerical Analysis Seminar CAAM 699.
Coupling Continuum Model and Smoothed Particle Hydrodynamics Methods for Reactive Transport Yilin Fang, Timothy D Scheibe and Alexandre M Tartakovsky Pacific.
Some problems of computational geophysics Yu.M. Laevsky, B.G. Mikhaylenko, G.V. Reshetova Institute of Computational Mathematics and Mathematical Geophysics.
Introduction to numerical simulation of fluid flows
A Bezier Based Approach to Unstructured Moving Meshes ALADDIN and Sangria Gary Miller David Cardoze Todd Phillips Noel Walkington Mark Olah Miklos Bergou.
IMA Workshop on Compatible Discretizations
Peyman Mostaghimi, Martin Blunt, Branko Bijeljic 11 th January 2010, Pore-scale project meeting Direct Numerical Simulation of Transport Phenomena on Pore-space.
Numerical Porous Media KAUST SRI Center Modeling and simulation of multiscale problems N Ahmed, VM Calo, Y Efendiev, H Fayed, O Iliev, Z.Lakdawala, K.Leonard,
1 Internal Seminar, November 14 th Effects of non conformal mesh on LES S. Rolfo The University of Manchester, M60 1QD, UK School of Mechanical,
REVIEW. What processes are represented in the governing equation that we use to represent solute transport through porous media? Advection, dispersion,
Parallel Mesh Refinement with Optimal Load Balancing Jean-Francois Remacle, Joseph E. Flaherty and Mark. S. Shephard Scientific Computation Research Center.
Preliminary Assessment of Porous Gas-Cooled and Thin- Liquid-Protected Divertors S. I. Abdel-Khalik, S. Shin, and M. Yoda ARIES Meeting, UCSD (March 2004)
Wolfgang Kinzelbach with Marc Wolf and Cornel Beffa
Multi-Scale Finite-Volume (MSFV) method for elliptic problems Subsurface flow simulation Mark van Kraaij, CASA Seminar Wednesday 13 April 2005.
Chamber Dynamic Response Modeling Zoran Dragojlovic.
Solutions to the Advection-Dispersion Equation
LES of Turbulent Flows: Lecture 3 (ME EN )
MCE 561 Computational Methods in Solid Mechanics
Frontiers and Future of Multiphase Fluid
Upscaling, Homogenization and HMM
Direct and iterative sparse linear solvers applied to groundwater flow simulations Matrix Analysis and Applications October 2007.
Numerical methods for PDEs PDEs are mathematical models for –Physical Phenomena Heat transfer Wave motion.
© 2011 Autodesk Freely licensed for use by educational institutions. Reuse and changes require a note indicating that content has been modified from the.
BIOPLUME II Introduction to Solution Methods and Model Mechanics.
Solute (and Suspension) Transport in Porous Media
A Hybrid Particle-Mesh Method for Viscous, Incompressible, Multiphase Flows Jie LIU, Seiichi KOSHIZUKA Yoshiaki OKA The University of Tokyo,
Modelling Flow through Fractures in Porous Media Holzbecher Ekkehard Wong LiWah Litz Marie-Sophie Georg-August-University Göttingen, Geological Sciences,
A conservative FE-discretisation of the Navier-Stokes equation JASS 2005, St. Petersburg Thomas Satzger.
The Geometry of Biomolecular Solvation 2. Electrostatics Patrice Koehl Computer Science and Genome Center
Contaminant Transport CIVE 7332 Lecture 3. Transport Processes Advection The process by which solutes are transported by the bulk of motion of the flowing.
Geometry Group Summer 08 Series Toon Lenaerts, Bart Adams, and Philip Dutre Presented by Michael Su May
Advection-Dispersion Equation (ADE)
© Fluent Inc. 11/24/2015J1 Fluids Review TRN Overview of CFD Solution Methodologies.
© IFP Controlled CO 2 | Diversified fuels | Fuel-efficient vehicles | Clean refining | Extended reserves Écrire ici dans le masque le nom de votre Direction.
HEAT TRANSFER FINITE ELEMENT FORMULATION
Engineering Analysis – Computational Fluid Dynamics –
A Dirichlet-to-Neumann (DtN)Multigrid Algorithm for Locally Conservative Methods Sandia National Laboratories is a multi program laboratory managed and.
Controlled CO 2 | Diversified Fuels | Fuel-efficient Vehicles | Clean Refining | Extended Reserves © IFP IEA Collaborative Project on EOR - 30th Annual.
Discretization Methods Chapter 2. Training Manual May 15, 2001 Inventory # Discretization Methods Topics Equations and The Goal Brief overview.
C enter for S ubsurface M odeling Coupling of MFE or Mimetic Finite Differences with Discontinuous Galerkin for Poroelasticity Mary F. Wheeler Ruijie Liu.
Distributed Resistances and Fan Models Chapter 4.
CO 2 maîtrisé | Carburants diversifiés | Véhicules économes | Raffinage propre | Réserves prolongées © IFP Écrire ici dans le masque le nom de votre Direction.
1 Simulation of the Couplex 1 test case and preliminary results of Couplex 2 H. HOTEIT 1,2, Ph. ACKERER 1, R. MOSE 1 1 IMFS STRASBOURG 2 IRISA RENNES 1.
Environmental Engineering Lecture Note Week 10 (Transport Processes) Joonhong Park Yonsei CEE Department CEE3330 Y2013 WEEK3.
Motivation – Why deal.II?  Adaptive Mesh Refinement (AMR)  Start with solving on coarse grid  Compute error  Refine mesh until error < tolerance 
Model Anything. Quantity Conserved c  advect  diffuse S ConservationConstitutiveGoverning Mass, M  q -- M Momentum fluid, Mv -- F Momentum fluid.
Computation Geometry Governing Equations Boundary Conditions Initial Conditions Parameters, dimensionless groups Constraints, other requirements Discretization/mesh.
Gauge/gravity duality in Einstein-dilaton theory Chanyong Park Workshop on String theory and cosmology (Pusan, ) Ref. S. Kulkarni,
Algorithm of the explicit type for porous medium flow simulation
Hasan Nourdeen Martin Blunt 10 Jan 2017
Free vs. Forced Convection
Chamber Dynamic Response Modeling
Dual Mesh Method in Dynamic Upscaling
Convergence in Computational Science
ArcEOR A posteriori error estimate tools to enhance the performance of
Investigators Tony Johnson, T. V. Hromadka II and Steve Horton
Comparison of CFEM and DG methods
Ph.D. Thesis Numerical Solution of PDEs and Their Object-oriented Parallel Implementations Xing Cai October 26, 1998.
Presentation transcript:

