,rt—O—Cs STELLINGEN Behorende bij het proefschrift ‘A Mechanics based Computational Platform for Pavement Engineering van A. Scarpas 1, Vakken behorende tot het ‘Master of Science’ diploma zouden onder de bevoegdheid van de Nederlandse onderzoekscholen moeten vallen, gezien hun doctorale karakter. Dit zou een doorgaande bron van financiering voor de Scholen verzekeren en tevens een betere codrdinatie van de onderzoekstructuur bewerkstelligen. 2. Hogere vakken in de studie Civiele Techniek vereisen wiskundige kwaliteiten op het niveau van gevorderden, Het is van nationaal belang dat de huidige trend van minimalisatie van het aantal gevorderde wiskundevakken wordt teruggedraaid. De onderzoekscholen moeten hierin een belangrijke rol gaan spelen door het beschikbaar stellen van de benodigdheden en de interuniversitaire expertise. 3. Het positieve van het ‘Poldermodel’ is, dat het discussie stimuleert apriori een beslissing wordt ‘genomen, en niet posteriori ten behoeve van haar minst pijnlike implementati, 4, Boekhoudkwaliteiten zijn nodig maar kunnen niet de enige kwalificatie van een leider met visie zijn. Samenwerking moet worden uitgelokt, niet afgedwongen. 6. Als meer Europese universiteiten de Bologna Conventie implementeren, zullen de voordelen zwaarder gaan wegen dan de implementatickosten. 7. Het individuele carriére-ontwikkelingsplan, toegepast door de Amerikaanse en de meeste Europese universiteiten buiten Nederland, is bedoeld om personeelsmotivatie te bevorderen en stagnatie van carriére-ontwikkeling te voorkomen, 8. Inhet artikel van J.C. Simo “Algorithms for static and dynamic multiplicative plasticity that preserve the classical return mapping schemes of the infinitesimal theory”, Computer Methods in Applied Mechanics and Engineering, Vol. 99, 1992, pp.61-112, de correcte formulering van formule (3.12b) is als volgt: ew 8f =exp[-24,0,0(74.4)|b4 In hetzelfde artikel, de correcte formulering van formule (3.16,) is tevens: ctr ie i exp|-249,F [CW ,]F2 | of 9. Inhet artikel van T. Roje, “On a mixed approach to the finite element solution of large strain elastoplastic problems”, Engineering Computations, Vol. 15, 1998, pp. 150-161, formule (5) is niet fysisch correct. De correcte formulering is: tag Sa . Te 0, T te te 10. In Paragraaf 2.6 van J.C. Simo & TI.R. Hughes, Computational Inelasticity, Springer, 1998, formule (2.6.4) is, om consistent te zijn met de rest van de formulering in dat hoofdstuk: KP ={yeL(B)\1>0} Tevens, formule (2.6.5) moet zij ce ( rye?) 11. In dezelfe referentie als bovenstaand, de correcte formulering van formule (8.3.6a) is: Tiel +a f(r) eT lnta=Fnteensifata Deze stellingen worden verdedigbaar geacht en zijn als zodanig goedgekeurd door de promotoren Prof. dr. it. J. Blaauwendraad and Prof. N. Aravas, BSc., MSc., PhD.PROPOSITIONS Associated with the thesis A Mechanics based Computational Platform for Pavement Engineering of A. Scarpas. 1. Considering their post-graduate nature, studies for the Master degree should be under the jurisdiction of the Dutch schools of Graduate Studies, now commonly known as Research Schools. This would ensure a continuous source of financing for the Schools and a more efficient coordination of the research infrastructure. 2. Advanced studies in civil engineering require advanced mathematical skills. It is of national impottance that the current shrinkage of advanced mathematical courses in engineering curricula is reversed. The Research Schools must play an important role in this aspect by providing the means and the necessary inter-university expertise. 3. The positive aspect of the “Polder model’ is that it encourages discussion apriori to a decision being made and not posteriori for its least painful implementation. 4, Accounting skills are necessary but cannot be the only qualification of a visionary leader. Cooperation should be encouraged not dictated. 6. As more European Universities implement the Bologna convention, its benefits will outweigh the implementation costs. 7. The personal career development plan adopted by American and most European Universities outside ‘The Netherlands is meant to encourage staff motivation and prevent career development stagnation, & In the article of J.C, Simo, “Algorithms for static and dynamic multiplicative plasticity that preserve the classical return mapping schemes of the infinitesimal theory”, Computer Methods in Applied Mechanics and Engineering, Vol, 99, 1992, pp.61-112, the correct form of Equation (3.12b) should read: ne a\]3et? 2.4,0,0(78,4)|O¢ Also, in the same article, the correct form of Equation (3.16) is: vig ol ]pett Pion}? |e bg =exp| 9, In the article of T. Rojec, “On a mixed approach to the finite element solution of large strain lastoplastic problems”, Engineering Computations, Vol. 15, 1998, pp. 150-161, Equation (5) is not physically correct. It should read: ta toad + f ta]RT=R,,,]t+ f oP DaeRT ty th R, | °, ta nt at] 10. In Section 2.6 of J.C. Simo & TJ.R. Hughes, Computational Inelasticity, Springer, 1998, for consistency with the rest of the developments in that Chapter, Equation (2.6.4) should read: KP = {yeL?(B)|4>0} Also, Equation (2.6.5) should read: 22 (rope? = rie? tafe 11. In the same reference as above, the correct form of Equation (8.3.62) is: a Fatoenifata ‘These propositions are considered defendable and as such have been approved by the supervisors Prof. dt. ir. J. Blaauwendraad and Prof. N. Aravas, BSc., MSc., PhD.TR diss 4406 A Mechanics based Computational Platform for Pavement Engineering Proefschrift ter verkrijging van de graad van doctor aan de Technische Universiteit Delt, op gezag van de Rector Magnificus prof. dr. ir. J.T. Fokkema, voorzitter van het College voor Promoties, in het openbaar te verdedigen op maandag 21 februari 2004 om 10:30 uur door Athanasios SCARPAS Master of Civil Engineering University of Canterbury, New Zealand Geboren te Athene, GriekenlandDit proefschrift is goedgekeurd door de promotoren: Prof. dr. ir, J. Blaauwendraad Prof. N. Aravas, BSc., MSc., PhD. Samenstelling Promotiecommissie: Rector Magnificus, voorzitter Prof. dr. ir. J. Blaauwendraad Prof. N. Aravas, BSc., MSc., PhD. Prof. C.S. Desai, BSc., MSc., PhD. Prof. dr. ir. A.A.A. Molenaar Prof. P. Selvadurai Prof. J. Uzan, BSc., MSc., PhD. Prof. dr. ir. F. Molenkamp Copyright © 2004 by A. Scarpas ISBN 90-9019040-6 Printed in the Netherlands Technische Universiteit Delft, voorzitter Technische Universiteit Delft, promotor University of Thessaly, promotor University of Arizona Technische Universiteit Delft McGill University Technion University Technische Universiteit Delft, reserve lidTo Johan Blaauwendraad my teacher, my mentor, my friendAcknowledgements Prof. Dr. ir, André Molenaar and Prof. dr. ir. Johan Blaauwendraad were the first who identified the need for the development of a computational platform focused on pavement engineering and who offered me the opportunity to develop it in what now constitutes the CAPA-3D finite element system. They established a spirit of close cooperation between the Road & Railroad Laboratory and the Section of Structural Mechanics of Delft University of Technology and entrusted me to create the conditions and the environment for successful joint research in the field of mechanics of pavements. In the course of the work I consider myself privileged for having had the opportunity to work closely for many years with dr. ir, Arian de Bondt and dr. ir, Sandra Erkens at the time when they were both working towards the completion of their own PHD research work at the Road & Railroad Laboratory. Many of the current features of CAPA-3D were inspired and developed in response to their research needs. In the process, they also became coffee addicts. Since the mid-nineties, ir. C. Kasbergen joined what has become known as the CAPA-3D development team, bringing with him expertise in computer programming and mathematics. He soon took over control of the design/maintenance of the user interface and the coordination of the programming efforts. Cor has contributed the most in designing a finite element system which enables many individuals to work concurrently in their own domain without interfering with each other’s development work. He has also been our unequivocal ambassador with the several guests we host every year, The fact that many of them come back to visit him is a testimony to his efforts for which I am grateful. Dr. Xueyan Liu also joined the development team since the beginning of his PhD thesis at the Section of Structural Mechanics. Over the years he concentrated on constitutive model development and implementation. This brought him into contact with many of our other PhD researchers and research guests and encouraged him to work with a great variety of engineering materials, constitutive models and field cases. By now, his name appears in the acknowledgements of at least as many PhD theses as mine... lam thankful for his support in outreaching to so many different disciplines and individuals. ‘A special thank you note is due to the efforts and contribution of ing. Frank Custers. Despite the continuously increasing number of hardware and intemaV/external users, Frank always ‘managed the seamless integration and operation of all resources. Several other of my students and colleagues have contributed to the verification and the improvement of the system and can be traced in the corresponding references of our joint publications. Among many, the contributions of Prof. Andrew Collop, dr. ir. Rien Huurman and of Prof. Jacob Uzan deserve special mention. Over the years, I have learned and benefited a lot from the writings of my co-promotor, Prof. Nikos Aravas, The plasticity reduction algorithms in CAPA-3D are clearly based on concepts that he initiated. For the time and effort he spent in meticulously reviewing and correcting early versions of this document, I shall always be grateful. Maureen, my wife, has always been a member of the CAPA-3D development team and is familiar with and to so many individuals within the pavement mechanics engineering community. This is the best recognition and reward for her contribution. ‘Tom Scarpas Delfi, November 2004List of Contents Chapter 1: Chapter 2: Introduction Fundamentals of Classical Continuum Formulation 2.1 Introduction 2.2 Kinematics 2.2.1 Volume and Area Transformations 2.2.2 Strain 2.2.3 Polar Decomposition 2.2.4 Volumetric/Deviatoric Multiplicative Split 2.2.5 Space and Time Derivatives 2.2.6 Rate of Deformation Tensors 2.2.7 Lie Time Derivative 2.2.8 Objectivity 2.2.9 Volume Changes 2.3 Stress tensor 2.3.1 Rate of Stress Tensors 2.4 Mechanical Balance Laws 2.4.1 Balance of Momentum 2.4.2 Balance of Moment of Momentum. 2.5 Variational Concepts 2.5.1 Virtual Displacements and their Variations 2.5.2 Virtual Work 2.5.3 Clausius-Planck Inequality 2.6 Linearized Equilibrium Equation 2.6.1 Linearized Internal Virtual Work 2.6.2 Linearized External Virtual Work 2.7 Rate Constitutive Equations 2.8 Discretization 2.8.1 Element Geometry Interpolation pas 2.8.3 Discretized Spatial Equilibrium Equation eld Variables Interpolation 2.8.4 Discretization of the Linearized Equilibrium Equation page 10 12 1B 14 16 16 7 17 19 20 20 = 23 23 24 a 27 27 30 32 EE) 33 34 35 38Appendix 2.1 Appendix 2.2 Appendix 2.3 ‘Appendix 2.4 Appendix 2.5 Appendix 2.6 PartI Constitutive Theories Chapter 3: Hyperelasticity 3.1 Introduction 3.2 Isotropic Hyperelastic Material Response 3.2.1 Elasticity Tensor 3.3 Isotropic Elasticity in Principal Directions 3.3.1 Elasticity Tensor Appendix 3.1 Appendix 3.2 Chapter 4: Elastoplasticity 4,1 Elementary Model of Elastoplastic Response 4.2 Three Dimensional Elastoplastic Model 4.2.1 Constitutive Framework 4.2.2 Hardening Response 4.2.3 Response Degradation 4.3 Material Parameter Determination 43.1 Hardening Response 4.3.2 Response Degradation 44 Algorithmic Aspects 4.4.1 Hardening Response 4.5 Response Degradation 4.5.1 Desai Surface 4.5.2 Hoffman Surface 4.6 Large Strains Formulation 4.7 Utilization Appendix 4.1 2 4B 45 48 49 51 533 54 57 = 63 67 7 ” 4 5 17 79 79 82 87 87 87 88 90 91Appendix 4.2 Appendix 4.3, Appendix 4.4 Appendix 4.5 Chapter 5: Viscoelasticity 5.1 Elementary Model of Viscoelastic Response 5.1.1 Generalized Linear Viscoelastic Model 5.1.2 Incremental formulation 5.2 Generalized Nonlinear Elastic Formulation 5.3 Three Dimensional Nonlinear Formulation 5.4 Burger’s Model 3.4.1 Elementary Model 5.4.2 Three Dimensional Model 5.5 Utilization Appendix 5.1 Appendix 5.2 Appendix 5.3, Appendix 5.4 Chapter 6: Elasto-Visco-Plasticity 6.1 Introduction 6.2 Multiplicative Decomposition 6.3 Generalized Model Local Dissipation 6.4 Plastic Component Integration Procedure 6.4.1 Stress Reduction Procedure 6.4.1.1 Trial Elastic State 6.4.1.2 Flow Rule Discretization 6.4.1.3 Retum Mapping Procedure 6.4.2 Characteristics of the Plastic Component 6.5 Viscoelastic Component Integration Procedure 6.5.1 Stress Reduction Procedure 6.5.1.1 Trial Elastic State 6.5.1.2 Flow Rule Discretization 6.5.1.3 Return Mapping Procedure 6.6 Utilization Appendix 6.1 94 7 104 WW lps 113 14 7 us 18 19 121 122 124 125 127 129 130 132 133 134 134 135 137 138 139 ee 139 140 141 142 144Appendix 6.2 Appendix 6.3 Appendix 6.4 Appendix 6.5 Appendix 6.6 Appendix 6.7 Part II Finite Elements Technology Chapter 7: Cubic Finite Element Overview 7.1 Introduction 7.2 Element Formulation 7.2.1 Element Geometry Interpolation 7.2.2 Field Variables Interpolation 7.3 Numerical Integration 7.4 Constitutive Law 7.5 Utilization Chapter 8: Interface Finite Element Overview 8.1 Introduction 8.2 Element Formulation 8.2.1 Element Geometry Interpolation 8.2.2 Field Variables Interpolation 8.3 Element Stiffness Matrix 8.4 Constitutive Law 8.4.1 Nonlinear Material Response 8.4.2 Parameter Determination 8.5 Algorithmic