Local Approach To Fracture

Local approach to fracture
Written by : Clotilde Berdin Jacques Besson Stphane Bugat Rodrigue Desmorat Frdric Feyel Samuel Forest Eric Lorentz Eric Maire Thomas Pardoen Andr Pineau Benot Tanguy
Published under the supervision of Jacques Besson
Les Presses de lcole des Mines Paris, 2004
cole des Mines de Paris, 2004 60, Boulevard Saint-Michel, 75272 Paris CEDEX 06 FRANCE email : Presses@ensmp.fr http://www.ensmp.fr/Presses ISBN : 2-911762-55-X Dpt lgal : juillet 2004 Achev dimprimer en juillet 2004 (Corlet, Cond-sur-Noireau) Tous droits de reproduction, dadaptation et dexcution rservs pour tous les pays
Contents
Introduction Global approach to fracture . . . . . . . . . . . . . . . . . . . . . . . 1.1 Small scale yielding Large scale yielding . . . . . . . . . . 1.2 Linear elastic fracture mechanics (LEFM) . . . . . . . . . . . 1.3 Non linear fracture mechanics (NLFM) . . . . . . . . . . . . 1.3.1 The -Integral . . . . . . . . . . . . . . . . . . . . 1.3.2 as stress-intensity factor: the HRR eld . . . . . 1.4 stress and factor . . . . . . . . . . . . . . . . . . . . . 2 Local approach to fracture . . . . . . . . . . . . . . . . . . . . . . . 2.1 Main principles of the local approach to fracture . . . . . . . 2.2 Examples of fracture behavior that the local approach can explain 2.2.1 Size and geometry effects . . . . . . . . . . . . . . 2.2.2 Warm pre-stress effect . . . . . . . . . . . . . . . . 3 Organization of the book . . . . . . . . . . . . . . . . . . . . . . . . 4 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Physical mechanisms of damage Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ductile fracture: metals and polymers . . . . . . . . . . . . . . . . . 2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Cavity nucleation . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Cavity growth . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Cavity coalescence and criteria for ductile fracture . . . . . . 3 Brittle fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Cleavage. Metals and ceramics . . . . . . . . . . . . . . . . . 3.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . 3.1.2 Is cleavage fracture nucleation or growth controlled ? 3.1.3 Scatter and size effects . . . . . . . . . . . . . . . 3.1.4 Microstructural effects . . . . . . . . . . . . . . . . 3.2 Intergranular fracture Metals and ceramics . . . . . . . . . 4 Crazing in polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Craze initiation . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Craze widening . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Craze breakdown . . . . . . . . . . . . . . . . . . . . . . . . 5 Quasi-brittle fracture Ceramic composite materials and concrete . . 5.1 Ceramic composite materials . . . . . . . . . . . . . . . . . . 5.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . 5.1.2 Fracture of brittle matrix composites . . . . . . . . 1 2 1

15 16 16 17 20 20 20 21 22 22 25 25 27 28 30 33 33 33 33 34 37 38 40 40 40 43 43 45 45 47 48 48 49 50 50 50 51
II
Local approach to fracture 5.2 Concrete . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ductile to brittle transition . . . . . . . . . . . . . . . . . . . . . . . 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Ductile to brittle transition in metals . . . . . . . . . . . . . . 6.2.1 Transition in smooth specimens . . . . . . . . . . . 6.2.2 Effect of notches on the ductile-brittle transition . . 6.2.3 Ductile to brittle transition in precracked specimens 6.3 Ductile to brittle transition in polymers . . . . . . . . . . . . Creep fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Introduction to creep deformation . . . . . . . . . . . . . . . 7.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3 Cavity nucleation . . . . . . . . . . . . . . . . . . . . . . . . 7.4 Diffusion controlled cavity growth . . . . . . . . . . . . . . . 7.4.1 Grain boundary control . . . . . . . . . . . . . . . 7.4.2 Surface diffusion control . . . . . . . . . . . . . . 7.4.3 Constrained diffusional cavity growth . . . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Quantitative measurement of damage Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Indirect measurement . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1 Principle of the measurement . . . . . . . . . . . . 2.1.2 Experimental difculties . . . . . . . . . . . . . . . 2.1.3 Examples in the literature . . . . . . . . . . . . . . 2.1.4 Discussion . . . . . . . . . . . . . . . . . . . . . . 2.2 Youngs Modulus . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Principle . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Experimental difculties . . . . . . . . . . . . . . . 2.2.3 Examples in the literature . . . . . . . . . . . . . . 2.2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . 2.3 Acoustic methods . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Propagation of ultrasonic waves . . . . . . . . . . . 2.3.2 Acoustic emission (AE) . . . . . . . . . . . . . . . 2.4 Miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . . Direct measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Different imaging methods . . . . . . . . . . . . . . . . . . . 3.1.1 Surface microscopy (light and scanning electron) . 3.1.2 Near surface microscopy (acoustic) . . . . . . . . . 3.1.3 X-ray imaging . . . . . . . . . . . . . . . . . . . . 3.2 In situ / ex situ character of the experiments . . . . . . . . . . Image processing and analysis . . . . . . . . . . . . . . . . . . . . . 4.1 Image processing . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Image analysis . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Stereology . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Measurement of the damage parameter from images . . . . . 4.5 Measurement of intervoid distance and void size and shape from images . . . . . . . . . . . . . . . . . . . . . . . . . . Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 56 56 57 57 58 61 62 64 64 66 68 68 69 70 71 73 79 79 80 80 80 81 81 82 83 83 84 84 86 86 86 88 88 89 89 89 90 93 96 97 97 98 99 100 102 104 105
8 III 1 2
5 6
Contents IV
Testing 109 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 Homogeneously deformed type tests . . . . . . . . . . . . . . . . . . 110 2.1 Tensile test . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 2.1.1 Geometries and notations . . . . . . . . . . . . . . 111 2.1.2 Mechanical test . . . . . . . . . . . . . . . . . . . 111 2.1.3 Simplied representations . . . . . . . . . . . . . . 114 2.2 Compression tests . . . . . . . . . . . . . . . . . . . . . . . 115 2.3 Notched tensile specimens . . . . . . . . . . . . . . . . . . . 115 2.3.1 Geometries and notations . . . . . . . . . . . . . . 116 2.3.2 Mechanical test . . . . . . . . . . . . . . . . . . . 118 2.3.3 Elastic analysis . . . . . . . . . . . . . . . . . . . 119 2.3.4 Bridgman analysis . . . . . . . . . . . . . . . . . . 120 3 Structural specimens . . . . . . . . . . . . . . . . . . . . . . . . . . 121 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 3.2 Notched specimens . . . . . . . . . . . . . . . . . . . . . . . 121 3.2.1 Geometries . . . . . . . . . . . . . . . . . . . . . . 121 3.2.2 Mechanical test . . . . . . . . . . . . . . . . . . . 123 3.3 Precracked specimens . . . . . . . . . . . . . . . . . . . . . 126 3.3.1 Geometries . . . . . . . . . . . . . . . . . . . . . . 126 3.3.2 Precracking procedures . . . . . . . . . . . . . . . 126 3.3.3 Mechanical test . . . . . . . . . . . . . . . . . . . 128 3.3.4 Application to brittle fracture: plane-strain fracture , determination . . . . . . . . . . . 128 toughness, 3.3.5 Application to ductile fracture, curves determination . . . . . . . . . . . . . . . . . . . . . . . 130 3.3.6 Size dependence of macroscopic fracture toughness 134 3.3.7 Application to the lower ductilebrittle transition study135 3.3.8 Fracture testing of weldments . . . . . . . . . . . . 137 4 Extensometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 4.1 Digital Speckle interferometry . . . . . . . . . . . . . . . . . 140 4.2 Digital Image Correlation . . . . . . . . . . . . . . . . . . . 141 5 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 1 2 Damage evolution laws and fracture criteria Micro-mechanics based models for ductile fracture . . . . . . . . . . 1.1 Nucleation . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.1 Nucleation criteria . . . . . . . . . . . . . . . . . . 1.1.2 Discussion . . . . . . . . . . . . . . . . . . . . . . 1.1.3 Particle at a free surface . . . . . . . . . . . . . . . 1.2 Void growth . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Basic problems . . . . . . . . . . . . . . . . . . . 1.2.2 Strain hardening effect . . . . . . . . . . . . . . . . 1.2.3 Void interaction . . . . . . . . . . . . . . . . . . . 1.3 Coalescence and fracture . . . . . . . . . . . . . . . . . . . . 2 Phenomenological damage evolution law . . . . . . . . . . . . . . . 3 Models for brittle fracture . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Scale effects and weakest link assumption . . . . . . . . . . . 3.2 Fracture criteria: Batdorf, RKR and Beremin models . . . . . 3.2.1 Batdorf and other criteria for bulk ceramic fracture .
147 147 147 147 150 151 153 153 155 155 156 157 158 158 161 161
Local approach to fracture 3.2.2 RKR criterion and Beremin model . . . . . . . . . 162 3.3 Non-local models . . . . . . . . . . . . . . . . . . . . . . . . 164 3.4 Parameter estimation . . . . . . . . . . . . . . . . . . . . . . 165 Numerical implementation . . . . . . . . . . . . . . . . . . . . . . . 167 4.1 Time integration . . . . . . . . . . . . . . . . . . . . . . . . 167 4.2 Space discretization . . . . . . . . . . . . . . . . . . . . . . 168 4.2.1 Crack tip mechanical elds . . . . . . . . . . . . . 168 4.2.2 Inuence of Element type and mechanical assumption 168 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 175 176 176 176 177 178 179 179 179 181 182 184 184 185 186 192 193 194 194 194 196 197 197 197 198 198 199 202 202 203 204 207 207 209 210 211 214 214 214
5 VI
Examples of application Geometry and size effect . . . . . . . . . . . . . . . . . . . . . . 1.1 Industrial background . . . . . . . . . . . . . . . . . . . 1.2 Experimental testing . . . . . . . . . . . . . . . . . . . . 1.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Local approach analysis . . . . . . . . . . . . . . . . . . 2 Rupture close to interfaces . . . . . . . . . . . . . . . . . . . . . 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Stress and strain state close to an interface . . . . . . . . . 2.3 Brittle fracture . . . . . . . . . . . . . . . . . . . . . . . 2.4 Ductile fracture . . . . . . . . . . . . . . . . . . . . . . . 3 Warm Pre-Stress Effect : application of a modied Beremin model 3.1 Industrial background . . . . . . . . . . . . . . . . . . . 3.2 Description of the modied Beremin model . . . . . . . . 3.3 Application of the model . . . . . . . . . . . . . . . . . . 4 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
VII Phenomenological constitutive damage models 1 Thermodynamics framework . . . . . . . . . . . . . . . . . . . 1.1 Thermodynamics Variables . . . . . . . . . . . . . . . . 1.2 State and Evolution Laws: generalized damage model . 1.3 Positivity of the Intrinsic Dissipation . . . . . . . . . . 1.4 Application to Metals, Elastomers, Concrete . . . . . . 2 Elasto-plasticity coupled with damage . . . . . . . . . . . . . . 2.1 Lemaitre damage model . . . . . . . . . . . . . . . . . 2.1.1 Thermodynamics Potentials . . . . . . . . . . 2.1.2 Constitutive Equations coupled with Damage . 2.1.3 Uniaxial tension with repeated unloadings . . 2.1.4 Identication of the Damage Law . . . . . . . 2.2 Induced damage anisotropy . . . . . . . . . . . . . . . 2.2.1 Elastic potential . . . . . . . . . . . . . . . . 2.2.2 Evolution laws . . . . . . . . . . . . . . . . . 3 Damage models for quasi-brittle materials . . . . . . . . . . . . 3.1 Marigo model . . . . . . . . . . . . . . . . . . . . . . . 3.2 Mazars model . . . . . . . . . . . . . . . . . . . . . . . 3.3 Induced damage anisotropy . . . . . . . . . . . . . . . 4 A damage model for elastomers . . . . . . . . . . . . . . . . . 5 Application to fatigue . . . . . . . . . . . . . . . . . . . . . . . 5.1 Damage threshold in fatigue . . . . . . . . . . . . . . . 5.2 Calculation of Manson-Cofn curve for metals . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
Contents 5.3 Calculation of the fatigue curve for elastomers Application to creep . . . . . . . . . . . . . . . . . . . 6.1 Secondary and Tertiary Creep of metals . . . . 6.2 A Word on the Model Identication . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 215 215 216 217 218 221 221 221 224 224 226 227 231 231 232 232 233 234 234 235 235 237 238 238 239 241 241 243 243 243 244 245 248 249 250 253 258 260 265 265 265 266 266 266 267 267
VIII Micromechanics-based constitutive models of ductile fracture 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 General introduction to the chapter . . . . . . . . . . . . . . . 1.2 Motivations from FEM void cell simulations . . . . . . . . . 1.2.1 Principes of void cell simulations . . . . . . . . . . 1.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . 2 A model for void growth: the Gurson model . . . . . . . . . . . . . . 3 Extensions of the original Gurson model . . . . . . . . . . . . . . . . 3.1 Thermoelasticity . . . . . . . . . . . . . . . . . . . . . . . 3.2 Improved modelling of the plastic ow behaviour of the matrix 3.2.1 Strain hardening . . . . . . . . . . . . . . . . . . . 3.2.2 Viscoplasticity . . . . . . . . . . . . . . . . . . . . 3.2.3 Plastic anisotropy . . . . . . . . . . . . . . . . . . 3.2.4 Kinematic hardening . . . . . . . . . . . . . . . . 3.3 Adjustments on unit cell calculations ( , , ) . . . . . . 3.4 Void shape . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Other extensions . . . . . . . . . . . . . . . . . . . . . . . . 4 Other void growth models . . . . . . . . . . . . . . . . . . . . . . . 4.1 Elliptic model . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Rousselier model . . . . . . . . . . . . . . . . . . . . . . . . 5 Void nucleation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Strain controlled nucleation . . . . . . . . . . . . . . . . . . 5.2 Stress controlled nucleation . . . . . . . . . . . . . . . . . . 5.3 A model combining particle cracking and decohesion . . . . . 6 Void coalescence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 Simple approach of coalescence: a phenomenological acceleration of the porosity rate . . . . . . . . . . . . . . . . . . . . 6.2 A micromechanical approach of coalescence as a localized plastic ow . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Linking microstructure, ow properties and fracture resistance . . . . 7.1 The ductile fracture model . . . . . . . . . . . . . . . . . . . 7.2 Ductility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3 Fracture toughness . . . . . . . . . . . . . . . . . . . . . . . 8 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX Cohesive zone models Introduction: some limitations of classical fracture theories 1.1 Crack initiation . . . . . . . . . . . . . . . . . . . 1.2 Time and space evolution of cracks . . . . . . . . 2 Cohesive zone models . . . . . . . . . . . . . . . . . . . 2.1 Barrenblatts model . . . . . . . . . . . . . . . . . 2.2 Dugdales model . . . . . . . . . . . . . . . . . . 2.3 Needlemans model . . . . . . . . . . . . . . . . .

. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
6 3 4 5
Local approach to fracture Link between CZM and fracture mechanics . . . . . . Solutions uniqueness of models involving CZM . . . . Finite element discretization . . . . . . . . . . . . . . 5.1 Cohesive elements . . . . . . . . . . . . . . . 5.2 Partition of unity methods . . . . . . . . . . . Some numerical problems related to CZM . . . . . . . 6.1 Mesh size dependence . . . . . . . . . . . . . 6.2 Choosing the appropriate CZM . . . . . . . . 6.2.1 Ductile failure: CZM and triaxiality . 6.2.2 CZM and fatigue crack growth . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268 269 270 270 272 273 274 275 275 276 277 279 279 280 280 281 282 284 286 286 289 290 291 294 294 295 295 296 297 298 299 301 302 303 304 305 306 306 308 311 311 312 312 314 316 318 319 321
7 X 1 2
3 4
5 6
8 XI 1
Numerical implementation of constitutive models Finite element method and material behavior . . . . . . . . . . . . . General methods for the implementation of constitutive equations . . 2.1 Rate formulation of the constitutive equations . . . . . . . . . 2.2 Local explicit integration . . . . . . . . . . . . . . . . . . . . 2.3 Local implicit integration . . . . . . . . . . . . . . . . . . . . Finite strain formulation . . . . . . . . . . . . . . . . . . . . . . . . Application to micromechanical models . . . . . . . . . . . . . . . . 4.1 Explicit/Implicit generic implementation . . . . . . . . . . . 4.2 Adiabatic heating . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Plastic anisotropy . . . . . . . . . . . . . . . . . . . . . . . . Application to phenomenological models: Anisotropic damage model Numerical treatment of fracture . . . . . . . . . . . . . . . . . . . . . 6.1 Convergence of the implicit integration method . . . . . . . . 6.2 Checking material failure . . . . . . . . . . . . . . . . . . . . 6.3 Choosing the element type . . . . . . . . . . . . . . . . . . . 6.4 Removing broken elements . . . . . . . . . . . . . . . . . Mesh and mesh size dependence . . . . . . . . . . . . . . . . . . . . 7.1 Axisymmetric notched bars . . . . . . . . . . . . . . . . . . 7.2 CT specimens . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3 Cracks under small scale yielding (SSY) . . . . . . . . . . . . 7.4 Role of element interpolation . . . . . . . . . . . . . . . . . . 7.5 Role of symmetry . . . . . . . . . . . . . . . . . . . . . . . . 7.6 Crack paths . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.7 Element shape and crack tip . . . . . . . . . . . . . . . . . . 7.8 3D calculations . . . . . . . . . . . . . . . . . . . . . . . . . 7.9 Mesh size as a material parameter . . . . . . . . . . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Localization and regularization Strain localization criteria in elastoviscoplastic solids . . . . . . . . . 1.1 The localization phenomenon: a rst insight . . . . . . . . . . 1.1.1 Onedimensional example . . . . . . . . . . . . . 1.1.2 Consequences on the modeling of real-life structures 1.2 Bifurcation modes in elastoplastic solids . . . . . . . . . . . . 1.2.1 Formulation of the boundary value problem . . . . 1.2.2 Loss of uniqueness; general bifurcation modes . . . 1.2.3 Existence of velocity gradient discontinuities . . . .
Contents 1.2.4 Bifurcation analysis in elastoplasticity . . . . . . . 1.2.5 Stability . . . . . . . . . . . . . . . . . . . . . . . 1.2.6 Localization criteria . . . . . . . . . . . . . . . . . 1.3 Application to tensile tests in isotropic incompressible and compressible associated elastoplastic materials . . . . . . . . 1.3.1 theory: tensile test . . . . . . . . . . . . . . . . 1.3.2 Elliptic potential . . . . . . . . . . . . . . . . . . . 1.4 Rate-dependent constitutive behavior . . . . . . . . . . . . . 1.5 Strain localization indicators in numerical simulations at nite strains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 The need for characteristic lengths . . . . . . . . . . . . . . . Regularization methods . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 A tool to explore nonlocality . . . . . . . . . . . . . . . . . . 2.1.1 Generalized standard materials . . . . . . . . . . . 2.1.2 Introduction of a dissipation potential . . . . . . . . 2.1.3 Extension to the scale of the structure . . . . . . . . 2.1.4 Time integration: a minimum principle . . . . . . . 2.2 Spatial regularization of the mechanical variables . . . . . . . 2.2.1 Regularization operators . . . . . . . . . . . . . . . 2.2.2 An attempt toward a classication . . . . . . . . . . 2.2.3 Numerical application . . . . . . . . . . . . . . . . 2.3 Enhancement of the kinematics: the micromorphic continuum 2.3.1 Balance and constitutive equations . . . . . . . . . 2.3.2 Application to strain localization phenomena . . . . 2.3.3 Finite element simulations with the micromorphic continuum . . . . . . . . . . . . . . . . . . . . . . 2.4 Gradient of internal variables approach . . . . . . . . . . . . 2.4.1 Introduction of the gradient of internal variables . . 2.4.2 Relation with gradient models of the literature . . . Case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Numerical application of gradient models to ductile plasticity 3.2 Deformation and fracture of aluminum foams . . . . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7 323 327 327 328 328 330 331 334 336 338 339 339 340 341 342 343 343 344 345 347 347 351 354 355 355 358 359 359 361 367 373 373 373 373 374 375 375 376 377 377 378 380 380 382 382
4 XII
Model identication Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data input for identication: global and local . . . . . . . . . . . . 2.1 Global data input . . . . . . . . . . . . . . . . . . . . . . . 2.2 Local data input . . . . . . . . . . . . . . . . . . . . . . . . 3 Automatic identication . . . . . . . . . . . . . . . . . . . . . . . 3.1 Building a cost function . . . . . . . . . . . . . . . . . . . 3.2 Search methods . . . . . . . . . . . . . . . . . . . . . . . . 4 Guidelines for identication . . . . . . . . . . . . . . . . . . . . . 4.1 General guidelines . . . . . . . . . . . . . . . . . . . . . . 4.2 Guidelines for the Gurson/Rousselier models . . . . . . . . 4.3 Guidelines for the phenomenological models . . . . . . . . 4.3.1 Guidelines for the Lemaitres type damage models 4.3.2 Guidelines for the quasi-brittle damage models . . 5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2
. . . . . . . . . . . . . .
8 5.1
Local approach to fracture Identication of the nucleation kinetics using microstructural observations . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 Hydrided Zircaloy4 alloy . . . . . . . . . . . . . . 5.1.2 Embrittled duplex stainless steel . . . . . . . . . . 5.1.3 16MND5 (A508) steel . . . . . . . . . . . . . . . . 5.2 Different models Same t . . . . . . . . . . . . . . . . . . 5.3 Aluminum plates: a model adjustement for cracking only . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
382 382 385 386 386 388 392 395 395 396 396 397 397 398 399 400 400 401 401 402 402 405 406 407 410 411 411 412 412 413 413 414 415 416 416 417 418 419 420 422
XIII Applications of damage models 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Spinning Cylinder Test . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Description of the NESC1 PTS experiment . . . . . . . . . . 2.2 Analysis of the experiment with respect to Local Approach . . 2.2.1 Parameters tting . . . . . . . . . . . . . . . . . . 2.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Conclusions of the study . . . . . . . . . . . . . . 3 Evaluation of the Charpy Transition Curve . . . . . . . . . . . . . . . 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Predictions in the brittle domain . . . . . . . . . . . . . . . . 3.3.1 Constitutive equations . . . . . . . . . . . . . . . . 3.3.2 Numerical Modeling . . . . . . . . . . . . . . . . . 3.4 Predictions in the ductile to brittle transition domain . . . . . 3.4.1 Constitutive equations . . . . . . . . . . . . . . . . 3.4.2 Numerical Modeling . . . . . . . . . . . . . . . . . 3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Ductile rupture of heterogeneous materials . . . . . . . . . . . . . . . 4.1 Presentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Modeling material heterogeneities . . . . . . . . . . . . . . . 4.3 Material and nature of the heterogeneities . . . . . . . . . . . 4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.1 Scatter and reduction of ductility . . . . . . . . . . 4.4.2 Notched tensile bars: comparison with experiments 4.4.3 CTspecimens: comparison with experiments . . . 5 Ductile tearing of aluminum sheets . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Test samples . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Fracture surfaces . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Finite element simulation . . . . . . . . . . . . . . . . . . . . 5.5 Modeling large crack advance . . . . . . . . . . . . . . . . . 6 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Preface PREFACE
Models for the prediction of failure of cracked structures and components were rst proposed more that 50 years ago, in the framework of linear fracture mechanics, and then extended to plasticity and viscoplasticity through nonlinear fracture mechanics about 30 years ago. Although these methods, which derive from a global approach to fracture, are extremely useful, frequently used and still improving, they have been proved to suffer from several limitations; at the same time, more and more industrial situations have been understood as needing new methodologies and tools to be analyzed realistically and mastered practically. This need for new methods, combined with the development of physically-based models of mechanical behavior and micromechanical treatments, has given rise in the early 80s to the so-called local approach to fracture. Unlike the global treatment, which makes the fracture resistance of a component mainly depend on a single global parameter, whatever the damage and deformation mechanisms of the specic material under study, the local approach emphasizes these material specicities: it combines a detailed experimental analysis of the considered materials and of their specic damage mechanisms, a realistic modeling of these mechanisms and the implementation of these models into a numerical simulation of the response of the structural components under investigation. This approach has known a remarkable development during the last two decades and it has achieved success in a number of complex cases, which were out of reach of the classical global approach: welding, assembling, thermo-mechanical and non-proportional loading, etc. Nevertheless, despite an intensive research activity, especially in Europe (France and Germany), USA and Japan, the organization of many conferences and a profuse scientic and technical literature, no really detailed, thorough and pedagogically conceived and realized book was still available in this essential eld of modern mechanics of materials and structures. This gap is now fortunately and successfully lled with this collective monograph. Actually, even though it does not deal with some very specic domains such as dynamic fracture, stress-induced corrosion,. . . and it lays the main (but not exclusive) stress on metals and alloys, this book offers an outstanding and extensive description of the current state of the art of the local approach to fracture. All the main physical and mechanical, conceptual, methodological, numerical and practical ingredients of this approach are presented in detail; both phenomenological and micromechanical damage models, for brittle as well as for ductile fracture, under monotonous as well as fatigue loading, are carefully explained, illustrated and discussed. Already classical as well later results are reported, including recent developments in the eld of ductile fracture, strain localization, model identication or numerical implementation. Special attention has been paid to the presentation of experimental tests and to the discussion of practical cases and industrial applications. Penetrating suggestions are also made for future directions of research in this eld. It is not really surprising for this fortunate initiative to originate from France: since the pioneer initial contributions of Beremin, with special support of the french electronuclear industry and of CNRS (e.g. through its Large strains and Damage research network in the 80s), a vigorous research activity has been continuously developed in several french teams and labs on various aspects of this approach, in cooperation with more and more groups from other countries. Although this book is connected with courses of a Summer School which are delivered in french, it is a very good thing that it has been written in english, in view of its direct international diffusion. It is also very comforting to notice that the editorial team, which
10
expresses different skills and feelings within the same overall scientic philosophy, gives concrete expression to the baton passed from the founders, represented here by my old accomplice and friend, Andr Pineau, to the new generation: I cannot e ignore the pleasant fact that many members of this young and talented group, led by Jacques Besson, the inspirer of this venture, were our common graduate students of the Mechanics and Materials DEA not so long ago... No doubt that this successful and timely publication promises to be considered as a reference book for a long time by a number of graduate students, research engineers and researchers !
Andr Zaoui e Laboratoire de M canique des Solides Ecole Polytechnique e Professor at Ecole Polytechnique Member of the French Academy of Engineering Associate Member of the French Academy of Sciences
The authors THE AUTHORS
11
Clotilde Berdin graduated from Universit Paris XI (Master) and obtained a PhD in e Material Sciences and Engineering at Ecole des Mines de Paris in 1993. She is now working at Ecole Centrale Paris (Lab. Mecanique des sols structures et mat riaux) as e an assistant professor. Her main research activity is the structural integrity assessment based on physical damage mechanisms. Jacques Besson, graduated from Ecole des Mines de Paris in 1986 and obtained a PhD in Material Sciences and Engineering in 1990. He is Charg de Recherche for e CNRS working at Ecole des Mines de Paris. His main research activity is the numerical simulation of rupture and he has interest in powder metallurgy, nite element method, object oriented programming in Engineering. He is the author of 50 peer-reviewed publications. St phane Bugat, graduated from Ecole des Mines de Paris in 1992 and obtained a e PhD in Material Sciences and Engineering in 2000. Since 2001, he is research engineer at EDF R&D, MMC Department. His research topics are: fracture mechanics, multiscale simulation, lifetime assessment of pressure vessels and internals. Rodrigue Desmorat is Professor at Ecole Normale Sup rieure de Cachan (ENS e Cachan) and is head of the Civil Engineering Department. He obtained is PhD in 1996 and a Habilitation a Diriger des Recherches in 2000 (both in mechanics and at ` Universit Paris VI). He is a former student of ENS-Cachan (1988-1992). He obtained e a french agr gation in mechanics in 1991. His research topics are: damage, fatigue, e failure of materials and structures Fr d ric Feyel graduated from Ecole des Mines de Paris and obtained a PhD in e e Material Sciences and Engineering in 1998. He is now working at ONERA , structure and damage mechanics Dept as a senior research scientist. His main research activities are: Multiscale modeling , Parallel computing , Non linear behaviors. Samuel Forest, graduated from Ecole des Mines de Paris in 1992 and obtained a PhD in Material Sciences and Engineering in 1996. He is Charg de Recherche e for CNRS working at Ecole des Mines of Paris His research topics are: Mechanics of generalized continua, Homogenization methods, Microstructural Mechanics. Eric Lorentz graduated from Ecole Polytechnique (1992) and Ecole Nationale des Ponts et Chauss s (1994). He obtained a PhD in Solid Mechanics (1999, Univ. Paris e 6). He works at Electricit de France R&D as head of the section Theoretical and e Applied Mechanics . His scientic work is dedicated to the numerical simulation of fracture processes to assess the safety of power plant components, with special focus on continuum nonlocal theories. Eric Maire graduated from the Institut National des Sciences Appliqu es (INSA) e in Lyon in 1990 as a materials science engineer. He obtained his Ph.D. at INSA for a thesis on the properties of metal matrix composites. After a 1 year post-doctoral visit at Mc Master University in Hamilton, Ontario canada, he was appointed Research Associate by the CNRS, in the GEMPPM laboratory, INSA de Lyon, France. He is currently involved in studying damage in heterogeneous materials using X ray tomography. E. Maire has been elected FEMS lecturer for 2002-2003 by the Federation of European Materials Societies. Thomas Pardoen is Belgian, born in the U.S. in 1971. He graduated from the Universit catholique de Louvain in 1994 as a materials science engineer, in 1996 with e a degree in philosophy, and was awarded a Ph.D. from the same university in 1998 for a the thesis on the micromechanics of ductile fracture. After a 2 years post-doctoral visit at Harvard University in the solids mechanics group of the Division of Engineering
12
and Applied Sciences, T. Pardoen joined the faculty of the Universit catholique de e Louvain in the Department of materials and processes sciences in 2000. He is currently supervising researches and teaching in the eld of the mechanics and micromechanics of materials and microsystems from both an experimental and modelling point of view. T. Pardoen has been elected FEMS lecturer for 2002-2003 by the Federation of European Materials Societies. He is a member of the Research Center in Micro and Nanoscopic Materials and Electronic Devices of the Universit catholique de Louvain. e Andr Pineau, graduated from Ecole des Mines of Nancy in 1965, obtained a PhD e in Physical Sciences in 1969 and is member of the french Academy of Engineering. He is Professor at Ecole des Mines of Paris since 1978 where he is teaching physical and mechanical metallurgy and is in charge of the Undergraduate students of the Material Sciences and Engineering department. His main research activities are: physical and mechanical metallurgy, including plasticity, fatigue and local approach to fracture. He is the author of more than 250 peer-reviewed publications dealing with different types of materials including steels, Aluminium, Zirconium and Nickel-based alloys. Benot Tanguy graduated from Ecole Centrale de Nantes in 1996 and obtained a PhD in Material Sciences and Engineering at Ecole des Mines de Paris in 2001. He is Charg de recherche working at Ecole des Mines de Paris. His main research e topics are: fracture mechanics, behavior and fracture modelling of metallic materials, structural integrity assessment.
Acknowledgements ACKNOWLEDGEMENTS
13
This book was rst written as a course for the Summer School M canique de e lEndommagement et Approche Locale de la Rupture (MEALOR) organized in July 2004. The authors want to acknowledge the nancial support of CNRS, ESIS (C. SainteCatherine), EdF (G. Rousselier) and Framatome (P. Gilles).
14
Chapter I
INTRODUCTION
by A. Pineau and J. Besson
The assessment of the mechanical integrity of any awed mechanical structure requires the development of approaches and methodologies which can deal not only with simple situations, such as small-scale yielding (SSY) under pure mode I isothermal loading, but also with much more complex situations, including large scale plasticity, mixedmode cracking, and non-isothermal loading. Two types of approaches have been developed for this purpose, as schematically depicted on gure I.1. The rst approach, referred to as the global approach is essentially based on the extensive development over the past few decades of linear elastic fracture mechanics (LEFM) and then no linear mechanics (NLFM). In this approach which was historically the rst one to be developed, it is assumed that the fracture resistance can be measured in terms of a single parameter, such as , or crack tip opening displacement (CTOD). Rules and standards uniquely based on the mechanical conditions of test conditions have been established for valid fracture toughness measurements, without paying attention to the failure micromechanisms. These rules are summarized in chapter IV. This global approach is extremely useful and absolutely necessary, but has also a number of limitations, in particular when large scale yielding conditions are met or when non isothermal conditions are encountered. Another limitation is the size effect which is usually observed when structural ferritic steels are tested in the brittle domain and in the ductile-to-brittle transition regime. It is now well established that fracture toughness is specimen size dependent even in the lower shelf regime of the ductileto-brittle transition. The specimen size requirements in an NLFM test to measure a valid fracture toughness, is also problematic. This raises the important problem related to the transferability of laboratory test results to large components tested under in-service conditions. The second approach, called the local approach was developed later in the 80s, although it can be considered that the models proposed by McClintock (McClintock, 1963; Mc Clintock, 1968) were already local approaches devoted to fatigue failure. In this methodology the modeling of fracture toughness is based on local fracture criteria established most often from tests on volume elements, in particular notched specimens. Then these criteria are applied to the crack-tip situation.The development of this methodology requires that, at least, two conditions are fullled, as illustrated in two conferences devoted to this topic (NED 1987 ; Euromech-Mecamat 1996): (i) Micromechanistically based models must be established; (ii) A perfect knowledge of the stress-strain eld ahead of a stationary and a propagating crack is required. This has been made possible, thanks to the advent of analytical and numerical solutions.

15
16
In particular, the rapid development of nite element methods (FEM) has largely contributed to the advance of this methodology. There has been considerable research on local approaches to fracture over the past two decades. These approaches either assume conventional material behavior supplemented by models of local failure processes, which is the situation for brittle fracture (see chapter V), or use continuum damage mechanics (CDM) involving coupled constitutive equations with a softening effect due to damage, which is the case for ductile fracture, as explained in detail in chapters VII and VIII. In this book a state of the art of the so-called Local Approach to Fracture is presented. Actually, the global and the local approaches to fracture are more complementary than contradictory. In particular the local approach to fracture, which necessitates detailed numerical FEM calculations, must be essentially used to deal with difcult situations involving complex loading conditions.
1960
1970
1980
1990
2000
Linear Elastic Fracture Mechanics Non Linear Fracture Mechanics Local Approach to Fracture
1981, ICF5 1992 AIME Seminar 1999 PVP 1986 1 Seminar 1996 1 Seminar

Figure I.1: Historical landmarks for the Local Approach to Fracture
1 GLOBAL APPROACH TO FRACTURE

1.1 SMALL SCALE YIELDING LARGE SCALE YIELDING
The distinction between small scale yielding (SSY) and large scale yielding (LSY) is important with respect to the modeling of fracture (gure I.2). SSY corresponds to case where the size of the plastic zone is much smaller than the size of the specimen or structure. In that case the overall behavior of the structure remains linear. LSY corresponds to the opposite case: a large portion of the specimen is plastically yielded and the overall behavior is non linear. SSY can be analyzed using linear elastic fracture mechanics (see 1.2); LSY is analyzed using NLFM (see 1.3). Most of the testing standards require the specimens to remain under SSY to be valid. These requirements are however more and more difcult to meet as materials fracture and
I. Introduction
17
large scale yielding
small scale yielding
Figure I.2: Overall behavior of a cracked structure. plastic zone is schematically shown in gray.
: displacement,
Figure I.3: Energy release rate in the linear case. toughness properties are continuously improved. Using the local approach to fracture is particularly interesting in the case of LSY.
1.2 LINEAR ELASTIC FRACTURE MECHANICS (LEFM)
In the case of a structure whose global response is linear, the link between the global behavior and crack advance can be obtained using the variation of the stiffness or compliance, , of the structure. This variation is assumed to be caused by crack extension only. In that case the variation of the stored energy corresponds to the energy dissipated by cracking. The stored elastic energy is given by: where is the prescribed displacement. Considering a constant value for (gure I.3) the energy released by the structure, , for an increase of the crack surface, is equal to:
%
corresponds to the energy release rate. The value of can be related to the applied loading using the expression of the stress tensor close to the crack tip. Under mixed mode loading (I, II and III) stresses
!
' &
# $"
' &
: force. The
! # ' 1& ' ) 0
!
# " (& " & !
(1)
@
)
#! ( ) I" ) ( ) 3 ) ( H) )
@B A9
# # # # #
11 42
) )
)
#! $"7)
) )
)
) #! $" ) )
#! $" )
#! #! ( ) & $% ) $" ) & #! ) 53 ) $% ) & 3 ) ) #! #! ( ') & $% ) $" (0) & ) ) ) ) & #! #! (') '$% ) $"
)
Displacements are given by:
(gure I.4) are given by:
18
crack
Figure I.4: Local coordinate system at the crack tip.
plane strain plane stress
(11)
(10)
(12) (8) (9) (7) (6) (5) (4) (3) (2)
G

G
)
)
G 2
#! ( ) $% )
G 2
& @ @ & 9 # P Q G #! $" ) E C ) ) ) E ) C #! ) $% ) E D) C # $!" ) E ) C E C ) FD)

) #
6 86 6 1 6 4
1 42
11 2
I. Introduction The corresponding energy release rate is given by:
19
The crack can propagate if the released energy equals the energy corresponding to the creation of two free surfaces ; that is , where is the surface energy. corresponds to the critical energy release rate which is a material characteristic. The corresponding mode I plane strain fracture toughness is computed using equation 13:
Typical values for are of the order of magnitude of 1 J/m which corresponds to a fracture toughness MPa (with GPa and ). This value is indeed very small and using equation 17 is only valid for very brittle materials (e.g. glass, ice, . . . ). In practice the actual value for is (much) larger than . In many materials, dissipation at the crack tip takes place in a given volume (so called process zone). This dissipation process originates from micro-cracking or phase transformation in though ceramics or ceramic matrix composites. In metals, dissipation corresponds to plasticity around the crack tip. The above calculation of , based on the compliance variation with crack advance (equation 1), is valid for elastic materials only. As mentioned above plasticity occurs at the crack tip. As long as small scale yielding prevails, it is possible to apply Irwin plastic zone correction (Irwin, 1957). Using this correction a crack of length in an elastoplastic medium is considered as equivalent to an effective crack of length in a purely elastic medium. is given by, under plane stress conditions:
where is the yield strength of the material. Using this model, the plastic zone size, , is equal to . The safety analysis can then be performed as for an elastic medium using instead of . It should also be added that under plane strain conditions the plastic zone size is three times smaller than that corresponding to plane stress conditions, i.e. .
It may be useful to express the stresses in the cylindrical reference frame (the terms remain unchanged):
( H)
)
&
"#B
#
"
3

&
B )
!
@ #! I"
)
!"
# ("
( 1
)
"
@
!
&
)& ) #! ( H) $"
" #
E #
1

)
!
# ( '&% $B
1
@
1 #! $%
& #

# "
D) )
D) )
D) )
"
1
1 D)
# # # # !
(13)

(14) (15) (16)
66 82 1 # 1
0 2 1
!0
(17)
(18)
20

(b)
Figure I.5: Energy release rate in the non linear case.