Shuyu Sun Earth Science and Engineering program KAUST Presented at the 2009 annual UTAM meeting, 2:05-2:40pm January 7, 2010 at the Sutton Building, University of Utah, Salt Lake City, Utah

Energy and Environment Problems

Single Phase Flow in Porous Media Continuity equation – from mass conservation Thermodynamic model For impressible fluid (constant density): Still need one more equation

Relate velocity with pressure: Darcy's law Experiment by Henry Darcy (1855–1856)

Darcy's law Can be derived from the Navier-Stokes equations via homogenization. It is analogous to –Fourier's law in the field of heat conduction, –Ohm's law in the field of electrical networks, –Fick's law in diffusion theory. In 3D:

Incompressible Single Phase Flow Continuity equation Darcy’s law Boundary conditions:

Transport in Porous Media Transport equation Boundary conditions Initial condition Dispersion/diffusion tensor

Numerical Methods for Flow & Transport Challenge #1: Require the numerical method to be: –Locally conservative for the volume/mass of fluid (flow equation) –Locally conservative for the mass of species (transport equation) –Provides fluxes that is continuous in the normal direction across the entire domain. Methods that are not locally conservative without post-processing –Point-Centered Finite Difference Methods –Continuous Galerkin Finite Element Methods –Collocation methods –……

Example: importance of local conservation

Numerical Methods for Flow & Transport Challenge #2: Fractured Porous Media –Different spatial scale: fracture much smaller –Different temporal scale: flow in fracture much faster Solutions: –Mesh adaptation for spatial scale difference –Time step adaptation for temporal scale difference

Example: flow/transport in fractured media

Locally refined mesh: FEM and FVM are better than FD for adaptive meshes and complex geometry

Example: flow/transport in fractured media CFL condition requires much smaller time step in fractures than in matrix: adaptive time stepping.