Aspects 8.6 The Reinforcement Component 8.7 Utilization Chapter 9: Infinite Boundary Finite Element Overview 9.1 Introduction 9.2 Element Formulation 9.2.1 Element Geometry Interpolation 9.2.2 Field Variables Interpolation 146 148 150 153 155 156 157 159 159 159 161 163 164 164 165 165 166 168 169 169 172 es 174 176 178 179 181 181 1829.3 Element Utilization Appendix 9.1 Chapter 10: 10.1 Introduction lent Boundary Finite Element Overview 10.2 Element Formulation 10.2.1 Field Variables Interpolation 10.3 Constitutive Law 10.4 Virtual Work 10.5 Utilization Summary & Conclusions References Samenvatting Curriculum Vitae 185 187 191 192 193 194 195 195 197 199 205 206viChapter 1 ——s _ Introduction Most current design methodologies against fatigue and permanent deformation of road pavements rely on experience and are based mainly on information obtained from simple, mostly uniaxial, laboratory tests. Nevertheless, it is common knowledge that when a traffic load is imposed on a pavement, a non-uniform displacement field develops giving rise to a ‘complicated state of triaxial stresses, Fig. 1.1 (exaggerated). Fig. 1.1 Stress states in a pavement due to traffic load Triaxiality has been known to significantly influence the response of pavement materials. For example, as shown schematically in Fig. 1.2 for the case of a cohesive material, in the presence of increasing amounts of lateral tension both, the stiffness and the compressive strength of the material in the perpendicular direction decrease. Fig. 1.2 Influence of lateral tension on compressive strength and stiffnessIntroduction Fig. 1.3 Influence of temperature and strain rate on material response Also, for a wide range of pavement engineering materials temperature and strain rate have influences on both the strength and the stiffness characteristics, Fig. 1 3. Most of the current pavement design methodologies ignore the above fundamental features of ‘material response and are based on simplified elastic theories. In spite of this inconsistency, fairly good pavements could be constructed until recently. This was mainly because the lack of appropriate material models could be backed up by practical experience. Relying on experience is however not acceptable anymore because of the rapidly changing conditions of the international road network. First of all, the number of trucks has increased beyond expectation. Next to that there is a strong increase in wheel loads as well as in tyre pressures as a consequence of the need to increase the road transport efficiency for environmental reasons. Furthermore, the expected introduction of aramide reinforced truck tyres in combination with foreseeable changes in wheel configurations will result in extremely high wheel pressures. All these, together with the need to use secondary recycled building materials, the need for “maintenance free" roads and the fact that pavements must increasingly fulfil non-structural demands (eg. noise levels, water permeability) result into the question how to design our future roads and asphalt mixes in order to make them resistant to the very heavy loads that will occur, Because these developments are far beyond our experience, one needs to address these questions by applying proper design and evaluation models, which are based on a sound engineering approach towards material behaviour. For this reason, interest in mechanics based approaches for road engineering design has recently grown considerably, both nationally and intemationally. This is partially due to the increased availability of computing power, which has led to the increased popularity of versatile computational tools such as the finite element method. By accounting for the above mentioned idiosyncrasies of material response and by enabling the visualisation of the internal distributions of stresses and strains in the body of a pavement, the finite element method constitutes a valuable tool in understanding the mechanisms and the processes leading to pavement deterioration. In addition, the method enables the quantification of the interaction between the material and the geometric characteristics of a pavement. Unfortunately, because of the dependence of pavement engineering materials on the state of stress, on the rate of loading and on the temperature, they constitute some of the most difficult materials for finite element simulation, Nevertheless, when such models are available, utilisation of the method can result to significant time and financial savings in laboratory and field-testing.Introduction Since the early 90s, the Section of Structural Mechanics and the Laboratory of Road & Railroad Engineering of the Faculty of Civil Engineering and Geosciences of TU Delft have been closely cooperating towards the development of tools and procedures capable of addressing realistically the response characteristics of a wide range of pavement engineering materials. In the framework of this cooperation, CAPA-3D has been developed as a finite element based platform to serve the computational needs of the joint research team at TU Delf, and of some international teams which cooperate with Delft. Over the years, CAPA-3D has evolved into a fully fledged finite element system for static or dynamic analysis of very large scale three dimensional pavement and soil engineering models. It consists of a sophisticated user interface, a powerful band-optimising mesh generator, high quality user controlled graphical output, several material and element types, and a variety of specialised algorithms for the more efficient analysis of pavement constructions. Among others, these include a moving load simulation algorithm and a contact algorithm. Ever since its inception in the early 90s, the system has been under continuous update and development. Invariably developments were dictated by the research needs of the progressively growing groups of rescarchers/users. In order to minimise interference between independently working groups, a hierarchical structure has been chosen for the system layout. This has enabled the utilisation of the system as the computational platform of several concurrent PhD theses within TU Delft and internationally. In pavement engineering very few, if any, realistic situations can be encountered which are truly two-dimensional. Nevertheless and despite the recent developments in computer software and hardware, the use of three-dimensional Finite Element models is both, time and resource consuming. Especially so if the non-linear nature of the materials and the processes concemed is considered, In the development of the CAPA-3D system the need for powerful, albeit expensive, computational facilities has been addressed by segmenting the system into three subsystems: 1. the ‘input subsystem’, consisting of a user interface for structured input of the geometric and the material data, a band minimising mesh generator and a graphical facility for mesh visualisation. 2. the ‘computational engine’, in which the main finite element operations are performed (e.g. matrix handling operations, state of stress determination). The system can perform static or dynamic analysis of very large scale three dimensional models as those typically encountered in pavement and soil engineering. 3. the ‘output subsystem’, consisting of a user interface for the exploitation/visualisation of the results of the finite element analyses. Al three subsystems can run in an Intel based personal computer, However, if optimum performance is to be obtained more powerful computers can be utilised to run the computational engine subsystem. Access to these computers can be attained either directly or via the Intemet. As such, research teams who do not have their own powerful computing facilities can still access the system. All subsystems provide their own integrated Windows based user guides. This monograph focuses exclusively on those aspects of continuum mechanics that were necessary for the development of a range of generic constitutive models and finite element types. These can be utilised by the user for the development of additional materials and element types.Introduction Over the years, many of my students and colleagues contributed to the verification and the improvement of the system. Among many, the contributions of A. de Bondt, C. Kasbergen, X. Liu, A. Collop, M. Huurman and J. Uzan deserve special mention and can be traced in the corresponding references of our joint publications. In this context, several of the significant characteristics of the system, like solution algorithms, the simulation of contact and of multiphase media will not be addressed. For these, the reader is referred to the various PhD theses of my past and current students. This document is structured as follows. In Chapter 2, the necessary fundamental non-linear continuum mechanics theories and their implementation in the context of the Finite Element Method are presented. This Chapter is of paramount importance for understanding the material included in the following Chapters. Most of the terminology is also established in this Chapter. The rest of the document is divided in two sections. In Section I, several generic theories of material constitutive response are presented in progressively increasing complexity. The last Chapter of this section, Chapter 6 provides a methodology which enables the combination of the earlier presented generic theories for the development of more complex constitutive models. In Section Il of the document, the characteristics of a variety of finite element types, necessary for efficient analyses of pavement and soil engineering problems, are discussed. The domain of element application and its proper utilisation are addressed. As mentioned earlier, current pavement design is based primarily on empirical rules. However, the advent of powerful computational hardware systems, will make tools like CAPA-3D accessible to a wider research and engineering audience. This will open the way for a new generation of pavement design techniques based on rational mechanics principles and known as ‘mechanistic’ based techniques, Yoder and Witezak {1975}. A typical example of this change in design philosophy is the recent NCHRP 1-37A [2004] report on future pavement design, in which the finite element method in combination with advanced material constitutive models and characterisation techniques constitute the backbone of the whole design process.Chapter 2 Fundamentals of Classical Continuum Formulation 2.1 Introduction In classical continuum mechanics several physical quantities are defined on the basis of the “representative volume” element (RVE). Typical examples are the density, the porosity etc. Furthermore, physical measures of deformation or force are defined by considering an averaging procedure over such a “representative volume”. representative volume element Fig. 2.1 Representative volume element A deformable body having volume V and surface area A can be assumed to be consisted of an assemblage of “representative volume” elements, Fig. 2.1. The classical definition of a 0. continuum is obtained by allowing (v. 2.2 Kinematics In a classical continuum, the location of a point before deformation is determined by the position vector X defined in a Cartesian basis system _ Fig. 2.2. After deformation, the coordinates of point P are located by position vector x defined wert. an altemative Cartesian basis system {e, ;i=1,2,3}. The motion of a particle from the reference to the current position can be viewed as a vector mapping x= $(X) 21Fundamentals t=t, 1 current _ configuration . t=0 x igura 1 reference configuration Fig.2.2 Reference and current configurations of a deformable body Without loss of generality, in the remainder of this document, the two base systems will be assumed to coincide, however, notational differences will be maintained. Capital letters will be utilised to indicate quantities in the undeformed, or as it is commonly termed, the “reference” configuration at time t=0 and, lower case letters to indicate the deformed, or as it is commonly termed, the “current” configuration at time t Eq. 2.1 can be generalized to read x=6(Xt) 22 The relative position vector dX of two material points P and Q in the reference configuration, Fig. 2.3, can be related to the relative position vector dx in the current configuration as dx=FdX 23 in which F is known as the deformation gradient tensor defined by 24Fundamentals t= reference configuration configuration 29 X2 Fig. 2.3 Mapping of relative position vector from the reference to the current configuration. In the mechanics literature the operation indicated by Eq. 2.3 is termed as a push forward operation. The formalism ax ~6, aX] 2.5 is frequently encountered. The condition detF > 0 and the existence of a unique inverse of F express the physical requirements of continuity, indestructibility and impenetrability of matter, Malvern [1969]. Then it is also valid aX =F'ldx 2.6 ‘The operation defined by Eq. 2.6 is termed as pull back. Formally it is indicated as aX = 47" [ax] 27 2.2.1 Volume and Area Transformations The individual vector components of the relative position vector dx can be expressed by means of the current configuration basis vectors e, and the components of the relative position vector dX as dx, = 28 J=l,...3] aX, ‘The differential volume dv in the current configuration is du= dx, -(dx, xdx,) 29Fundamentals Substituting for the individual vector components on the basis of Eq. 2.8 it results du = det (dX,dX,dX,)=J av 2.10 By means of the postulate of mass conservation, the material density p, in the reference configuration can be related to the density p in the current configuration via dm =p, dV =pdv=JpdV 241 in which dm is an element of mass. On the basis of