1.3 NON LINEAR FRACTURE MECHANICS (NLFM)
1.3.1
The -Integral
Path-independent integrals have been introduced into fracture mechanics by Cherepanov (1967) and Rice (1968). Budiansky and Rice (1973) also showed that . This relation this -integral is identical with the energy release rate, i.e.: has become a common technique to calculate stress-intensity factors in linear elastic fracture mechanics (LEFM). The -integral is dened as a contour integral around the crack tip:
! #
where
is an arbitrary contour around the tip of the crack, is the unit vector normal to . is the stress acting on the contour: . Because of its path independence (Rice, 1968), the -integral can be calculated in the remote eld and characterizes also the near-tip situation. This establishes its role as a fracture parameter. However it should be noted that the path independence does only hold if the following conditions are met: (i) time-independent processes, no body forces ( ), (ii) small strains, (iii) homogeneous hyper-elastic material ( ) , (iv) plane stress-strain eld, i.e. no dependence on and (v) straight and stress-free crack borders parallel to .
1.3.2
as stress-intensity factor: the HRR eld
Besides its identication as energy release rate for hyperelastic materials in non linear fracture mechanics (NLFM), also plays the role of an intensity factor like in the case of LEFM. Hutchinson (1968a,b) and Rice and Rosengren (1968) derived the singular stress-strain elds at a crack tip in a power-law hardening material (called the HRR-eld). The hardening law is given by:
(21)
"!
# # # ( #
) 0( %
' (# %&
#
(19)
(20)
I. Introduction
21
ds
crack
Figure I.6: The where .

"
contour integral.
is the reference yield stress, the reference yield strain which is equal to is a material constant. Stresses are given by:
Deformations are given by:
Displacements are given by:
The functions , et as well as depend on the hardening exponent (Shih, 1983). depends weakly on . It should be emphasized that the HRR eld is valid for a non linear elastic material. Considering that equation 21 represents a plastic behavior is only valid for monotonic loading. The ratio is proportional to the crack tip opening displacement, CTOD, in the HRR theory. The CTOD is given by:
Considering the similarity between the roles played by the stress intensity factor and by the -integral (i.e. link with the energy release rate and determination of the
1.4
STRESS AND
FACTOR

where strongly depends on the work hardening exponent, Under plane strain conditions ( for conditions ( for ).
(McMeeking, 1977). ). Under plane stress
&
# #

& B ' B ' $B ) # # # # # # % "B % $B $B &
(
( H
# )
& 1#
( H
% % ' & # % '
%% ' %
%
) ' % % ' % #
&

&
' % ) )
&
(22)
(23)
(24)
(25)
22
stress/strain elds), a fracture criterion for NLFM (analogous to ) was proposed as: where is assumed to be a material characteristic. However experiments on various specimen geometries and specimen size have demonstrated that this relation does not hold (see below 2.2.1). One reason for this problem is that the above mentioned solutions for stress (equation 4 or equation 22) only describe the near tip eld and consider only the rst term of a Taylor expansion. In the linear case, the second term in the Taylor expansion corresponds to the so called stress which acts in the direction parallel to the crack advance direction. The mode I stress eld becomes:
where is the Kronecker delta. The function is given by equation 4. A positive (resp. negative) stress correspond to higher (resp. lower) principal stresses and a more conned (resp. less conned) plasticity. Therefore will favor brittle fracture. The stress effect is often quantied using the non dimensional factor:
where is the crack length. In the case of NLFM, the factor was introduced by ODowd and Shih (1991, 1992) to account for difference between the HRR stress eld and nite element results. The factor corresponds to an additional hydrostatic pressure. The modied stress eld is obtained as: (28)
where
is given by equationequation 22. In practice
is computed as: (29)
where is the stress obtained using nite element calculations using ne rened for . Positive values of meshes. is usually evaluated at a distance will favor brittle fracture and enhance ductile void growth. 2 LOCAL APPROACH TO FRACTURE Local approaches and micromechanical modeling of damage and fracture (Pineau, 1981) have found increasing interest. The main advantage, compared with classical fracture mechanics, is that, in principle, the parameters of these models depend only on the material and not on the geometry. Thus these approaches guarantee a better transferability from specimens to structures. They can still be used when only a small amount of the material is available which is the situation met with the standards applied in LEFM and NLFM. The identication and determination of the micromechanical parameters require a hybrid methodology of combined testing and numerical simulation.
2.1 MAIN PRINCIPLES OF THE LOCAL APPROACH TO FRACTURE
The terminology local approach to fracture originates from the very principles of the methodology. Stresses, strains, damage and other state variables are estimated
& &
& %
%

#
) #

(26)

&
(27)
I. Introduction
23
locally in parts which failure has to be assessed. This usually corresponds to locations of stress and strain concentrations such as at the root of a notch or at a crack tip. One of the rst application of the methodology is the RKR model for brittle fracture of ferritic steels (Ritchie et al., 1973). In this model, the stressstrain eld is evaluated using the HRR model. Failure occurs when the maximum principal stress, , exceeds a critical value, over a critical distance (see gure I.7). Many of the ingredients of the local approach to fracture are already present in this model: The precise evaluation of the local material loading: In 1973, stresses and strains were evaluated using the HRR model. Nowadays the nite element method is very often preferred to account in particular for crack blunting effect. A failure criterion: The RKR model was dealing with brittle fracture and proposed a simple criterion. Nowadays, the method is applied to brittle, ductile and creep fracture. Finding an appropriate failure criterion is still an active research eld. Coupled models (see e.g. (Gurson, 1977) or (Lemaitre, 1992)) integrating the interaction between softening due to damage growth and plastic hardening are now widely used. Need for a characteristic length: The need for a characteristic length is probably the most difcult problem to cope with for both brittle failure (see e.g. (Kroon and Faleskog, 2002)) and ductile failure (see e.g. (Rousselier, 1987; Liu et al., 1994; Lorentz and Andrieux, 1999)). This research eld is very active in particular with respect to the development of non local constitutive models 1 .
HRR eld
Figure I.7: Stress eld at failure initiation using the RKR model. The principal stress is higher that the critical stress over a distance larger than . The application of the local approach is based on four main steps (see gure I.8): Experimentation Pertinent experiments must be designed to characterize the material and its failure behavior. This indeed involves mechanical tests but also the observation of the material before and after failure. Quantication of damage is of great importance. Newly available in situ experiments may be used. Modeling A material model must be dened to represent the experimental data. The model should be physically based: i.e. designed to represent physical phenomena (i.e. void growth, material cracking, oxidation,. . . ) that have been observed experimentally. Using such a model possibly allows to make
1 The terminology non local does not mean global in this context. It implies that the material behavior at a given location, , does not only depend on the material state variables at but also on the material state around (e.g. in the sphere dened by , where is a material characteristic length.)
!

24
Local approach to fracture extrapolations from the experimental data base. Purely phenomenological models should be only used to interpolate experimental data.
Identication The material parameters introduced by the employed model have to be identied. This is usually done by comparing the experiments with simulations. This raises the problem of the strategy which has to be used to determine these parameters and of the uniqueness of their values. Simulation Finally, the material model must be used to perform simulation of tests parts and eventually actual industrial components. This implies the use of simulation tools (e.g. the nite element method) in which the models have to be implemented.
I: EXPERIMENTS mechanical tests microstructural observations determination of the pertinent micromechanisms
II: MODELING constitutive equations damage modeling
III: IDENTIFICATION
IV: SIMULATIONS OF LABORATORY SPECIMENS
IV : SIMULATION OF STRUCTURAL COMPONENTS Figure I.8: Flow chart showing the methodology of local approaches. These different steps can be briey illustrated in the case of the Beremin model for brittle fracture (Beremin, 1983): I Experiments: Mechanical tests are carried out on notched specimens with various notch radii in order to change the stress triaxiality ratio. The mechanical analysis of these tests is relatively straightforward. A large volume of material is tested so that they can be regarded as volume elements. An active effort has been made to standardize these tests (Schwalbe, 1998) (see also chapter IV). However the standardization level remains much lower compared to that reached for or measurement (i.e. global approach). Another experimental strategy relies upon the use of fracture toughness measurements made on cracked specimens. This procedure has also been discussed in some detail by Ruggieri et al. (2000). II Model: The model (Beremin, 1983) introduces the Weibull stress, , as a probabilistic fracture parameter. A key feature of the Beremin approach is that
I. Introduction
25
follows a two parameter Weibull distribution is terms of a Weibull shape factor, , and a scale parameter, (see chapter V for details). The Weibull stress is considered to be the crack driving force which can be used in transferring the results on notched specimens to cracked specimens and to cracked structures (see e.g. (Ruggieri and Dodds Jr., 1996; Ruggieri et al., 1998)). The control feature in this methodology adopts the simple axiom that unstable cleavage crack propagation occurs at a critical value of the Weibull stress corresponding to a given probability to failure. III Identication: The Weibull modulus, , plays a major role in the process to correlate the effects of constraint loss for varying crack congurations and loading modes (tension, bending. . . ). Therefore, robust schemes to calibrate become a key element in fracture assessment the Weibull parameter procedures based on Weibull stress (Bakker and Koers, 1991; Moinereau et al., 1996). A similar situation arises for ductile fracture. IV Simulations: When implemented in a nite element (FE) code, the Beremin model predicts the variation of the Weibull with the applied macroscopic load to dene conditions leading to (local) material failure. It then becomes possible to predict global parameters at fracture initiation such as or . In the following, two examples illustrating the benet from the local approach methodology are briey presented. Both examples deal with brittle cleavage fracture.
2.2 EXAMPLES OF FRACTURE BEHAVIOR THAT THE LOCAL APPROACH CAN EXPLAIN
2.2.1
2
Size and geometry effects
An experimental study by Sumpter and Forbes (1992) and Sumpter (1993) on SENB and CCP 3 specimens has evidenced a strong effect of the ratio ( : crack length, specimen width) on the value of the integral at crack initiation. The study was carried out at C on an A36 steel (brittle failure). As is increased in SENB specimens, strongly decreases as shown on gure I.9. These experimental results clearly showed that using as a failure criterion was not appropriate. The use of a two parameters global failure criterion using and was then proposed by ODowd and Shih (1991, 1992). It is then possible to dene a unique failure locus as shown on gure I.10. The observed dispersion is inherent to brittle failure. The observed failure behavior can be qualitatively explained by the RKR model (Ritchie et al., 1973). The main result is that the local stress triaxiality varies depending on the crack depth. For low values of the crack length, the factor is negative so that the local stress triaxiality is less than for long cracks. A higher load level (i.e. a higher value of ) must be applied to locally reach the critical stress, . Similar considerations explain the differences between SENB and CCP specimens. More quantitatively, the Beremin model was used to reproduce these trends together with the experimental scatter (Bauvineau, 1996) (see gure I.9).
2 SENB: 3 CCP:
Single Edge Notch Bending center crack panel
26
300 250
data Beremin model t
200 (kJ/m )

150 100 50 0 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Figure I.9: Evolution of as a function of for SENB specimens for a type A36 steel at C (Sumpter and Forbes, 1992; Sumpter, 1993). Bars indicate the 10%90% failure prediction obtained with the Beremin model (Bauvineau, 1996).
200 (kJ/m )
SENB
150
100
50
0 -1.5
-1.0
-0.5
Figure I.10: Evolution of as a function of for SENB and CCP specimens for a type A36 steel at C (Sumpter and Forbes, 1992; Sumpter, 1993; ODowd et al., 1995).
& $B # & # $B
CCP CCP

% $B #
250
# B $B
B
0.0
I. Introduction
27
(MPa
( C)
Figure I.11: Principle of the warm-prestress effect. The gray area indicates the isothermal rupture toughness range.
2.2.2
Warm pre-stress effect
The Weibull statistical model has also been successfully applied to warm pre-stress effect (see e.g. (Beremin, 1981; Roos et al., 1998; Lefevre et al., 2002)). This effect plays a key role in the assessment of nuclear pressure vessel integrity during a pressurized thermal shock (PTS), for example a loss of coolant accident (LOCA). Two thermomechanical cycles are shown in gure I.11 to illustrate this phenomenon. In both cases, a precracked specimen is rst loaded in the upper shelf of the transition curve at a temperature and at value of or which is below that corresponding to fracture ( or ). It is well to remember that the fracture toughness transition curve is always determined under isothermal conditions. In the rst type of cycle, the specimen is slowly cooled at constant load down to a lower temperature which corresponds to the lower shelf regime. The transition curve can be crossed without observing any fracture. This is simply due to the fact that plastic deformation is necessary to initiate cleavage fracture (see chapter II). As the yield strength of the material increases when the temperature is decreased no plastic deformation takes place during the cooling process, in particular when the transition curve is crossed. At temperature , a higher value of the stress intensity factor has to be applied to initiate cleavage fracture. This cycle, called loadcoolfracture (LCF), produces a signicant increase of the fracture toughness, compared to the value determined under isothermal conditions at temperature, . The second type of thermomechanical loading (load unloadcoolfracture or LUCF) is a little more complex since it involves a rst loading, followed by unloading at temperature , then the cooling step and the nal test to fracture. Here again an apparent increase of the fracture toughness is observed at temperature . The results of tests performed on a pressure vessel steel which exhibited a nil ductility temperature (NDT) of C are shown in gure I.12 and I.13(Roos et al., 1998). The test results corresponding to the LCF and LUCF cycles are given in gure I.12 and gure I.13, respectively. In both cases, signicant increases in apparent

