parallel wave propagation and topological operators for fragmentation simulation
Post on 30-Dec-2015
30 Views
Preview:
DESCRIPTION
TRANSCRIPT
Parallel Wave Propagation and Topological Operators for Fragmentation Simulation
Glaucio H. Paulino
Professor, Faculty CEE, MechSE, CSE
Donald Biggar Willett Professor of Engineering
6th Annual Workshop on CHARM++ and its Applications
4/15/2008 2http//cee.uiuc.edu/paulino paulino@uiuc.edu
Acknowledgments
CEE COLLABORATOR:
Mr. Kyoungsoo Park
CS COLLABORATORS:
Prof. Laxmikant V. Kale (UIUC)
Dr. Celso L. Mendes (UIUC)
Dr. Terry L. Wilmarth (UIUC)
Mr. Aaron Becker (UIUC)
Mr. Isaac Dooley (UIUC)
Prof. Waldemar Celes (PUC-Rio)
Mr. Rodrigo Espinha (PUC-Rio)
4/15/2008 3http//cee.uiuc.edu/paulino paulino@uiuc.edu
Stress waves
Waves
Rosakis AJ, Samudrala O, Coker D, 1999, Science 284
4/15/2008 4http//cee.uiuc.edu/paulino paulino@uiuc.edu
Outline
Waves Wave Propagation
Rayleigh wave Parallel computing: ParFUM Results: Geological simulation
Dynamic Fracture Cohesive zone modeling Node Perturbation & Edge-Swap Operation Results: Fracture & Compact compression tests
Summary Future Work
4/15/2008 5http//cee.uiuc.edu/paulino paulino@uiuc.edu
Wave Propagation: Rayleigh Wave
Surface Wave
Lord Rayleigh, 1885
Seismology, Geology, Material Science, etc
Homogeneous & Orthotropic materials (2005)
Large-Scale 3D Analysis for Graded media
• Rayleigh L. 1885, On waves propagated along the plane surface of an elastic solid, Proc. R. Soc. Lond. A 17, 4-11
• Vinh PC, Ogden RW. 2005, On the Rayleigh wave speed in orthotropic elastic solids, Meccanica 40, 147-161.
4/15/2008 6http//cee.uiuc.edu/paulino paulino@uiuc.edu
Parallel Computing: ParFUM
Finite Element Analysis
Bill Gropp: Very “Easy to write code that
scales and performs poorly.”
Time Integration
Central difference method
Communications
Update Rint
Shared-node summation operation
21
12n n n nt tu u u u+ = +D + D& &&
1 int1 1 1( )ext
n n n-
+ + += -u M R R&&
1 1( )2n n n n
t+ +
D= + +u u u u& & && &&
Lawler OS, Chakravorty S, Wilmarth TL, Choudhury N, Dooley I, Zheng G, Kale LV, 2006, ParFUM: a parallel framework for unstructured meshes for scalable dynamic physics applications, Engineering with Computers 22, 215-235.
4/15/2008 7http//cee.uiuc.edu/paulino paulino@uiuc.edu
Machine Specification
Dell Cluster [Abe] Peak FLOPS: 89.47 TF Number of Blades (nodes): 1200 Number of CPUs (cores): 9600 Processor: Intel 64 2.33GHz dual socket quad core 8 MB L2 cache (2 MB) Memory: 8GB (1GB) Total: 9600 GB
Dell Xeon Cluster [Tungsten] Peak FLOPS: 16.38 TF Number of nodes: 1280 Number of processors: 2560 Processor: Intel Xeon 3.2 GHz (32-bit) Memory: 1.5 GB Total: 3840 GB
4/15/2008 8http//cee.uiuc.edu/paulino paulino@uiuc.edu
0
2000
4000
6000
8000
10000
32 64 128 256 512 1024
Runtime Performance
Number of processors
Par
alle
l ru
nti
me
(sec
) # of Elements: 0.4 million
91%
90%
81%62% 43%
Dell Cluster [Abe]
4/15/2008 9http//cee.uiuc.edu/paulino paulino@uiuc.edu
Geology Simulation: Rayleigh Wave
Layer #4
Layer #3
Layer #2
Layer #1 (Graded)
4km
2km
2km
2km
10 km
10 km
Layer #4
Homogeneous
10km
10 km
10 km
z
P(t) P(t)
( ) sin(2 ) (0 2)
( ) 0 ( 2)
P t t t
P t t
p= £ £
= >
V. Pereyra, E. Richardson, S. E. Zarantonello, Large scale calculations of 3D elastic wave propagation in a complex geology, Proceedings of the 1992 ACM/IEEE conference on Supercomputing, Minneapolis, Minnesota, 301-309.
