Meshfree methods and simulations of material failures
1 / 171

Meshfree Simulations of Adiabatic Shear Band and Spall Fracture - PowerPoint PPT Presentation

  • Uploaded on

Meshfree Methods and Simulations of Material Failures Shaofan Li Department of Civil and Environmental Engineering, University of California at Berkeley Collaborators Dr. Bo C Simonsen, Technical University of Denmark; Dr. Daniel C. Simkins,

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
Download Presentation

PowerPoint Slideshow about 'Meshfree Simulations of Adiabatic Shear Band and Spall Fracture' - issac

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Meshfree methods and simulations of material failures l.jpg

Meshfree Methods and Simulations of Material Failures


Department of Civil and Environmental Engineering,

University of California at Berkeley

Collaborators l.jpg

Dr. Bo C Simonsen,

Technical University of Denmark;

Dr. Daniel C. Simkins,

University of South Florida;

Dr. Sergey N. Medyanik,

Northwestern University

Table of contents l.jpg
Table of Contents

  • Introduction: What is Meshfree Method

  • Simulations of Large Deformations

  • Simulations of Strain Localizations

  • Simulations of Dynamics Shear Band Propagations

  • Simulations of Ductile Fracture

    6. Simulations of Impact and penetrations

    7. Conclusions

Meshfree methods l.jpg
Meshfree Methods

  • Smooth Particle Hydrodynamics (SPH)

  • Element Free Galerkin (EFG)

  • Reproducing Kernel Particle Method (RKPM)

    1. Unknown is represented by convolving a smooth kernel with dependent variable

    2. Discretize by evaluating integral via nodal integration

Rkpm kernel l.jpg
RKPM Kernel

  • f(x-y) is a smooth compactly-supported function, e.g. cubic spline

  • PT(x) = [1 x y z xy xz …] is vector of monomial terms

  • b(x) (called the normalizer) is used to regain discrete partition of unity

Slide8 l.jpg

Moment Equation

Note and P(0) = (1, 0, 0, …, 0) then

Compare fem shape function with meshfree shape function l.jpg
Compare FEM shape function with Meshfree shape function

(a) FEM ; (b) Meshfree

Compare fem shape function with meshfree shape function11 l.jpg
Compare FEM shape function with Meshfree shape function

(a) FEM ; (b) Meshfree

Meshfree discretization and meshfree shape functions l.jpg
Meshfree discretization and Meshfree shape functions

Slide13 l.jpg

(a) Shape function

(b) Derivative

(c) Derivative

(d) Derivative

“The Cloud”: 3-D Meshfree shape function and its first derivatives generated

by the tri-linear polynomial basis, P(X) = (1, x1 ,x2 , x3 , x1x2 ,x2x3 , x3x1 , x1x2x3 )

Slide14 l.jpg

1.2 A Few Virtues of Meshfree Methods

Convergence Property

as as

Reproducing Property

For That is

Non-local Interpolation (Discrete Convolution)

Slide15 l.jpg

* Large Support Size

It enables meshfree discretization/interpolation to endure large mesh distortion and

sustain computation without remeshing;

is deformation map

Slide17 l.jpg

Example 2.1 : Compression of A Rubber Cylinder


50% Compression

65% Compression

85% Compression

90% Compression

Slide21 l.jpg

Example 2.2 A pinched cylindrical shell

Material Properties

Computational Parameters

Number of Particles: 30300

Time Step:

Slide23 l.jpg

Example 2.3 Hemispheric Shell with Pinched Load


Material: Elastic-plastic material;

Geometry: Hemisphere shell with radius of 1 inch, thickness of 0.04 inch.

Particles: 12,300

Slide24 l.jpg

(c) t = 3.010-3s

(b) t = 1.510-3s

(a) t = 0.510-3s

(f) t = 7.510-3s

(d) t = 4.510-3s

(e) t = 6.010-3s

Slide25 l.jpg

Example 2.4 The snap-through of a conic shell

Material Properties

Computational Parameters

Number of Particles: 12300

Time Step:

Slide26 l.jpg

The snap-through of a 3D conic shell

(a) (b) (c)

(c) (d) (e)

Slide28 l.jpg

Excessive mesh distortion (hourglassing)



Slide30 l.jpg

Meshfree Simulation of Strain Localization



From Reid, Gilbert, and Hahn [1966]

Meshfree methods element free galerkin efg l.jpg
Meshfree Methods: Element-free Galerkin (EFG)

Fleming and Belytechko [1997]

Slide33 l.jpg

Shearband Path for a Plate with 31 Holes (FEM vs. Meshfree Methods)

(b) 60  90 mesh

(c) 90  60 mesh

(a) 60  60 mesh

(d) 60  90 particles

(e) 90  60 particles

(c) 60  60 particles

Slide36 l.jpg

FEM Methods)Meshfree

(a) 20  20

(b) 30  20

(c) 40  20

(d) 50  20

Slide37 l.jpg

4. Meshfree Multiscale Computations Methods)

= Scale 0 + Scale 1 + Scale 2 + …...