LCF
LUCF
28
fracture toughness are observed. Two main factors contribute to this increase (see (Beremin, 1981)): (i) crack tip blunting effect, and (ii) compressive residual stress for the LUCF cycle. The results of the application of the Beremin model accounts reasonably well for the variations of fracture toughness with temperature under isothermal conditions in the lower shelf regime (not shown). Moreover it is observed that the model accounts also for the LCF test results, but it tends to underestimate the fracture toughness results obtained for the LUCF loading path. This might be related to the effect of large plastic strains which tend to increase the apparent brittle fracture stress as indicated elsewhere (chapter V). In the numerical of the WPS effect, the importance of the constitutive equations, in particular the relative importance of kinematic and isotropic hardening should also be mentioned. For the details on the application of the local approach to the WPS effect, see also chapter VI and the recent publication by Lefevre et al. (2002). 200 180 160 140 120 100 80
LCF
MPa
Temp rature ( C) e
Figure I.12: Results of WPS experiments for LCF cycles Comparison based on the Beremin model (see (Roos et al., 1998) for details).
3 ORGANIZATION OF THE BOOK The organization of the book follows the different aspect of the local approach to fracture. Some basic features of global approaches have been presented in this introduction. Physics of damage and experimentation Chapter II recalls the physical mechanisms leading to progressive damage up to failure. Chapters III and IV deal with experiments; chapter III presents the techniques used to determine and quantitatively measure damage; chapter IV deals with mechanical testing. Failure criteria Chapter V presents models for both brittle and ductile fracture which can be used as post-processors of standard elastoplastic calculations (e.g.
B
B B
B B )
60 40 20 0 experimental scatterband LCF exp. LCF calc.
I. Introduction
29
200 180 160 140 120 100 80
LUCF
MPa
Temp rature ( C) e
Figure I.13: Results of WPS experiments LUCF cycles. Comparison based on the Beremin model (see (Roos et al., 1998) for details). Beremin model). Examples of applications are presented in chapter VI. Material constitutive equations for damage Chapter VII and chapter VIII deal with constitutive equations allowing for a coupling between plastic hardening and damage softening (e.g. Lemaitre, Rousselier or Gurson models). Chapter IX deals with cohesive zone models which describe interfacial cracks. Computational fracture mechanics Chapter X deals with the numerical implementation of the different constitutive models. Chapter XI presents the numerical problems encountered while using these models (e.g. localization) as well as some remedies. The problems related to model identication are treated in chapter XII. Examples of applications are presented in chapter XIII.
B
B
B B
B
B ) B
60 40 20 0 experimental scatterband LUCF exp. LUCF calc.
30 4 BIBLIOGRAPHY
Bakker, A., Koers, R., 1991. Prediction of cleavage fracture events in the brittle ductile transition region of a ferritic steel. In: Defect Assessment in Components Fundamentals and Applications. Mechanical Engineering Publications, London. Bauvineau, L., 1996. Approche locale de la rupture ductile : application a un acier ` carbone-mangan` se. Ph.D. thesis, Ecole des Mines de Paris. e Beremin, F., 1981. Cavity formation from inclusions in ductile fracture of A508 steel. Met. Trans. 12A, 723731. Beremin, F. M., 1983. A local criterion for cleavage fracture of a nuclear pressure vessel steel. Met. Trans. 14A, 22772287. Budiansky, E., Rice, J., 1973. Conservation laws and energy release rates. J. Applied Mech. 40, 201203. Cherepanov, G., 1967. Crack propagation in continuous media. J. Applied Math. Mech. 31 (3), 476488. Gurson, A., 1977. Continuum theory of ductile rupture by void nucleation and growth: Part I Yield criteria and ow rules for porous ductile media. J. Engng Mater. Technology 99, 215. Hutchinson, J., 1968a. Plastic stress and strain elds at a crack tip. J. Mech. Phys. Solids 16, 337342. Hutchinson, J., 1968b. Singular behaviour at the end of a tensile crack in a hardening material. J. Mech. Phys. Solids 16, 1331. Irwin, G., 1957. Analysis of stress and strains near the end of a crack traversing a plate. J. Applied Mech. 24, 361364. Kroon, M., Faleskog, J., 2002. A probabilistic model for cleavage fracture with a length scale inlfuence of material parameters and constraint. International Journal of Fracture 118, 99118. Lefevre, W., Barbier, G., Masson, R., Rousselier, G., 2002. A modied Beremin model to simulate the warm prestress effect. Nucl. Eng. Design 216, 2742. Lemaitre, J., 1992. A course on Damage Mechanics. Springer. Liu, Y., Murakami, S., Kanagawa, Y., 1994. Meshdependence and stress singularity in nite element analysis of creep crack growth by continuum damage mechanics approach. Eur. J. Mech. 13A (3), 395417. Lorentz, E., Andrieux, S., 1999. A variational formulation for nonlocal damage models. Int. J. Plasticity 15 (2), 119138. Mc Clintock, F., 1968. A criterion for ductile fracture by the growth of holes. J. App. Mech. 35, 363371. McClintock, F., 1963. Fracture of solids. John Wiley, Ch. On the plasticity of the growth of fatigue cracks, pp. 65102. McMeeking, R., 1977. Finite deformation analysis of crack-tip opening in elasticplastic materials and implications for fracture. J. Mech. Phys. Solids 25, 357381. Moinereau, D., Frund, J., Brochard, J., Marini, B., Joly, P., Guichard, D., Bhandari, S., Sherry, A., Sanderson, D., France, C., Lidbury, D., 1996. Local approach to fracture applied to reactor pressure vessel : synthesis of a cooperative programme between EDF, CEA, Framatome and AEA. In: Les Editions de Physique (Ed.), 1st european mechanics of materials conference on local approach to fracture, Journal de Physique IV. Vol. 6. Fontainebleau, 911 Septembre. ODowd, N., Shih, C., 1991. Family of crack-tip elds characterized by a triaxiality parameterI. Structure of Fields. J. Mech. Phys. Solids 39 (8), 9891015. ODowd, N., Shih, C., 1992. Family of crack-tip elds characterized by a triaxiality
I. Introduction
31
parameterII. Fracture applications. J. Mech. Phys. Solids 40 (8), 939963. ODowd, N., Shih, C., Dodds, R., 1995. The role of geometry can crack growth on constraint and implications for ductile/brittle fracture. In: Constraint effects in fracture: theory and applications. Vol. 2. Rice, J., 1968. A path independent integral and the approximate analysis of strain concentration by notched and cracks. J. Applied Mech. 35, 379. Rice, J., Rosengren, G., 1968. Plane strain deformation near a crack tip in a power-law hardening material. J. Mech. Phys. Solids 16, 112. Ritchie, R., Knott, J., Rice, J., 1973. On the relationship between critical tensile stress and fracture toughness in mild steel. J. Mech. Phys. Solids 21, 395410. Roos, E., Alsmann, U., Els sser, K., Eisele, W., Seidenfuss, M., 1998. Experiments an a warm prestress effffect and their numerical simulation based on local approah. In: publishing, E. (Ed.), Fracture from defects ECF12. Shefeld, 1418 Septembre. Rousselier, G., 1987. Ductile fracture models and their potential in local approach of fracture. Nucl. Eng. Design 105, 97111. Ruggieri, C., Dodds, R., Wallin, K., 1998. Constraint effects on refference temperature for ferritic steels in the transition region. Engng Fract. Mech. 60, 1936. Ruggieri, C., Dodds Jr., R., 1996. A tranferability model for brittle fracture including constraint and ductile tearing effects: a probabilistic approach. Int. J. Frac. 79, 309 340. Ruggieri, C., Gao, X., Dodds, R., 2000. Transferability of elastic-plastic fracture toughness using the weibull stress approach: signicance of parameter calibration. Engng Fract. Mech. 67, 101117. Schwalbe, K.-H., 1998. ESIS P6 98: Procedure to measure and calculate material parameters for the local approach to fracture using notched tensile specimens. Tech. rep., GKSS Research Centre. Shih, C., 1983. Tables of HutchinsonRiceRosengren singular eld quantities. Tech. rep., MRL E-147, Brown University. Sumpter, J., 1993. An experimental investigation of the t stress approach. In: Constraint effects in fracture, ASTM STP 1171. Sumpter, J., Forbes, A., 1992. Constraint based analysis of shallow cracks in mild steel. In: TWI/EWI/IS International conference on shallow crack fracture mechanics, Toughness Tests and applications.
32
Chapter II
PHYSICAL MECHANISMS OF DAMAGE
by A. Pineau
INTRODUCTION
This chapter is devoted to the study of failure micromechanisms in various kinds of materials. There are two main reasons for this chapter. The rst one is related to the necessity of a sound understanding of the physical mechanisms of damage before dealing with modeling these complex processes. The second one, which is essentially that one advanced by the materials scientists, is that it is the only way to design new materials, to improve the mechanical properties of existing materials, and to predict the behaviour of these materials under in-service conditions which cannot be easily obtained with laboratory tests, in particular long-term tests. The chapter is divided into 5 main parts. The rst part is devoted to ductile fracture in metals and polymers. The emphasis is then laid on cavitation phenomenon in plastically deformed materials. In the second part, brittle cleavage fracture is briey described, starting with cleavage fracture in metals and ceramics, continuing with intergranular fracture (essentially in metals, although this failure mode is often observed in ceramics), and nishing with crazing phenomenon which is a fracture mode typical of those encountered in polymers. In the third part, the emphasis is laid on quasi-brittle fracture observed in composite materials, in particular in ceramic composite materials and in concrete. The fourth part is devoted to ductile-to-brittle transition in metallic materials with a BCC or HCP crystallographic structure. In the last part the micromechanisms of creep damage in metallic materials are described. 2 DUCTILE FRACTURE: METALS AND POLYMERS
2.1 BACKGROUND
Ductile fracture results from the nucleation, growth and coalescence of cavities. This failure mode occurs in metals and in polymers provided that these materials are loaded at sufciently high temperatures and strain rates. Large improvements in the resistance to ductile fracture and in fracture toughness could be achieved if cavity nucleation could be postponed. It is therefore important to investigate the material variables and the mechanical factors controlling this rst step of ductile fracture. Cavity growth and coalescence are also briey discussed. A fuller account of these steps preceding nal rupture is given in chapters V and VIII. Ductile fracture is essentially connected to glide mechanisms which are not 33
34
(b) (a)
Figure II.1: Development of a single shear band at the tip of a deeply notched tensile specimen made of a non-hardening material; (b) More diffuse plastic shear deformation in a hardening material (source: (Francois and Pineau, 2001)).
necessary crystallographic. This is illustrated in gure II.1 corresponding to the extreme case of a deeply notched tensile specimen made of a non-hardening material. A single glide band develops inclined at 45 degrees to the tensile axis. In the case of a hardening material, the plastic deformation takes place on several glide systems (gure II.1b) and is more diffuse than in the preceding case. In both cases the separation of the specimen in two pieces is due to the full extent of these tearing mechanisms. Considering a smooth tensile specimen, at the maximum load a local necking appears when the strain hardening of the material is not matching anymore the reduction of the load bearing section of the specimen owing to plastic deformation. This plastic instability leads to the separation of the specimen in two pieces, in some cases by complete extent of the neck down to a point. In most cases, however, the fracture starts sooner by the development of a crack in a mid-section of the neck, ending again by a shearing mechanism. The characteristic cup and cone fracture results (Figure II.2). Looking down at the crack surface through a scanning electron microscope shows characteristic dimples (Figure II.3). Thus the fracture in the midsection of the test piece results from the growth of cavities which join together.
2.2 CAVITY NUCLEATION
In metallic materials and in particle-reinforced polymers, cavity initiation sites are usually associated with second phase particles or non-metallic inclusions. However in very pure single phase metals, like Ti alloys, there is some evidence showing that, under certain conditions, cavities can be homogeneously nucleated (see e.g. (Thompson and Williams, 1977)). A good review of the theoretical and experimental results related to the micromechanisms of cavity nucleation has been published by Goods and Brown (1979). As a general rule cavity nucleation results from the inhomogeneity in plastic deformation between the matrix and the inclusions. Several approaches to this problem have been proposed, based either on dislocation theory (for crystalline materials) or on pure continuum mechanics for all kinds of materials. As pointed out by Goods and
II. Physical mechanisms of damage
35
(b) (a)
Figure II.2: a) Section through the tensile axis of a broken specimen showing the typical cup and cone fracture; (b) View at a higher magnication of the longitudinal midsection through the fracture surface showing cavities. Brown (1979), the applicability of one approach or the other one is mainly a function of particle size. According to these authors the critical radius of particles above which continuum mechanics can apply is of the order of 1 m, depending on the workhardening rate of the material. These particles correspond to carbides in spheroidized steels or non-metallic inclusions in metallic materials such as steels and Al alloys, while the second type of particles is associated with precipitation strengthening phases in steels and Al alloys, or other metallic materials. Cavity formation cannot occur unless the elastic energy released from the particle by interfacial separation is at least equal to the surface energy created (energy criterion). A criterion based on the achievement of a critical stress at the interface or inside the particle, , must also be satised. For small particles ( 1 m) which are most often strongly bonded to the matrix, it is necessary to nucleate voids by subjecting the particles to high stresses from dislocation loops left around them. The interfacial critical stress can be approximately written as:
where is the shear modulus, the radius of the particles and the Burgers vector. This equation shows that for very small particles the strain to initiate cavitation, , is inversely proportional to particle size. This model yields satisfactory results. The application of the dislocation approach to cavity nucleation shows also that, except for very small particles ( 1 m) the energy criterion is always satised before the stress criterion. For bigger ( 1 m) and widely spaced particles a continuum mechanics approach can be applied to calculate the stresses inside the inclusions. It is clear that a stress criterion based on this approach leads to a critical strain independent of particle size. Several types of calculations have been proposed to determine the local stress-strain eld as a function of the far-eld applied to the material. The theory of inclusions and inhomogeneities by Eshelby (1961) has been used by Tanaka et al.
(1)
36
100 m Figure II.3: Fracture surface of pipeline steel (X100). Small inclusions are observed within dimples. (1970). This theory is based on an elastic analysis. Hence it is strictly valid only when the applied strains are not too large. A nucleation model based on a critical stress derived from continuum plasticity has been proposed by Argon et al. (1975) Their model incorporates also certain microstructural features of the deformation processes in dispersion strengthened alloys. Argon and co-workers have also attempted to account for particle interactions. Moreover, since it is well established that the hydrostatic component of the applied stress has an effect on cavity formation (see e.g. (French and Weinrich, 1974)), their criterion was written as:
where is the equivalent von Mises stress and the hydrostatic stress. In equation (2) the inhomogeneity in plastic deformation between the matrix and the inclusions does not appear explicitly. This is one of the reasons why a slightly different criterion was introduced (Beremin, 1981):

where is the maximum principal stress, and are temperature independent material parameters which are a function of particle shape while is the yield strength of the material. Equation (3) was derived from an extension of the Eshelby theory to plastically deformed materials (Berveiller and Zaoui, 1978). Typical values for are 1700 MPa for Fe C particles in a spheroidized steel (Argon et al., 1975). Smaller values ( 800 MPa) were found when cavity nucleation occurred by the decohesion between the matrix and manganese sulde (MnS) inclusions. The interfacial strength between second phase particles and the matrix is dependent on the local chemical composition. The segregation of impurity elements similar to those which induce intergranular embrittlement (see section 3.2) can reduce the interfacial resistance. Hydrogen induced ductility losses in low strength steels could also be at least partly explained in this way. A reduction in by a factor of about 2 between uncharged and hydrogen charged specimens was reported (Cialone and
1 H
#
1
(2)
(3)
37
Asaro, 1979). For recent modeling of hydrogen-induced decohesion at particle/matrix interfaces, see also Liang and Sofronis (2003). In many circumstances nucleation is not instantaneous but occurs continuously during plastic deformation, that is the fraction of cavities is an increasing function of plastic strain. Moreover there is still a lot of debate in the literature to know whether continuous nucleation is a strain- or stress-controlled phenomenon. A number of authors (see e.g. (Kwon and Asaro, 1990)) have concluded that an interfacial stress controlled nucleation criterion was more realistic than a straincontrolled criterion to characterize the nucleation behavior of spheroidized steel. On the other hand, results obtained on Al alloys (Walsh et al., 1989) and on a cast duplex stainless steel (Joly and Pineau, 1991; Bugat et al., 1999) have shown that in these materials cavity nucleation was essentially strain-controlled. The answer to this question is likely very much dependent of materials, and can only be obtained through detailed metallographic observations. Another important aspect of cavity nucleation is the variability since cavities are not usually homogeneously distributed. This inhomogeneity can be due to a statistical distribution of the matrix/particle cohesion stress about a mean interfacial stress (see e.g. (Kwon and Asaro, 1990)) or to local microplasticity effects in relation with crystallographic details (see e.g. (Bugat et al., 1999). This inhomogeneity in local nucleation rate which plays a key role in modeling ductile rupture (see chapter XIII) was thoroughly investigated in duplex stainless steel (Pineau and Joly, 1991; DevillersGuerville et al., 1997; Joly et al., 1990) using interrupted tests, as shown in gure II.4 where clusters of cavities represented by Vorono cells are clearly observed. The histogram of the cell sizes shows that a very small fraction of the surface area in gure II.4 is leading to extremely large local nucleation rates compared to the mean nucleation rate. This clustering effect, particularly pronounced in this material, is another feature of cavity nucleation which should be kept in mind when modeling ductile rupture.
2.3 CAVITY GROWTH
Considerable progress in the understanding of hole growth has been made through theoretical models (Berg, 1962; Mc Clintock, 1968; Rice and Tracey, 1969). In these models based on continuum mechanics the voids grow under the combined effect of the applied plastic strain, , and that of the mean stress, , or stress-triaxiality ratio, . These models are based upon a number of simplifying assumptions which do not necessarily reect the actual behavior of real materials. In particular, they assume that no interaction takes place between two neighboring cavities, and that crystallographic aspects can be neglected (which is not necessarily the case in materials like Ti or Zr exhibiting large differences in critical resolved shear stress between different slip planes (see e.g. (Cr pin, 1995)). In the Rice and Tracey (1969) e model, the cavity growth rate is expressed as:
where is the cavity radius and is the increment of plastic strain. This relation holds for a spherical cavity located inside on innite perfectly plastic material. Experiments on a low alloy steel containing initial cavities initiated from spherical ceramic particles have shown that this relation was rather well obeyed (Marini et al., 1985). In equation (4), the pre-exponential factor of 0.283 was later corrected (Huang,
( )&

&
& $B #
0
(4)
38
(a)
Figure II.4: Cast duplex stainless steel. (a) Initiation of cavities produced by the formation of cleavage microcracks in the ferrite phase; (b) Vorono cells illustrating the heterogeneity in the distribution of cavities initiated from cleavage microcracks. The formation of damage clusters is noted (source: (Devillers-Guerville et al., 1997)). 1991). It should also be added that the work of Mc Clintock has shown that an increase in the Norton exponent, , for a viscous medium decreased the cavity growth rate (Mc Clintock and Argon, 1966). This was conrmed by Budiansky et al. (1982).
2.4 CAVITY COALESCENCE AND CRITERIA FOR DUCTILE FRACTURE
Cavity coalescence is still very poorly understood and requires a large research effort. Many approximations are often made to simplify this problem. Two broad classes of behavior must be distinguished. In the rst class which is observed in a number of materials with a single microstructure, cavity growth occurs continuously until the voids impinge together. In this case only one population of well-formed dimples are observed on the fracture surfaces (gure II.5a). In the second class which corresponds to many structural materials with complex microstructures, cavity growth process initiated from large second phase particles is bypassed by the formation of shear bands between growing cavities. In these localization bands, smaller cavities are initiated from strengthening precipitates (carbides in steels, precipitates in Al alloys) (gure II.5b). The mechanical behavior of a material containing a double population of cavities has been studied by a number of authors (see e.g. (Perrin, 1992)). Moreover it is also clear that in this coalescence phenomenon, statistical aspects must be considered (e.g. clustering). The rst and the simplest criterion for ductile rupture was introduced by Mc Clintock (1971). This criterion has also been developed by Beremin (1981). It is assumed that a material fails for a critical void volume fraction, which is independent of stress triaxiality ratio. Equation (4) shows that for a stress history during which the stress triaxiality is maintained constant, failure will occur when the following equation
25 mm
(b)
39
Inclusions
Precipitates
(a)
(b)
Figure II.5: Sketch showing two mechanisms of cavity coalescence (a) by impingement; (b) by the formation of shear bands and associated population of cavities initiated from precipitates. is satised.
where is the initial void size and the initial void volume fraction. It was experimentally shown that equation (5) is relatively well satised and, in particular, that the strain to failure, , decreases exponentially with stress triaxiality ratio (Pineau, 1981, 1992). In many situations, as stated earlier, continuous cavity nucleation must be considered. If one assumes that the cavity volume fraction is only due to continuous nucleation and that nucleation is linearly related to plastic strain, i.e.:
is a material parameter including cavity shape, in adding the two contribuwhere tions to the volume fraction of cavities, it can easily be shown that the volume fraction is simply given by:

This expression is plotted in gure II.6 for different values of , where it is observed that for large values of the nucleation rate, , i.e., small values of the ratio , the strain to failure, is no longer strongly dependent on stress triaxiality. Although the proposed model is largely oversimplied it clearly indicates that a close examination to the slope of the curves gives some useful information on the micromechanisms of ductile fracture, in particular on the relative importance of cavity nucleation and void growth in controlling ductile fracture.
'
! "
'
'

where volume fraction of cavities, as:
. Assuming that ductile fracture occurs for a critical , the variation of the strain to failure is simply expressed (8)
( )&

&
& "#B
) #

% % )

(2
' 1#

'

&
%
(

&
& "#B
(5)
&
'
% )
(6)
'
(7)
40
1.0 strain to failure 0.8 0.6 0.4 0.2 0.0 0.0
0.2
0.4
0.6 0.8 1.0 stress triaxiality
1.2
1.4
Figure II.6: Variation of the strain to failure as a function of stress triaxiality ratio for various values of the nucleation rate, (equation 8). Comparison with experiments on Cast duplex stainless steel (Pineau, 1992). A large research effort has been made over the last past decade to develop more sophisticated criteria for ductile fracture (See chapter V and (Pineau and Besson, 2000)). In particular the mechanics of porous damaging materials has been largely developed. In these models (Gurson, 1977; Rousselier, 1987), the softening effect due to growing cavities is introduced in the constitutive equations of the materials. Moreover it has been proposed (Zhang and Niemi, 1994; Benzerga et al., 1999) to use the model of coalescence introduced by Thomason (Thomason, 1968, 1985b,a). The scatter and the size effect observed in ductile fracture due to the inhomogeneity in the distribution of cavity initiation sites have also been reproduced using Monte Carlo type simulations based on these most advanced models for ductile fracture (DevillersGuerville et al., 1997; Decamp et al., 1997) 3 BRITTLE FRACTURE Scientists and engineers are very much concerned with brittle fracture modes observed in metals and ceramics, i.e. transgranular cleavage fracture and intergranular fracture. In metals, brittle intergranular fracture is essentially associated with the segregation of impurities along the grain boundaries.
3.1 CLEAVAGE. METALS AND CERAMICS
3.1.1
Introduction
In crystalline materials, cleavage fracture occurs preferentially over welldened crystallographic atomic planes: 100 in BCC metals 0001 in metals with an HCP structure, like Mg, Zn, Ti, Be, etc. . . (gure II.7). It can easily be shown that the

B &
1.6
#$B ) P $B P $B % ) "B ) # # P # # & # # "B & 'B $B ) P B $B #

&
1.2
'
'

010
41
001
(a)
100
121
(b)
(c)
Figure II.7: Cleavage fracture (a) Scanning electron micrograph showing rivers and the presence of a tongue on the surface formed by cleavage of the ferrite phase in a cast duplex stainless steel. The faces correspond to (001) cleavage deviated from the principal plane shaded in (b) by a mechanical twin; (c) Formation of rivers in a polycristal of zinc broken by cleavage, note the occurrence of rivers across the grain boundary and their direction of ow, which shows the direction of local propagation of the crack (source: (Francois et al., 1998)). theoretical cleavage stress in a crystalline solid is given by:
where is the Youngs modulus, , the Burgers vector and is the surface energy. This theoretical cleavage stress is much higher than that usually found (about 1000 MPa for steels). The reasons for this large difference are examined in the following. Whatever the damage mechanism, its initiation requires the breaking of atomic bonds. The expansion of a small cavity corresponds to an increase of the surface energy whereas the hydrostatic stress produces a work to increase the volume. The classical theory of nucleation shows that the critical radius of a cavity nucleus is given by:
For the smallest possible nucleus, without thermal activation,
"
)
(
#
"
"
(9)
(10) being of the order of
42
On the other hand, the calculation of the stress needed to pull apart two halves of a crystal yields equation (9). The two expressions are compatible with , which is the right order of magnitude. The same is found again by a cohesive zone model at the tip of a crack, in a purely linear elastic calculation. This atomic bond strength is usually high compared with the fracture strength of usual materials, which means that fracture is due to large stress concentrations. These stress concentrations can be due to the presence of geometrical defects (pores, surface indentations, etc..) which is the situation met with ceramics or to plastic deformation in metals. In these materials, heterogeneous slips and mechanical twins form another source of stress concentrations.
Figure II.8: Cast duplex stainless steel. Cleavage initiated in the ferrite phase from the intersection of two mechanical twins. The planes of the twins and of the cleavage microcrack are indicated (source: (Francois et al., 1998)). Several theories have been advanced to explain cleavage in metals (Stroh, 1954; Cottrell, 1958; Smith, 1966). In the Cottrell theory cleavage cracks are initiated at the intersection between two slip planes or two mechanical twins. An example of this mechanism is given in gure II.8. In this theory it is shown that the fracture strength is given by: (12)
being the material parameter in the HallPetch law relating yield strength, grain size, :

