A Posteriori Error Estimation for
Predictive Models
James R. Stewart [email protected]
Engineering Sciences Center
Advanced Computational Mechanics Architectures Dept (9143) Sandia National Laboratories
Error Estimation Example
•
Linear elastic bar with
g
ra
v
ity
lo
a
d
in
g
•
8-node hexahedral
elements
This is among the
simplest problems to
solve using FEA
Q. Can we estimate
the error
accurately
?
A
. Y
e
s
…
Error Estimation Example
What does it depend on?
•
The type of error
estimator
•
The mesh resolution
(possibly)
•
The engineering
“quantity of interest”
– Average vertical displacement » On a side surface» On the bottom surface
– Average normal stress on the the bottom surface
Rigorous
Definition of Error
• Exact error
is the exact solution to the math problem
is the numerical solution (finite-element, finite-volume, finite-difference, etc.)
Non-Rigorous
Definitions of Error
•
Difference between fine and coarse meshes
•
Difference between higher-order and lower-order
meshes
Definitions of Error (cont.)
Many (most) error estimators work directly on
with the assumption that
is small
or
•
The exact error can be written as
where
•
Remark
More Definitions
•
Quantity of interest
– Functional of solution
– Local engineering output
» Average vertical displacement on a surface
» Displacement at a point
Error
Estimates
vs. Error
Bounds
•
Error
estimates
– Attempt to quantify the error (notice the use of )
– Quality of estimate often given by the effectivity index
•
Error
bounds
– Upper and lower bounds
– Can easily convert bounds into estimates
Remarks
•
For Uncertainty Quantification, error
bounds
are more
useful than error
estimates
•
Error
estimates
provide stopping criteria for adaptive
mesh refinement
•
Converting an estimate into bounds is, in general, not
trivial
Error Estimation Example (cont.)
•
Quantity of interest
– Average vertical displacement
– Case 1: is a side surface
– Case 2: is the bottom surface
Lower and Upper Bounds:
Case 1 (Side Surface)
Valid upper
bound on
Valid lower
bound on
• Bounds are with respect to
• Quality of bounds depends on H
Lower and Upper Bounds:
Case 2 (Bottom Surface)
H
2
• Solution ( ) is exact for all H (note that is nodally exact)
• Error estimate (average of bounds) is exact for all H
1 0.5
0
.
1
8
06
.
2
e
8
06
.
2
e
2
.
06
e
8
8
06
.
2
e
8
06
.
2
e
8
06
.
2
e
1
.
0
8
06
.
2
e
8
06
.
2
e
8
06
.
2
e
1
.
0
Classes of Error Estimators
•
Recovery methods
•
Residual methods
Recovery Methods
•
Super-convergent Patch Recovery
(ZZ-SPR or ZZ)
– Zienkiewicz and Zhu
– Post process gradients of FE solution on patches of neighboring elements
– Gives global energy-norm estimates (under stringent assumptions)
– Does not lend itself to rigorous mathematical analysis
•
Polynomial Preserving Recovery
(PPR)
– Zhang, et al.
– Post process nodal values of FE solution on patches of neighboring elements
– Gives global energy-norm estimates
Residual Methods
•
Traditional categorizations
– Explicit residual methods» In reality, a misnomer » Provides error estimates
– Implicit residual methods
» Can provide error estimates and bounds
•
Original ideas of
Babuska and Rheinbolt
[1978-81]
with extensions by
Babuska and Miller
[1984-87]
Explicit
Residual Methods
Error Representation Formula
“Weak” residual estimate:
• Solve dual problem for
• Substitute into (3)
Remarks:
• Dual problem must be solved on either or
Explicit
Residual Methods
“Strong” residual estimate:
where and are stability factors
Remarks:
• (4) implies that the error can be large even if the residuals are small
• The stability factors are properties of the pde
Explicit
Residual Methods
Nonlinear Operators
Explicit
Residual Methods
Sample Applications
• Chalmers (Sweden) Group (Johnson, Eriksson, Estep, Hansbo, et al.)