Eq. 2.10 an interrelationship can be also obtained between an infinitesimal clement of area dA in the reference configuration and da the corresponding infinitesimal element of area in the current configuration da=JF7dA * 2.12 2.2.2 Strain Let dS denote the magnitude of the relative position vector dX. at the reference configuration and ds the magnitude of the same in the current configuration. Then a measure of the change of length between the two configurations can be expressed as (as) -(dS)? =dx-dx-dX-dX =dX-FTFdX —dX-dX 2.13 =dX-(FTF—I)dX =dX-(C-I)dX The tensor C=F'F 214 is known as the right Cauchy-Green deformation tensor Malvern [1969], Bonet & Wood [1999], Holzapfel [2000]. On the basis of C the Lagrangian-Green strain tensor is defined as E=H(C-1) 245 Defining x=Xt+u 2.16 in which u represents the displacement vector from X to x, on the basis of Eq. 2.4 ax OX 217 al * The notation FTFundamentals while C can be expressed as C=F'F= iz ax +0} {24} du] [du], [au x *laxl* ox! In case of small displacements, the second order terms in Eq. 2.18 can be dropped and the familiar, small strain tensor €,, is obtained 1 [jou (| =) + ( 219 afl +133 Alternatively, a measure of strain can be obtained on the basis of the current configuration by expressing the product dX-dX as 2.18 iT du 0 aX-dX =(F ‘dx)' (Fax) = dx" (FTP dx 2.20 =dx-b"!dx Tensor b is commonly known as the left Cauchy-Green or Finger tensor, Malvern [1969], Bonet & Wood [1999], Holzapfel [2000]. Eq. 2.13 can now be expressed in terms of dx dx-dx—dX-dX = dx-dx—dx-bo' dx 2.21 =dx-(I- bo! )dx ‘The Eulerian-Almansi strain tensor is then defined as e=+(I-b"') 2.22 ‘The push forward }, and pull back ;' operations that were defined in the above can also be extended to the tensorial measures of strain. From Eq, 2.13 and Eq. 2.21 it results aX, -Ed, 2.23 = dx,-ed: hence E=FeF ; e=F EF! 2.24 The first of Eq, 2.24 can be viewed as the mapping of the strain tensor from the current to the reference configuration and, the second of Eq. 2.24 as the reverse procedure. In terms of the operator , : e=6,(E] 2.25Fundamentals 2.2.3, Polar Decomposition The deformation gradient F is a second order tensor. As such it can be decomposed as the product of an orthogonal tensor R. times a symmetric tensor U F=RU 2.26 In literature U is known as the material stretch tensor and Ras the rotation tensor. On the basis of Eq. 2.13 C=FTF=U'R™RU =U 2.27 since R'R=I and U7 =U. Because C is symmetric and positive definite, by means of the spectral decomposition theorem 4 C= LOL, 2.28 ia in which A? >0 are the eigenvalues and Z, are the corresponding orthonormal eigenvectors of C. Also it can be shown that 2.29 The eigenvalues A, are also known as the material stretches along the principal directions L, . On the basis of Eq. 2.27 it is also valid 4 U=DA LoL, 2.30 = Hoger & Carlson [1984] have shown that once the eigenvalues A, are known, U can be computed directly without need for computation of the principal directions L,. Details are provided in Appendix 2.1. Then R=FU" 231 A physical interpretation of the role of tensors U and R. in transforming the vector dX from the reference configuration to dx in the current can be obtained by combining Eq. 2.3 and Eq. 2.26 dx=R(Uax) 2.32 In this, vector dX is first stretched to UdX in the reference configuration and, subsequently rotated to the current configuration by R., Fig. 2.4 10Fundamentals Fig. 2.4 Polar decomposition F can be also decomposed as the product of a symmetric tensor V times the same as above tensor R. so that dx =FdX=VRdX 2.33 In this, the material vector dX is first rotated via R. and then stretched to dx. The symmetric Finger tensor b can now be expressed as b=(VR)(VR)" = 2.34 Also on the basis of Eq, 2.32 and Eq. 2.33 it results Vv=RUR" 2.35 By means of the spectral theorem 3 b= NV LOL, 2.36 a in which Z, are the orthogonal eigenvectors of b and , the spatial stretches.Fundamentals 4 Hence V=>,404 237 Combining Eq. 2.35 and Eq. 2.30 the relationship between the material and spatial stretches and the corresponding principal directions is obtained as 238 In anticipation of developments in later Chapters, an expression can be obtained for the Lagrangian-Green strain tensor in terms of the principal material stretches by means of Eq. 2.15 and Eq. 2.28 as B=SoH(A?-1) E61, 239 iat 2.2.4 Volumetric/Deviatoric Multiplicative Split In Eq, 2.26 the deformation gradient was decomposed as the product of an orthogonal Rand a symmetric tensor U.. Another possible decomposition can be constructed if it is assumed that the overall deformation of a solid can be decomposed into volumetric and deviatoric (hence volume preserving) components. Starting from the reference configuration, let =R,, dX 2.40 Kiso denote the volume preserving (commonly termed as isochoric) part of the total deformation. Also, let dx=F,,, dx, 241 vol FXiso denote the deformation necessary to recover the total deformation dx = F dX. From the above it results F=F 01 Fiso 2.42 Eq. 2.42 is typical of a whole range of F decompositions typically known as multiplicative splits, or multiplicative decompositions, Simo & Hughes (1998}, Belytschko et al. [2000]. They shall constitute in later Chapters the foundation for the development of some of the inelastic material formulations of CAPA-3D. From Eq, 2.10, for a volume preserving deformation detF,,, = J =1. For this condition to be satisfied FB =J¥F 2.43 which provides a convenient expression for the computation of the deviatoric components of deformation. Then from Eq. 2.42 it becomes evident that BA aJeF 2.44Fundamentals Once F,,, and F,,, are known, associated deviatoric and volumetric measures of deformation can be defined e.g. — Css =4(c,.-1] 2.45 2.2.5 Space and Time Derivatives A linear approximation to the increment of a nonlinear scalar valued function W(X) in the direction of vector u can be obtained by means of a truncated Taylor series as UR +en)= 0K) + 20 (K+ en} =¥(K)+D, W(X) 2.46 ie lo The differential term on the right-hand side of Eq. 2.46 is known from calculus as the directional derivative of U(X) in the direction of w. A pictorial representation in a 2 dimensional space (X,,X,) is shown in Fig. 2.5 Fig. 2.5 Linearisation of a function ‘The material time derivative of an operator V in the reference configuration is defined as DYXt) V(X.) pete I 247 in which the subscript X indicates differentiation exclusively with respect to time. 1BFundamentals If the operator is defined in the current configuration then, by means of the chain rule of differentiation Dut) U(x,t)= - Y _2vG50) | 2ub56) 6.264 vas at Ox ot Ik Ix=o; ix] _ 9¥(x,t) oo: In the above :)(x,t)/At is commonly termed the spatial time derivative of >) and v is termed the velocity vector. 2.2.6 Rate of Deformation Tensors The spatial velocity gradient tensor 1 is defined by considering the relative velocity dv between two particles in the current configuration dv (ES ac = veax=tax 2.49 ox 1 can also be defined on the basis of the deformation gradient F . Considering the material time derivative of F DF(X,t Dt 2.50 so that 1=FF* 2.51 The term material velocity gradient is commonly used for F , Holzapfel {2000} A useful relationship can be obtained by associating the material time derivative of F with the concept of the directional derivative. Applying the concept of directional derivative to the deformation gradient along the direction of the velocity vector v a{xtev) afax, | _ oO 0 d D,F(X)= ox ac|aX OX) OX 0 Comparison of the above with Eq. 2.50 it results #(X)=D, F(X) 2.53Fundamentals In Section 2.2.2 strain was defined by the change of length of a vector between the reference and the current configurations. Similarly, the sirain rate can be defined as the time change of the magnitude of vector dx in the current configuration D ce ples) l= pelt a) 2.54 Expressing dx on the basis of Eq. 2.3 (and noting that dX is time independent) Ples¥] =dX.(E7 P+ETF)aX =dX-Cax 2.55 =2dX-BdX in which the tensor F+FT#) 2.56 is commonly known as the Lagrangian material strain rate tensor. It expresses the rate of change of dx on the basis ofits definition in the reference configuration. Alternatively, E can be expressed in terms of the principal material stretches A,,i= 12,3 by considering the time derivative of Eq. 2.39 3 a B= PAA Let, + 4A; (i,01,+L,00) 287 r= ist The rate of change of the magnitude of dx in the current configuration can be obtained by substituting in the last of Eq. 2.55 dX = F'dx 2 ((asy|=2ax re Br" ax =dx-[FT ET + PP Jax 2.58 =2dx-ddx The tensor a=FTEF" 2.59 is commonly known as the rate of deformation tensor . For a rigid body motion d =0. Substituting F from Eq. 2.50 into Eq. 2.56 and utilizing Eq, 2.59 it results that the symmetric component of the velocity gradient tensor constitutes the rate of deformation tensor d=3(l4") 5 d?=d 2.60 while the antisymmetric component constitutes a tensor commonly known as the spin tensor wat) 5 wT 2.61 1sFundamentals Its components are frequently defined on the basis of the associated angular velocity vector T uo, as 2.62 For use in later Chapters on material response simulation, an explicit relation between the spin tensor and the rotation tensor can be obtained by substituting Eq. 2.51 in Eq. 2.61 w=RR™+4R(OU"'-U'U)R™ 2.63 which, for rigid body motions reduces to w 2.64 2.2.7 Lie Time Derivative The Lie time derivative £ is a powerful conceptual tool, which will be frequently utilized in the following Sections in operations involving the transformation of various mechanical quantities between the current and the reference configurations. If g=g(x,t) denotes a function in the current configuration the Lie time derivative is obtained by means of the following operations: i, perform a pull back operation on g to obtain G(X,t)=4, "[z(x.t)] ii, compute the material derivative G(X,t) iii, perform a push forward operation on G to obtain £(g) =, [G(X,t)] Since, in the reference configuration, the material derivative is equivalent to the directional derivative along the velocity vector v (ie. Eq. 2.53) it holds G(X =D, G(X) 2.65 so that £(g) can also be expressed as £(e)= 4,[6(X.0]=9,[D, 6X0] 2.66 2.2.8 Objectivity Objectivity constitutes one of the most fundamental principles of mechanics, Malvern [1969], Simo & Hughes [1998], Quantities which are used to describe material behavior must be objective. The notion of objectivity can be physically illustrated by imposing an orthogonal transformation R. on a vector dx =F dX. The resulting vector can be expressed as dz=Rdx 2.67Fundamentals By means of Eq. 2.13 it can be shown that the magnitudes of dx and dx are equal. In the mechanics literature, vectors which under rigid body rotations transform according to Eq. 2.67 are termed objective. In contrast to many second order tensors whose behavior under orthogonal transformations will be examined in the following, it is worth noticing that the deformation gradient tensor F under orthogonal transformations transforms according to Eq. 2.67 since a% R)_ |e a= —=|—RF 2.68 ee 3 The notion of objectivity can be extended to tensors of any order. In particular, if an orthogonal transformation R. is imposed to a second order tensor $= dx, @ dx,, the rotated tensor § = d&, @ dk, can be expressed by means of Eq. 2.67 as 2.69 =R(dx, @dx,)R™ =RSR™ In the mechanics literature second order tensors which transform according to Eq. 2.69 are called objective, Simo & Hughes [1998], Holzapfel [2000]. Several of the deformation ‘measures introduced in the above like C, E, d and e are objective. In Appendix 2.2 a proof for the objectivity of e is included. Similar steps can be followed for other tensors. Examples of non-objective quantities are the velocity vector v and the velocity gradient tensor 1. More examples will be encountered in subsequent sections. 2.2.9 Volume Changes ‘The rate of volume change can be computed by differentiating with respect to time Eq. 2.10 7 0 2.124 W is known as the Helmholtz free-energy. In the special case where W is solely a function of F it is also known as the strain energy or the stored energy. "ie. ifthermal effects are ignored 25Fundamentals A mechanical process is termed reversible if D=0. This condition is true for elastic materials, that is, materials for which the stress at any time is a function only of the state of deformation F (X) (and the temperature). If D>0, the process is termed irreversible and the corresponding material dissipative. Plastic materials are typical examples. In contrast to the non-dissipative case, W is not, anymore, exclusively a function of F but it depends also on a number of additional variables associated with the memory properties of the material. They represent the irreversible physical processes within the microstructure of the material, which occur throughout the deformation history and which are the causes of nonlinearity. Then V can be typically expressed as W=(F(X).£,.8,,.--£4) 2.125 The term internal variables is commonly utilised for the €s. Depending on the particular characteristics of a material, their number n and their individual nature can vary.” Substituting in Eq. 2.124 with ¥=0( F(X). d 2.126 eae) For D to be non-negative for every choice of F the necessary conditions, Coleman & Noll [1963], Coleman & Gurtin [1967], are ov OF P 2.128 In order to describe the development of the irreversible processes in the material, evolution Jaws must be specified for each of the internal variables €, = ©( F(X),€, 855-80) 2.129 In general they form a system of ordinary first order differential equations. The dissipation inequality will be utilised in later Chapters for the derivation of some fundamental characteristics of elastoplastic and viscoelastic materials. 