Numerical Methods for Flow & Transport Challenge #3: Sharp fronts or shocks –Require a numerical method with little numerical diffusion –Especially important for nonlinearly coupled system, with sharp gradients or shocks easily being formed Solutions: –Characteristic finite element methods –Discontinuous Galerkin methods

Example: Comparison of DG and FVM Advection of an injected species from the left boundary under constant Darcy velocity. Plots show concentration profile at 0.5 PVI. Upwind-FVM on 40 elementsLinear DG on 40 elements

Example: Comparison of DG and FVM Flow in a medium with high permeability region (red) and low permeability region (blue) with flow rate specified on left boundary. Contaminated fluid flood into clean media.

Example: Comparison of DG and FVM Advection of an injected species from the left. Plots show concentration profiles at 3 years (0.6 PVI). FVMLinear DG

Numerical Method for Flow & Transport Challenge #4: Time dependent local phenomena –For example: moving contaminant plume Solutions: –Dynamic mesh adaptation Based on conforming mesh adaptation Based on non-conforming mesh adaptation

Adaptive DG methods – an example Sorption occurs only in the lower half sub- domain, SIPG is used.

Adaptive DG example (cont.) Anisotropic mesh adaptation

Adaptive DG example (cont.) Estimators using hierarchic bases

Adaptive DG example (cont.) L2(L2) Error Estimators

A Posteriori Error Estimators Residual based –L2(L2) –L2(H1) Implicit –Solve a dual problem, can give estimates on a target functional –Disadvantages: computational costly and not flexible –Advantages: More accurate estimates Hierarchical bases – Brute-force: difference between solutions of two discretizations (most expensive) – Local problems-based –Advantage: can guide anisotropic hp-adaptivity Superconvergence points-based –Difficult for unstructured and non-conforming meshes

A posteriori error estimates Residuals –Interior residuals –(Element-)boundary residuals

A posteriori error estimate in L2(L2) for SIPG Proof Sketch: Compare with L2 projection; Cauchy- Schwarz; Properties of cut-off operator; Approximation results; Inverse and Gronwell’s inequalities; Relation of residue and error

Dynamic mesh adaptation with DG Nonconforming meshes –Effective implementation of mesh adaptation, –Elements will not degenerate unless using anisotropic refinement on purpose. Dynamic mesh adaptation –Time slices = a number of time steps; only change mesh for time slices. –Refinement + coarsening  number of elements remain constant.

Concentration projections during dynamic mesh modification Standard L2 projection used –Computation involved only in elements being coarsened L2 projection is a local computation for discontinuous spaces –This results in computational efficiency for DG –L2 projection is a global computation for CG L2 projection is locally mass conservative –This maintains solution accuracy for DG –Interpolation or interpolation-based projection used in CG is NOT locally conservative

Adaptive DG example (quads) L2(L2) Error Estimators for SIPG

Adaptive DG example (quads) L2(L2) Error Estimators for SIPG

Adaptive DG example (quads) L2(L2) Error Estimators for SIPG

Adaptive DG (with triangles) L2(L2) Error Estimators on Triangles Initial mesh

Adaptive DG (with triangles) L2(L2) Error Estimators on Triangles T=0.5 T=1.0

Adaptive DG (with triangles) L2(L2) Error Estimators on Triangles T=1.5 T=2.0

Adaptive DG example in 3D L2(L2) Error Estimators on 3D T=1.5 T=2.0 T=0.1T=0.5 T=1.0

ANDRA-Couplex1 case Background –ANDRA: the French National Radioactive Waste Management Agency –Couplex1 Test Case Nuclear waste management: Simplified 2D Far Field model Flow, Advection, Diffusion-dispersion, Adsorption Challenges –Parameters are highly varying permeability; retardation factor; effective porosity; effective diffusivity –Very concentrated nature of source concentrated in space concentrated in time –Long time simulation 10 million years –Multiple space scales Around source / Far from source –Multiple time scales Short time behavior (Diffusion dominated) Long time behavior (Advection dominated)