1-D Meshfree Multiscale Basis Functions for Polynomial Basis P = ( 1, x, x2, x3 )

Slide39 l.jpg

[0,0] [1,0] [0,1]

[2,0] [1,1] [0,2]

Hierarchical Partition of Unity of Quadratic Basis

Multiscale analysis l.jpg
Multiscale Analysis [1,0] [0,1]

Spectral Meshfree Adaptive Simulations

Multiscale analysis42 l.jpg
Multiscale Analysis [1,0] [0,1]

(a) Total Scale(b) Low Scale(c ) High Scale(d) Adaptive Pattern

Total Particles: 3321

Particles used for adaptive calculation: 361

Slide43 l.jpg

Example 4.2 Multiresolution Analysis [1,0] [0,1]

Total Scale

Low Scale

High Scale

Slide44 l.jpg

Tension of A Bar with A Hole [1,0] [0,1]

Meshfree (particles)

Zoom in



Mode I

4272 particles/layer

with 3 layers in

thickness direction

Mesh-based (element distortion)

Slide45 l.jpg

Symmetry slip line solution [1,0] [0,1]

Shear Plane Development for a bar with A Hole in Tension (I)

Simulation of anti symmetric slip line solution l.jpg
Simulation of anti-symmetric [1,0] [0,1]slip line solution

How to simulate curve shearband l.jpg
How to simulate curve shearband [1,0] [0,1]

Curved Shear Band Formation in Double-Notched Bar in Tension

13364 particles

Reference 1: Ewing, D.J.F. and Hill, R. J. Mech. Phys. Solids, 15, 115 (1967)

Slide52 l.jpg

Remark: [1,0] [0,1]FEM’s regularized strong discontinuous element technique may be able to simulate a shearband, only if one knows the shearband path a priori. If the shearband path has unknown and evolving curvature, it will be almost impossible to employ such technique.

3d simulation capacity l.jpg
3D Simulation capacity [1,0] [0,1]

(a) (b) (c)

(d) (e) (f)

Necking of a thin plate l.jpg
Necking of a thin plate [1,0] [0,1]

0.0 s 1.010- 5 s 2.010- 5 s

3.010- 5 s 4.010- 5 s 5.010- 5 s

Slide56 l.jpg

Experimental Results on Multiple Shear Band [1,0] [0,1](Anand and Spitzig, 1982)

250 m

0.1 mm

1 mm

Computational Results by Meshfree Method

Iv dynamic shear band simulations l.jpg
IV: Dynamic Shear-Band Simulations [1,0] [0,1]

Recently, we have done some massive

parallel computations on simulations of

stress collapse inside the adiabatic shear band.

Slide58 l.jpg

(a) [1,0] [0,1]


Morphology of adiabatic shear band: (a) scanning electron micrograph of a

Shear band; (b) Details of a shear band (After Leech, P. W, (1985)).

Slide59 l.jpg

Meshfree Simulation of Dynamic Shear Band Propagation [1,0] [0,1]

Experiment on impact-loaded plate

Kalthoff & Winkler [1987], Mason, Rosakis and Ravichandran [1994], and Zhou, Rosakis and Ravichandran [1996a,b]

Failure Mode Transition

Kalthoff problem l.jpg
Kalthoff Problem [1,0] [0,1]

(a) V < Vc

(b) V > Vc

Slide62 l.jpg

Freund-Hutchinson Model ([1 [1,0] [0,1]985]


Zhou-Rosakis Model ([1996])



Slide63 l.jpg

Propagation of Shear Band and Crack at Intermediate Impact Speed

Experimental observation (Rosakis et al., 1999), Impact speed 23.4m/s, C-300 steel

Computation by meshfree method (Contour of effective stress)

Slide64 l.jpg


Image Acquisiton Rate: 1 MHz (1 million frames per second)

Response Time: 750 ns

Noise: 2 mV or ~2o temperature resolution

1.1 mm

Complete System


Slide65 l.jpg

Numerical Techniques Simulations Speed

of Fracture and Crack Growth

  • Erosion Algorithm (Rashid [1968])

  • 2. Traditional FEM remeshing (Wawrzynek and Ingraffea [1987]);

  • 3. Cohesive FEM (Xu-Needleman [1994], and

  • Camacho and Ortiz [1997]; )

  • 4. Meshfree Methods (Belytschko et al [1994]);

  • 5. X-FEM and Level set method (Mose et al [1999],

  • and Belytschko et al [2001]).