2 Without a loss of generality and for reasons of notational convenience only, the internal variables in this Section have been assumed as second order tensors. 26Fundamentals 2.6 Linearized Equilibrium Equation The equilibrium expression in the form of Eq. 2.121 is, in general, a nonlinear equation. Nonlinearities may arise from the material response and/or the geometry. Within the context of the nonlinear analysis, the equilibrium configuration ¢ is sought by means of an iterative type technique. Over the years several variants of Newton’s technique have been developed. A. common characteristic of all these techniques is the replacement of the nonlinear equation with a linearized one in the vicinity of the expected solution. According to Section 2.2.5, in the vicinity of a trial solution x=x, of 5W«), a linear approximation to 8W in the direction of an increment Au is SW(x,) + D,,8W(x,)=0 2.130 From Eq. 2.121 Da, SW=D,, 8W,,, — Da Woy 2431 an in which 8W,,, is defined in Eq. 2.122 and 8W,,., in Eq, 2.123. In the following, the steps necessary for computation of the terms of Eq. 2.131 will be presented. 2.6.1 Linearized Internal Virtual Work On the basis of Eq. 2.122 the linearization of 6W,,, along Au leads to D4, = fD,,(S:88)ar vam = [Dg 8:8bav + J 8:Dau dav Considering the linearization of $ along Au D4,8=—2{5(B(o+-Au) -8 ee 2.133 The term (98/08) represents the well known 4" order material or Lagrangian elasticity tensor C 2.134 with individual terms 2.135 then D,,S=C:D,,E 2.136 nFundamentals From Eq. 2.56 2.137 Since the current configuration is kept constant during the velocity field variation du éF=D, |—|=0 2.138 «a so that SB=4(8FT P+ ET 2.139 Additionally 6 =D, (= a alee) ae 2.140 ti jy, Ma Cea oe a {ox} del Ox ox 0" = Hence for the term D,,, 88 of Eq. 2.132 it holds ely (get pa pt Dy, SE=5D,, (FTF +FT OF) : 141 From Eq. 2.140 . abv Day FP =Daul gx |=0 2.142 because the virtual velocities field is independent of the displacement field. On the basis of the above and Eq. 2.140, Eq. 2.141 is simplified to D4, 86 =3]6#"D,,F+D,, Fai] ‘i = silvery V,Au+(VAu)” ety] 2.143 =sym|(V,u)" v.64] 136 and Eq, 2.143 into Eq. 2.132 and utilizing the properties of the trace tr(AB") and the fact that $ is symmetric D, EdV +f S:sym{(V,Au)" V,or]av 2144 The 1" term of Eq. 2.144 can be elaborated further. On the basis of Eq. 2.140 8P=V,ov=D,,F 2.145 hence, from Eq. 2.139, 8B can be also expressed asFundamentals [D.FTF+F'DF=D,,£ 2.146 so that Eq, 2.144 can be written as : it TO a) DyuoWine =f, Div BiC:Dy, BAY + fi S:sym[(V,du) V,bvjav 2.147 with Dy,B=5[D,,FF+FTD,,F] =3 (van F+ETY,Au| 2.148 =sym|F"V, Au] Linearization in the current configuration can be elegantly performed by simple push forward and pull back operations of the various terms of Eq. 2.147, Bonet & Wood [1999], Holzapfel (2000). Applying a push forward operation on DE and utilizing the relation V,Au= VAuF ‘au [PaB]=F" Da EF . Se "(eau +V,AuF"| 2.149 =3[(vAuf+v,u]=2 > D,B=FTeF Similarly, the push forward operation of D,, E is P. E|=4, [8B|=F * SEF * =}[r Tee? +8hF] 2 | 7 : 2.150 =a (V,8%)" +(V,8v)F =lvevyt = 3|vew) +Viv] However, from Eq. 2.60 td= aw bv)" + Viv 2.151 Hence B|=F TD EF! 2.152 =F TSF! =8d > D E=F" dF 29Fundamentals In Appendix 2.5 it is shown that on the basis of Eq. 2.149 and Eq. 2.152, the term in the first integral of the linearized virtual work expression in the reference configuration can be expressed in the current configuration as D,E:C:D,,EdV = td:credv 2.153 in which c is the spatial or Eulerian elasticity tensor. Regarding the second integral of Eq, 2.147, the reference configuration gradient terms can be replaced by equivalent terms in the current configuration v,au=VAuF ; — V,ov=VévF 2154 while $ can be substituted from Eq, 2.83 and dV from Eq, 2.10 resulting to 8,|(v,Au)" (V,6v)]av = o:[(V au)" (vov)]av 2.155 By means of Eq. 2.153 and Eq. 2.155 the linearized internal virtual work expression can be expressed in the current configuration as Dy, SW, = fSdicredy + f'o:|(VAu)" (Vbv)]dv 2.156 2.6.2 Linearized External Virtual Work On the basis of Eq, 2.123 the linearization of SW,,, along Aw leads to D,.bWeus =Day [fF-bvdu + ft-bvda — fpa-bvde | 2.157 =DaWerce + Dan’ Went — ParbWorea Utilizing the relation p, = Jp the body forces component can be expressed as Ware = fl dvdv= fog-bvdv= fp g-svav 2.158 : 7 v Then Dg,bWores =P f Pagls-t¥ av =0 2.159 7 because both g and Sv are independent of the displacement field Au For the extemal forces component 8W,,,, itholds Wage = [t-dvda= f'pn-bvda 2.160 in which p is the applied pressure over an infinitesimal area da with normal n.. A follower type external force is assumed.Fundamentals Simo et al. [1991] have proposed an elegant methodology for evaluation of Eq. 2.160. It consists of defining a local set of coordinate axes (1) on the basis of the boundaries of the surface area a and in a manner analogous to that typically utilized in isoparametric elements. Then it holds Gxiox adn lax dx| . S aga 2.161 lax dx jag auf" and hence 2.162 The directional derivative is D,,oW, Dao = : = 2.163 fot Se] +, Utilizing the aforementioned condition Dy, ov =0 and since ' a(x+eAu) aAu Dg, (2 ]=4 afe+ea) —— 2.164 i el Oe a lo and similarly 2.165 Eq. 2.163 can then be written as Dy, OW, att = 2.166 © On f ica aan dx (24u bv ; Eq. 2.166 is unsymmetric hence discretization will result to an unsymmetric tangent matrix, Simo et al. [1991], Holzapfel [2000]. By utilizing the relations between surface and line integrals Bonet & Wood [1999] propose an alternative but symmetric expression for Dy dWoree aS 31Fundamentals ani DysbWore = af? fee Geo} +(e xan| Au aby ~ 4 — x bv] 4] 2" x aul tata oe. [ »| +(e “| edn It is this expression which will be utilized in the following for derivation of the contribution of surface forces to the stiffness matrix. 2.7 Rate Constitutive Equations The relation between the time rates of the second Piola-Kirchhoff stress tensor $ and the Lagrangian strain tensor E can be evaluated by means of linearization of $ along Av because then oteaw)|—FEF8 ED 2.168 but, on the basis of Section 2.2.5, the directional derivatives of S and E along Av are equal to their time rates D,,S=8 ; Dy, B=E 2.169 and, on the basis of Section 2.6, the 4" order tensor OS/GE corresponds to the elasticity tensor C, hence 2.170 or in indicial form Cie 2.171 Kr In the current configuration an equivalent to Eq. 2.170 relation can also be obtained. From Section 2.3.1, the relation between the Truesdell rate of Cauchy stress tensor 6 and $ is 6q,=0" th Sy Fy 2.172 Also from Section 2.2.6 Ba = x da Fy, 2.173 ro Substituting Eq. 2.171 and Eq, 2.173 into Eq. 2.172 32Fundamentals eye" Dh Cmt Er Lem [Death 24 7 FFF Comat] LKL A comparison with Eq. A.2.6.4 indicates that nt Fx Fy 2.175 i.e. the spatial elasticity tensor. Hence 2.176 or 2177 Pinsky et al. [1983] comment that the Truesdell rate of Cauchy stress is the only stress rate having a linear dependence on the rate of deformation tensor through the spatial elastic modulus tensor which is a function only of the deformation. 2.8 Discretization In the following some aspects of the finite element discretization of the linearised equilibrium equation will be presented. 2.8.1. Element Geometry Interpolation Following a typical isoparametric formulation, the matrix N of interpolation functions for an n-noded element can be defined as: =[X. N, + | 2.178 in which Ny@ande=(& & &)- If the vector of nodal coordinates X, is defined as: Kk . Xs] 2.179 then the initial configuration of any point within the element can be interpolated in terms of, the corresponding nodal coordinates as: 33Fundamentals 2.180 Anticipating an isoparametric formulation, at any subsequent time, the current configuration x is interpolated in terms of the corresponding current nodal quantities x, as: 2.181 2.8.2 Field Variables Interpolation Similarly, the displacements can be expressed in terms of the current nodal displacements u, : 2.182 The deformation gradient tensor F is interpolated over an element by differentiating Eq. 2.181 with respect to the reference system coordinates. Utilizing indicial notation 2.183 In order to evaluate the individual Cartesian derivatives VjN, in Eq. 2.183, the following transformation can be utilized 34Fundamentals ON, + = I" VN, 2.184 (ej oe, in which J is known as the coordinate Jacobian matrix: OX, OX, OX, 2.185 On the basis of Eq. 2.180 the individual terms of J can be computed as: o% Eyork x, 2.186 ag) i a8; " In CAPA-3D J and J? are evaluated at every integration point of the element. Then, by means of Eq. 2.184 ON, aN, ( x= 2.187 ON) _ 5-0 (ON) (ox oe and hence the individual terms of the deformation gradient tensor F can be computed. In similarity to Eg. 2.185 the Jacobian on the basis of the current configuration is defined as 3. Ox j= Viziewe, 2.188 ‘Once F is known several other deformation measures can be computed. 2.8.3 Discretized Spatial Equilibrium Equation ‘The expression for the spatial virtual work is given by Eq. 2.116 as sW= forsdde- ff-8vde ~ ftsvaa + foa-bvdv 2.189 In this Section the above will be utilized to express the virtual work contributed by a finite element whose nodes are subjected to the virtual velocity field 6v. In doing so, the various tensors will be substituted with their discretised equivalents. By summing up the contribution of all elements, a statement will be obtained for equilibrium of the continuum in discretised form. The rate of deformation tensor d has been defined earlier in Eq. 2.60 as a=i[vey +04] 2.190 35Fundamentals In terms of the element nodal values and the element shape functions. 2.191 faite Ox, aN, =Sv, gk Enom hence 2 r a De ovn,| + Dv, VN, _ = 2.192 12. =sdIPNent4evM] from which SPN, o6y, + 6v, 2VN,] 2.193 Also dv= ON, bv, 2.194 Substituting into Eq. 2.189 jielem 2 awl = fo:dyolun, o6y, +8v, @VN, [dv a : . 2.195 = PP SON, dy, dv = PtSON, dy, da + foa-SON, by, dv > wl ist + By exploiting the symmetry of the Cauchy stress tensor it holds So[vN, e6v, +6v, 99N,| 2.196 ‘Substituting in Eq. 2.195, a typical term representing the contribution of element node k =a to the total virtual work of the element due to the application of a single virtual velocity bv, at node a is 36Fundamentals Povnav~ fN, fav tda + for,adn v 2.197 =6y,+ ( pier _ pice ) in which Egle f oN, aviv, 2.198 Riclem fr fdv +f tda ~ foN,adv The virtual work contributed by all elements connected at node a is = ee . } | icin _ pice (2 2.199 =v. (E, in which kelem is the number of elements connected at node a, E,, is the total nodal force at node due to internal element stresses and F, is the total external force applied at node a. It then follows that the virtual work contributed by all nodes is nando elem —— tem | nnodes BW Se SS sw[ |= "Soy, -(B, 2.200 fot Tietee io in which nnodes is the total number of nodes. Eq. 2.200 constitutes the discretized equivalent of Eq. 2.114 in Section 2.5.2. At equilibrium gods > by, -(B,-B)=0 2.201 oi 8w and since Eq. 2.201 must be satisfied for any arbitrary set Sv, , it can be concluded that E, V k=1...unodes 2.202 In nonlinear analysis, determination of the nodal displacements, which correspond to the equilibrium solution, is commonly achieved by means of a Newton type iterative technique. As mentioned in Section 2.6, a common characteristic of these techniques is the replacement of the non-linear equation, in the vicinity of a trial solution x= x,,, by a linear approximation in the direction of a displacement increment Au SW(x,) + Dy, SW(x,)=0 2.203 with Day SW=Dy, SW, ~ Dy Wane 2.204 37Fundamentals and §W,,, defined by Eq. 2.122 and 6W,,, by Eq. 2.123. Itis the directional derivative term in Eq, 2.203 which provides the commonly known tangent stiffness operator, a key ingredient in Newton-Raphson type techniques. This can be demonstrated by presuming that a single displacement increment Au, is applied at node b. Then, the directional derivative of the work contributed by node a is, according to Eq. 2.199 Dy, WH =a, [ov.-(B . B= 8¥.-Day, (E, - 2.208 =F,. Then R,(x,+8,cAu,}]] = = = 2.206 The differential term in Eq, 2.206 expresses the change in forces at node a due to changes in the current position of node b which is precisely the definition of a component k,, of the stiffness matrix in finite element analyses D, =by,-k,, Au, 2.207 ‘an, OW haa It can thus be concluded that the sangent stiffness operator necessary for Newton-Raphson type iterative techniques results from the linearization of the virtual work expression. In the following Section, the linearised virtual work expressions determined in Section 2.6 will be iscretised so that they can be utilized for computation of the components of the stiffness matrix of the discretised continuum. 