thoughts on modeling and simulation in the doe environment
TRANSCRIPT
Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia
Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department
of Energy's National Nuclear Security Administration under contract DE-AC04-94AL85000.
Thoughts on Modeling and Simulation
in the DOE environment
David Womble
Sandia National Laboratories
2
Three phases in the design cycle have
different modeling requirements
Requirements definition Design Qualification
Le
ve
l o
f fi
de
lity
3
Three phases in the design cycle have
different modeling requirements
Requirements definition Design Qualification
Le
ve
l o
f fi
de
lity
High fidelity
Multiple scales and physics
Full environment
4
Three phases in the design cycle have
different modeling requirements
Requirements definition Design Qualification
Le
ve
l o
f fi
de
lity
High fidelity
Multiple scales and physics
Full environment
Low fidelity
Rapid turnaround design studies
Component level
Often single physics and scales
5
Three phases in the design cycle have
different modeling requirements
Requirements definition Design Qualification
R238
1.5kQ28
MMBT2907
Q6
MMBT2222
V2
3Vdc
M3
MTB30P06V
R239
1K
0
0
RF_PWR3_TXPA
Q62
BFS17A
DA_BATTERY3
R207
1k
R236
1k
RL3
10
0
V8TD = 70ms
TF = 3ns
TR = 3ns
V1 = 3
V2 = 0
Q63
MMBT2907
R242
10R251
499
0
D50
MMSZ5236BT1
0
R204
4.99K
V1
-8Vdc
FET_BIAS
VDD
R241
200
R240
100
RF_PWR3_ENB_N
R206
1K
C1
5uF
R226
4.99k
R237
1k
R46
1k
V7
TD = 0
TF = 1ms
TR = 50ms
V1 = 0
V2 = 10.8
0
C40
1uf
0
R210
10K
Le
ve
l o
f fi
de
lity
High fidelity
Multiple scales and physics
Full environment
Low fidelity
Rapid turnaround design studies
Component level
Often single physics and scales
High fidelity
Multiple scales and physics
Rigorous V&V
6
How much computing is required
for different phases?
Crash scenario
(no practical testing)
Combined abnormal environment
(no testing)
Re-entry (admiral’s test)
Re-entry with a radiation event
(no testing)
Carry/release (limited testing)
Laydown (limited testing)
Launch accelerometer
(limited testing)
Timing and fuzing operations
(limited testing)
Exascale
P
eta
scale
Te
rascale
Normal electrical performance
(with testing)
Co
mp
on
en
ts
Sys
tem
s
En
viro
nm
en
ts
7
How much computing is required
for different phases?
Crash scenario
(no practical testing)
Combined abnormal environment
(no testing)
Re-entry (admiral’s test)
Re-entry with a radiation event
(no testing)
Carry/release (limited testing)
Laydown (limited testing)
Launch accelerometer
(limited testing)
Timing and fuzing operations
(limited testing)
Exascale
P
eta
scale
Te
rascale
Re
qu
irem
en
ts b
ou
nd
ing
eve
nts
Pe
rform
an
ce
eve
nts
Normal electrical performance
(with testing)
Co
mp
on
en
ts
Sys
tem
s
En
viro
nm
en
ts
8
How much computing is required
for different phases?
Crash scenario
(no practical testing)
Combined abnormal environment
(no testing)
Re-entry (admiral’s test)
Re-entry with a radiation event
(no testing)
Carry/release (limited testing)
Laydown (limited testing)
Launch accelerometer
(limited testing)
Timing and fuzing operations
(limited testing)
Exascale
P
eta
scale
Te
rascale
Re
qu
irem
en
ts b
ou
nd
ing
eve
nts
Pe
rform
an
ce
eve
nts
Normal electrical performance
(with testing)
Co
mp
on
en
ts
Sys
tem
s
En
viro
nm
en
ts
Desig
n
Qu
alific
atio
n
9
Is physics-based predictive?
Is predictive physics-based?
Experiments
Empirical models
and table look-ups
Simulations
Results
Verification
and
Validation
Observational M&S
• Models derived from observation
• Not predictive
• Cannot validate
10
Is physics-based predictive?
Is predictive physics-based?