ANDRA-Couplex1 case (cont.) 200k years 2m years

Compositional Three-Phase Flow Mass Conservation (without molecular diffusion) Darcy’s Law

Numerical Modeling for Flow & Transport Challenge #5: Importance of capillarity –Capillary pressure usually ignored in compositional flow modeling –Even the immiscible two-phase flow or the black oil model usually assumes only a single capillary function (i.e. assuming a single uniform rock)

Two-dimensional 400x200m^2 domain Contains a less-permeable (K=1md) rock in the center of the domain while the rest has K=100md. Isotropic permeability tensor used. Porosity = 0.2 Densities: 1000 kg/m^3 (W) and 660 kg/m^3 (O) Viscosities: 1 cp (W) and 0.45 cp (O) Inject on the left edge, and produce on the right edge Injection rate: 0.1 PV/year Initial water saturation: 0.0; Injected saturation: 1.0 Example: Reservoir Description

Relative permeabilities (assuming zero residual saturations): Capillary pressure Reservoir Description (cont.) K=100md K=1md

Discretization DG-MFE-Iterative Pressure time step: 10years / 1000 timeSteps Saturation time step = 1/100 pressure time step Mesh: 32x64 uniform rectangular grid:

Comparison: if ignore capillary pressure … Saturation at 10 years: Iter-DG-MFE With nonzero capPres With zero capPres

Numerical Modeling for Flow & Transport Challenge #6: Discontinuous saturation distribution –Saturation usually is discontinuous across different rock type, which is ignored in many works in literature –When permeability changes, the capillary function usually also changes! Solutions: –Discontinuous Galerkin methods

Saturation at 3 years Iter-DG-MFE Simulation Notice that Sw is continuous within each rock, but Sw is discontinuous across the two rocks

Saturation at 5 years Iter-DG-MFE Simulation Notice that Sw is continuous within each rock, but Sw is discontinuous across the two rocks

Saturation at 10 years Iter-DG-MFE Simulation Notice that Sw is continuous within each rock, but Sw is discontinuous across the two rocks

Water pressure at 10 years Iter-DG-MFE Simulation (pressure unit: Pa) Notice that Pw is continuous within the entire domain.

Capillary pressure at 10 years Iter-DG-MFE Simulation (pressure unit: Pa) Notice that Pc is continuous within the entire domain.

Numerical Modeling for Flow & Transport Challenge #7: Multiscale heterogeneous permeability –Fine scale permeability has pronounced influence on coarse scale flow behaviors –Direct simulation on fine scale is intractable with available computational power Solutions: –Upscaling schemes –Multiscale finite element methods

Recall: DG scheme for flow equation Bilinear form Linear functional Scheme: seek such that

DG on two meshes Fine mesh Coarse mesh

Space decomposition Introduce Solution

Closure Assumption Introduce Two-scale solution

Implementation Multiscale basis functions: For each Multiscale approximation space: Two-scale DG solution

Other Closure Options Local problems for solving multiscale basis functions need a closure assumption. In previous derivation, we strongly impose zero Dirichlet boundary condition on local problems. Other options: –Weakly impose zero Dirichlet boundary condition on local problems. –Strongly impose zero Neumann boundary condition on local problems. –Weakly impose zero Neumann boundary condition on local problems. –Combination of zero Neumann and zero Dirichlet.

Comparison with direct DG Memory requirement –Direct DG solution in fine mesh: –Multiscale DG solution Computational time –Direct DG solution in fine mesh: –Multiscale DG solution

Example Conductivity: Boundary conditions: –Left: p=1; Right: p=0; top & bottom: u=0. Discretization: –R=r=1; –Coarse mesh 16x16; Fine mesh 256x256

Example (cont.) Coarse DG solution Brute-force Fine DG solution

Example (cont.) Multiscale DG solution Brute-force Fine DG solution

Future work Multiscale DG methods for compositional multiple-phase flow in heterogeneous media, Stochastic PDE simulations, Multigrid solver for DG (including p-multigrid), Other future works: –Automatically adaptive time stepping, –Implicit a posteriori error estimators, –Fully automatically hp-adaptivity for DG, –A posteriori estimators for coupled reactive transport and flow.