2.8.4 Discretization of the Linearized Equilibrium Equation The linearised internal virtual work expression in the current configuration has been expressed in Section 2.6.1 as Dy 8, = fadicredv + fa:|(VAu)" (vev)]dv 2.208 On the basis of the isoparametric formulation presented in Section 2.8.2 Vu and VEv can be interpolated as VAu=)> Au, @ VN, 2.209 and Viv = Sov, @ VN, 2.210 ro 38Fundamentals Also em FLIPS, Au, +Aa,00N,| 2211 a By means of these discretised expressions the individual integral expression of the virtual work equation Eq. 2.208 can be written as foaccredo= [LS IOn, oo, +5y, @VNj] a o Piet Ae 2.212 Ba lrNeas, +Au,@VNildv Atypical term of the product under the integral of Eq. 2.212 is Q,,=4/9N, 86%, +8, 00, ¢:[VN, Au, + Au, @VN] 2.213 Expressing fv, @VN, and VN, # Au, in index notation bv, @VN, = D> bVamVNan Om BCn 2.214 3 VN, @Au, = D> VN, Au, en en 2.215 mast the double contraction of the 4" order tensor ¢ with VN, @Au, produces a second order tensor which can be written as H(enoe) Performing also the double contraction of sv, ® VN, with the above tensor it results to 3 [ven] DS b¥anV Nan Cian VNYAY,, (¢2@ey](e,@)) silat ao17 = DE BVGVNg, Cas VN Oth, lea ‘On the basis of Eq. 2.217 Qn in Eq. 2.213 can be written as 13 Qy HAE YD vy VN, Cyo PNG, 2.218 igkIeL which can be recast as 39.Fundamentals Quy = va KS, wy, 2.219 with (Kak, = SEPNcuPNy de 3 ie [L,2,3] 2.220 The term Ké,, has been termed the constitutive component matrix of the element tangent stiffness matrix Holzapfel [2000], Belytschko [2000]. Discretization of the second integral term of Eq. 2.208 leads to a Tf [Ex em] [Se.orn,] dv 2.221 ict ist Jovan) (wovJav= fo: A typical term of the product within the brackets of Eq. 2.221 is (Au, @UN,)" Gv, @VN,)=(6v, Au,) VN, @VN, 2.222 hence bv,-Au,) fo:(VN,@VN,)dv 2.223, Jo:[(ov,-Au,)¥N, oN, Noticing that the integral term in the above equation is a scalar, the dot product can be further rearranged to read (8v,-Au,) fo:(VN, @YN,)dv = bv, + Au, 2.224 [fo:[VN, @ VN, } ae The term K= 2.225 fo:[vN, @VN,]a0) associating nodes a and b of the element/structure has been termed the geometrical stress component matrix of the element tangent stiffness matrix Holzapfel (2000], Belytschko (2000). ‘The linearized external virtual work expression in the current configuration has been expressed in Section 2.6.2 as . 1 ¢ [dx |/oau abv Dy PWoree z fol a+ Gee ’ 2.226 dx |(AAu aby - | Bet exons Tt holds fv= oN ov, ; Au=)°N, Au, 2.227 tet it 40Fundamentals hence _ ony vs 2.228 O% ty dau a and eet yea : 2.229 OR far so that 2.230 A typical term of the product under the integral of Eq. 2.226 is Ox{ON, ON, fy, xu, }- Sole =| Me = Sx,] - 2.231 as] déd fe edn, The term 1p fax(an,. Kt poll 2.232 Ox(AN, an associating nodes a and b of the element/structure has been termed the external force component matrix of the element tangent stiffness matrix Bonet & Wood [1999], Belytschko et al. [2000] The element tangent stiffness matrix can now be assembled by the contributions of the constitutive, stress and force component matrices as © EK? 4 Kt Ky, = Ke ae Ky ae Ki, 2.233 For computing purposes matrix notation (formally known also as Voigt notation) is frequently utilized in finite element codes. The transformation rules from tensor to matrix notation are presented in Appendix 2.6. atFundamentals Appendix 2.1 Polar Decomposition Hoger & Carlson [1984] have shown that U~* can be computed on the basis of the following methodology: 1, Compute (a) the right Cauchy-Green deformation tensor C= FF and, (b) P =c'C 2. Compute the eigenvalues A? ; AZ ; A} of C?. 3. Compute the invariants of the right stretch U. On the basis of Eq, 2.27 these are =A, +A, +A, T,=A\A, +A,A,+A,A, A2LL 1, =A,A,Ay 1. Compute U~! by means of the equation uv =e,[q1+e,0+¢,0"| A212 in which q= =y| [QL -%)) 5 -1,(t ; A213 aFundamentals Appendix 2.2 Objectivity of e A typical example of a tensor which transforms according to Eq. 2.69 is the aforementioned Eulerian/Almansi strain tensor e. Reformulating Eq. 2.21 in terms of the rotated vector d& (as)' -(as)’ =ax-ax—ax-ax = dx-dx—dx" (FT F!)ax A221 Comparison of the above with Eq, 2.20 it reveals that F-7 F-! Utilizing Eq. 2.68 Tp. =RF7F'RT A222 Rb'RT which can be recast in the form required by Eq. 2.69 b=RbR™ A223 ‘The magnitude of the rotated vector dx implied by Eq. A.2.2.1 can be further elaborated as, 2 Poa (as -(a8) = dx-ax~ax" (FT Fax A224 =dx-(I- FTF Jax which is identical to Eq. 2.21. Therefore it can be concluded that utilization of the tensor b as a deformation measure produces results that are independent of orthogonal transformations and hence objective. 8Fundamentals Appendix 2.3 Cauchy Stress Symmetry ‘Symmetry of the Cauchy stress tensor can be established by expanding Eq. 2.107 4 _ 7 E07 Fe Bos ee 7 7 = (Gaeta + insta) + 1 7 (Es + Farsi )eo + 1 r (E129 + 621821) ¢s A231 825) + (S15 ~ %51) 82 + (S21 ~ 42) 85 Imposing the condition € : o™ = 0 establishes the symmetric nature of the Cauchy stress tensor. 44Fundamentals Appendix 2.4 Virtual Work Before recasting Eq. 2.115, the virtual work expression, to a form suitable for finite element implementation some mathematical preliminaries will be reviewed. If o represent the second order stress tensor and 6 the virtual velocity vector, the divergence of ( tv) is defined as div(osv) = V(obv) :1 A241 Itholds alos V(o8v) eae, Be, A242 I= x Bin @ Cn A243 Uma hence 6,6 1m? jam aby, oi =| bmFundamentals Wi = dive-iv + 0: Viv Utilizing Eq. A.2.4.4, the term dive -8v in Eq. 2.115 can be replaced by dive bv =div(a8v) - 6: Vov to yield awe fdiv(osv)ba — fo: VEviv + fF-svev — foa-svov=0 By means of Gauss theorem Jf aiv(osv)ov = fn-covea so that sw= fnotvia = fo: vevee + ft-svbv — foa-tvso=0 On the basis of Eq. 2.110 the gradient of the virtual velocity can be rewritten as Viv =8l A244 A245 A246 A247 A248 A249 Also, on the basis of Eq. 2.73 and by taking into account the symmetry of Cauchy stress t? =n"? =n" so that the term (n- ov) can be expressed as n-o6v=n"obv=t" bv=t-bv Substituting these into Eq. A.2.4.8 swe ft-bvba forbid. + [t-dvov - foasviv=0 From Eq. 2.60 and Eq, 2.61 A24.10 A241] A2412but o:bw=tr(o7 bw)= Sey bw, i because o is symmetric and w antisymmetric. Hence Eq. A.2.4.12 can be expressed as fosdasy = ff-ovbu + ft-bvba - fpa-svéu a7 Fundamentals A2413 A2414 A2415Fundamentals Appendix 2.5 Spatial Elasticity Tensor Expressing the first term of Eq. 2.153 in indicial notation D,,E:C:D,,BdV = oa Di Exy Ciyxt, PawEx AV From Eq. 2.149 and Eq. 2.152 PyBw= Faby ly + Paden = Tha tu 5 Substituting into Eq. A.2.5.1 pbc: gba = 51885 }eon|EFeaFafo ae = Fou PBs Fich, Co feat Setting oe # Sa =F BRR Fi Cuan it results to D,B:C:D, dv = 224s cean de = bd:c:edv u 48 A251 A252 A253 A255Fundamentals Appendix 2.6 Matrix Notation The transformation from tensor to matrix notation is accomplished by the index transformations shown in Table A. Matrix Index 1 2 3 4 5 6 Tensor | 1 22 33 12 23 13 Index 2 32 3 Table A.2-] Transformation of indices from tensor to matrix notation. Accordingly 6d can be expressed as fd=(64,, d,, fd,, 264 no I 9 7 r 2 28dy, 28d,,) A26.1 In Eq. A.2.6.1 a factor of 2 has been included in the off-diagonal terms to ensure equivalence between double contraction tensor operations and the corresponding vector/matrix multiplication operations. On the basis of matrix notation Sd can be computed from 8d= 7B, oy, t =Biv A.2.6.2 with B, the familiar in finite element literature matrix: 49Fundamentals A263 The small strain tensor €, (Eq.2.211) can be expressed as or by means of matrix notation in vector format as, 7 5 Roy og 2.) A264 B, u, x — A265 =Bu Applying the transformation of indices from tensor to matrix notation indicated in Table A.2-1 to the elements of the constitutive tensor ¢ of Eq. 2.153 it results to the more familiar format PCr ne ras Cue + Cun Cua Cun Cm +Cnn ome rs, Coe + Cm Czas + Cra: Sana + Cras 1 sy9 Caoir + Caan Conan + Casa Camis +Casm. P= Crom Car Cuaze +Eas2 Cin +Cann — sym Craze +532 Sona + Cass ino + Cram In terms of the vector/matrix expressions for the tensors the first term of Eq. 2.208 can be written as (ey) ([oarcie,dv= f(6d)" De, d [BT DBdel u A267 50.sl Section | Constitutive Theories2Chapter 3 Hyperelasticity 3.1 Introduction In Chapter 2 it was stated that a material is termed elastic if the stress at any time is a function only of the state of deformation (and the temperature). It was also stated that elastic materials have zero internal dissipation for every admissible process. Introducing the additional requirement that the Helmholtz free energy W is solely a function of the deformation gradient F so that av(F) v or F = then, from the Clausius-Planck relation for a non-dissipative material 32 Since F is arbitrary it results “OF 38 Elastic materials for which the stress tensor can be defined on the basis of Eq. 3.3 constitute a subclass of elastic materials termed hyperelastic. Considering that the double contraction P: F expresses work per unit reference volume, it can be concluded that, for a hyperelastic material, the strain energy function corresponds to the work done by the stresses from the initial to the final configuration 34 The strain energy W resulting from the motion x=(X) must be objective. According to Section 2.2.8 this implies that for every orthogonal rotation tensor Q W(F)=¥(QF)=0(F) 35 Since Q is arbitrary, it can be replaced by the transpose of the rotation tensor R resulting from the polar decomposition of F w(F)=4) RTF) RTRU)=¥(U) 3.6 33Hyperelasticity This relation indicates the necessary and sufficient condition for the strain energy to be objective, Holzapfel [2000]. Then, on the basis of Eq. 2.26 and Eq. 2.14 it is also valid W(F)= ¥(C) = ¥(E) 37 Considering the time derivative of Eq. 3.7 OY OF _ Oe ae 38 _ dv ac rp 7 =2—:F 8G at = 5G! F250 from which OY op OE Bo oF ac aw \d hencs P=2F— 3.10 andenes ac Then on the basis of Eq. 2.80 and Eq. 2.81 s-2e | gaay wer 3.1 ac 3.2 Isotropic Hyperelastic Material Response ‘The requirement of objectivity of the strain energy function as expressed by Eq. 3.6 is a general requirement necessary for the response of any hyperelastic material. The notion of isotropy expresses the observation that, many engineering materials exhibit during testing the same physical response irrespective of the direction of testing. Isotropy imposes limitations on the nature of the strain energy function. Let x = >(X) denote the motion of a point of a hyperelastic continuum. Another motion is now considered consisting first of a rigid body rotation of the continuum X = QX and then, of a motion which moves the continuum to the same location as ¢(X) i.e. such that x=0(X)=4(X) * 3.12 Ox _ Ox Itholds Fa 3 = O-FQ 3.13, A hyperelastic material is termed isotropic with respect to the reference configuration if the following relation holds ¥(E)=¥(F)=¥(FQT) 3.14 In similarity to Eq. 3.71 it can also be written * X can be viewed as a new reference system with deformation gradient F = 0x/A%. 34Hyperelasticity F)= v(QF"FQ")=¥(QcQ") 3.15 so that from Eq. 3.14 ¥(C)=¥(6)= ¥(acQ") 3.16 By replacing Q with the orthogonal rotation tensor R,, the strain energy function may also be expressed in terms of the left Cauchy-Green tensor v(C)=¥(RCRT 3.17 Functions fulfilling Eq. 3.16 for any symmetric tensor C and orthogonal tensor Q are termed scalar valued isotropic tensor functions, Rivlin and Ericksen [1955]. They posses the important characteristic that they can be expressed equivalently via the invariants of their arguments i.e. W(C)=Y(I, (C),1, (©),1, (C)) = (I, (b), 1, (b),1, (b)) 3.18 with I, (C)=C:1=df +3 +4 =b:I=1,(b) 1,(€)=4{f- XP + AD +A = HE -b: I, (C)=detC=J? = Xf 033 = det =I, (b) 3.19 in which \, i=1,2,3 are the principal stretches. Several forms of strain energy functions can be encountered in literature. The Blatz-Ko function originally proposed for a compressible material Blatz & Ko [1962] was based on theoretical arguments and was substantiated by experimental data. It has the form 1 | qe _y +l 3.20 in which and j, are the Lameé material constants and v is the Poisson ratio. By setting in the above c=1 the strain energy function describing what has become known as the compressible neo-Hookean model is obtained Lurie [1990], Lubarda [2002] lio » ¥(I,1,45) El, -3)+2(5 -1 + OTe "The definition 1, (C)=C:C=4 +3 +04 =I, (b) is also encountered in literatureHyperelasticity In the previous Section, the various stress tensors were expressed in terms of the tensor awac. On the basis of Eq. 3.18 ¥(C)_ HOWL byly) a, ac Le 3.22 For computation of the third derivative, 9, (detC), the concept of the directional derivative can be utilized, Bonet and Wood [1999], Holzapfel [2000]. Accordingly it results 2¢ d¢(dere}= Pc 3.24 Introducing the above into Eq. 3.11, enables the stress tensor to be evaluated as S$=2| [o,v+1a,¥I- a,ve+1,a, veo! XY A 1 14s, C+s,€ in which the coefficients s,,5),5, for the Blatz-Ko and the neo-Hookean compressible isotropic hyperelastic models are shown in Table 3-1. I Blatz-Ko Neo-Hookean 1 4 of +0-o}] " 1, : p(lc) 82 ag 0 oe L oS 83 +O- P-E -nl, Table 3-1 Material model coefficients for Eq. 3.25 An altemative expression for the Cauchy stress tensor can be obtained by means of Eq. 3.112 as “1 1 o=25 [toa,¥+1.2,,9) +8, ¥b—-1,9,,¥b 3.26 56Hyperelasticity where the relation FFTPF" = bb =b? has been utilized. 3.2.1 Elasticity Tensor In Chapter 2 the 4" order elasticity tensor C was defined as 3.27 F(T Ly, CCK ae 3.28 Expanding Eq. 3.28 (Appendix 3.1), an expression convenient for computation and numerical implementation is obtained for the elasticity tensor in terms of the invariant set (I,,1),1 [eFot+e,dec+Cette,(lec'+cot+ eC @C+e,(C@O"+C 1 @C}+e,01aeT+ 3.29 eC 1c t+. with Gn Wh 2 \ (eye +28] x 292 W421, Of W409, 8422, ¥) {re y2 = (10%), 9 +11, 2%, 9] 3.30 1,0,0 I is the 4" order identity tensor defined as Ty, = dyyc yz, The 4" order tensor C™ is defined, Holzapfel [2000], Lubarda [2002], as 1k ‘IL cloct act 5 —Ieg-k ( KCL + CL chy) 381 ocHyperelasticity Coefficients for the Blatz-Ko and the neo-Hookean compressible isotropic hyperelastic models can be found in Table 3-2 ‘Table 3-2. Material model coefficients for Eq. 3.29 tis worth noticing that in the case of small strains when I,—1, the coefficients c, and c, of the neo-Hookean material are associated with the Lamé constants of small strain elasticity. Once the material elasticity tensor is available, the spatial elasticity tensor can be computed as shown in Chapter 2. 38Hyperelasticity 3.3 Isotropic Elasticity in Principal Directions From Eq. 3.19 it becomes apparent that scalar valued isotropic tensor functions like the strain energy can also be expressed in terms of the principal values of their arguments i.e. W(C)=¥(d,,Ay.A3) 3.32 This enables re-expressing the stresses in terms of the principal stretches , Substituting into Eq. 3.25 the identity 1,2,3. 3 I=DLe@L, 3.33 isl in which L, represent the orthonormal eigenvectors of C and Eq. 2.28 and Eq, 2.29 it results 3 . 