@ "

where is grain size and
is a material parameter given by: (13) , to (14)
B
)
"
)
)

the atomic distance, the corresponding stress is the atomic bonds fracture stress: (11)
43
In structural steels there is now a growing body of evidence showing that cleavage cracks may also nucleate from microstructural defects, such as inclusions. This situation is most frequently observed when cleavage fracture occurs in the ductilebrittle transition regime (see e.g. (Roseneld et al., 1983)).
(a)
(b)
(c)
(d)
Figure II.9: Sketch showing the different steps in the initiation and the propagation of a cleavage crack initiated from an inclusion or a second phase particle (a) and (b), arrested at grain boundaries (c), propagating across the grain boundaries (d).
3.1.2
Is cleavage fracture nucleation or growth controlled ?
Another key question in modeling cleavage fracture is to know whether this failure mode is initiation or growth controlled. Arrested microcracks initiated from nucleation sites such as carbides in ferritic steels are often found at grain boundaries or bainitic packet boundaries (see e.g. (Lambert-Perlade et al., 2003) and gure II.10). This indicates that, in this case, cleavage fracture is essentially growth controlled, and must be modeled with several steps, as shown schematically in gure II.9. The rst step involves crack initiation at particles. The microcrack reaches immediately the particle/matrix interface whose strength is (step 2). The microcrack propagates into the matrix across the particle/matrix interface and reaches the rst strong matrix/matrix interface of strength (step 3). The last step (equation 4) involves crack propagation across the matrix/matrix interface leading to nal fracture. Depending on the relative strength of these microstructural barriers, cleavage fracture may be either nucleation controlled (step 1 is the most critical event) or growth controlled (step 4 is the critical event). A number of observations reported in the literature (see e.g. (Lambert-Perlade et al., 2003; Martin-Meizoso et al., 1994)) have shown that, at low temperature, cleavage fracture is essentially nucleation controlled, while, at higher temperature, it is rather growth controlled. This may partly explain why, in a number of circumstances, it has been reported that the cleavage stress was increasing with temperature (Margolin et al., 1997; Rossoll et al., 2002; Hausild, 2002; Tanguy, 2001).
3.1.3
Scatter and size effects
Brittle fracture test results in metals and ceramics are largely scattered. Moreover a size effect is observed: the lower the cleavage fracture stress is, the larger is the test piece. Rather surprisingly, although the scatter in cleavage stress measurements is well established, it is only rather recently that models have been proposed to account for these effects (for a review, see e.g. (Wallin, 1991)). More details on this aspect
44
(a)
(b)
(c)
Figure II.10: Arrested cleavage microcracks obtained with interrupted tests on notched specimens using acoustic emission analysis (Lambert-Perlade et al., 2003). (a) Fracture surface after subsequent fatigue crack propagation ( C). Some EBSD indexations of crack arresting boundaries are given. (b) EBSD map of a polished cross section (( C). Thin lines and thick lines denote low angle and high angle boundaries respectively. (c) Scanning electron micrograph and disorientation analysis along the crack path showing crack arrest at high angle boundaries (same area as (b)).
&
) B
45
of brittle fracture are given in chapter V. Here it is enough to say that the Weibull law is the most widely used to describe both the scatter and the size effects. Writing the cumulative failure probability, of a specimen of volume, , submitted to an homogeneous stress state, , as:
where is an arbitrary volume, is the average strength of that unit volume, while is the shape factor of the Weibull distribution, it is usually found that in conventional ceramics , in high strength ceramics, while in metals. 3.1.4 Microstructural effects
In ferritic steels which are materials prone to cleavage fracture at low temperature and high strain rates, many metallurgical factors may inuence the cleavage strength. This includes the grain size (see e.g. (Curry and Knott, 1978)), the packet size in bainitic/martensitic steels (Naylor and Krahe, 1974; Brozzo et al., 1977), the misorientation between bainitic packets (Bouyne et al., 1998; Lambert-Perlade et al., 2003), the presence of local brittle zones (Martensite/Austenite constituents in welded microstructures), etc. . . As a rule, the dislocation mobility controls the fracture strength. Factors affecting this mobility, such as impurities or irradiation defects, may strongly affect the cleavage stress (see e.g. (Al Mundheri et al., 1989)). A number of authors have also shown that the cleavage fracture stress may be affected by large strains (see eg. (Beremin, 1983)). In particular a strain-induced anisotropy effect was evidenced, the cleavage stress being increased in the direction of the maximum principal strain. It should also be added that a criterion for nucleationcontrolled cleavage fracture must include other terms of the stress or strain tensor than the maximum principal stress. Modeling cleavage fracture under these circumstances bears, to some extent, a strong analogy with cavity nucleation in ductile rupture (see e.g. equation (3)). An attempt in this direction has been recently proposed (Bernauer et al., 1999).
3.2 INTERGRANULAR FRACTURE METALS AND CERAMICS
Intergranular fracture is often observed in ceramics. In ferritic steels, brittle fracture normally occurs by cleavage at low temperatures or high strain rates. However in these materials, the segregation of a number of solute elements, in particular impurities, such as P, Sn, Sb or As, to grain boundaries can change the brittle fracture mode to intergranular. Since this mode of failure operates at lower stresses, brittle fracture can then be observed at higher temperatures where failure by ductile rupture would otherwise take place. The classical example of this transition in failure mode is temper-embrittlement in low alloy steels (gure II.11). Such embrittlement is evidently caused by a lowering of intergranular cohesion, the magnitude of which depends on the quantity and the nature of segregated elements. The physical basis of this embrittlement effect has not yet been studied in a very quantitative manner. However the studies by Cottrell (Cottrell, 1989, 1990b,a) should be mentioned. Very few well documented studies have been performed to determine quantitatively the variation of the critical intergranular fracture stress with test parameters using procedures similar to those used to measure the cleavage stress (see chapter IV). Most often, the effect of temper-embrittlement is investigated by determining the shift in the
)
B )
(15)
46
Figure II.11: Intergranular fracture observed in a low alloy steel (16MND5) submitted to temper embrittlement (source: (Yahya, 1997)). ductile-to-brittle transition temperature measured vith Charpy V notched specimens. However the work by Kameda and Mc Mahon (1980) should be mentioned. These authors showed that the critical local stress to fracture a grain boundary was directly related to the Sb concentration segregated on that boundary. A similar work has been made on a steel with a composition representative of that found in the ghost lines in the shells of pressurized water reactors (Naudin et al., 1999). The results reported in gure II.13 clearly show that the intergranular fracture stress is a decreasing function of the amount of phosphorus segregated along the grain boundaries. In many situations, brittle cleavage and intergranular fracture are competing modes. This is illustrated in gure II.13 (Amar, 1986) where the fracture surface of a low alloy steel tested at C is shown. This specimen broke at an anomalous low value of the stressintensity factor. Figure II.13 shows that the fracture surface is essentially covered by cleavage facets, except in a small initiation area which is intergranular. More research work should concentrate on the respective scatter of these competing fracture modes (see e.g. (Naudin et al., 1999, 2001)). The results obtained on a C MnNiMo steel shown in gure II.14 indicate that not only the intergranular fracture stress is lower than the cleavage fracture stress but the scatter measured by the value of the shape factor, , in the Weibull law is also larger. Similar results have been reported on a 2.25 Cr-1 Mo steel (Holzmann et al., 1991). Another feature which deserves also more attention is the respective strain-rate sensitivity of both brittle failure modes. Intergranular fracture involves also plastic deformation, but there is no reason to believe that the variation of the dissipated energy, , as a function of strain rate and temperature are the same for both mechanisms. This might produce interesting results as schematically shown in gure II.15 where it is suggested that a
47
2400 2300 Fracture stress (MPa) 2200 2100 2000 1900 1800 1700 intergranular fracture 1600 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 Phosphorus monolayer coverage Figure II.12: 16MND5 Steel. Cleavage fracture stress and intergranular fracture stress as a function of the amount of phosphorus segregated along the grain boundaries (source: (Naudin et al., 1999)). brittle cleavage crack could be initiated by intergranular fracture (see also (Jaeckels et al., 1994)). 4 CRAZING IN POLYMERS cleavage fracture
Glassy polymers typically exhibit two types of failure: shear yielding and crazing. Under special conditions, such as compressive loading, both thermoplastics and thermosets can be made to shear, but under tension, many polymers fail in a brittle manner, especially when they are loaded at high strain rates. Polymer blends are systems in which a dispersion of small rubber particles have been added with the intent to toughen the material, by suppressing crazing and promoting shear yielding. In many blends, such as high impact polystyrene, crazing and shear yielding occur simultaneously. In these materials, as well as in composites, multiple crazing occurs and is controlled by the microstructure. Examples of crazes are shown in gure II.16 and II.17 (Lefebvre, 2002). In polymers the cleavage planes that result from failure of the weak Van der Weals bonds are linked by laments consisting of molecular chains. These microcracks, whose presence is revealed by the materials whitening, are known as crazes (gure II.16); the laments, forming bridges between the two faces are extracted from the matrix as the crack opens (gure II.17). Craze formation involves three successive stages: (i) nucleation, (ii) propagation of the craze front, (iii) widening of an existing craze in a direction normal to its plane. It seems to be established that the widening of a craze is to a large extent determined by surface drawing: the craze widens by continuously pulling new polymer material into the craze structure, as schematically depicted in gure II.18. Crazes in amorphous polymers generally reach lengths in the order of tenths of millimeters, whereas the
48
Figure II.13: Initiation of cleavage fracture from intergranular fracture in an axisymmetrically precracked specimen (source: Amar (1986)). width of the craze remains in the order of several micrometers. In semicrystalline polymers such as polyethylene, crazes can become substantially wider but that involves additional mechanisms that are not taking place in the amorphous polymers. The numerical simulation of crazing has been undertaken at various length scales: (i) the molecular level (Han et al., 1998), (ii) higher scale with a spring network model to explicitly model the brils in the craze (Sha et al., 1997), (iii) continuum level (Brown, 1991; Knauss, 1993; Xu and Needleman, 1994; Tijssens et al., 2000).
4.1 CRAZE INITIATION
The physical mechanism for craze initiation is not yet fully understood and multiple criteria have been proposed. It has been formulated that crazing occurs when the following condition is satised (Sternstein and Myers, 1973).
in which and are the maximum and minimum principal stresses, respectively, and 0 are temperature dependent constants and is three times the hydrostatic stress under plane stress conditions. Following Argon (1973) it can be assumed that the parameters and vary with temperature, , according to an Arrhenius expression:

4.2 CRAZE WIDENING
Once a craze has initiated, widening is assumed to be a process of drawing in new polymer material from the craze/bulk interface, as schematically shown in gure II.19.
B
'

'
(16)
B
'
'
'
(17)
49
1.0
#
Probability
0.6 Cleavage Critical stress 1850 MPa

) %) # #
0.4
0.2
0.0 1000 1500 Fracture stress (MPa) 2000
Figure II.14: Probability to failure as a function of the applied stress in 16MND5 steel breaking either under cleavage or intergranular fracture (source: (Kantidis, 1993)). The highly stretched network of molecules in the brils strongly resists to further elongation, thereby pulling new amorphous material into the bril. As these brils continue to form, the initial craze voids grow into highly prolate toroidal space around the brils. Craze material is a complex structure in which long cylindrical brils of polymer material are interconnected by crosstie brils which give the craze some tangential load carrying capacity when it widens. The basic assumption (Tijssens et al., 2000) is that the viscoplastic ow into the brils is the ratelimiting process. These authors proposed the following widening law in terms of the normal separation rate, as a function of the normal stress :
4.3 CRAZE BREAKDOWN
For a long time, widening of crazes was thought to be the result of creep deformation of the briller material. Breakdown of crazes was therefore assumed to occur at the mid rib of the craze, which is the oldest portion of the craze material. However crazes have been shown, to break at the craze/bulk interface. Although it is now recognized that the crosstie brils play an importance role in the lifetime of a craze, the theoretical framework on craze breakdown is far from complete. A number of authors have proposed to describe craze fracture using cohesive surface models. In these models it is usually assumed that craze breakdown occurs as soon as the plastic craze opening
where , and are material constants. values as high as .
is the separation rate when
)'
) % #
0.8
Intergranular Critical stress
1325 MPa
B 4
)
'
(18) reaches
50
Cleavage Effective surface energy ( )
Intergranular
Strain rate, crack velocity
Figure II.15: Schematic diagram showing the effect of strain rate or crack velocity on the effective surface energy for both cleavage and intergranular fracture.
Figure II.16: Light micrograph obtained with polarizer and analyser. Craze formation in Polyamide (PA)11. (Source Lefebre, 2002). Observation at the free surface. attains a critical value, temperature and loading rate.
5 QUASI-BRITTLE FRACTURE CERAMIC COMPOSITE MATERIALS AND CONCRETE

5.1 CERAMIC COMPOSITE MATERIALS
5.1.1
Introduction
Ceramic materials are becoming more and more interesting for engineering applications where good wear, erosion and corrosion resistance or high strength at elevated temperatures are required. The most advanced ceramics materials are silicon nitride, zirconia, silicon carbide, alumina. Ceramics are brittle materials. With ber reinforcement, however fracture toughness of ceramics can be increased to a level approaching that of hard metals, the fracture being no longer brittle. Using ceramic
!
)
which is taken to be constant with respect to time,
51
Figure II.17: Scanning electron micrograph showing the bottom of two crazes in PA11 (source: (Lefebvre, 2002)).
matrix composites (CMC) the fracture energy can reach levels several hundred times that of monolithic ceramics. The general principle of ber reinforcement of ceramics is, therefore, not so much to reinforce or stiffen (which is the case for polymer or metal matrix composites), but rather to overcome poor reliability and brittleness by an appropriate fracture leading to energy absorption through crack growth controlling mechanisms. These mechanisms can only operate if the ber/matrix bond-strength is sufciently low to prevent cracks from propagating through the bers, and results in a fracture behavior where the bers remain undamaged in spite of multiple matrix cracking. Even with extensive matrix cracking the composite retains a signicant load bearing capacity and the bers hold the component together in one piece. In this way, catastrophic failure can be avoided by the use of ber reinforcements in ceramics. 5.1.2 Fracture of brittle matrix composites
The elastic properties of continuous ber brittle matrix composites are readily predicted by a simple rule of mixtures (Bretheau et al., 2001). Since the Youngs moduli rarely differ much between ber and matrix for ceramic composites, little stiffening is achieved. The prediction of ultimate strength is complicated by the low strain to failure of the matrix. On stressing the composite axially, both ber and matrix exhibit the same strain which leads to matrix cracking occurring at some level which depends on a wide range of materials parameters including ber content and ber/matrix elastic moduli. If the matrix is well bonded to the bers, matrix cracking will lead to ber fracture and hence overall composite fracture with linear elastic stress-strain behavior (gure II.20). If, however, the ber/matrix interfacial bond is sufciently weak to prevent matrix cracks from penetrating neighboring bers, considerably high levels of strength can be achieved with a non-linear elastic stress-strain behavior, as shown, in gure II.20. In this case, the full potential of the reinforcing bers is utilized. To obtain the derived fracture behavior it is extremely important to tailor the ber/matrix bond strength, . Depending on , different toughening mechanisms operate, as illustrated in gure II.21: (a) ber pull out; (b) ber debonding; (c) debonding followed by ber

52
(a) main bril bulk material
crosstie bril
(b) cohesize surface
Figure II.18: Schematic modeling of a craze (a) by a cohesive surface, (b) characterized by a traction and a separation over this surface. fracture and pull-out, (d) crack bridging, (e) microcracking and crack branching, (f) crack deection. Since the bers are stronger and usually stiffer than the matrix, load transfer to the bers may reduce the stress concentration of an advancing crack by redistributing the load over a larger volume of the material. If the bers have a higher thermal expansion, , than the matrix , the matrix is precompressed during cooling from sintering temperature resulting in a higher matrix tensile strength and a higher strain for the rst matrix cracking. Such prestressing also requires load transfer and a strong ber/matrix bond. Thermal expansion mismatch also plays an important role in crack deection. This mechanism has been treated extensively for particulate composites and models have been developed to describe and explain the toughening observed (see e.g. (Faber and Evans, 1983)). An important mechanism operating in addition to crack deection but after the crack has passed is wake toughening or crack bridging (Swanson et al., 1987). Each bridging ber acts to shield the crack tip from the applied stress so that the stress intensity factor, , at the crack tip is less than the apparently applied . Since the magnitude of the wake toughening will depend on the number of bridging bers, the resistance to crack extension will increase with crack length, i.e. R-curve behavior will be observed (Freiman, 1988). For long ber composites an important toughening mechanism is debonding followed by ber fracture and pull-out. A long ber can break some way from the fracture surface due to the presence of a defect. This is particularly likely if the ber

'
'

initially amorphous polymer material in active zone failure occurs here drawing stress
53
brils
highly stretched molecules in brils
Figure II.19: Illustration of the surface drawing mechanism in craze formation.
has debonded because then the stress in the ber increases further back along the ber. After fracture the ber may pull out thus adding to the fracture energy. Even a debonded ber can have a large pull-out energy because after it has broken it expands by the Poisson effect thus pushing against the matrix. The magnitude of these different energy dissipating mechanisms is difcult to predict, since properties such as interfacial bond strength, residual thermal mismatch stress and ber strength after manufacture often are unknown. However, orders of magnitude for the different toughening processes are given in table II.1 (Lundberg, 1989). For short ber composites the toughening is obtained mainly by crack deection, crack branching, crack bridging, load transfer or pre-stressing, whereas for continuous ber composites mechanisms like ber debonding and pull-out are the most important.
Pre-stressing Crack deection Crack bridging Debonding Pull-out Debonding and Pull-out
12 24 5 10 10 15 100 200 500 1000
Table II.1: Estimated relative fracture energy increase (Lundberg, 1989)

Toughening mechanism
Approximate toughening
54
ber
composite (strong bond)
composite (weak bond) matrix cracking
Stress
matrix
Strain
Figure II.20: Schematic diagrams showing the stressstrain response of composite materials with either strong or weak matrix/ber bonds.
5.2 CONCRETE
From the viewpoint of the materials scientists, concrete can be considered as a particulate strengthened ceramic-matrix composite material (gure II.22). The sand and stone are the dispersed particles in a multiphase matrix of cement paste. Reinforced concrete can then be considered as a ber - reinforced composite with the reinforcement bars as the bers adding toughness. One fundamental difference, however, between this material and other engineering ceramic based composites is that the composition, and hence the properties of the cement paste, do not remain constant after processing but vary both with time and with temperature and relative humidity over the normal ambient ranges (Hansson, 1995; Acker, 2001). A second difference is that the pores which, at normal relative humidity, are lled with a highly alkaline solution can be regarded as a separate phase and play a major role in the strength of the concrete. Moreover, the second phases in concrete span 10 orders of magnitude which is much larger than in other advanced industrial materials. These second phases range from nanometer-sized pores and gel particles to reinforcement bars which can be tens of meters in length, and with paste, sand and stone, particles of every magnitude in between. The constituents of ordinary Portland cement concrete are hydraulic cement, sand, stone and water. While the properties of the sand and stone-shape, size, density and porosity - all inuence the performance of the concrete, they are determined by nature. The cement paste is the most important component for the materials scientists. The cement clinker consists of several anhydrous oxides, primarily tri calcium silicate (C S) and di-calcium silicate (C S) with smaller amount of tri calcium aluminate (C A) and calcium alumino-ferrite (C AF). When mixed with water, the clinker hydrates to form cement paste, an interconnected network of solid phases which gives the concrete its strength and stability. The principal constituent of the paste (50 to 60% by volume