– Advection-diffusion; General nonlinear parabolic operators
– Mostly global norms (done in early 90’s) • Estep, Larson, and Williams
– Nonlinear reaction-diffusion systems
– Coupled parabolic pde’s and (singular) ode’s in time
– = average value in domain
• Heidelberg Group (Becker and Rannacher; Bangerth) – DWR (Dual Weighted Residual) Method
– Variety of nonlinear fluid and solid mechanics problems
– “Improved” error indicators for mesh adaptivity
– (Time-domain) acoustic wave equation; elastic wave equation • Barth and Larson
– Extension to finite volume methods
Implicit
Residual Methods
• Can give error estimates and bounds
• Simplest form (for error estimates): Recall
• Substitute and rearrange; solve for
Remarks:
• Recall ; Therefore, must be solved in a higher-order subspace
• For efficiency, is solved on the broken space • Element residual method
• Subdomain residual method
Implicit
Residual Methods
The
Broken Space
• Broken space can also consist of (overlapping or non-overlapping) patches
Implicit
Residual Methods
Quantity-of-Interest
Bounds
•
Various derivations have been published (beyond the
scope of this talk…), e.g.,
– Babuska and Strouboulis
– Peraire and Patera
– Oden and Prudhomme
•
Main result (broken space dependence made explicit)
•
Subtract to get
error
bounds
• Requirement: (i.e., set of functions defined on also contains functions defined on )
Implicit
Residual Methods
General Bound Procedure
•
Following presentation of
Peraire and Patera
(the
“Truth”
Error Bounds)
•
Step 0:
Solve FE problem
– Find such that
•
Step 1:
Solve global (linear) dual problem on
Implicit
Residual Methods
General Bound Procedure
•
Step 2:
For each patch in , solve
– Local primal error problem: Find such that
– Local dual error problem: Find such that
where is the jump bilinear form defined by
Implicit
Residual Methods
General Bound Procedure
•
Step 3:
Compute the bounds
“Truth” Error Bounds
Software Design and Implementation
Quasi-statics code
Adagio
SIERRA design hierarchy
Domain
Procedure (time step control) Region A
(single step of physics A)
Mechanics
Mesh and Fields
Region B
(single step of physics B)
Mechanics
Mesh and Fields
Adagio
“Truth” Error Bounds
Software Design and Implementation
•
Observations
– We require solution of (1+2N) auxiliary PDE’s (where N is the number of patches in )
» Each has the same lhs (for linear, self-adjoint operator) » Each has, in general, a different rhs
•
The
SIERRA Framework
helps manage some of this
complexity
Adagio
“Truth” Error Bounds
Software Design and Implementation
Primal “Patch” Region
Fields Fields
Dual “Patch” Region
3. Region copy-subset For each patch in
“Global Dual” Region
Mesh and Fields
1. Region copy-subset
Adagio Procedure Adagio Region
Mechanics
Mesh and Fields
Mechanics
Residual Residual
2. Solve dual problem
4. Create patch mesh
Mesh Mesh
5. Transfer field values 6. Solve local problems 7. Transfer field values
8. Local update inner products and norms 9. Compute bounds
“Truth” Error Bounds in
Adagio
Case 1 (Side Surface)
Valid upper
bound on
Valid lower
bound on
Implicit
Residual Methods
Sample Applications
TICAM Group
(
Babuska and Strouboulis
;
Oden and
•
Variety of linear and nonlinear
elliptic
problems
– Elasticity problems
» Local and average displacements and stress components
– Heat equation with nonlinear and orthotropic materials
» Local temperature
– Burgers’ equation
» Local velocity
– Incompressible Navier-Stokes
» Kinetic energy of flow
– Helmholtz
» Local amplitude
– Eigenvalue problems
Extrapolation Methods
•
Richardson Extrapolation
– Applies in the asymptotic convergence region
– Assume is computed with two mesh sizes, and , where
– Using a known convergence rate
– Can now eliminate to get a very accurate approximation of
1
h
h
2 1 2h
h
α
H.O.T.
)
(
)
(
1 1 αch
u
l
u
l
Q h QH.O.T.
)
(
)
(
2 2 αch
u
l
u
l
Q h Qc
H.O.T.
1
)
(
)
(
)
(
1 2 α αr
u
l
r
u
l
u
l
h Q h Q Qh
r
Extrapolation Methods
•
Richardson Extrapolation, cont.
– If is not known (almost always), then a third solution must be obtained on where 2 3
h
h
α
r
e
e
h hlog
)
/
log(
2 1α
)
(
)
(
1 2 1 h Q h Q hl
u
l
u
e
)
(
)
(
2 3 2 h Q h Q hl
u
l
u
e
(Difference in successive meshes)
constant
1 2 2 3h
h
h
h
r
Extrapolation Methods
Converting
Estimates
into
Bounds
•
Grid Convergence Index
(GCI) (
Roache
)
•
Essentially a
factor of safety
•
For two-grid extrapolation,
•
For three-grid extrapolation,
•
(Approximate) quantity-of-interest bounds:
s
F
)
(
grid]
fine
[
fine h Q RE su
l
e
F
GCI
3
sF
25
.