2) 49729 wr? 8=Do[2a,.¥+20,.¥(h, Ni} +29 OY? | OL, 3.34 which shows that for isotropic materials, the principal directions of the 2" Piola-Kirchhoff stress tensor and the Lagrangian strain tensor coincide. Furthermore, substituting into Eq. 3.34 the following relations (which are obtained on the basis of Eq. 3.19) a1 a FA, i=b ; U-N=-3 a= 3.35 an? eM ON, the stress tensor can be expressed in terms of its principal components as 3 s 2, gh +20),¥ Bah +20, Pals Le, 3 =) 3.36 us a Alternatively from Eq. 3.11; ow $=2-—= 337 aC From a comparison of Eq, 3.36 with Eq. 3.37 it can be concluded that the magnitude of the i~ th principal stress component is ov _ ov = 3.38 an? AX; X,Hyperelasticity Also, a comparison with Eq. 2.28 and/or Eq. 2.39 reveals that for isotropic materials, $ has the same principal directions as C and E. For the Cauchy stress it holds o=J'FSFT 3.39 which shows that for isotropic materials, the principal directions of the Cauchy stress tensor and the Eulerian strain tensor coincide. 3.3.1 Elasticity Tensor In Chapter 2, E was expressed as 3 3 B=D AA, L024 O44} (94, +L,04) 3.40 a a According to Eq, 2.382 each of the orthonormal eigenvectors can be expressed on the basis of the orthonormal Cartesian basis vectors E, as L,=RE 3.41 in which R. is an orthogonal tensor. Introducing the identity R™R =I and considering the fact that the basis vectors are fixed in space (RRT)RE, 3.42 i from which 3.43 1 orthogonality of L,-s. Utilizing Eq, 2.38 it results FL,= A,1, ea ARL, because of the 3 * Note: FL= nur=Rda,0,04) - 60Substituting Eq. 3.42 into Eq. 3.42 and considering that w is a skew tensor a 3 B= DAA, LOL, + $04} (wL,0L, +101, ist oi 3 4 =DAALOL,+ 4047 (-whet-Lelw) * i i= DAA Leb, ~ $(o™C+Cu) cat As shown in Appendix 3.2, the last term of Eq. 3.44 can be further simplified to : oT C+Cw=- Leuti (hoz, +404) ir Hyperelasticity 3.44 3.45 Substituting this into Eg. 3.44, the tensor E is obtained in terms of its diagonal and off- diagonal components as, a 3 E=D AA LOL, + 2] 0 45(¢ il By iis ]h @h, The time derivative of $ can also be cast in a similar format, From Eq. 3.37 of a $=)8,L,0L,4)8(i91,+1e4,) a a in which the material derivative of the principal stress §, is Also, in similarity to Eq. 3.44 Vs (e401, +L,@0h)= ~ = w8,(L,22, +1;2 2) isl fae ’ By means of Eq. 3.48 and Eq. 3.49, $ can then be written as “Note: L@0L, =1,(oL,)" =1, Io" =- 1, w=- Lele 6 3.46 3.47 3.48 3.49Hyperelasticity 4 8= ee Behhok+ Ya, (5-824, 3.50 im? ij it Examination of the particular forms of and $ reveals that both are expressed as a sum of their diagonal and off-diagonal components. By exploiting this particular characteristic, the Lagrangian elasticity tensor Cfor a material whose response is described in terms of the principal stretches is obtained in Appendix 3.2 as as, LA, LOLOL 24, +23, i(8,-$,} 2,9 1,9 1,8, 3.51 a jet Once C is available, the spatial or Eulerian elasticity tensor can be computed via Eq. 2.181 oHyperelasticity Appendix 3.1 Expansion of Eq. 3.28 leads to 9 [ow)_ 2) 0¥ Ot, ov OL, | ay at AB.L1 ac|ac} ac) ar, ac * “aI, aC — First term a |ow ay) _ ar, jaw), aw a [a _ 30 {ara 30° oe fa1,|" a1, aclac i \ov|_ ew Ol aw AL ay Al, aC aI,| a1,01, dC ALal, OC 1,1, eC : A313 oe ——1 v {i1-c) ot a “O1ar, * a1,aT, "4 1,41, * a {2%,) m= aeloc|~aell=° A314 Hence 2 | av A) _ ay av a [ah iC} a1, ac} aC aC r ‘1, C|aC Second term A316 8Hyperelasticity aa! Pu Oh , PY A, oH Ay C| a1,| A1,a1, AC” 1,01, OC A1,1, aC apo n ABLT ey aw == 1+ (1-0) + 1," 81,01, — 1,01, nt aly) a ae aolt-4] coed AB.L8 19 ac Hence [dv A1,|_ AL | ow oy _ —?e|——I+——_(I,1- Cc} + -——__ =—1,c°" +2ue1- q 7 AL, OC] dC |d1,a1, aor, ) aah t | ey ey =(I1-C)9|-—_1+—_(11-¢) I-e) 1,81, 1,01, \* *ret-y I, ey 2 OW ew =, 191-—cal+? —_1a1-,—_l@ 1 91al, 7 aE — 191,01, / TOL, d1, ay ey 7 1 Cel CeC+]1,——1ec! TOL, Al, ° on +h 381,01, ow “hor, olga 2, =, oe L Pe, Net -| oe +> OY Noo ddI, 1 d1,01, 1, 1,81, 11,01, eu eu . 1eC+ cect Iec! aat, °* a1at, aah — 14, coe ,Ol, “Ol, Third term 2 oe Its A3.L10 ac| a1, acHyperelasticity ay a, ew dl 1,1, 3C *O1,i, AI, AB.1.11 ey ea 1 31,01, t aes) aot 2 1,90" . A3112 ac 3 ac — =Le- 1 tec =1,C1ec"1-1,¢-10c TF ctec aw A3.1.13 acme Toc! ay Pv lt 2 Fw oe = 41, e914 [P41 2 lee Saal, 13 d1,01, 391,01, _ * al, 7 Finally 65Hyperelasticity ror-2* toc+, #y Ol, aly ow aw ~(aral, aL aT, | av), al, ad Talal, ° 1 a1,d1, aw -i——1ec+——Cc Pe 'aLpy ° a ee ie 2, ar b FY, pow Iac} 2. -1,2*% c@c-I, ao" oc? 1,01, ‘wy fu, ey i» ew. ow _ — I 4 a 101,31, ‘oral, anar,* al, ne * aw IeC+Cel + -- ngear|te oy a ay ge Py. oot saan, an (tect+eel) + aren, °°° + A314 1 [at (cec+c ac) + av a : av g I, ctac | 31,1, | al, hee? + 2 OL, a aw onlyHyperelasticity Appendix 3.2 The last term of Eq. 3.44 can be simplified as follows 3 3 3 i DVM LeL, + EAne L, dlhex) Fi a i=l : 3 2p ety at 2p TE aT DNB ET ET + LAE ELE if 3 3 Oars T 2 it DAME LL + OAL LoL A321 i= i 3 2 VwjAp Led, + Soya) ATL,@ L, ij iF Voy (Her, +1,0L) ia A diagonal term of E can be computed as £, =LeL:8 3 = DAA Leh Le, +4 > M- A) Ee LLL, A322 jal ay ie AA, LOL, also, an off-diagonal term of E can be computed as B,=L0L,:8 3 Ae : Lhe wher +4 Ley -A\L@L,: Lal, A323 = i ijt 2_ 42 $4 AP A?) @ Dy Visj In explicit form the contraction 8=C: E is written as 63 B= Cy) Bma(Nj@N,ON, ON,}:(Nq @Nq) ijkjbmn=1 A324 ‘i 3.2. = Cuba (Non) i,gkl=1 Atypical component is then Cyn Eyy + Cire Eqg +---Cijgo Ego + Cig Bas, : A325 a, 8 a, , Sie = +l By te + op 12 32 33 ory OE 2 OB yy OF 33 For i=j vin Py, + Cina Byg +---Cyigo Bag + Chigg By _ A comparison of Eq. 2.6 with Eq. 3.50 shows that C,.,=0. This implies that in the kel contraction $C: only terms of the 4° order tensor C involving tensor products with the form (N; @N, @N, @N,) participate. Hence the diagonal component of the constitutive matrix can be written as 3. as, = DZ A OL eLeL, A327 OA; For ij a comparison with Eq. 3.50 shows that C,,)=0. This implies that in the kel :B, only terms of the 4" order tensor C involving tensor products with the form (N,@Nj®N, @N, \ixj participate. The conditions i+ j and k=l also enable the kel contraction $= reduction of the contraction operation free indices from 4 to 2 so that a typical tensor product term can be written as (N, @N,@N,@N,). Hence the off-diagonal component of the constitutive matrix can be written as 8 L818 L, A328 Summing the two 4" order tensors, the Lagrangian elasticity tensor C for a material whose response is described in terms of the principal stretches is obtained as.Hyperelasticity 3. aS, ; C=C,+C,4= Ln Whoneker, ip PAy 3 A329 +2) u,(8,-S)LoLeLeL, eaHyperelasticity 70Chapter 4 Elastoplasticity 4.1 Elementary Model of Elastoplastic Response Elastoplastic materials are typical examples of dissipative materials with history dependent response controlled by a number of internal variables such as the plastic strain and variables controlling isotropic and/or kinematic hardening. In this Section, an elementary rheological model will be utilised to establish the fundamental aspects of elastoplastic material response. In subsequent Sections these will be extended to the three dimensional case. Considering first the case of an ideal elastic perfectly plastic model, from Fig. 4.1, after yield, the stress in the spring of the elementary plastic model is o=E(e-e,)=Ee, 41 E °y va Ee Sp Aa ot € 2 Fig. 4.1 Elementary elastoplastic model Since the stress appears to be a function of the elastic strain, it is natural to postulate the existence of a Helmholtz free energy function of the form V=V(eQ,...) 42 so that (see Chapter 2) at least for the elastic phase of response it is valid that aw Oe 43 Considering now the case of a hardening elastoplastic model, an internal parameter € can be introduced to control the post-yield response of the model, Fig. 4.2 V=V(e,,6) 44 nElastoplasticity § Fig. 4.2 Hardening post-yield response Then, the small strain uniaxial version of the Clausius-Planck inequality for a dissipative material states fee an0n Oe o-¢-W=0-¢-— Se =o - Ue? -S 43 ov ov. ow ap our ads 7 a Be, Poe 29 which according to the arguments of Coleman and Nol [1963] results to a constitutive relation for the stress | 46 Be and the inequality oeptae 20 47 with q=-00/as. In anticipation of the generalisation of the model to the three dimensional case, the notion of a domain of admissible stresses in stress space is defined as w+{o:£(0,q)<0} 48 in which f is the yield surface. A schematic of £ in the principal stress space is shown in Fig. 43. nElastoplasticity Fig. 4.3 Schematic of yield surface The principle of maximum plastic dissipation has played a crucial role in the modern mathematical formulation of plasticity, Teman [1985], Simo & Hughes [1998]. It states that for a given set of (é,,£), among all possible sets (0,q) satisfying the condition of Eq, 4.8, the actual one is the one which maximizes the argument of inequality 4.7, that is, the one for which eeu : max(o-é, +a-6) 49 Utilizing the formal notation of Appendix 4.1, Simo [1992] has pointed out that the above two relations can be recast as the following constraint minimization problem minimize —(0-2) +q:6) 4.10 subject to f(0,q)<0 According to Appendix 4.4 this is equivalent to 9(0,4) —é, +h: =0 2 00 af (oa) _ ~— 2 oe 4 d£(,4)=0 d>0 Eq. 4.11; expresses the well known notion of normality of the plastic strain to the yield surface. The corresponding plasticity models are termed associative. The conditions expressed by Eq. 4.113 and Eq, 4.114 are known in optimization theory as the Kuhn-Tucker conditions, Luenberger [1984]. In plasticity theory, the Lagrange multiplier \ is known as the plastic multiplier. Its computation will be presented in a following section. BElastoplasticity 4.2 Three Dimensional Elastoplastic Model In the following a model for simulation of the response of a wide variety of engineering ‘materials ranging from soils to conerete to rocks and to asphaltic materials will be presented. It is based on the theory of rate dependent consistent plasticity. By retaining the fundamental to classical plasticity notions of flow surface, decomposition of strains, hardening and/or softening, the theory has emerged as an attempt to provide a realistic, unified, phenomenological approach for materials exhibiting strain rate dependent inelastic deformations. 4.2.1 Constitutive Framework In similarity to inviscid plasticity, in the theory of rate dependent consistent plasticity, the plastic rate is defined as ce ao 4.12 whith \ a constant of proportionality’, and f is a response surface associated with a locus of states of stress @ corresponding to a certain magnitude of inelastic response. ‘Two main phases of material response are distinguished by the formulation, Fig. 4.4: (a) “hardening”, spanning the range from zero stress to ultimate response and, (b) “softening”, spanning from ultimate response to response annihilation, o Fig. 4.4. Schematic of main phases of material model response ‘The standard Kuhn-Tucker conditions are imposed: X20, <0 , X-f=0 413 Evolution of plastic flow is determined by the consistency condition: * For Perzyna type viscoplasticity 4, =T'-{() with T° the material Quidity and (®) an overstress function 4Elastoplasticity i(o,8Tx)=0 414 in which & is the deformation rate, T is the temperature and x is some measure of hardening/softening. 4.2.2, Hardening Response The model utilizes a single flow surface for hardening response simulation. Flow surface characteristics The flow surface proposed by Desai [1990], has been chosen. In stress invariant space it is defined as: J 3 FF 4AS Pa in which: +R), (1,+R) ( ] of } 4.16 Pa Pa 1/2 R= (1-B-cos 30) : 417 4.18 1, J, and J, are stress invariants and p, is the atmospheric pressure. Eq, 4.15 represents a closed surface in the (T,,./. 8) space, Fig. 4.3, alleviating thus the need for definition of additional cap surfaces along the hydrostatic axis. The trace of the surface on the meridian and the deviatoric planes are shown schematically in Fig. 4.5. Fig. 4.5 Trace of the Desai surface in the meridian and the deviatoric planes 8Elastoplasticity The trace of the surface on the octahedral plane is determined by parameter 3 . For B=0 the trace is circular. As B increases, the trace progressively becomes triangular. ‘The hardening response of the material is controlled by parameter a. As a. decreases, the size of the flow surface increases, Fig, 4.6. By defining cas a decreasing function of plastic strain or plastic work, the characteristics of the hardening response of the material can be simulated. Desai surface hardening for decreasing a Fig. 4.6 Parameter « controls hardening of the Desai surface R is the triaxial tensile strength of the material. The slope of the ultimate response surface in the I, — JJ, plane is controlled by parameter ~). Its attained for 0 =0. [B =0.089 y= 0.442 n=25 cos30=1,| 25 on — Est & Eowt 5 0 os 8 ww 8 w 11 Mpa] Fig. 4.7 Hardening response of Desai surface Parameter n determines the apex of the surface on the I, — J plane. It defines the state of stress beyond which the material begins to dilate, Fig. 4.8. 