55
Figure II.21: Various mechanisms contributing to the increase of fracture toughness of composite materials; (a) ber pull-out; (b) ber debonding; (c) debonding + ber fracture + pull-out; (d) crack bridging; (e) microcracking and crack branching; (f) crack deection. of the solids) is calcium-silicate-hydrate (CSH), which is largely responsible for the strength and cementing properties of the paste. Calcium hydroxide (C H), a product of the calcium silicates hydration, occupies 20 to 25 pct of the solid volume, but does not contribute signicantly to the strength of the paste. The rate at which the clinker constituents react with water varies with time. The reactions start within seconds but may not reach completion for many years. Details of the actual mechanisms of hydration are not yet fully understood. One region which is the interface between the aggregate particles and the cement past plays an important role. This interface which is relatively weak is porous and contains a higher proportion of the calcium hydroxide and aluminum sulfates and less CSH than the bulk paste. Consequently, the interface region provides an easy path to fracture. The volume variations during hydration produce local stresses and more importantly, the formation of cracks. Concrete can therefore be considered as a multi cracked ceramic based material at various scales from the smallest scale corresponding to CSH platelets up to larger aggregates. A critical fracture stress is associated to each crack. However the formulation of the fracture criterion at this scale is not easy because of the complexity in the crack array. In particular, as concrete is usually loaded under compression, the crack mouth displacement is limited by friction effects. The most critical cracks start to propagate during loading but their propagation is arrested by the presence of aggregates along their path, as in CMC materials. To grow further, the cracks must deviate. The cement paste is brittle but concrete is much less brittle because of the stabilizing effect of granulates which favor multicracking. This effect plays a minor role in tension but a major one in compression, when the critical cracks propagate under the action of shear stresses (Modes II and III) and deviate, tending to move into Mode I. In a pure compression test, these deviations orient the cracks parallel to the axis
56
Figure II.22: Light micrograph illustrating the various scales in concrete material.
(a)
(b)
(c)
Figure II.23: Development of microcracks in concrete under increasing compressive loading from (a) to (c). of compression, in which direction they are no longer subjected to the load and are therefore stabilized (gure II.23). However in concrete the stress eld is not uniform at the scale of the aggregate and locally there can be tensile stresses sufcient to cause crack propagation. Rupture, resulting from very diffusive cracking mechanisms, is progressive and absorbs much more energy than it would be under tension when abrupt stress localization would generate instability. This is why concrete is much stronger in compression than in tension. 6 DUCTILE TO BRITTLE TRANSITION
6.1 INTRODUCTION
Ductile to brittle (DB) transition of fracture in steel structures is familiar to most mechanical engineers. Many cases starting with the famous molasses tank fracture in Boston in the winter of 1919, the welded Liberty ships breaking apart in World War II
57
and after, the ever continuing discussions on the Titanic, the collapse in earthquakes of highway superstructures, all keep the problem current. One important insight into the ductile or brittle phenomenon of fracture was due to Rice and Thomson (1974). These authors distinguished materials which are intrinsically ductile and not subject to a DB transition and those which are intrinsically brittle that can exhibit brittle behavior at low temperatures and high strain rates but may become ductile at higher temperatures. In the Rice and Thomson (1974) model, if an atomically sharp mode I-crack in a solid can emit dislocations from its tip that can then freely multiply before the crack propagates by cleavage, the solid is identied to be intrinsically ductile under all conditions. Crystalline materials capable of such response are relatively few and are restricted to the metals, Ag, Al, Au, Cu, Ni, Nb, Pb, Pt and Ta. According to this classication, all other metals and compounds, including inorganic glasses and possibly all polymers are intrinsically brittle. Nevertheless, they can exhibit ductile behavior above a certain temperature, for a given rate of loading. In engineering practice the principal familiarity in structural applications is with the DB transition. However, under certain circumstances the reverse process is also important. This has been investigated in detail in Si and also in steels when the conditions of crack arrest in a thermal gradient have been studied (see e.g. (Kanninen and Popelar, 1985; Di Fant et al., 1990; Bouyne et al., 2001)). The BD fracture transition involves a crack-tip initiated thermally-activated process of emission of dislocations followed by their rapid multiplication in the overstressed crack-tip environment. In this part we will only concentrate on the DB transition.
6.2 DUCTILE TO BRITTLE TRANSITION IN METALS
This transition is generally investigated using notched or pre-cracked specimens. Among the notched specimens the most popular is the Charpy V notch test which was invented more than a century ago and introduced in the late 1940s as a standard material test for welded steel construction. It was not until the 1960s, however, that Charpy test requirements were introduced into standards for bridges, buildings, low temperature pressure vessels and storage tanks (Folch and Burdekin, 1999). See chapter IV for more details on testing procedures.
6.2.1
Transition in smooth specimens
Consider rst a tensile test performed on a smooth test piece of a ferritic steel. If the temperature is lowered the yield strength will increase (gure II.24), the cleavage stress, will remain approximately constant and the two curves will meet at a temperature . When loaded at a little above this temperature the sample will begin to deform plastically when the stress reaches the yield strength, . The metal will then work-harden and cleavage will occur at increasingly higher temperatures the elongation at failure, , increases as the difference between and increases. Below T fracture does not occur before the yield strength is reached. The elongation is then null and the fracture strength is the same as the yield strength. The results obtained by Low (1954), who compared the tensile fracture strength with the compressive yield strength for a steel at low temperature, are given in gure II.25 and show this effect clearly. is the nil-ductility temperature for smooth specimens.
1 H
'
58
Figure II.24: Diagram showing the ductile-to-brittle transition for a smooth tensile specimen. 6.2.2 Effect of notches on the ductile-brittle transition
Below the nilductility temperature (NDT), at which temperature the yield strength is greater than the cleavage stress, a notched test piece will break when the former is reached (gure II.26). Fracture initiates at the root of the notch, where the stress is maximum before any overall plastic deformation occurs. Since for a notched cylindrical bar, the load at fracture is given by:
where is the radius of the minimum section of the notched specimen. Above this temperature, fracture occurs when the maximum stress, at the front of the plastic zone is then larger than the initiated from the notch, reaches the cleavage stress. Since yield strength (see chapter IV), the fracture load increases in proportion to the reduction in yield strength, caused by the increase in temperature. This continues until the limit load of the specimen is reached, i.e. for axisymmetrically notched specimens when the stress is maximum on the axis where, within the Bridgman approximation, it has the value:
averaging the stress distribution over the minimum section it can easily be shown that the fracture load is given by:
Above the temperature corresponding to these conditions , the maximum stress, always on the axis, continues to increase as a result of work-hardening. The fracture load
%)

'
1 #
1 #
cleavage ductile fracture (19)
(20)
(21)

59 (mm) 0.16 0.11 0.08
0.44 800 Yield stress or fracture stress (MPa) 700 600 500 400 300 200
0.25
0 1.0
1.5
2.0

2.5 (mm

3.0 )
Figure II.25: Comparison of tensile stress at fracture with yield strength in compression for mild steel tested at C, for various grain sizes (source: Low (1954)).
will remain constant since it has been assumed to be independent of temperature. However, the deformation of the test piece at the minimum section continues to increase and cleavage fracture is replaced by ductile fracture with dimples when the deformation on the specimen axis, where the stress triaxiality ratio is highest, reaches the corresponding critical fracture value (see equation (5)). Cavities rst initiate and then grow in the center of the specimen. After the BD transition temperature the variation of the fracture load parallels that of the limit load, the difference between the two depending on the elongation at fracture. This difference decreases with stress triaxiality which itself is a function of the specimen geometry. Figure II.26 is a schematic representation of the variation of fracture loads with temperature, derived from the above considerations. This gure applies to axisymmetrically notched specimens (NT) with a diameter in the minimum section of 10 mm and with different values of the notch radius (10 and 2 mm), NT2 specimens giving rise to much higher values for the stress triaxiality factor than NT10 specimens. Figure II.27 concerns the Charpy V notch specimen which is described in more detail in chapter IV and to which similar considerations can be applied. However, in this specimen when tested under impact conditions the situation is much more complex. The high strain rate and the resulting adiabatic heating conditions may combine to raise the ductile-brittle transition temperature. This acts as a factor of safety in the choice of a steel that will have to withstand a certain minimum temperature in working conditions. The choice will be made so that the transition temperature is below the service temperature. Many metallurgical factors may affect the DB transition temperature. This transition is associated with a change in the physical process of fracture. It does not exist in FCC metals, for which there is no cleavage. On the other hand, it does exist in BCC and CPH metals. In those materials, grain size is the most important
100
compressive yield strength ( tensile fracture stress (
C) C)
3.5
4.0
60
th
NT
10 NT 2
Figure II.26: Smooth and notched specimens. Variation with temperature of (a) fracture load and limit load, (b) strain to failure for smooth and two notched specimens. is the elongation at rupture for a smooth fully ductile specimen.
(b)
(a)
oo
sm
/NDT
smooth
NT10 NT2
/NDT
) ) %

)

61
Load
Load to fracture
Limit load
Notch root deformation
Temperature
wing formation
fracture at limit load
Figure II.27: Charpy V-notched specimens. Variation with temperature of limit load, fracture load and notch root deformation. The active plastic zone appears in light gray.
metallurgical parameters. Reducing the grain size raises the yield strength (Hall Petch relation, equation (14)) and the cleavage stress even more (equation (12)), resulting in a lowering of the transition temperature, as shown in gure II.28. 6.2.3 Ductile to brittle transition in precracked specimens

Fatigue precracked specimens are used to determine the tearing resistance curves or crack growth resistance, Rcurves (see chapter IV). In materials which can give rise to a ductile-brittle transition, the effect of a crack on this transition in fracture modes can be studied. Figure II.29 is a schematic representation of the situation observed when testing these materials. On the vertical axis we have plotted the ratio since, it can be shown that, within a rst approximation, this quantity is independent of test temperature. At a given temperature, after some ductile crack growth, , cleavage fracture is abruptly initiated. This is mainly due to the intensication of the tensile stresses ahead of a propagating crack and to the increase in the amount of material which is sampled by the crack (see size effect in brittle fracture, section 3.1.3). These amplication effects are such that the cleavage (or Weibull) stress is reached. However,

1 H
62
80 70 60 50 40 80 Transition temperature ( C) 60 40 20 0 -20 -40 1.2
1.4
1.6
Figure II.28: Effect of grain size on the ductile-to-brittle transition temperature for a 0.11% carbon steel. as indicated previously (section 3.1.3), cleavage fracture is a statistical phenomenon with some scatter. This is illustrated in gure II.29 where the curves corresponding to failure probabilities of 10% and 90% are drawn at a given temperature. When the temperature is increased, brittle fracture becomes less and less likely and occurs after increasing crack growth, , as shown schematically on this gure. A number of empirical relations have been proposed in the literature to relate the DB transition temperatures determined with Charpy V notched specimens and precracked specimens. These empirical correlations are useful for engineering purposes, but they are not based on a solid physical and mechanical analysis of the phenomena involved in these specimens loaded at different rates. This strongly limitates their applicability. However, more recently, a large research effort has been made to model both types of specimens and, using the local approach methodology, to relate the results obtained on both types of test (Francois and Pineau, 2002).
6.3 DUCTILE TO BRITTLE TRANSITION IN POLYMERS
Many aspects of the ductile to brittle transition observed in metallic materials, including the temperature, strain-rate effect and the notch effect, do apply also to polymers. However there are a number of effects specic to polymers. The yield strength of polymers is much more sensitive to strain rate and temperature than metals. In those materials a typical gure would be a factor of 10 over the temperature range C to +20 C, compared to a factor of about 2 in medium strength steels. There is also a further complication in varying strain rate in polymers, which is due to the low thermal conductivity of these materials. This can produce an isothermal-adiabatic transition when the strain rate is sufciently increased. Adiabatic heating does not affect the yield strength and does not also affect too much the ductile to brittle transition but it does cause a considerable reduction in fracture energy. This may

C

( m) 15 12 10
30 25 20
'%
1.8 2.0 ( in mm)
2.2
2.4
63
10% 90% (mm)

10%
10% y ilit 90% bab pro re ilu fa
fatigue precrack

Ductile crack growth,
Figure II.29: Schematic diagram showing the variation of integral with crack growth and the occurrence of cleavage fracture at increasing temperatures. be the situation occurring in impact tests, even if brittle fracture does not occur. It has therefore been suggested that are two critical impact velocities at which the fracture energy drops sharply as the strain rate is increased. First, there is the isothermaladiabatic transition, and secondly, at higher strain rates, the ductile to brittle transition. As we would expect, changes in testing temperature have very little effect on the position of the isothermal-adiabatic transition, but a large effect on the ductile to brittle transition. In polymers the molecular weight does not seem to have a strong inuence on the yield strength, but it is suggested that the fracture strength is related to the number average molecular weight, by the relationship.
This relationship holds to a rough approximation for the brittle strength of several polyethylenes (PE), polymethyl methacrylates (PMMA) and polystyrenes (PS). The yield strength of these polymeric materials could differ appreciably with the degree of branching which affects the crystallinity so that the temperature of the ductile to brittle transition would be a complex function of molecular weight and branch content. It appears that in polymers considerations similar to those employed to explain the ductile to brittle transition in metals can be applied. In polymers it can also be assumed that, within a rst approximation, the fracture stress, , is not very dependent of test temperature and strain rate (gure II.30). On the other hand, as indicated previously the yield strength of polymers is largely dependent of temperature and strain rate, as illustrated in gure II.30, and, as observed for instance in PMMA (Vincent, 1964). This would suggest that the shift in the ductile-to-brittle transition temperature would
90%

ductile

cleavage
(mm)
(22)
64
fracture stress
stress
yield stress
Temperature
Figure II.30: Schematic diagram showing the effect of strain rate on the variation with temperature of the yield strength and the fracture stress in polymers. be larger in polymers than in metals, but as indicated previously, this may be largely complicated by adiabatic heating effects. The ductile-to-brittle transition behavior of polymers is also affected by notch effect. However modeling of this effect is much less developed than in metallic materials due to the complexities specic to polymers. A diagram has been proposed by Vincent (1964). This diagram is shown in gure II.31 where the circles represent and at C, the triangles and at C. These two temperatures were chosen with the idea that the C values would give a rough indication of the behavior in impact at C, i.e. lowering the temperature by C is assumed to be equivalent to increasing the strain rate by a factor of about . From the known behavior of 13 polymers shown in gure II.31, two characteristic lines can be drawn. Line A divides polymers which are brittle unnotched from those which are ductile unnotched but brittle notched, while line B divides polymers which are brittle notched but ductile unnotched from those which are ductile even notched. However, in spite of the engineering interest of these distinctions, the modeling efforts have not yet been sufciently developed to give a fully comprehensive description of these transitions in failure modes of polymers. 7 CREEP FRACTURE
7.1 INTRODUCTION TO CREEP DEFORMATION
Creep of materials is classically associated with time-dependent plasticity under a , where constant stress at an elevated temperature, often larger than roughly is the absolute melting temperature. The plastic deformation under these conditions is schematically described in gure II.32 for constant stress (a) and constant strain
$B #
B B P
) B
B )
) B
) B
65
300
PC
N POM
Brittle fracture stress (MPa)
250 200
LDPE PP
PVC PTP CA PMMA
SAN PB
150 100 50 0 0
PTFE
PS
20
40
60 80 100 120 140 160 180 Yield strength (MPa)
Figure II.31: Brittle fracture stress at about C versus yield strength at C or 20 C (0) for various polymers. Line A divides polymers which are brittle unnotched from those which are ductile unnotched but brittle notched, and line B divides polymers which are brittle notched, but ductile unnotched, but from those which are ductile even when notched. PMMA: Polymethylmetracrylate, PS: Polystyrene, SAN: Copolymer of styrene and acrylonitrile, N: Nylon 66, POM: Polyoxylethylene, PC: Polycarbonate, PVC: Polyvinyl chloride, PTP: Polyethylene terephthalate, CA: cellulose acetate, PP: Polypropylene, LDPE: Low density polyethylene, PB: polybutene-1, PTFE: Polytetrauoroethylene (source: (Ward, 1985)).

rate (b). Three regions in gure II.32a are delineated: (i) Stage I or primary creep where the creep rate decreases with increasing plastic strain or time; (ii) Stage II or secondary, or steady-state creep; (iii) Stage III, or tertiary creep where an increase of the strain rate (gname II.32a) or a decrease of the ow stress (gname II.32b) is observed. Sometimes, under constant stress, Stage I leads directly to Stage III and an inexion in the curves is not observed. Thus, care must be often taken in concluding on the existence of a steady state regime. The micromechanisms controlling creep deformation involve not only dislocation glide but also dislocation climb and, at low stresses, diffusion controlled mass transport along the grain boundaries. It is therefore not surprising to observe that, in many cases, the activation energy, , appearing in the Norton law:
is close to that corresponding to diffusion. In this expression, is a constant while the Norton exponent varies between 1 (diffusion controlled creep) to 46 for thermally activated dislocation creep.
B )
B
&
(23)
66
strain
II
III
II time
time
Figure II.32: Constant true stress and constant strain rate creep behavior in metals.
Figure II.33: Scanning electron micrograph illustrating the formation of wedgetype intergranular cracks in austenitic stainless steel (source: (Auzoux, 2003)).
7.2 BACKGROUND
Creep deformation can lead to tertiary or Stage III creep and failure. For a recent review, see eg. (Kassner and Hayes, 2003). It has been suggested that creep fracture can occur by w or wedgetype cracking, illustrated in gure II.33, at grain boundary triple points. A number of authors have suggested that wtype cracks form most easily at higher stresses (lower temperatures) and larger grain sizes. Others have suggested that the wtype cracks nucleate as a consequence of grain boundary sliding. Another mode of failure has been associated with rtype (round) cavities illustrated in gure II.34. It should be noted that the wtype cracks may be brittle in origin or simply as resulting from an accumulation of rtypes voids (Chen and Argon, 1981a,b)). Inasmuch as w type cracks are related to rtype cavities, it is enough to devote this short summary of creep fracture to cavitation. Many reviews on creep fracture have been published (see eg. (Cocks and Ashby, 1982; Nix, 1988; Needleman and Rice, 1980; Riedel, 1987; Evans, 1984)). In this summary only a short account of the mechanisms responsible for creep fracture is given. Creep fracture in uniaxial tension under constant stress is usually described by the MonkmanGrant relationship (Monkman and Grant, 1956) which states that creep
const.
const.