4/15/2008 10http//cee.uiuc.edu/paulino paulino@uiuc.edu
Homogeneous Media
Material Properties
Dynamic Responses
CP (km/s)
CS (km/s)
ρ(kg/m3)
E (MPa) v
Media 2 1.2 2 7.02 0.219
Tim
e (sec)
5
4
3
2
1
( ) 1/f r r
1
1.01
CR = 1.1 km/s
4/15/2008 11http//cee.uiuc.edu/paulino paulino@uiuc.edu
Graded & Layered Media
( ) 1/f r r
( ) 1/f r r
CP (km/s)
CS (km/s)
p(kg/m3)
E (MPa) v
Graded Layer #1
23.6
1.22.16
23.6
7.0240.94
0.2190.219
Layer #2 3.5 2.1 2.3 24.7 0.336
Layer #3 4.5 2.1 2.3 27.6 0.361
Layer #4 5.5 2.1 2.3 28.7 0.430
Tim
e (sec)
5
4
3
2
1
1
1.25
4/15/2008 12http//cee.uiuc.edu/paulino paulino@uiuc.edu
Waves Wave Propagation
Rayleigh wave Parallel computing: ParFUM Results: Geology simulation
Dynamic Fracture Cohesive zone modeling Node Perturbation & Edge-Swap Operation Results: Fracture & Compact compression tests
Summary Future Work
4/15/2008 13http//cee.uiuc.edu/paulino paulino@uiuc.edu
Dynamic Fracture: Cohesive Zone
Cohesive Zone Model
Computational Simulation TRULY Extrinsic cohesive surface elements
Several claims of extrinsic simulations in the literature are NOT truly extrinsic (e.g. activated elements are not extrinsic)
Ce-basedTi-basedXi XK et al., 2005,, Physical Review Letters, 94, 125510
4/15/2008 14http//cee.uiuc.edu/paulino paulino@uiuc.edu
Topology-based Data Structure
Complete Topological Data
Reduced Representation
Support for Adaptive Analysis
• W. Celes, G.H. Paulino, R. Espinha, 2005, Efficient handling of implicit entities in reduced mesh representations, Journal of Computing and Information Science in Engineering 5 (4), 348-359.
• W. Celes, G.H. Paulino, R. Espinha, 2005, A compact adjacency-based topological data structure for finite element mesh representation, IJNME 64(11), 1529-1556
• G. H. Paulino, W. Celes, R. Espinha, Z. Zhang, 2008, A general topology-based framework for adaptive insertion of cohesive elements in finite element meshes, EWC 24, 59-78
4/15/2008 15http//cee.uiuc.edu/paulino paulino@uiuc.edu
Entity Enumeration
Model
Entity information
Elapsed time (s)
Topologicalentity
Number of entities
Titan IV model (linear hexahedral
mesh)
Element 1,738,240 0.097
Node 1,845,640 0.046
Facet 5,321,600 0.219
Edge 5,429,000 0.292
Vertex 1,845,640 0.186
W. Celes, G.H. Paulino, R. Espinha, 2005, Efficient handling of implicit entities in reduced mesh representations, Journal of Computing and Information Science in Engineering 5 (4), 348-359.
4/15/2008 16http//cee.uiuc.edu/paulino paulino@uiuc.edu
4K Structured Mesh
Mesh Orientation Dependence 4 direction Maximum error: 45º
8 direction Maximum error: 22.5º
Undesirable crack pattern
4/15/2008 17http//cee.uiuc.edu/paulino paulino@uiuc.edu
Node Perturbation
Edge Swap
Proposed Remediation
0.0 0.1 0.3
4/15/2008 18http//cee.uiuc.edu/paulino paulino@uiuc.edu
Crack Length Convergence
Simulation Outline Find the shortest path
(e.g. ) Node Perturbation (NP) Factor: 0.3 Edge Swap Square 4K structured mesh Element size: 0.1 Simulate 100 randomly perturbed
meshes for each node perturbation factor
1
2.4
1tan (2.4) 67.38
2.6ExactL
4/15/2008 19http//cee.uiuc.edu/paulino paulino@uiuc.edu
Results
NP 0.3
Avg. Error = 5.5%
NP 0.3 & Edge Swap
Avg. Error = 4.5%NP0, Error = 8.2%
4/15/2008 20http//cee.uiuc.edu/paulino paulino@uiuc.edu
Crack Angle Convergence
67.38
NP Factor = 0 NP Factor (0.3) & Edge swap
4/15/2008 21http//cee.uiuc.edu/paulino paulino@uiuc.edu
Effect of Element Size & Edge-SwapA
ng
le (
º)
Element Size
Activate Edge Swap (ES)
Without Edge Swap (ES)
NP Factor = 0.3Given Angle (α)
0 10 20 30 40 50
50.2
53.1
56.3
59.7
63.4
67.4
71.6
76.0
An
gle
(º)
The number of appearance
Element Size = 0.1
45
50
55
60
65
70
75
80
0 0.05 0.1 0.15 0.2
with edge-swap
without edge-swap
4/15/2008 22http//cee.uiuc.edu/paulino paulino@uiuc.edu
Fracture Test (Verification)
0 0.0124
16 mm
4.2 mm
4K structured mesh (80 X 21)
Material PropertiesE = 3.24 GPav = 0.35ρ = 1190 kg/m3
GI = 352 N/mσmax = 129.6 MPa
4/15/2008 23http//cee.uiuc.edu/paulino paulino@uiuc.edu
[mm]
Compact Compression Specimen
Incident bar Transmitter bar
60
70
2035
16
E = 5.76 GPav = 0.42ρ = 1182 kg/mm3
GI = GII = 4800 N/mσmax = 105 MPa
striker
4/15/2008 24http//cee.uiuc.edu/paulino paulino@uiuc.edu
Summary
Large-Scale Parallel Wave Propagation 1024 processors
Rayleigh Wave Speed in 3D Functionally Graded Media
Crack Path Representation thru Topological representation Node perturbation & Edge swap operators
Adaptive Dynamic Fracture Simulation V & V
4/15/2008 25http//cee.uiuc.edu/paulino paulino@uiuc.edu
Future Work
Wave propagation for complex geology systems Provide guidance to estimate Rayleigh wave speed in
smoothly graded heterogeneous media
Incorporate data from geological surveys
Parallel Dynamic Fracture Simulation Parallel adaptive insertion of cohesive surface element
Dynamic adaptive load balancing
There is a lot of exciting work to do !
top related