16Elastoplasticity Evol Fig. 4.8...Parameter n determines the point of material dilation As shown in Appendix 4.2, by computing the values of the stress invariants at the state of, stress at which dilation initiates, the value of n can be determined from the expression p= 4, 4.19 (1, +R}'+(1~8eos30) ” 4.2.3. Response Degradation For «=0 the ultimate response of the material is attained. In the model, for deformation levels beyond those corresponding to a = 0, two independent mechanisms are activated for controlling the subsequent response. +) degradation An isotropic measure of response surface degradation has been introduced in the current implementation by means of specifying y, after response degradation initiation, as a decaying function of the monotonically increasing post fracture plastic work, the deformation rate 6 and the temperature T : 4.20 Only work contributed by compressive stress paths is taken into account. After response degradation initiation, in all subsequent steps, the principal values of the plastic strain vector are computed. The increment of post fracture plastic work is constructed then as 3 aw =D dey: dey <0 421 in which o, is the stress component along the principal strain direction I. nElastoplasticity Isotropic softening is meant to simulate, primarily, the overall material response degradation that is observed experimentally as a result of compressive loading. In the model it is complemented by an additional softening mechanism acting on specific material planes as described in the following. Cracking In contrast to the diffused nature of compressive damage, in materials like concrete, rocks and asphalts at low temperatures, tensile states of stress can lead to tensile damage whose nature is much more discrete and which tends to concentrate along planes of cracking. ‘An additional characteristic of these materials is that, for cyclic states of stress in which the material is subjected to alternating stress paths, tensile cracking of the material in one part of the loading cycle does not weaken its compressive strength after stress reversal. On the other hand, as mentioned earlier, compressive loading does seem to weaken the tensile strength of the material even in the cross direction introducing thus some degree of isotropic softening. In order to incorporate both of these aspects of material response into the model formulation, an additional, independent of -y softening, softening criterion has been introduced whose main purpose is the simulation of the tensile softening response. Along the lines of the classical notion of fixed cracking, for states of stress exceeding the magnitude of the flow surface (as determined by the current value of parameter ~ ), a plane of cracking is introduced perpendicular to the principal tensile stress direction. On the crack plane, a Hoffman type criterion similar to that utilized by Schellekens [1992] and Scarpas & Blaauwendraad [1992] is specified to control the subsequent softening response: oP +a(2 +72 )= (82%) — in which o is the normal stress on the crack plane, Fig. 4.9(a), 7, and 1, are the shear stress components, f, the uniaxial tensile stress after crack initiation and x some measure of softening. @) ) Fig. 4.9 (a) Stresses on crack plane, (b) Schematic of Hoffman surface on crack plane 8Elastoplasticity A schematic of the Hoffman surface is shown in Fig, 4.9(b). Up to three orthogonal cracking planes can be introduced at a material point, Fig. 4.10(a). Compatibility of shear stresses along ‘orthogonal crack planes is ensured, Fig. 4.10(b). y (a) ) Fig. 4.10 (a) Orthogonal cracking planes, (b) Shear stresses compatibility on cracking planes 4.3 Material Parameter Determination Evaluation of the model parameters is based on the results of appropriate laboratory investigations. Ideally, results from triaxial tests at different stress paths, rates and temperatures are required if the full simulation capabilities of the model are to be relished. Nevertheless, these are not always easy to obtain, In the following determination of the most significant model parameters will be demonstrated on the basis of uniaxial tests only. Laboratory tests from uniaxial tests on an asphaltic concrete mix of type 0/16 with 6% bitumen 80/100 will be utilised. 4.3.1 Hardening Response Parameters 6, >) and R At ultimate, the hardening parameter c= 0. Hence Eq. 4.15 can be written as: —y(1-Beosas) (1, +R) 4.23 If uniaxial test results are available only, the influence of Lode angle @ on the response can not be included. In terms of the Desai surface, this is equivalent to setting = 0. For determination of the values of the parameters (,-y and R. on the basis of uniaxial tests, for any given combination of temperature and displacement rate values, the ultimate strength values of compressive and tensile tests were plotted on the I,—/Jy space. R was then determined by the intersection of the line connecting these two points with the T, axis. The ”Elastoplasticity value of ~) was determined subsequently by substituting either of the ultimate strength values into Eq. 4.15, By means of this procedure, for the range of temperatures and deformation rates the following relation was determined: vB (1.8) in which {,,(T,6) and £,, (7,8) are the values of the laboratory measured monotonic uniaxial compressive and tensile material strengths. In CAPA-3D, on the I, -/J, plane, R= R(T,4) is computed as the intersection of the line joining the points {,, (7,8) and £,, (1) with the I, axis. In case that triaxial test results at constant deformation rate and temperature are available, the values of I,, J and cos30 at various states of stress, can be computed. Rearranging the os38 (1, +R)'| Desai surface equation, a nonlinear equation can be obtained with and +y” as the unknowns 2 6 | 5 4.25 | 7 ‘An overdetermined system of equations can then be set up for the available sets of I,, J, and cos 30 values (al(s oP B] 4.26 which can be solved via a least squares technique. Dilation parameter n ‘This constant is associated with the state of stress beyond which the specimen begins to dilate. Its value can be determined from 4.27 n (+R) (2 Beosa0) ” in which all invariants are computed for the state of stress at initiation of dilation. 80Elastoplasticity Parameter a For a given deformation rate 6 and temperature T, parameter a determines the size of the successive dynamic flow surfaces and can therefore be utilized as a hardening measure. On a physical basis it can be associated with some measure of plastic deformation, plastic work W,, ete In Scarpas et al. (1997) the postulate was made: a=a(T.6,W,) 4.28 For any test, at a given stress level, Eq. 4.15 can be solved in terms of a. At the same stress level, W,, can be computed from force, 4.29 on 0.075) 0.025)". °. See AA 0 0.0375 0.075 0.1125 0.15 Wp Fig. 4.11 variation for all temperatures and strain rates. Eq. 4.31 indicated by solid line Fig. 4.11 summarizes the experimental results over all tested temperature and displacement rate ranges. The following relationship was found to represent adequately the dependence of ©. on W, over the range of test data 99 — Wy sim Wp l+e-W, 430 in which W, j,.i8 the experimentally determined plastic work at maximum response and c a degradation constant. Non-linear curve fitting of Eq, 4.30 over the available data sets resulted to 0.1-0.67-W, 431 14250-W, > It is compared in Fig. 4.11 with the experimental data. Currently models based on the work of Erkens et al. (2002), Erkens (2002) and Dunhil (2002) for asphaltic materials and Liu (2003) for sand are also implemented. In these the postulate was made: 81Elastoplasticity 4.32 in which t= f (aeP: ae) 433 Details can be found in the respective references. Identification of the initial value a.) of « at which initiation of plastic response occurs can be done on the basis of the ratio t. wa a) in Fig. 4.8. a is computed at that state of stress corresponding to the point of divergence of this ratio from linearity. 4.3.2 Response Degradation ‘Two different sets of model parameters control response degradation. The first is associated with the shrinking size of the Desai flow surface. The second refers to the Hoffman surface. Parameter > During response degradation, for a given deformation rate 6 and temperature T parameter ~ determines the size of the successive Desai dynamic flow surfaces. In Scarpas et al. (1997) the degradation of ~) was associated with W,,; the amount of work expended during the post fracture plastic deformation of the material. Some typical curves are shown in Fig. 4.12. T=25°% 04 0.075 > 0.05 0.025 0 025 05 075 4 Wor Fig. 4.12 Variation of ~ with W. On the basis of the available experimental evidence the following relationship was derived: 2 (vious -k-wip 434 82Elastoplasticity in which ~j,ax represents the value of ») the moment of response degradation initiation and is a degradation parameter which depends on both, deformation rate and temperature. For T=25 °C non linear fitting of the experimental data resulted to 0.32 k(T=25% 2b 435 Models based on the work of Erkens et al. (2002) and Erkens (2002) for asphaltic materials and Liu (2003) for sand are also implemented, Details can be found in the respective references, Cracking By assuming the tensile stress direction as the reference direction, coefficient q of the Hoffinan surface can be computed as: Jee 436 a et in which the uniaxial tensile strength f,,, and the shear strength T, are computed by means of the Desai surface at the moment of crack plane introduction. Isotropic softening of the Hoffiman surface is postulated in the form: Wmnax ow x oof | 437 in which c is a degradation constant, Wag is the crack width at stress annihilation and w is the experimentally measured crack opening on the crack plane, Fig. 4.13. |a oi Wi. & Fig. 4.13 Uniaxial response of specimen in tension 83Elastoplasticity It is important that out of the total measured change in specimen length Al, only that part corresponding to the true crack opening w be utilised in Eq. 4.37. As shown in Fig. 4.13, the remaining component Al, corresponds to the unloading deformation of those parts of the specimen outside the fracture zone. Erkens et al. (2000) have proposed special testing and instrumentation techniques for measurement of the crack opening after crack initiation, Details can be found in Erkens (2002). In addition to Eq. 4.37, the following expression proposed by Erkens (2002) for asphaltic materials is also implemented in CAPA-3D 1 f =f 438 w 1+|———_ [eatin with es =a 439 Z + ghexp{-t4 ) in which a,b,c,f,g and h are laboratory determined material parameters and T is the temperature in Kelvin. 4.4 Algorithmic Aspects Implementation aspects of the specific stress reduction algorithms as they relate to the Desai and Hoffman response surfaces are presented in the following, 4.4.1 Hardening Response The model considers a yield condition of the form: Nn 2 J, 1s) of “al 2 440 Pe ce in which formalistic expressions for a,-; and R_ are determined experimentally as indicated in the previous Section. Integration of Eq. 4.12 with a backward Euler procedure yields: tr At t4AtAeP = At (= 2) 442 da, For the Desai surface: of Of aoe, ao OL,” OYa, 2p 4.43Elastoplasticity in which I=(111000) and ¢ 8518p 2tay »2t ye +22). Substituting in Eq, 4.42 and defining: HAE oe har JS 4.44 ar| that Ack =XAt 4.45 “a a, the increment of plastic strain ‘tStAeP can be expressed in terms of volumetric and deviatoric components as: “Ate? = Ach 14g Ach 4.46 Stress update procedure The implemented stress update procedure is shown in Fig. 4.14. A trial state of stress is defined as: oti to a De Ac 4.47 te trAte Fig. 4.14. Stress update procedure The updated stress state at t+ At is computed by means of a stress correction: tHAtg — gtrial pe trata ep 4.48 Utilizing Eq. 4.42-4.45 it is shown in Appendix 4.3 that the term D®‘*4*Ae? can be expressed in terms of the hydrostatic and deviatoric plastic strain components and the elastic bulk K and shear G moduli. Then, the updated stress state can be written as’Elastoplasticity tet trAtg = gttitl 3K API — GAR 449 which indicates the process of stress correction along the hydrostatic and the deviatoric axes. A schematic of this process is shown in Fig. 4.15. - Fig. 4.15 Stress update along the hydrostatic and the deviatoric axes teat ‘ . a By substituting the vector 6 Ni, ) which defines the return direction on the deviatoric rial plane, with the known vector (& //J, J" evaluated atthe trial stress state, Aravas (1987), proposed a formulation which reduces the order of the nonlinear system of equations that have to be solved to two only. Iti this formulation that has been implemented in CAPA-3D. Rewriting Eq. 4.46 on the basis of the Aravas postulate, the increment of viscoplastic strain can be expressed by means of two unknowns only, namely Ae? and Ae?: trial tt+Ataep _AacPy + Lach |_& | ‘AeP=AcPT + dAc 4.50 ¥ _— Sy Ac® and AP can be computed by means of a Newton-Raphson iterative procedure set up at local material level. Newton-Raphson methodology Eliminating the term \ At between Eq. 4.44 and Eq. 4.45 : 86Elastoplasticity tet tht Ae — at ass, at 2) nce =o 451 | - 2 This relationship together with the consistency condition: "AtG (I), fT Ack, Ache, 452 constitute a system of two nonlinear algebraic equations in the primary variables Ac? and Ach. The system can be solved by means of a Newton-Raphson iteration process. Details are included in Appendix 4.4. Once Ac? and Ac! are known, the updated state of stress can be computed from Eq. 4.49. 4.5 Response Degradation As mentioned in Section 4.2.3, both flow surfaces are assumed active during the response degradation phase of material response. 4.5.1 Desai Surface The algorithm of Section 4.4.1 is applicable without any significant modifications for controlling the softening response of the Desai surface. 45.2 Hoffman Surface On the basis of the implicit stress reduction procedure of Fig. 4.14: oti ty ym ttatag _, trate, =2(o! _ rm) 453 E Also it holds toa tt Atag 4.54 = (44 - traeace) from which trAtae, — tAtaep 4.55 Substituting Eq. 4.53 into Eq, 4.55 after rearrangement it results: Oty y BH AtAEP o gttal 456 ‘Similarly for the shear stresses on the crack plane: Oty, gGthat yer = zttlal 451 tet GQ ttAt yee = sal aos 87Elastoplasticity The values of plastic strains on the crack plane can be evaluated as ofy tt

