See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/277324166

Upper Bound Shakedown Analysis of Elastic-Plastic Bounded Linearly

Kinematic Hardening Structures.

Thesis · July 2011

CITATIONS READS

7 317

1 author:

Tình Phú Phạm

Hanoi Architectural University
14 PUBLICATIONS   66 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Biomechanics of abdominal organs View project

Einstellbares alloplastisches Schlingensystem zur minimal-invasiven Therapie der Belastungsinkontinenz bei Frauen [Adjustable alloplastic sling system for the
minimally invasive treatment of stress incontinence in women] View project

All content following this page was uploaded by Tình Phú Phạm on 02 June 2015.

The user has requested enhancement of the downloaded file.

 Upper Bound Limit and Shakedown Analysis of Elastic-Plastic
Bounded Linearly Kinematic Hardening Structures

Von der Fakultät für Maschinenwesen

der Rheinisch-Westfälischen Technischen Hochschule Aachen
zur Erlangung des akademischen Grades eines
Doktors der Ingenieurwissenschaften
genehmigte Dissertation

vorgelegt von

Phu Tinh Pham

Berichter: Univ.-Prof. Dr.-Ing. Dieter Weichert

Prof. Dr.-Ing. (griech.) Christos D. Bisbos

Tag der mündlichen Prüfung: 28. Juli 2011

Diese Dissertation ist auf den Internetseiten der Hochulbibliothek online verfügbar.
 Dedication

To my parents
To Loan, Phượng, and Hồng Hà

Phạm Phú Tình

 Acknowledgements

I would like to thank the Government of Vietnam, which granted me a scholarship

by the decision number: 7197/QĐ-BGDĐT, dated 04.12.2006 by Ministry of Education
and Training of Vietnam.

I would like to thank Aachen University of Applied Sciences, which supports me

financially and with good conditions for my research activities.

I will always be grateful to Prof. Dr.-Ing. Manfred Staat, Aachen University of

Applied Sciences, for not only giving me very clear orientation and guidance, but also the
best conditions and encouragement. The time of working on the Ph.D. thesis has been very
effective.

I would like to thank Prof. Dr.-Ing. Dieter Weichert, Rheinisch-Westfälische

Technische Hochschule Aachen (RWTH), for having kindly accepted my registration and
for his kind assistance and supervision.

I would like to thank Professor Dr.-Ing. Christos Bisbos, Aristotle University

Thessaloniki, Greece, for having kindly accepted to review this thesis.

I would like to thank Dr.-Ing. Vũ Đức Khôi and Dr.-Ing. Trần Thanh Ngọc for their
kind help and discussions.

I would like to say thank you to the open source software Code_Aster, which is the
solver of the problem.

Jülich, Germany, 2011

Phạm Phú Tình

The lower pictures have been illustrated by Phạm Phú Tình, dual to the upper pictures, copyright by Bill Waterson.
Abstract
This thesis develops a new FEM based algorithm for shakedown analysis of structures
made of elastic plastic bounded linearly kinematic hardening material. Its concept can be
briefly described as:

Hardening law is simulated using a two-surface plastic model. One yield surface is
the initial surface, defined by yield stress  y , and the other one is the bounding surface,
defined by ultimate strength  u . The initial surface can translate inside the bounding
surface without changing its shape and size. The subsequent yield surface is bounded by
one of the two following conditions: (1) it always stays inside the bounding surface, or (2)
its centre cannot move outside the back-stress surface, where the back-stress surface is
defined by    u   y  . Both ways of bounding are equivalent. The subsequent yield
surface may touch the bounding surface, it means ratchetting occurs and benefit of
hardening is quite clear, or it may not touch the bounding surface, it means alternating
occurs, and there is no effect of hardening. If  y   u , the two-surface model becomes
elastic perfectly plastic model, and if  u  2 y , the model becomes unbounded kinematic
hardening model. Since the two-surface model bases only on yield stress and ultimate
strength, so it does not depend on the hardening curve, consequently it is linear kinematic
hardening.

Direct methods lead to plastic limit and shakedown bounds directly. They help to
reduce considerably computing costs and numerical errors, and make the solution simpler.

Mathematically, the shakedown problem is considered as a nonlinear programming

problem. Starting from upper bound theorem, shakedown bound is the minimum of the
plastic dissipation function, which is based on von Mises yield criterion, subjected to
compatibility, incompressibility and normalized constraints. This constraint nonlinear
optimization problem is solved by combined penalty function and Lagrange multiplier
methods.

Key words: shakedown, ratchetting, two-surface plasticity, bounded kinematic hardening.