Predictive modeling and simulation must be physics-based and validated
Experiments
Empirical models
and table look-ups
Simulations
Results
Verification
and
Validation
Observational M&S
• Models derived from observation
• Not predictive
• Cannot validate
Experiments
Physics-based
models
Simulations
Results
Verification
and
Validation
“Physics”
Physics-based M&S
• Models based on “science”
• Can be predictive
• Incorporates multiple scales
• Can validate
11
There are many engineering challenges that will
rely on predictive modeling and simulation
• Large Scale/Complex Structural Response: Structural response due to
severe mechanical insult including shock and penetration
• Predictive Flight Test Simulation: Simulated re-entry body flight
environments including structural load propagation to internal component
mechanical response
• Engine design on a laptop: Turbulent reacting flows in combustion,
predictive simulation code for high-speed mixing & fluid jet break-up
• Thermal Mechanical Failure: Abnormal thermal environments with
applicability to ship fire, plane fire, …
• Energetic Material Initiation: Predictive capability for simulating explosives
• Electromagnetic Effects: Lightning with EM coupling
• Aging: Predictive capability for physical effects such as hardening,
embrittlement, degradation
• Manufacturing: Including flows, coating with the ability to predict as-built
performance
There are many research challenges needed to support a predictive engineering capability
Driven by engineering challenges
• Validated, predictive modeling approach(es) for ductile material crack initiation,
propagation and fracture
• Determining pressure and velocity field turbulent aerodynamic flow fluctuations with
structural dynamic load coupling models & testing (e.g. 6-degree of freedom shaker)
• Improved high resolution, time resolved diagnostics: i) imaging fluid break up, ii)
characterizing explosive initiation, iii) measuring spatial and temporal plasma
properties
• Computationally tractable approaches for multi-scale (10 m – 10 mm) mechanics
simulations (e.g. shock/blast to structural dynamics)
• Coupled thermal-mechanical simulation with time/temperature dependent materials
properties
• Determining operative lightning-induced failure mechanisms in weapon systems
• Rapid problem set up for large models with different levels of detail
Further out
• Computational design of engineered materials/structures at the
continuum scale – with materials science
• Inverse analysis capabilities for failure assessments – algorithm
development with computer sciences
• Community geomechanics model that couples
thermal/mechanical/hydraulic/chemical response – with geosciences
CAD
Model Mesh
13
V&V should be an intrinsic part of all elements
of modeling and simulation
• V&V/UQ is a part of every analysis
• Aspects of “intrinsic V&V” include
– SQA standards and rigorous testing methodology
– Automatic code and feature coverage reports in each log file
– Dynamic test suites designed around key applications
– Support for solution convergence for various quantities of interest by users
– Embedded sensitivities and UQ where possible linked to sample-based UQ
– Integrated workflow that includes support for computing margins
– Support for DoE and coupling to validation experiments
– Code expects models and input to include uncertainties and propagates these
uncertainties through simulations
– Integrated post-processing (e.g., visualization) includes techniques for
understanding and presenting uncertainties and margins
• Customers demand it, analysts perform it, scientists, engineers and
developers enable it
14
Sandia’s X-Prize is an attempt to test to predictivity of
fracture and failure modeling and simulation capabilities
4 Modeling Paradigms Were Used • XFEM – diplacement and discontinuities embedded
in elements
• Peridynamics – meshless method
• Tearing Parameter – plastic strain evolution integral
• Localization Elements – focused on element
interfaces
Three challenge problems were defined to
assess failure initiation and crack
propagation methods
Summary • Predicting ductile failure initiation and crack propagation
remains an extremely difficult problem.
• Wide variation in modeling results suggest our methods
are not yet predictive.
• Limits in experimental capabilities were also observed
L
r
d
F
L
r
d
L
r
d
F
3.5mm 3.5mm 3.5mm
2.5 mm
2.5 mm
2.5 mm
A B C
D
E
F
D E F
A B C
Experimentally Observed
Paths
15
Closing Thoughts
• Modeling and simulation can impact all phases of the
design/qualification cycle, but computing needs are very
different.
• “Predictive” simulation is a high bar.
• The goal is really to be able to use modeling and simulation as
the basis for decisions.