Slide66 l.jpg

Ductile-to-Brittle Transition Speed

(a) Computational Configuration; (b) “Mesh” or Particle Distribution

Macroscale picture of shear bands in rosakis model l.jpg
Macroscale Picture of Shear Bands in Rosakis Model Speed



With Heat Conduction

Slide69 l.jpg

Comparison with Experiment at High Impact Speed Speed


Impact speed 29.4 m/s, C-300 steel

(Computation V = 37 m/s)

Dynamic shear band propagation with heat conduction l.jpg

Dynamic Shear Band Propagation Speed with Heat Conduction

  • Motivation

  • Heat Conduction may help to regularize the solution

Formulas for shear band width estimation:

Bai(1986) :



Where k is thermal conductivity, ac(~10-3/K), m (~10-2) are softening and rate sensitivity

parameters, =1-10, and =105 – 106/sec are strain and strain rate inside band

  • Look inside the shear band, see its microstructure (“hot spots” were found in experiments)

  • 3) Measure parameters of shear band.

Slide75 l.jpg

Detailed Shear Band Structure ( temperature signature) Speed

Using Reproducing Kernel Particle Method(RKPM), an numerical meshfree method, we were able to predict for the first time: the ductile-to-brittle failure mode transition; the character change of the curved shear band with the projectile impact speed; an intense, high strain rate region in front of the shear band tip, which we believe to cause a stress collapse that drives the formation and propagation of a dynamic shear band.

Multiple Scale Analysis of Shear Band: comparison with experiments





200 m

Slide76 l.jpg

2.5mm Speed


DT (K)




Field of View

1.1 x 1.1 mm




Slide77 l.jpg

Comparison between Experiment and Computation Results (Temperature)

A. (t =12 s) B. (t = 32 s) C. (t = 72 s)

D. (t =12 s) E. (t = 36 s) F. (t = 72 s)

Experiment (A, B, C), Computation (D,E,F)

Mesh dependency of band width for adiabatic shear band l.jpg
Mesh Dependency of Band Width (Temperature)for adiabatic shear band

Inside of the shear band l.jpg
Inside of the Shear Band (Temperature)


(mesh dependency)

With Heat Conduction

(resolved band width)

Hot spots in a small model l.jpg
Hot Spots in a Small Model (Temperature)

With Heat Conduction solution is regularized, band width is about 20 microns.

Instability of viscous flow occurs inside the shear band.

Similar instabilities l.jpg
Similar Instabilities (Temperature)

1) Instability of plane Couette flow (viscous fluid)

  • Instability in a layer of thermo-viscoplastic material

  • “Structures in shear zones due to thermal effects”

  • (1991, Molinary and Leroy)

Slide85 l.jpg

Shear-Band Bifurcation (Temperature)

Slide86 l.jpg

Experimental Results. (Temperature)

by Wittman et. al. (1990)

Numerical Simulation

V simulation of ductile fracture l.jpg
V (Temperature). Simulation of ductile fracture

X fem l.jpg
X-FEM mesh-dependent

Difficulties and challenges l.jpg
Difficulties and Challenges: mesh-dependent

  • No remeshing and no adaptive discretization is allowed;

  • 2.Fracture criteria;

  • 3.Prediction and simulation main features of ductile failures;

Objectives l.jpg
Objectives: mesh-dependent

To Seek a simple, viable,

and accurate ductile crack growth numerical algorithm that can be used in both engineering computations and scientific research.

Slide105 l.jpg

Why is shape function discontinuous mesh-dependent

At the crack tip ?

Formulations l.jpg
Formulations mesh-dependent

  • Finite deformation thermo-mechanical system

Weak form l.jpg
Weak form mesh-dependent

Solution methodology l.jpg
Solution Methodology mesh-dependent

  • Lagrangian Finite Deformation

  • Explicit time integration

  • Operator split for heat conduction

Constitutive model l.jpg
Constitutive Model mesh-dependent

  • Additive decomposition in rate of deformation

  • Use the Jaumann rate to satisfy objectivity

Numerical simulations l.jpg
Numerical Simulations mesh-dependent

2D Plane strain

uniaxial tension test:

A plate with pre-notched

center crack subjected

prescribed velocity.