III
stress
"B #
#
$B #
67
wtype
(b) Figure II.34: Intergranular creep cavities in creeping metals; (a) schematic diagram; (b) wedgetype cracks and cavities observed in austenitic strainless steel (source: (Auzoux, 2003)). fracture is controlled by the steady-state creep rate:
where is referred to as the MonkmanGrant constant and is a constant typically . In many materials it has been observed that follows a S shape curve about when plotted as a function of strain rate, (gure II.35). The lower value of the MonkmanGrant constant corresponds to intergranular fracture (gure II.36) while the value corresponding to the upper shelf is associated with the fracture mode operating at higher strain rates, i.e. ductile rupture (see section 2). Another relationship to predict rupture times, utilizes the LarsonMiller (LM) parameter (Larson and Miller, 1952) described by:
where is the absolute temperature. The constant is empirically determined and is suggested to be about 20, independent of the material. One difculty with equations (24) and (25) is that the constants determined in a creep regime, with a given ratecontrolling mechanism, may not be reliable for extrapolation to the rupture times corresponding to another creep regime where the constants may actually change. This is why in creep fracture like in other fracture modes, it is extremely important to investigate the damage micromechanisms before
'

(a)
rtype
(24)
(25)
68
MonkmanGrant constant (
intergranular
strain rate ,
(s
Figure II.35: Schematic variation of the MonkmanGrant constant, C with creep strain rate in metals. Note the existence of a transition in fracture mode from transgranular at high strain rate to intergranular at low strain rate. formulating a sound model. The creep fracture mechanisms that will be shortly discussed are those resulting from the nucleation cavities followed by growth and interlinkage leading, as in ductile fracture (see section 2) to catastrophic failure.
7.3 CAVITY NUCLEATION
Creep cavities nucleate on grain boundaries, particularly on those transverse to a tensile stress. In commercial alloys, the cavities appear to be associated with second phase particles located along the grain boundaries. The nucleation micromechanisms fall into several categories that are illustrated in gure II.37: (a) grain boundary sliding leading to voids at the head of a triple point of a boundary or formation of voids by tensile grain boundary ledges; (b) vacancy condensation, usually at grain boundaries at areas of large stress concentration; (c) cavity formation at the head of a dislocation pile-up; (d) cavity formation from a particle. Creep cavitation is usually a continuous process, i.e. the density of creep cavities increases with creep strain. The application of the classical nucleation theory for the formation of cavities by vacancy condensation indicates that the size of critical-sized nuclei is about 25 nm, which are difcult to detect. Interestingly, observations of cavity nucleation not only suggest continuous cavitation but also no incubation time and that strain rather than time is more closely associated with nucleation.
7.4 DIFFUSION CONTROLLED CAVITY GROWTH
Three extreme cases must be distinguished depending on the mechanisms controlling cavity growth. In the rst case, growth occurs and is limited by vacancy diffusion. In
4 B
4 B
mi xed
B 4
transgranular
69
Figure II.36: Scanning electron fractograph showing intergranular fracture in austenitic stainless steel tested under creep conditions at 600 C (source (Auzoux, 2003)). the second case, cavity growth is controlled by surface diffusion along the faces of the cavities. The third case corresponds to the situation where diffusion cavity growth is controlled by matrix viscoplasticity. 7.4.1 Grain boundary control
The cavity growth process at grain boundaries at elevated temperature has long been suggested to involve vacancy diffusion. Diffusion occurs by cavity surface migration and subsequent transport along the grain boundary or within the grain, with either grain boundary diffusive mechanism or surface diffusion having been suggested to be rate controlling, depending on the specic or higher strain rates where cavity growth is controlled by plasticity (see section 2). Hull and Rimmer (1959)) were the rst authors to propose a mechanism by which diffusion leads to cavity growth of an isolated cavity in a material under an applied external stress, . The mechanism is schematically illustrated in gure II.38. The growth rate of the cavity of radius, a, is given by:
where is the intergranular diffusion coefcient, is the grain boundary width, is the atomic volume, while is the surface energy. By integrating equation (26) between the critical radius , below which sintering of the cavity occurs and (gure II.38), one obtains:
(27)
This is the rst relationship between stress and rupture time for unconstrained cavity growth. Improvements to the above relationship have been made by a number of authors but, in all cases, a linear stress dependence and an activation energy of grain boundary diffusion is predicted.

0
&
)
)
)
) &
&

(26)

) %

70 cavities
GBS tensile ledge
(a)
compressive ledge
(b)
Grain boundary Cavity GBS
particle
(c)
(d)
Figure II.37: Creep cavity nucleation mechanisms: (a) sliding leading to cavitation from grain boundary ledges and triple points; (b) vacancy condensation at a highly stressed region; (c) dislocation pile-up; d) formation of a cavity from a particleobstacle in conjunction with the mechanisms described in (a-c).
7.4.2
Surface diffusion control
A number of authors (Chuang and Rice, 1973; Needleman and Rice, 1980) suggested that surface rather than grain boundary diffusion may actually control cavity growth and that this assumption gives rise to a three-power stress relationship for cavity growth at low stresses:

The complication with these growth relationships (equations (26) & (27)) is that they are inconsistent with the MonkmanGrant relationship (equation (23)). In common materials with creep exponent equal to 5 ( in equation (22)), the MG relationship suggests that the cavity growth rate should be proportional to the stress to the fth power rather than 13 power. This may be associated to the importance of nucleation in the rate controlling process for creep cavitation failure. (see e.g. (Dyson, 1983)).
#
where
is the surface diffusion coefcient.
) &
GBS
GBS
(28)
71
Figure II.38: Cavity growth from diffusion across the cavity surface and through the grain boundaries due to a stress gradient. 7.4.3 Constrained diffusional cavity growth
Cavity formation may be heterogeneous, as schematically illustrated in gure II.39. Fracture could then be controlled by the viscoplastic creep rate in uncavitated regions that can lead to cavity nucleation. This leads to a better consistency with the MG relationship (Dyson, 1976; Rice, 1981)). Constrained diffusional growth which was originally suggested by Dyson (1976) is described by the following relationship (Riedel, 1987):

where is the fraction of cavitated grain boundaries, is a threshold stress, is the stationary creep stress while is the grain size. One notes that for higher strain rates, where the increase in volume from cavity growth can be easily accommodated by matrix viscoplasticity, the growth rate is primarily a function of the grain boundary diffusion coefcient similar to equation (26). It has been shown (Riedel, 1987) that if only certain grain boundary facets cavitate, then the time for coalescence, on these facets can be calculated:
However failure is not expected to occur by a simple coalescence of cavities on isolated facets. Additional time may be required to join facet-size microcracks. The mechanism of joining the facets may be rate controlling. It appears likely, however, that this model can explain the longer times to failure than those expected from unconstrained cavity growth. One must also consider that creep cavities are continuously nucleated. For continuous nucleation and unconstrained diffusive, it has been suggested (Riedel, 1984,
% & %
&
P ) B
# $
boundary diffusion
boundary diffusion
&
surface diffusion
&
& P B B B
"#
(29)
(30)
72
Figure II.39: Heterogeneous creep cavitation at especially transverse grain boundaries. 1987):
is the critical cavitated area fraction and is the nucleation rate. Assuming where that the nucleation rate is proportional to the strain rate, i.e.:
with equation (30) predicts that the time to failure is inversely proportional to the stress power , i.e.:

which is closer to the MG relationship than the previous relationships (equations (26) & (27)) ignoring continuous cavity nucleation.
'
%)
#
&

&
#
(31)
'
(32)
(33)
II. Physical mechanisms of damage 8 BIBLIOGRAPHY
73
Acker, P., 2001. Universit de tous les savoirs. Quest-ce que les technologies. Vol. 5. e Editions Odile Jacob, Ch. Les B tons, pp. 362371. e Al Mundheri, M., Soulat, P., Pineau, A., 1989. Irradiation embrittlement of a low alloy steel interpreted in terms of a local approach of cleavage fracture. Fatigue and Fract. Eng. Mater. Structures 12, 1930. Amar, E., 1986. Application de lapproche locale de la rupture a l tude de la transition ` e ductile-fragile dans lacier 16MND5. Ph.D. thesis, Ecole des Mines de Paris. Argon, A., 1973. Physical basis of distortional and dilational plastic ow in glassy polymers. J. Macromol. Sci. Phys. B 8, 573596. Argon, A., Im, J., Safoglu, R., 1975. Cavity formation from inclusions in ductile fracture. Met. Trans. 6A, 825837. Auzoux, Q., 2003. Ecole des Mines de Paris, France, private communication. Benzerga, A., Besson, J., Pineau, A., 1999. Coalescencecontrolled anisotropic ductile fracture. J. Eng. Mat. Eng. 121, 221229. Beremin, F., 1981. Cavity formation from inclusions in ductile fracture of A508 steel. Met. Trans. 12A, 723731. Beremin, F. M., 1983. A local criterion for cleavage fracture of a nuclear pressure vessel steel. Met. Trans. 14A, 22772287. Berg, C., 1962. The motion of cracks in plane viscous deformation. In: Proceedings of the 4th U.S. National Congress on Applied Mechanics (9th ed.). Vol. 2. Bernauer, G., Brocks, W., Schmitt, W., 1999. Modications of the Beremin model for cleavage fracture in the transition region of a ferritic steel. Engng Fract. Mech. 64, 305325. Berveiller, M., Zaoui, A., 1978. An extension of the selfconsistent scheme to plastically owing polycrystals. J. Mech. Phys. Solids 26, 325344. Bouyne, E., Flower, H., Lindley, T., Pineau, A., 1998. Use of ebsd technique to examine microstructure and cracking in a bainitic steel. Scripta Mater. 39, 295300. Bouyne, E., Joly, P., Houssin, B., Wiesner, C., Pineau, A., 2001. Mechanical and microstructural investigations into the crack arrest behaviour of a modern 2 Cr-1Mo pressure vessel steel. Fat. Fract. of Eng. Mater. Structures 24, 105116. Bretheau, T., Bornert, M., Gilormini, P. (Eds.), 2001. Homog n isation en m canique e e e des mat riaux, I & II. Hermes. e Brown, H., 1991. A molecular interpretation of the toughness of glassy polymers. Macromolecules 24, 27522756. Brozzo, P., Buzzichelli, G., Mascanzoni, A., Mirabile, M., 1977. Microstructure and cleavage resistance of low carbon bainitic steels. Metal Science 11, 123129. Budiansky, B., Hutchinson, J., Slutsky, S., 1982. Mechanics of solids The Rodney Hill 60th Anniversary Volume. Pergamon, Ch. Void growth and collapse in viscous solids, pp. 1345. Bugat, S., Besson, J., Pineau, A., 1999. Micromechanical modeling of the behavior of duplex stainless steels. Computational Materials Science 16 (14), 158166. Chen, I., Argon, A., 1981a. Creep cavitation in 304 stainless steel. Acta Metall. 29, 13211333. Chen, I., Argon, A., 1981b. Diffusive growth of grain boundary cavities. Acta Metall. 29, 17591768. Chuang, T., Rice, J., 1973. The shape of intergranular creep cracks growing by surface diffusion. Acta Metall. 21, 16251628. Cialone, H., Asaro, R., 1979. The role of hydrogen in the ductile fracture of plain
74
carbon steels. Met. Trans. 10A, 367375. Cocks, A., Ashby, A., 1982. Creep fracture by coupled powerlaw creep and diffusion under multiaxial stress. Metal Sci. 16, 465474. Cottrell, A., 1958. Theory of brittle fracture in steel and similar metals. Trans. AIME 212, 192203. Cottrell, A., 1989. Strengths of grain boundaries in pure metals. Mat. Science and Technology 5, 11651167. Cottrell, A., 1990a. Strength of grain boundaries in impure metals. Mat. Science and Technology 6, 325329. Cottrell, A., 1990b. Strengthening of grain boundaries by segregated interstitials in iron. Mat. Science and Technology 6, 121123. Cr pin, J., 1995. M canismes de d formation et dendommagement du zirconium e e e grade 702 trait . application aux cordons de soudure. Ph.D. thesis, Ecole e Polytechnique. Curry, D., Knott, J., 1978. Effect of microstructure on cleavage fracture stress in steel. Metal Science 12, 511514. Decamp, K., Bauvineau, L., Besson, J., Pineau, A., 1997. Size and geometry effects on ductile rupture of notched bars in a CMn steel: Experiments and modelling. Int. J. Frac. 88 (1), 118. Devillers-Guerville, L., Besson, J., Pineau, A., 1997. Notch fracture toughness of a cast duplex stainless steel: modelling of experimental scatter and size effect. Nuclear Engineering and Design 168, 211225. Di Fant, M., Genty, A., Pineau, A., 1990. Thermal-stock-induced crack arrest of two low alloy steels. High Temperature Technology 8, 105114. Dyson, B., 1976. Constraints on diffusional cavity growth rates. Metal Sci. 10, 349 353. Dyson, B., 1983. Continuous cavity nucleation and creep fracture. Scripta Met. 17, 3137. Eshelby, J., 1961. Prog. Solid Mech. Amsterdam-North-Holland. Evans, H., 1984. Mechanisms of creep fracture. Elsevier Applied Science. Faber, K., Evans, A., 1983. Crack deection processes i. theory. Acta Metall. 31, 565576. Folch, L., Burdekin, F., 1999. Application of coupled brittle-ductile model to study correlation between charpy energy and fracture toughness values. Engng Fract. Mech. 63, 5780. Francois, D., Pineau, A., 2001. Physical Aspects of Fracture. Nato Science Series, Ch. Fracture of metals, pp. 125146. Francois, D., Pineau, A., 2002. From Charpy to present impact testing. ESIS Publica tion 30, Elsevier. Francois, D., Pineau, A., Zaoui, A., 1998. Mechanical Behaviour of Materials. Kluwer Academic Publishers. Freiman, S., 1988. Brittle fracture behavior of ceramics. Am. Ceram. Soc. Bull. 67, 392402. French, I., Weinrich, P., 1974. The inuence of hydrostatic pressure on the tensile deformation of a spheroidised 0.5% c steel. Scripta Met. 8, 8790. Goods, S., Brown, L., 1979. Overview N I : The nucleation of cavities by plastic deformation. Acta Metall. 27, 115. Gurson, A., 1977. Continuum theory of ductile rupture by void nucleation and growth: Part I Yield criteria and ow rules for porous ductile media. J. Eng. Mat. Technol. 99, 215.
75
Han, H., McLeish, T., Duckett, R., Ward, N., Johnson, A., Donald, A., Butler, M., 1998. Experimental and theoretical studies of the molecular motions in polymer crazing: I. Tube model. Macromolecules 31, 13481357. Hansson, D., 1995. Concrete : the advanced industrial material of the 21 century. Metall. and Mat. Trans. 26A, 13211341. Hausild, P., 2002. Transition ductilefragile dans un acier faiblement alli . Ph.D. thesis, e Ecole Centrale Paris. Holzmann, M., Vlach, B., Man, J., 1991. The ductile-to-brittle transition of a pressure vessel steel embrittled by step cooling heat-treatment. In: Defect Assessment in ComponentsFundamentals and Applications. ESIS/EGF9, Mechanical Engineering Publications, London. Huang, Y., 1991. Accurate dilatation rates for spherical voids in triaxial stress elds. J. Appl. Mech. 58, 10841086. Hull, D., Rimmer, D., 1959. The growth of grain boundary voids under stress. Phil. Mag. 4, 673687. Jaeckels, H., Iung, T., Pineau, A., 1994. Dynamic crack propagation and crack arrest behaviour in relation to brittle intergranular and cleavage fracture. Fat. Fract. Eng. Mater. Structures 17, 12811293. Joly, P., Cozar, R., Pineau, A., 1990. Effect of crystallographic orientation of austenite on the formation of cleavage cracks in ferrite in an aged duplex stainless steel. Scripta Met. 24, 22352240. Joly, P., Pineau, A., 1991. Local versus global approaches to elastic-plastic fracture mechanics. Application to ferritic steels and a cast duplex stainless steel. In: Defect Assessment in ComponentsFundamentals and Applications. Mechanical Engineering Publications, London. Kameda, J., Mc Mahon, C., 1980. Solute segregation and brittle fracture in an alloy steel. Met. Trans. 11A, 91101. Kanninen, M., Popelar, C., 1985. Advanced fracture mechanics. Oxford University Press, New York. Kantidis, E., 1993. Rupture fragile intergranulaire dun acier faiblement alli . Ape proches globale et locale. Ph.D. thesis, Ecole des Mines de Paris. Kassner, M., Hayes, T., 2003. Creep cavitation in metals. Int. J. of Plasticity 19, 1715 1748. Knauss, W., 1993. Time dependent fracture and cohesive zones. J. Eng. Mat. Tech. 115, 262267. Kwon, D., Asaro, R., 1990. A study of void nucleation, growth, and coalescence in spheroidized 1518 steel. Met. Trans. 21A, 117134. Lambert-Perlade, A., Gourgues, A., Besson, J., Sturel, J., Pineau, A., 2003. Mechanisms and modelling of cleavage fracture in simulated heat affected zones microstructure of a high strength low alloy steel. to appear in Met. Trans. . Larson, F., Miller, J., 1952. A time-temperature relationship for rupture and creep stresses. Trans. ASME 74, 765775. Lefebvre, X., 2002. Fissuration fragile lente du polyamide 11 : m canismes et dur es e e de vie en uage. Ph.D. thesis, Ecole des Mines de Paris. Liang, Y., Sofronis, P., 2003. Toward a phenomenological description of hydrogeninduced decohesion at particle/matrix interfaces. J. Mech. Phys. Solids 51, 1509 1531. Low, J., 1954. Relation of properties to microstructure. ASM, Novelty, Ohio, Ch. The relation of microstructure to brittle fracture, pp. 163179. Lundberg, R., 1989. Fiber reinforced ceramic composite. Ph.D. thesis, Department of

76
Inorganic Chemistry G teborg. o Margolin, B., Shvetsova, V., Karzov, G., 1997. Brittle fracture of nuclear pressure vessel steelsi. local criterion for cleavage fracture. Int. J. Pres. Ves. and Piping 72, 7387. Marini, B., Mudry, F., Pineau, A., 1985. Ductile rupture of a508 steel under non radial loading. Engng Fract. Mech. 22, 375386. Martin-Meizoso, A., Oca a Arizcorreta, I., Gil-Sevillano, J., Fuentes-P rez, M., 1994. n e Modelling cleavage fracture of bainitic steels. Acta Metall. Mater. 42, 20572068. Mc Clintock, F., 1968. A criterion for ductile fracture by the growth of holes. J. App. Mech. 35, 363371. Mc Clintock, F., 1971. Fracture. Vol. 3. Academic Press , New York and London, Ch. Plasticity aspects of failure, pp. 47225. Mc Clintock, F., Argon, A., 1966. Mechanical Behaviour of Materials. AddisonWesley Pub., Reading. Monkman, F., Grant, N., 1956. An empirical relationship between rupture life and minimum creep rate in creep-rupture test. Proc. ASTM 56, 593620. Naudin, C., Frund, J., Pineau, A., 1999. Intergranular fracture stress and phosphorus grain boundary segregation of a MnNiMo steel. Scripta Mater. 40, 10131019. Naudin, C., Pineau, A., Frund, J., 2001. Toughness modelling of pwr vessel steel containing segregated zones. In: Proc. 10 Conf. on environmental degradation of materials in nuclear power systems-water reactors, Lake Tahoe, Nevada (USA). Naylor, J., Krahe, P., 1974. The effect of the bainitic packet size on toughness. Met. Trans. 6A, 16991701. Needleman, A., Rice, J., 1980. Plastic creep ow effects in the diffusive cavitation of grain boundaries. Acta Metall. 28, 13151332. Nix, W., 1988. Mechanisms and controlling factors in creep fracture. Mater. Sci. Eng. A 103, 103110. Perrin, G., 1992. Contribution a l tude th orique et num rique de la rupture ductile ` e e e des m taux. Ph.D. thesis, Ecole Polytechnique. e Pineau, A., 1981. Review of fracture micromechanisms and a local approach to predicting crack resistance in low strength steels. In: Advances in Fracture Research, ICF5. Pineau, A., 1992. Global and local approaches to fracture Transferability of laboratory test results to components. In: Argon, A. (Ed.), Topics in Fracture and Fatigue. Springer Verlag Inc., NY. Pineau, A., Besson, J., 2000. Some new trends in modeling ductile rupture of structural alloys. In: Continuous damage and fracture. Pineau, A., Joly, P., 1991. Local versus global approaches to elastic-plastic fracture mechanics. Application to ferritic steels and a cast duplex stainless steel. In: Defect Assessment in ComponentsFundamentals and Applications. ESIS/EGF9, Mechanical Engineering Publications, London. Rice, J., 1981. Constraints on the diffusive cavitation of isolated grain boundary facets in creeping polycristals. Acta Metall. 29, 675681. Rice, J., Thomson, R., 1974. Ductile versus brittle behaviour of crystals. Philos. Mag. 29, 7397. Rice, J., Tracey, D., 1969. On the ductile enlargement of voids in triaxial stress elds. J. Mech. Phys. Solids 17, 201217. Riedel, H., 1984. Cavity nucleation at particles on sliding grain boundaries. a shear crack model for grain boundary sliding in creeping polycrystals. Acta Metall. 32, 313321.

