Nov. 9, 2006 DDSim: A Next Generation Damage and Durability Simulator Presenting: John Emery Advising and Supporting: Prof. Tony Ingraffea, John Dailey.

Nov. 9, 2006

DDSim: A Next Generation Damage and

Durability Simulator

Presenting: John EmeryAdvising and Supporting: Prof. Tony Ingraffea, John Dailey Jr.,

Gerd Heber, Wash Wawrzynek

funded through NASA’s Constellation University Institutes Project.

2John Emery & Tony IngraffeaCornell University

ASME International Mech. Eng. Congress and Expo DDSim: A Damage and Durability Simulator

Outline for the Talk

The Big Picture & Overview DDSim Level I – Reduced-order filter

InputApproachResults & Performance

Level II – Automated crack insertion ApproachResults

Level III – Multiscale simulationStatistically accurate microstructural geometryMultiscale implementation

Conclusions



DDSim

Finite element model of structure including boundary/environmental conditions

Material system & pertinent microstructural statistics

Best available physics-based damage models

Random input

Time to failure, N

Probabilistic life prediction

The Big Picture

P



Overview of the Hierarchical Approach

Level I: A fast, analytical, reduced-order filter to determine life-limiting hot-spots in complex structures

Level II: Traditional continuum fracture mechanics, FRANC3D, to compute the life of the structure consumed by growth of microstructurally large cracks (NMLC)

Level III: Multiscale simulation to compute the life of the structure consumed by incubation, nucleation and propagation of microstructurally small cracks (NMSC)

Assumption: Ntotal = NMLC + NMSC

A multiscale approach with 3 hierarchical levels:



Models and Parameters for Fracture and Damage Mechanics

• Models for fatigue crack growth (NASGRO equation*)

• Statistical material data & initial damage size

Database from FEM without damage

• Mesh

• Field information

• Boundary conditions

Stress field contour plot: Rib-stiffened element

*Forman & Mettu, Fracture Mechanics: Twenty-second Symposium, Vol. 1, ASTM STP 1131

A (slide 6)

DDSim Level I: Input

B (slide 9)



Life prediction contour ploton original FE Mesh

(29,072 surface nodes, average ai=2.76e-4 in)

• Analytical solutions & field data from undamaged FEM used to estimate service life limited by damage at a large number of possible origins (mesh nodes).

Key Ideas for Level I:High Volume, High Automation, Probabilistic, &

Conservative First Order Analysis

• These damage origins do NOT become part of the geometrical model in Level I.

• These damage origins do NOT interact with each other.

• These simplifications readily allow parallel processing.

How to map:

Stress Life prediction?• Initial flaw size from statistical

distribution (eg. particle x-sectional area).

Stress field contour plot: x-section A (previous), Rib-stiffened element

DDSim Level I



Conservative SIF History

0.0

30.0

60.0

90.0

0.00 0.10 0.20

a_DDSim b_F3Db_DDSim a_F3D

Crack length, (in)S

tres

s in

tens

ity f

acto

r, (

ksii

n)Principal stress on a

thermomechanically loaded part (courtesy of FAC)

DDSim Level I is designed to provide conservative estimates of K (compared with FRANC3D here).

a

b



Under Fatigue spectrum Nodes (i.e. initial flaw locations): 63,974 Random initial flaws (from particle filter): 10,000 No. of processors @ 3.6GHz w/ 2GB RAM: 16 Min & Max computed life (cylces): 18542 - 99,999 Processing time (hr:mm): 5:48

Level I Results & Performance

Particle diameter, (in)

0

0.25

0.5

0.75

1

1.00E-05 1.00E-04 1.00E-03

Particles

Broken

Pro

babi

lity

of o

ccur

renc

e

Life prediction contour plot w/ 10,000 initial flaws



Fully 3D crack growth simulation at “hot spots”: • Explicit representation of crack surface in FE model geometry• Automatically inserted at “hot spots” determined by Level I analysis

Level I Life prediction contour plot (x-section B slide 5)

Automatically inserted, grown and remeshed crack

DDSim Level II



0

10000

20000

30000

40000

0.015 0.065 0.115 0.165

Level I, aLevel I, bLevel II, aLevel II, bLevel II, m

Level II Results

Crack paths

b

m

a

Low fidelity NMLC = 803 cycles, High fidelity NMLC = 4070 cycles

Crack length, (in)

N, (

load

cyc

les)

NMLC = 4070

Level I predictions



Representative digital microstructure

With a first order, probabilistic analysis completed, focus on the “hot spots” to increase the accuracy of the NMSC

prediction using:

• Representative digital microstructure • Best available physics • Multiscale simulation• High performance parallel computing

DDSim Level III: Multiscale Simulation

Life contour plot from initial prediction

Focusing on a “hot spot”, rectangular void for PC model



Crack Incubation Crack Nucleation Crack Propagation

Level III: Microstructurally Small Damage

Important geometrical features: Grains Particles

Damage processes and events: (a) Crack incubation process – damage accumulation until the

particle cracks (b) Crack nucleation event – (c) Microstructurally small crack propagation – process of

crack growth within grains and across grain boundaries

a

b

c

10 m



Level III: Current Microstructural Geometry Models

The lumber model (right) approximates the average grain size and aspect ratios of AA 7075 in a randomly assorted stack

The rolled Voronoi model (right) is our most statistically accurate geometry for rolled AA 7075, approximating grain morphology and average size.