Example i gurson tvergaard needleman model l.jpg
Example I: mesh-dependent Gurson-Tvergaard-Needleman model

Crack surface configuration i l.jpg
Crack Surface Configuration (I) mesh-dependent

(a) t=6.0µs, (b)=14.0µs

Crack surface configuration ii l.jpg
Crack Surface Configuration (II) mesh-dependent

(a) t = 18.0 µs (b) t=22.0µs

Size effects l.jpg
Size-effects mesh-dependent

Example ii thermal viscoplastic model l.jpg
Example II: mesh-dependent Thermal-viscoplastic model

Crack propagation in an adiabatic shearband l.jpg
Crack Propagation in an Adiabatic Shearband mesh-dependent

Thermo-mechanical coupling

Thermo-elasto-viscoplastic material constitutive

Example iii johnson cook l.jpg
Example III: Johnson-Cook mesh-dependent

  • elasto-visco-plastic

Damage model l.jpg
Damage Model mesh-dependent

  • Johnson-Cook

Di are parameters

D>1 implies fracture

Operator split for energy equation l.jpg
Operator Split for Energy Equation mesh-dependent

Adiabatic heating, solved in constitutive update:

Heat conduction, solved with momentum:

2d results cont l.jpg
2D Results (cont) mesh-dependent

Crack modeling 3d l.jpg
Crack Modeling-3D mesh-dependent

  • Still through cracks

  • Crack tip fiber - generalized crack tip

  • Cracks propagate from fiber to fiber

  • Damage averaged over nodes on fiber

  • Crack panels - natural extension of 2D

    • not necessarily planar!

  • Visualization by brute force 3D meshing

Crack morphology 3d l.jpg
Crack Morphology - 3D mesh-dependent


Visibility 3d cont l.jpg
Visibility-3D (cont) mesh-dependent

  • Compute intersection of lines with panels

  • Two cases:

    • If panel is parallel to z-axis or x-y plane, treat as a planar quadrangle

    • otherwise, use non-linear iteration to find intersection

Line intersect plane l.jpg
Line Intersect Plane mesh-dependent

Solve for l

Plane intersection if

Parametric visibility algorithm l.jpg
Parametric Visibility Algorithm mesh-dependent

Pick l, Define:




Convergence when

3d results cont l.jpg
3D Results (cont) mesh-dependent

3d results con t l.jpg
3D Results (con mesh-dependent ’t)

Slide144 l.jpg

Front Half mesh-dependent

Slide145 l.jpg

Inside View mesh-dependent

Explosion of a cylinder l.jpg
Explosion of a Cylinder mesh-dependent

Slide151 l.jpg

Simulation of a thin cylinderical shell mesh-dependent

under thermal-mechanical loading

V i simulation of impact and penetration l.jpg
V mesh-dependent I. Simulation of Impact and penetration

Slide155 l.jpg

Impact and Penetration mesh-dependent



Brittle Failure Shear Localization

Slide157 l.jpg

mesh-dependent A multiple-scale constitutive law with length scale dependent cohesive model

where is the flow stress. are material constants.

 In the multi-scale formulation, the flow stress is proposed as

 Inclusion of strain rate and temperature effect (The modified Johnson-Cook model)


the multi-scale length scale dependent cohesive model is embedded in 0 .

 = 0.005 , T0 = 293K ,  = 0.01 , 0 = 0.002, 0.002 /sec

Localization-induced Cohesive Law

Slide158 l.jpg

A Demonstration of the Model mesh-dependent

Strain Rate Effect Temperature Effect

Slide159 l.jpg

Comparison of yield surface between proposed smooth cap model and Drucker-Prager model

Effect of compression strength


Effect of void volume fraction






Slide160 l.jpg

Problem Statement model and Drucker-Prager model

Area of interest

Geometric Parameters:

W = 50.8 cm T = 15.24 cm L =50.8 cm

d = 1.09 cm h = 1.35 cm

Material Properties:

Projectile = 8770 Kg/m3 , Target = 2300 Kg/m3

ETarget = 2058 Mpa mtarget = 90.456 Kg mprojectile = 0.011 Kg

Target = 0.3

The Projectile is modeled as a rigid body in the simulation

v = 1666 m/sec

Time step dt = 3e-7 sec

Total steps 17400




Details of the Projectile



Slide161 l.jpg

Single Projectile Movie - Stress model and Drucker-Prager model

Slide163 l.jpg

Single Projectile Movie - Damage model and Drucker-Prager model

Slide164 l.jpg

Damage Evolution model and Drucker-Prager model

Slide165 l.jpg

Multiple Projectiles - Stress model and Drucker-Prager model

Conclusions l.jpg
Conclusions model and Drucker-Prager model

  • Thank You !

Slide170 l.jpg

Bifurcated Energy Release Solutions model and Drucker-Prager model

Meshfree methods171 l.jpg
Meshfree Methods model and Drucker-Prager model