77
Riedel, H., 1987. Fracture at high temperatures. SpringerVerlag, Berlin. Roseneld, A., Shetty, D., Skidmore, A., 1983. Fractographic observations of cleavage initiation in the ductile-brittle transition region of a reactor-pressure vessel steel. Met. Trans. 14A, 19341937. Rossoll, A., Berdin, C., Prioul, C., 2002. Determination of the fracture toughness of a low alloy steel by the instrumented Charpy impact test. Int. J. Frac. 115, 205226. Rousselier, G., 1987. Ductile fracture models and their potential in local approach of fracture. Nuc. Eng. Design 105, 97111. Sha, Y., Hui, C., Ruina, A., Kramer, E., 1997. Detailed simulation of craze bril failure at a crack tip in a glassy polymer. Acta Mater. 45, 35553563. Smith, E., 1966. The nucleation and growth of cleavage microcracks in mild steel. In: Physical basis of Yield and Fracture, Conf. Proceedings Inst. Phys. and Phys. Soc. London. Sternstein, S., Myers, F., 1973. Yielding of glassy polymers in the second quadrant of principal stress space. J. Macromol. Sci. Phys. B8, 529571. Stroh, A., 1954. The formation of cracks as a result of plastic ow. Proc. Roy. Soc. 223, 404414. Swanson, P., Fairbanks, C., Lawn, B., Mai, Y., Hockey, B., 1987. Crackinterface grain bridging as a fracture resistance mechanism in ceramics. I :Experimental study on alumine. J. Amer. Ceram. Soc 70, 279289. Tanaka, K., Mori, T., Nakamura, T., 1970. Cavity formation at the interface of a spherical inclusion in a plastically deformed matrix. Phil. Mag. 21, 267279. Tanguy, B., 2001. Mod lisation de lessai Charpy par lapproche locale de la rupture. e Application au cas de lacier 16MND5 dans le domaine de la transition. Ph.D. thesis, Ecole des Mines de Paris. Thomason, P. F., 1968. A theory for ductile fracture by internal necking of cavities. J. Ins. Metals 96, 360365. Thomason, P. F., 1985a. A threedimensional model for ductile fracture by the growth and coalescence of microvoids. Acta Metall. 33 (6), 10871095. Thomason, P. F., 1985b. Threedimensional models for the plastic limitloads at incipient failure of the intervoid matrix in ductile porous solids. Acta Metall. 33 (6), 10791085. Thompson, A., Williams, J., 1977. Nuclei for ductile fracture in titanium. In: Proc. ICF4. Vol. 2. Tijssens, M., Van der Giessen, E., Sluys, L., 2000. Modeling of crazing using a cohesive surface methodology. Mechanics of Materials 32, 1935. Vincent, P., 1964. Strength of plastics. conclusion yield stress and brittle strength. Plastics 29, 79. Wallin, K., 1991. Statistical modelling of fracture in the ductile to brittle transition region. In: Defect Assessment in ComponentsFundamentals and Applications. ESIS/EGF9, Mechanical Engineering Publications, London. Walsh, J., Jata, K., Starke, E., 1989. The inuence of Mn dispersoid content and stress state on ductile fracture of 2134 type Al alloys. Acta Metall. 37, 28612871. Ward, I., 1985. Mechanical properties of solid polymers. 2 Edition. John Wiley & Sons. Xu, X., Needleman, A., 1994. Numerical simulations of fast crack growth in brittle solids. J. Mech. Phys. Solids 42, 13971434. Yahya, O., 1997. Approche locale de la rupture fragile intergranulaire de lacier 16MND5. Ph.D. thesis, Ecole des Mines de Paris. Zhang, Z., Niemi, E., 1994. Analyzing ductile fracture using dual dilational constitutive

78 equations. Fat. Frac. Eng. Mater. Structures 17, 695707.
Chapter III
QUANTITATIVE MEASUREMENT OF DAMAGE
by Eric Maire
INTRODUCTION
As explained in details in the second chapter of this book, damage occurs by the nucleation of voids or cracks. Most of the time this nucleation takes place in special sites of the microstructure which can be referred to as inclusions. The presence of these inclusions has been observed in most of the materials. Their content can for instance be calculated in steels from the knowledge of the chemical composition in oxygen (%O), mangan` se (%Mn) and sulfur (%S) using the formula given in e Franklin (1969): %S %O
This kind of useful formula is not known for every material but in any case, the nature and content of second phase particles can be determined by different methods, some of which being described in this chapter. These inclusions play a key role on damage, acting both as weak points and/or stress concentrators. Damage initiation by inclusion, interface or matrix fracture can be observed depending on the kind of material studied or on the kind of loading applied to the sample. Damage can have a very strong inuence on the global behavior of a material especially in the case of composites where the amount of inclusions is very important. Its importance is even stronger at a local scale, as damage is the phenomenon responsible for the rupture of materials. Therefore an experimental characterization of damage is required. This characterization helps to understand qualitatively the damage mechanism which then guides a physical modelling of this process. Damage quantication also helps to validate the proposed theoretical tools. The number of industrial and academic studies devoted to the development of damage in materials which can be found in the literature has always been very large. The majority of this book presents the result of these studies mainly in terms of modelling methods feeded and then validated with key experiments. The present chapter summarizes the experimental methods available to qualify and then to quantify damage. These methods can be classied in two categories:
indirect methods which measure a physical or a mechanical parameter modied by the presence of damage, 79
%Mn
) )

# B B "B

B B B
"#
P B B
#
# $
(1)
80
Local approach to fracture direct methods which give an image of the microstructure in which damage can be detected.
2 INDIRECT MEASUREMENT Several methods can give an indirect information about the amount of damage generated by the thermomechanical loading of a material. These methods are based on the modication of a physical or mechanical property induced by the presence of the discontinuities associated to damage. If this modication can be measured with a sufcient precision and if the relation between the amount of damage and its effect on the parameter measured is well known, it becomes possible to indirectly quantify one from the measurement of the other. Several of these properties have been used over the years to detect and measure damage. These measurements can be used to assess the initial amount of damage. In this case, standards of the material, for which the value of the physical or mechanical parameter in the absence of damage is well known, are required. These standards are not easy to obtain. As a consequence, in reality, these methods are more likely used to follow the evolution of damage induced by a given loading (which is most of the time, a thermomechanical loading). In this latter case, the reference state is the one before loading assumed to be free or nearly free of damage and one measures the variation of the parameter induced by the loading. We will focus on this procedure in the present chapter. The two most popular measurements are those relative to the reduction of the macroscopic density and that of the Youngs modulus of a sample. These two methods will rstly be described in details in the following but this will lead us to the conclusion that in some cases these two parameters do not provide a sufcient information on the nature of damage. If this is the case, other physico-mechanical methods can be used. They will be described at the end of the present section.
2.1 DENSITY
It can be of great importance to experimentally follow the amount of porosity in a material. It is one of the key measurements in the eld of any approach to rupture. Some very popular models (for instance the Gurson model described in chapter VIII) have also been developed to predict this parameter and its inuence on the behavior. Rupture is believed in some approaches (Mc Clintock, 1968; Beremin, 1981) to be governed by a critical value of this parameter . Density uctuation induced by damage growth is certainly the oldest way of quantifying . 2.1.1 Principle of the measurement

Most of the structural materials are incompressible and then normally deform at nearly constant volume in the absence of damage. When damage occurs the volume of the material increases. A change in the amount of porosity leads to a corresponding increase in the volume: (3)
The density volume :
of a sample is equal to its total mass
divided by its total external (2)
III. Quantitative measurement of damage
81
Most of the time, doesnt change during the thermo-mechanical deformation of a material. The change in the amount of porosity can then be calculated from the change in density:

If the damage evolution is intended to be measured sequentially on a same specimen, measuring the volume and using equation 3 is sufcient. If the damage evolution is measured in parallel using several specimens, strained at different levels, equation 4 has to be used and then the density of each specimen has to be measured. 2.1.2
Experimental difculties
The precision required for this kind of measurement of can be of the order of . Measuring the mass of a sample is possible with a resolution of with conventional devices today. However, measuring a volume with a precision of is not so trivial. If parallelepipedic pieces can be cut from the deformed samples (destructive method), their geometry can be machined and then their dimensions can then requires samples be measured with a precision of 10 m but a resolution of with a characteristic length of 10 cm to be used. The precision in the measurement of the dimensions can be reduced down to 1 m and the required length to 1 cm but this requires a specic metrology equipment. If one wants to follow the evolution of non destructively, Archimedes principle can be used to obtain the volume of the sample more simply and with a better accuracy. The sample is precisely weighted in air plus in different liquids. The difference in weight is proportional to the volume of the sample to be measured provided that the density of the liquids is perfectly well known. This apparently very simple method is extremely difcult to apply correctly in practice. The experimental difculties have two main causes: The differences to measure are very small. A lot of uctuations can then affect the measurement. Thermal uctuations are the most difcult to control so generally, reference samples with a known volume, have to be measured each time together with the damaged samples, in order to correct for these uctuations. The measured samples have to be as big as possible to reduce these errors.
One has to be sure that the volume uctuation measured is only due to an evolution of damage. In metals, an increase in the dislocation density can for instance slightly modify the volume of a material. This effect can be reduced by applying a thermal treatment to the sample before the measurement of the density in order to eliminate the dislocations induced by plastic deformation. Examples in the literature
2.1.3
Ochsner et al. (2001) have quantied damage using the density uctuation in the case of cast aluminium alloys. Figure III.1 shows the results obtained by these authors. This kind of experiments is extensively used to quantify damage induced during the cold forming process of sheets of steel or aluminium. In the experiments developed for instance at the Pechiney Research Centre in Voreppe (Daniel, 2003), it has been shown using bulging experiments with three different strain paths followed by density
4 B
(4)
4 B 4 B
82
0.005
0.004 Void volume fraction

specimen slice
0.003
0.002
0.001
0.000 0.0
0.2
0.4 0.6 0.8 Equivalent plastic strain
1.0
Figure III.1: Void fraction as a function of plastic strain measured by means of the density of Al/Si materials (gure 2c in Ochsner et al. (2001)).
measurements (see gure III.2) that the strain to failure does not occur for a same value of the overall void fraction in the case of different strain paths. Other examples can also be found in the literature: Pardoen et al. (1998) have used the technique (together with a careful observation of the microstructure) in order to investigate damage during tensile testing of copper bars; Baudelet and Grange (1992), and more recently Courbon (2003) have also used it with success for quantifying damage in aluminium alloys. Whitehouse and Clyne (1993) have used it in the case of metal matrix composites.
2.1.4
Discussion
The advantages of the density evolution measurement have been mentioned previously. Although not straightforward, density measurement is feasible and provides an information which can be directly used to validate outputs of some widely used models. The information obtained is however only characterized by a scalar. It gives no information about the anisotropy of the damage created in the material. If they have the same volume, a spherical micro-void and a very at platelet shape micro-crack have the same effect on the global average density of the material. In reality, it can trivially be shown that these two voids have completely different effects on the behavior of the material. In the case of anisotropic damage, it is then of great importance to use other methods for a better description of damage. Other possible methods are described in what follows.
83
Variation of relative density
0.004
0.003
0.002
0.001
0.000 0 10 20 30 40 Equivalent plastic strain (%) 50
Figure III.2: Density evolution of an aluminium alloy during bulging experiments with three different strain paths (E1/E2= 1, 1.4 and 2). The volume fraction of pores measured at rupture is not the same for the three paths (Daniel, 2003).
2.2 YOUNGS MODULUS
2.2.1
Principle
Damage is responsible for a reduction in the load bearing capacity of a material. It has then a direct effect on the global mechanical behavior of a sample. This effect, the quantication of which is sometimes the goal of the analysis (see for example chapter VII of this book), can also be used indirectly to provide a damage measurement. The Youngs modulus in the elastic regime of a tensile test is for instance strongly affected by the evolution of damage. This has led to a number of quantication in the literature through the reduction in the Youngsmodulus of a sample. If damage occurs, the section of material available to bear the load is reduced from to . This can be used (Lemaitre and Chaboche, 1985) to dene a damage parameter d according to:
Note that this denition allows to be anisotropic. is dened through a ratio of the projections of the reduced surfaces remaining to bear the load over the initial surface of undamaged material. These projections are dened on particular planes, perpendicular to the considered direction of the loading. This ratio then depends on the morphology (isotropy and orientation) of the newly created surfaces so their projections are not the same along the different directions. This point is described in more details in section 4.4. Subjected to a same force , a newly damaged material, globally subjected to a
P # " " B ) B # " "

# "
!## #
0.005
"

(5)
84 stress
If and are the undamaged and the damaged Youngs modulii respectively, one can show, assuming an equivalence in deformation (see chapter VII), that:
The damage parameter can then be directly measured from measurements of and . In practice unloadingreloading cycles are applied to the sample during the deformation and the measurement of the modulus is performed during these elastic unloading sequences. 2.2.2 Experimental difculties
" "
Once again, the experimental determination of the Youngs modulus of a sample can seem to be a trivial problem. But this, in the case of damage, has to be performed with a good accuracy. The value of the deformation to measure in the elastic regime is very small. Strain gages glued on the sample, corrected for the temperature uctuation, are generally very good sensors allowing a sensitive measurement of the deformation. The tensile rig to be used has however to allow a precise control of this displacement. This method is then restricted to a variation of with a magnitude of more than . Note that this can be induced by an amount of porosity much less than this value of 1%, for instance if the appearing cracks have a very small aspect ratio. This is the case when damage initiates by cleavage of brittle inclusions or by interface decohesion. The quantity to be measured is the purely elastic stiffness of the newly damaged sample. When unloading an elastoplastic material, which is also in reality most of the time a little bit viscous, some physical parameters can strongly bias the value of the measured stiffness. These effects have been discussed in (Hamel et al., 1990) in the case of aluminium. At the beginning of the unloading, the apparent slope is increased by viscoplastic effects. When approaching the value of zero load, reverse plasticity strongly reduces the apparent slope of the unloading curve. Figure III.3 shows that the measured value of depends on the strain at which it is measured during the unloading. One has then to be cautious when doing these measurements. A practical rule is to perform the measurement between the rst and the second 1/3 of the total unloading curve. An important issue to keep in mind is the possible contribution of texture change on the evolution of the measured modulus. It should indeed be veried that the measured decrease of modulus is not simply due to the reorientation of the crystals during the deformation, which can lead to the modication of the intrinsic elastic stiffness of the crystallographic network of the material. 2.2.3 Examples in the literature
B
This kind of measurement has given very good results in the case of materials containing a high volume fraction of brittle inclusions. The more illustrative examples concern particle reinforced metal matrix composites. Llorca et al. (1993) and Prangnell et al. (1994) for instance, have shown a strong and linear reduction of the Youngs modulus of Al/SiC composites during a tensile straining. The evolution of the damage parameters in these materials then seems to be linear with strain. Prangnell et al. also showed that the reduction is negligible when the material is loaded in compression.
0

"
# ("
" "
will be locally subjected to a higher equivalent stress dened as: (6)
"
(7)
85
65 Youngs modulus (GPa)
55
50
Figure III.3: Youngs modulus measured during an unloading experiment as a function of the strain range for the measurement (Hamel et al., 1990).
B 4
25
50
75 (
60
100 )
s s s
) B B ) # B # B
#
!##$ '#
70
125
150
86
0.10 10% 17% 1024% 1024%
0.08
0.06
0.04
0.02
0.00 0.00
0.05
0.10 0.15 True plastic strain

0.20
Figure III.4: Evolution of the damage parameter measured from the reduction in Youngs modulus as a function of plastic strain during a tensile experiment for three Al/Cu alloys containing different volume fraction of Al Cu particles (Maire et al., 1997). It has also for instance been shown in (Maire et al., 1997) that the loss of modulus evolves much faster when the amount of brittle inclusions increases in model Al/Cu alloys containing different amounts of Al Cu inter-metallic particles (see gure III.4). This method has also been used to quantify damage of cast iron (Dong et al., 1997). The chapter VII of this book contains also some examples of such quantications. 2.2.4 Discussion

By denition, except in the case of isotropic damage, this damage parameter can not be straightforwardly linked to the amount of porosity in the sample. It must be regarded as a reduction of the section of the load bearing capacity through a reduction of safe section of the material projected in a plane perpendicular to the direction of the mechanical measurement. This method is based on the measurement of a deformation. It gives a global and directional information about damage in a given volume of material. The damage parameter measured by this method depends on the direction in which the load is applied. The method is only valid for a uniaxial loading. The eventual multiaxiality of damage is difcult to check using this technique but can be assessed using ultrasonic waves, as discussed below.
2.3 ACOUSTIC METHODS
2.3.1
Propagation of ultrasonic waves
Acoustic methods are based on the fact that the velocities of the different ultra sonic (US) waves propagating in a material depend on the elastic properties of the material
87
Figure III.5: Experimental device used in (El Guerjouma et al., 2001) to determine the velocities of ultrasonic waves in a material in different propagation planes.
itself. These methods have been widely used to measure elastic properties. They are then sensitive to the presence of damage for the reasons already explained in the preceding section 2.2. US waves velocities can rstly be measured along one direction. If the material is isotropic, the relative value of the velocities of transversal and longitudinal waves allow the two engineering elastic constants ( and ) to be calculated. The relative precision of these measurements can be of the order of . US waves velocities are even more interesting when measured in several directions. This kind of measurement can be made destructively by cutting samples along the appropriate directions but another experimental procedure has been proposed in (El Guerjouma et al., 2001). This latter procedure is non destructive and only requires two of the faces of the sample to be perfectly parallel. The device allows angular scanning in several planes of propagation through the thickness of the sample. The sample is immersed in water (or any coupling liquid) between two US transducers mounted on rotation stages (see gure III.5). From a set of measurement of the US waves velocities with different incidences on a particular plane of the material, one can retrieve the complete elastic constants tensor of an anisotropic material (El Guerjouma et al., 2001; Dalmaz et al., 2000). This requires an optimization iterative process including the minimization of an error function: elastic constants are xed a priori, allowing the velocities to be calculated in a rst step; the calculated velocities are then compared with the experimental ones and the error function is calculated; then the elastic constants are modied until the error function reaches a minimum value. Applying such a method, (Rokhlin et al., 1995) have measured the damage induced by a fatigue loading to a metal matrix composite. Figure III.6 shows the decrease of the measured Youngs modulus of these materials during fatigue life. Damage occurred mainly by decohesion of the bres perpendicular to the loading direction in this case. This method has been applied to quantify the amount of damage in ceramic matrix composites (Baste et al., 1992) and polymer matrix composites (El Guerjouma et al., 2001).
@ B
"

Local Approach To Fracture

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Local Approach To Fracture

Uploaded by

Copyright:

Available Formats

Local approach to fracture

Published under the supervision of Jacques Besson

Les Presses de lcole des Mines Paris, 2004