The Voronoi model (left) approximates random crystallographic structure of an unrolled alloy



DDSim Level III: Multiscale Simulation

Life contour plot from initial prediction

Focusing on a “hot spot”, rectangular void for PC model

Multiscale model = Continuum + structure!

Representative digital microstructure



DDSim Level I provides a high volume, highly automated, probabilistic, and conservative life prediction (Ntotal) for real structures & locates areas of high interest for the Level II & III simulations

Level II uses the current best practice fracture mechanics life predictions methodologies for high fidelity NMLC

The Level III microstructural models incorporate state-of-the-art physics and accounts for microstructural stochasticity for high fidelity NMSC.

DDSim, as a multiscale system, will provide microstructurally educated life predictions for real structures.

Conclusions

Our assumption is: Ntotal = NMLC + NMSC



Safety slides

Intentionally blank



Metallic Composite

0.6 mm

Dissimilar materials…similar microstructural geometrical features

Level III: Microstructural Geometry and Damage

50 mMetallic micrographs courtesy of A. Rollett, CMU.

Composite micrographs from: Nicoletto G., Enrica R., Composites: Part A, 35, 2004, 787 – 795, & S. Stanzl-Tschegg, personal communication



Level III: Particle Crack Incubation Criterion

Affect of grain orientation on particle stress, 3 categories:

High stress orientation Intermediate stress orientation Low stress orientation

xx

High orientation

Low orientation Intermediate orientation

• Particle aspect ratio• Grain orientation• Strain level

Pro

babi

lity Particle Tensile Stress (σxx)

Probability Density Function

Particle Tensile Stress

Pro

babi

lity



Currently, we have these loose ends:

Monte Carlo simulation is feasible for the low fidelity life prediction of DDSim Level I, however, it is NOT feasible for the multi-scale simulation

One microstructural model requires millions of DOF

Required number of samples makes MC intractable

Level II computes the life consumed by continuum length scale damage evolution

Level III computes the life consumed by micro-scale damage evolution

Putting It All Together

Recall our assumption was: Ntotal = Nmacro + Nmicro



Low fidelity life predictor:

DDSim Level I

Low fidelity life prediction

Damage site iterator:

DDSim Level II, Nmacro

Combine conditional probabilities

“predictor”

“corrector”

Multi-scale simulation:

DDSim Level III, Nmicro

Bayesian estimation

)@|( DSaNP imicro

Low fidelity “prior” conditional life cdf

nodes

iiitotalT aPaNPNP )()|()(

)@|( DSaNP imicro

High fidelity “post” conditional life cdf

Random input

High fidelity High fidelity life predictionlife prediction

Putting It All Together

N

P

N

P

N

P

N

P



3-D Continuum Model 3-D Continuum Field Analysis 3-D Realistic Microstructure Model

Analyze microstructure for to capture

microstructural damage evolution,

update continuum damage state and fields

Apply B.C.’s from Macroscale Model

Simulate damage evolution:steam enhanced delamination

(with models from collaborating CUIP IFST teams)particle debonding/cracking

crystal plasticity/cohesive constitutive modelsintra/intergranular microcracking

Level III: Multi-scale



Close-Up of Bolt-Holex (ksi)

Continuum Scale

Continuum Model

E=10,500 ksi

=0.33

45.17 ksi

Gather Boundary Conditions and Apply to Polycrystal Model

Polycrystal Model

Polycrystal Scale

x (ksi)

Calculate New Modulus for Each Gauss Point in the Continuum

Model

Grain Boundary Decohered

E=10,500 +/- 1,000 ksi

=0.33

tp=72.5 ksi

Update Stiffness

Level III: Multi-scale, 2D Example

x

y



x

Updated Continuum Model

Smeared Crack

Level III: Multi-scale, 2D Example

x

y

Nov. 9, 2006 DDSim: A Next Generation Damage and Durability Simulator Presenting: John Emery Advising and Supporting: Prof. Tony Ingraffea, John Dailey.

Documents

damage origins

generation damage

expo ddsim

damage mechanics models

hierarchical approach

durability simulator

durability simulator

durability simulator