0 and £(x) <0, Eq. A.4.1.2> is equivalent to the statement that a component of X can be nonzero only if the corresponding constraint is zero and vice versa. This observation enables the development of a solution procedure for the system of Eq. A.4.1.2 according to which, for any given combination of active constraints, the sign of the corresponding Lagrange multipliers is checked after solving the reduced system of equations. Only those combinations of constraints fulfilling Eq. A.4.1.25 are retained, In optimization theory, Eq. A4.1.2, and Eq. A.4.1.2; are known as the Kuhn-Tucker conditions. o1Elastoplasticity Appendix 4.2 The methodology for the derivation of an expression for coefficient n of the Desai surface on the basis of the state of stress is presented in the following. The volumetric plastic strain is A421 with, AA22 Itholds A423 Substituting successively into Eq. A.4.2.2 and into Eq. A.4.2.1 3 ot deb, =dd Lar ar 3 + oI 5454 —F8mnSam)) AA24 () Fig. A.4.2.1 Parameter n controls initiation of volumetric dilation 92Elastoplasticity At the point where the specimen inelastic response changes from contractant to dilatant (point A in Fig. A4.2.1(a)) dep, =0 or, equivalently, from Eq. A42.4, (9f/91,)=0. In the 1, — JJ, space of Fig. A.4.2.1(b) this condition corresponds to the apex point of the trace of the surface. From the definition of the Desai surface, it holds 2-1 ot +R L+R ; onl Pi =2y e (1 Boose A425 Considering that the term ( — cos 38) can not be zero 2y (1 +RP? St LjhtR AA26 on {py Substituting Eq. A.4.2.6 into Eq. 4.15 an expression for n can be obtained on the basis of the stress invariants A427 (1, +R) + (1 -¢0s38) 93Elastoplasticity Appendix 4.3 The methodology for the derivation of expressions relating the first stress invariant I, to the equivalent volumetric plastic strain increment Ae? and the deviatoric stress s to the equivalent deviatoric plastic strain increment Ac! is presented in the following. The stress-strain relation for an elasto-plastic material is: = de® 4deP de, = dey _ dey a ds; ot A431 16, + + ay —— -2 i 3G a ‘Also, the stress increment vector can be expressed as: do;;=ds,; + $d], 6, A432 Solving Eq. A.4.3.1 in terms of ds,, and substituting in Eq. A4.3.2: do;; =2Gde,, aang | = Blane, AA33 3 9K This can be further simplified as follows. From Eq. A.4.3.1, for dl, =a (te, -ad ‘nal A434 af Bonn Substituting into Eq. A.4.3.3 and employing the identity: de, = dey + Fey. 8; AA3s after some algebraic manipulation it results: do, =2Gde, +Kdey, §, af A436 = 2 ae a{(K -36) = Ban $262 For the chosen form of the hardening surface f(0;;) = f(I,, J, hence: at at | at_Ayoy 90;; ol, Boy “aie 80; wed _ ot 94Elastoplasticity so that in Eq. A.4.3.6 the term: A438 can be simplified as: 2¢)\ at 1 at : K— $6 an + 55 FR Summ | Senn 8,5 = ( 1 7 2], OT, | i or 2 ' (5-46) pt vole, A439 3(K -36) ay, Substituting the last of Eq. A.4.3.9 and Eq. A.4.3.7 into Eq. A.4.3.6, the differential increment of stress can be computed in term of an elastic trial stress increment and a stress correction: do, =2Gde,, + Kdey, 6, af G AA43.10 —dd j3K 6, + >= | ar, 8 Yi, 8 Fig. 4.14 portrays the reduction procedure postulated by Eq. A.4.3.10. In terms of finite stress increments: ° at Ao? = dd At) 3K—8,. + A43.11 fi a, %8 ‘As shown in Fig 4.14, itis also valid that tty — gttial _ agp Oj = OR Aol; A43.12 Define: A= Ar ar2t A43.13, a1, Ack =anar—2t A43.14 so that from Eq. A.4.3.11 and Eq. A.4.3.12: oo ee 2 oy, = off! 3K AEBS, — ae A4B.15 9sElastoplasticity Substituting Aly, and oye in the above with their corresponding expressions in terms of 1, and 5,5: 4(2%1,)8, +My Lata) tl gic ack i - Go tatg Ach 443.16 i i i ‘i from which it can be deduced: A43.17 and A4B.18 As mentioned in Section 0, this expression can be recast for stress reduction purposes as: tty ‘rial § a 2| Ga A43.19 ee] | 2Elastoplasticity Appendix 4.4 Local Newton-Raphson Iterative Process Non-linear system set up The magnitudes of the volumetric Ac? and the deviatoric Ae equivalent plastic strain increments can be computed on the basis of Newton-Raphson iterative process. In Section 4.4.1 the following system of nonlinear equations was set up: AAAI Ty, fy AeP, Ack. a6, dg, ddeh +— SL aac? Dace Yt AF A442 ai B2-aAek + ao ach dace A443 =I (I, as, os, BAe BAek] [AP] py a 8 : -] A444 & Om | Jarek!” [b aver OAR Analytical expressions for the individual derivatives are presented in the following. 97Elastoplasticity Computation of dg, /3A } a 7 —(t-1y E+ Ee“ |écdr 3.12 i G(t—réq@dr i j x with GW)=B,, + DB. A 5.13 5.1.2 Incremental Formulation On the basis of Eq. 5.12 the current stress is erat ‘tg Bf emdr + >) Si trae ; 7 ettrat 0 con 5.14 The first term of the above can be written as trae eae B,, f emdr=B,, feed + f ecdr 0 fo : 51SViscoelasticity For the integral of the second term of Eq. 5.14 it holds eae pf ecttkemar 0 teat ‘ = feted + fe (At ede 0 t t era . a Drét—~OC_C_COCO eam gender 0 ° t tat ‘Ath, feo é@mdr + if e t 0 ols gear By means of the mid-point rule the second term of the above can be integrated as, tat i eek sends 2 etal eon} At } ee 5.17 APS Ae Also let eos sed 5.18 Then substituting in Eq. 5.16 vat Petesntheear = e2%5th, + oA de 7 5.19 = tat, On the basis of Eq. 5.15 and Eq. 5.19 the total stress can be expressed as, Simo [1987], Simo & Hughes [1998], Holzapfel [2000] sdtgan, oi 4 Sop, 520 in which ‘*“*H is defined by Eq. 5.15 and ‘t*'h, by Eq. 5.19. 5.2 Generalized Nonlinear Elastic Formulation In the previous Section, the assumption of linearity was made on the elastic stress-strain response. In this Section, this assumption will be modified in order to enable more general hyperelastic models to be utilised. This can be achieved by recasting the formulation of the previous Section in such a way that the notion of strain energy function is introduced in the formulation. Then any suitable strain energy function can be utilised to express a linear or nonlinear elastic stress-strain response. The elastic stored energy in the springs associated with the model of Fig. 5.2 is 14Viscoelasticity x 2 ¥O=$B.,[e06] + FE [ee 08] 5.21 ia or in a notation more appropriate for latter developments in Chapter 6 N VO)= HO] + Yo, [e0.2400] 5.22 Then, the derivative with respect to strain <(t) is CO) . 7 Exelt) + LA KO-e0) x 5.23 =Ec(t) - Salt) in which 5.24 and 4; (O=E,e,; (t) 5.25 From a comparison of Eq, 5.9 and Eq. 5.23 it can be concluded that dv(t) y= 5.26 a(t) 9 Ifa new energy function is defined as wo =tEle(o], > 5.27 and 5.28 then Eq, 5.9 and Eq. 5.11 can be expressed as awe x o()=—— - Ya) 5.29 a(t) 4,()+—— oN a The solution of Eq. 5.29 is (Appendix 5.2) * Bg. 5.282 is a direct consequence of Eq. 5.24 and the definitions of Eq. 5.28, usViscoelasticity OU) | ay Pua t= - f « 5.30 a @=9,)=--fe mae Summing over N xf awit) ‘ pew FU Ya = ef. Wao Seat + oy ; oy = fap Att) aa dr ms Leta, . afe | However, in view of Eq. 5.282 it is valid PUD Pua = 5.32 Sx Gae Trae TO ol Fae al Substituting in Eq. 5.31 Lae Pe wy PVD daw- oe) ir Sue oo ar 5.33 aut) ni ag oe “be “jp ie _ and hence from Eq. 5.29; pore (=o Lire Dr0e dr 5.34 In order to obtain an expression convenient for numerical implementation, the stress at the current time step t + At can be rewritten as Ht Wer i: tga, as f ore 7 + ist 5.35 By means of the same procedure as in Section 5.1.2 the following recursive formula is obtained for computation of Sta HBtg i HOH Sao, at 1 a'wcr) HA UC) SSH tH+Ao ; tH dt; Ao= — 5.36 ) Orde Orde t Pun tp eA th te Ae 5 tha fet Sar i , 1 Orde 116Viscoelasticity 5.3 Three Dimensional Nonlinear Formulation In this Section the elementary multi-component model of Section 5.2 will be extended to the three dimensional case. In the currently implemented version of the model in CAPA-3D the strain energy function W is defined by wor=Fr(erT] there 537 in which ) and 4 are the standard elastic material constants. Any other function can be also implemented. In similarity to Eq. 5.29, the three dimensional stress tensor can be expressed as - Va) 5.38 The internal variables are postulated to be characterized by rate equations analogous to Eq.5.292 (t) 4, We) 4, (+32 5.39 : % de ‘Summing up over N an equation analogous to Eq. 5.33 is obtained N Out) 4 Sethe, FU(t) t= - + ; dn 5.40 ya = Fe J [r- y Ue drde Substituting in Eq, 5.38, the expression for the stress tensor at t+At is finally obtained as eat gPw(t) HBte (tye o() =" J oa At tvatr)y OW(t) oy % J Tbe 5.41 By means of the same procedure as in Section 5.1.2 the recursive formulae necessary for computation of ‘*4'a are N tty ay Sa th, ‘st “OHH +Ao ; tH,,= $.42 M7Viscoelasticity 5.4 Burger’s Model So far the assumption has been made that the Maxwell components comprising the constitutive model were acting in parallel. Nevertheless there is a wide range of engineering materials in which other arrangements of the individual components are more appropriate. A typical example are materials whose response is represented by the well known generalized Burger's mode! which consists of a number of Kelvin components acting in series together with a dashpot and an elastic spring. 5.4.1 Elementary Model ‘A schematic of the elementary generalized Burger’s model is shown in Fig. 5.3. In this type of model the stress is transmitted through each element and the strains are additive so that (t) 5.43 N E(t) =eq(t) +eyp()+ > i= where e¢, andéy, are the total, elastic, viscoelastic and viscoplastic strain al? Eves components. E, 'N Bo ny, ec] LYS on nN Fig. 5.3 Elementary Burger’s model The strain in the elastic component when a stress ois applied to the mechanical device of Fig. 5.3 can be readily calculated as Eq(t) =0(t)/ Ex. 3.44 in which E,, is the modulus of elasticity. On the basis of the observations made in Section 5.1, the stress in the i-th viscoelastic component can be computed from o(t)=n, Evei(t) + Bye, (t) 5.45 t) t) Setting x, = ao 5.46 im 7 a which can be solved by means of a Laplace transform (Appendix 5.3) to obtain 118Viscoelasticity a abo -fe Mh eenar 35.47 i ° Finally, the strain increment corresponding to a stress increment Ac is (Appendix 5.3) tat, > At 5.48 the fn Ao+|1-e % th, E with ay fe : seoas 5.49 0 The response of the viscoplastic component is described by t) ep (t= 22 5.50 Noo which can be solved by means of a Laplace transform (Appendix 5.4) to obtain 1 4 e,p(t)=—{ o(t) t— th, 551 sos } in which técrdt 5.52 Finally, the strain increment corresponding to a stress increment Ac is DEyp = ate ‘ev At Ao 5.53 =—| o(t)+— Noo 2 5.4.2, Three Dimensional Model In this Section the elementary model of the previous Section is extended to the three dimensional case. In three dimensions the equivalent incremental form of Eq. 5.44 is Ac, =C,,Ao 5.54 9Viscoelasticity v0 0 0 te 0 0 0 / ova 0 0 0 — 0 0 A+vy) 0 0 = © oo oO 0 21+vy) 0 0 0 0 0 0 21+v,) and v,, the Poisson ratio of the elastic component. Similarly, for the viscoelastic components and the viscoplastic component it holds -At a Ae,.;=Cyei4 [Ie vn ao+li-e A 'h, 5.56 1. Ae Ay =Cyp] 'o + At 357 in which C,,; and C,, have the same structure as C,, but with E,, substituted by 1, or Mo and Vv, bY ¥; OF VA. As indicated in Section 5.4.1, for a given increment of stress, strain compatibility requires N Ac=Ae, + Acyy + Deaer 5.58 ia Substituting in the above the appropriate expressions for the strain increments it results Ae= C, Ao + Ae Cyp} a+ At 5.59 ~At At l-e ve tHe “4 which can be further rearranged as N Loves Aot+ lage 5.60 with +O y+ 5.61 120Viscoelasticity N and Aé=Ae-)°C,, ial ree Cy, 'o At 3.62 Once Aq is known, the hereditary terms ‘**th, and ‘*4th., can be computed for use in the next time increment. 5.5 Utilization The above material models have been implemented in CAPA-3D and have been utilised extensively for simulation of the response of various structures. Typical applications include the simulation of the response of the mastic constituent of asphaltic materials in micromechanical studies of the response of chip-seals, Huurman et al. [2003], Milne [2004] and, the simulation of the response of asphaltic mixes under complex stress conditions, Collop et al, {2002}, Collop et al. [2003]. 121Viscoelasticity Appendix 5.1 The evolution law for the inelastic strain in the dashpot of Fig. 5.1 was determined in Section S.las at dat ASA a (On the basis of Laplace transforms it can be expressed as 1 5Q8)-Q, +498) =4Ks) ASA2 K . a Qo = + KL B®) ASA st— st ® ® Taking the inverse Laplace transform t . le LQ] =a, % + felt-s)—e % ds J 5 ‘ 5 1 ~t-9, = ayo % + fete % as Je =aye * +20 % fee ds ASA 6 0 low t =0 + =e % Peso ds ® ° =alt) Integrating by parts ! 1 feweras =emre%] — fokecds ® ° ASAS A =e(t)e% ~ (0) — feeds ° Substituting in Eq. A.5.1.4 122Viscoelasticity / a(t =aye™ +e Ke(tye% —e(0)— fede na ° ‘ = [a ~e(o]e% +ett)—f 0 ni =ay-fe" Keds ‘0 %eisrds AS.L6Viscoelasticity Appendix 5.2 The response of every one of the viscous components is described by Eq. 5.292 q(t) +, OWE) 4,() += K; ae Taking the inverse Laplace transform 7 + au at)=aye * +9 f——= ome =ae* +o f ° S -y pOW(t) =qe % +16 % fi eh _ Se ek dr aut) ae Integrating by parts aut) , V(t) , 1 Pw : OO Shae =O — fe*——ar H Owe _ uit) ok avo) _ de ae Pur ‘Substituting in Eq. A.5.2.2 a t o % Ut) & avo) % Avr) a)=ae* +16 2V0)| [ame =|a)— ——— 4 fw Ve ft Ge 7 OVO) fp wry POD ae aioe is the solution of Eq. 5.29 124 A521 AS5.2.2 A523Viscoelasticity Appendix 5.3 The evolution law for the strain in the viscoelastic component of Fig. 5.3 was determined in Section 5.4 as 5) _ oft) &()+ KOM AS31 with x, -2 On the basis of Laplace transforms the above equation can be expressed as SE) -B,(0) + +B) =+26) AS32 7 a E,(0) 0, T+ pso A533 s+— s+— R, a and hence Taking the inverse Laplace transform et) = 27 [B,9| =e(e% + fot Lema ¢(Oe + foe Oa Ir ‘ = AS34 =e % + forme Kar 0 a =e) +e Pormte% ar Oo * a Integrating by parts drsamel — fer sender 0 ( lo : AS3S =o(t)e* =0(0)= fe% simdr 6 Substituting in Eq. A.5.3.4 125