1
sF
)
(
)
(
fine fine h Q h Qu
GCI
l
u
l
Extrapolation Methods
Ideal vs. Reality
•
Ideal (monotonic convergence)
Example Problem
g
Elastic material (nonlinear) Rigid Body•
Explicit transient dynamics (DYNA)
Extrapolation Methods
Ideal vs. Reality
•
Reality
Num Elements
Extrapolation Methods
Ideal vs. Reality
•
Reality (how it was actually handled)
• GCI applied to middle three points (these were monotonic)
• Timestep errors were ignored
Areas of Active Research
•
Extension of bounds to include modeling and
uncertainty errors (
Oden and Prudhomme
, others…)
•
Harder problems
– Nonlinear parabolic problems
– Hyperbolic problems
– Multiphysics problems
– Extreme anisotropic materials
•
Bounds of exact error (
Babuska and Strouboulis
,
Peraire, et al.
)
– “Certificates” of precision
•
Errors due to operator splitting (
Estep
)
Areas of Active Research (cont.)
•
Nearby problems (
Hopkins, Roy
)
•
Discretization procedures
– Finite volume techniques
– Semi-discrete time integration (method of lines)
– Application to shells, other element types
– Stabilized methods
•
Adaptive error control
– Error indicators
Example of “Hard” Problem
Sandia Thermal Battery
•
Battery operation
– and are heated above melting temperature, then cooled
– Need to stay melted 1 hour
•
Problem features
– Highly transient
– Nonlinear materials (temp-dep)
– Nonsmooth data (read from table)
– Highly orthotropic materials
– Nonlinear BC’s (radiation, convection)
Error Estimation
Risk Assessment
Problem Class Sandia Code(s) Risk Research Issues
Elliptic
Adagio
Salinas (Freq dom) Calore (steady) Low Nonlinearities Nonsmoothness Anisotropic matls Parabolic Calore Aria Fuego Premo (subsonic) Medium Time errors Finite volume Turbulence Hyperbolic Presto
Salinas (Time dom) Premo (supersonic) High Explicit, lumped timestepping History-dep vars Multiphysics Calagio Fuego/Calore/Syrinx … Med-High Loose coupling – transfer operators
Current Limitations of
Error Estimation Impact
•
Can we identify anything that is limiting the impact
and potential of
a posteriori
error estimation?
(Answer: yes)
– Priorities (of the code customers and commercial vendors)
– Computational cost of the algorithms
– Complexity of implementation
Customer Priorities
•
Commercial customers (code end-users) and
commercial software vendors
1. Robustness 2. CPU cost
3. Memory cost 4. Accuracy
•
Accuracy (and knowledge of accuracy) is important,
but not most important
•
In the current marketplace, customers are not willing
to pay a high price for error estimation capabilities
Customer Priorities (cont.)
•
ASCI
– V & V program has elevated the importance of error estimation
» Solution verification » Model validation
– ASCI is driving the need to develop techniques for complex engineering problems
•
Potential to impact the commercial sector, and change
the way all engineering design and analysis is done
•
Success is not guaranteed!
Summary
•
Primary classes of error estimators
– Recovery methods
– Residual methods (quantity-of-interest estimates and bounds)
– Extrapolation methods
•
(Quantity-of-Interest) A Posteriori error estimation is
relatively mature mainly for elliptic problems
•
Can provide both error estimates and error bounds
(good for UQ)
•
A Posteriori error estimation proven very effective for
adaptive error control
Remark:
The
optimization community
has experience
Computational Cost of the
Algorithms
•
Extrapolation
estimators
– Many (at least three) solutions required
– All solutions must be sufficiently resolved, which increases the cost
•
Residual-based
estimators
– Global dual problems required (although linear)
» For estimates, must be solved on finer (h or p) mesh
» For bounds, additional local (element or patch) problems must be solved for error in dual solution
– Time dependent problems
» Dual problems are backwards in time!
Remark:
The
optimization community
has experience
Complexity of Implementation
•
Extrapolation
estimators
– Richardson extrapolation: Usually simple to implement
» Issues mainly with ability to obtain mesh refining or coarsening tools
– Usually can be thought of as a post-processing tool
•
Residual-based
estimators
– Generally intrusive to the code
– Must solve additional problems (pde’s)
» Compute residuals, additional right-hand-sides, etc » Must have a way of handling hybrid fluxes
• Equilibration (very complex to implement)
• Bank-Weiser projection (simpler but less accurate)
• Use subdomains or overlapping patches (potentially costlier)
Limitations on Applicability
of the Algorithms
•
Extrapolation
estimators
– Issue: applicability outside asymptotic convergence regime
•
Residual-based
estimators
– More algorithm research needed
» Solving the backwards-in-time dual problem
» Semi-discrete (method-of-lines) discretizations (instead of DG) » Hyperbolic problems
» Multiphysics problems (e.g., operator splitting…) » Etc.