Zusammenfassung
Diese Dissertation entwickelt einen neuen FEM-basierten Algorithmus für die
Einspielanalyse von Strukturen aus elastisch-plastischem, beschränkt linear kinematisch
verfestigendem Material. Ihre Konzeption kann kurz beschrieben werden:

Das Verfestigungsgesetz wird mit Hilfe eines Zweiflächenmodells der Plastizität

simuliert. Eine Fließfläche ist die durch die Fließspannung σy definierte Anfangsfließfläche
und die andere ist die durch Zugfestigkeit σu definierte Begrenzungsfläche. Die
Anfangsfließfläche kann innerhalb der Begrenzungsfläche verschoben werden, ohne ihre
Form und Größe zu ändern. Die Folgefließfläche wird von einer der zwei folgenden
Bedingungen begrenzt: (1) sie bleibt immer in der Begrenzungsfläche oder (2), ihr
Mittelpunkt kann sich nicht außerhalb der Rückspannungsfläche bewegen, wobei die
Bauschingerspannungsfläche durch π = σu- σy definiert ist. Beide Begrenzungen sind
äquivalent. Falls die Folgefließfläche die Begrenzungsfläche berührt, tritt Ratchetting auf
und Verfestigung erhöht die Belastbarkeit deutlich; die Folgefließfläche kann die
Begrenzungsfläche aber auch nicht berühren, sodass Wechselplastifizierung auftritt und
sich die Verfestigung nicht auswirkt. Für σy = σu, reduziert sich das Zweiflächenmodell auf
das ideal-plastische Modell und wenn σu ≥ 2σy, gleicht das Modell dem unendlich
kinematischen Verfestigungsmodell. Weil das Zweiflächenmodell nur auf Fließspannung
und Zugfestigkeit basiert, hängt es nicht von der Verfestigungskurve ab, folglich ist es
linear kinematisch verfestigend.

Direkte Methoden führen direkt zu plastischen Traglast- und Einspielgrenzlasten.

Sie helfen beträchtlich, Rechenkosten und numerische Fehler zu reduzieren und die
Lösung des Problems zu vereinfachen.

Mathematisch wird das Einspielproblem als ein nichtlineares Optimierungsproblem

betrachtet. Ausgehend vom oberen Schrankensatz ist die Einspielgrenze das Minimum der
plastischen Dissipationsfunktion, die auf der von Mises-Fließbedingung basiert, restringiert
durch Kompatibilität, Inkompressibilität und normalisierte Nebenbedingungen. Dieses
restringierte nichtlineare Optimierungsproblem wird durch eine kombinierte Straffunktion
und Lagrange-Multiplikatormethoden gelöst.

Schlüsselwörter: Einspielen, Ratchetting, Zwei-Flächen-Plastizität, begrenzte

kinematische Verfestigung.
Contents
Chapter 0 .............................................................................................................................. 1
Overview ........................................................................................................................... 1
Motivations and aims of the thesis .................................................................................... 3
Methods ............................................................................................................................. 4
Organization of the thesis .................................................................................................. 4
Original contributions ........................................................................................................ 5

Chapter 1 .............................................................................................................................. 7
1.1 Inelastic behaviour of material .................................................................................... 7
1.2 Yield function-yield surface ........................................................................................ 8
1.2.1 Von Mises yield criterion ..................................................................................... 9
1.2.2 Tresca yield criterion ............................................................................................ 9
1.3 Perfect plasticity material and the initial yield surface ............................................. 10
1.4 Hardening materials and the subsequent yield surfaces ............................................ 11
1.4.1 Isotropic hardening and its subsequent yield surface ......................................... 12
1.4.2 Bauschinger effect .............................................................................................. 13
1.4.3 Kinematic hardening and its subsequent yield surface ....................................... 13
1.5 Drucker’s postulate .................................................................................................... 17
1.6 Plastic dissipation functions ...................................................................................... 19
1.6.1 Initial surface ...................................................................................................... 19
1.6.2 Subsequent yield surface .................................................................................... 20
1.6.3 Bounding surface ................................................................................................ 20
1.6.4 Total internal dissipation energy of bounded linearly kinematic hardening ...... 21
1.7 Fundamental principles in plasticity .......................................................................... 21

Chapter 2 ............................................................................................................................ 23
2.1 Introduction ............................................................................................................... 23
2.2 Definition of licit fields ............................................................................................. 24
2.2.1 Licit stress field: is the field that satisfies the two following conditions: .......... 24
2.2.2 Licit velocity field: is the field that satisfies the two following conditions: ...... 24
2.3 General theorems of limit analysis ............................................................................ 24
2.3.1 Lower bound theorem......................................................................................... 24

i
 2.3.2 Upper bound theorem ......................................................................................... 25
Summary.......................................................................................................................... 27

Chapter 3 Shakedown theories......................................................................................... 29

3.1 Behaviour of a structure ............................................................................................ 29
3.2 Definition of load domain and fictitious elastic stress field........................................ 31
3.2.1 Load domain ....................................................................................................... 31
3.2.2 Fictitious elastic stress field ................................................................................. 32
3.3 Fundamental theorems of shakedown ......................................................................... 33
3.3.1 Static shakedown theorem (Melan) .................................................................. 33
3.3.2 Kinematic shakedown theorem (Koiter)............................................................. 34
3.4 Methods of calculating LCF limit and ratchetting limit ............................................ 36
3.4.1 Separated shakedown method ............................................................................ 36
Summary.......................................................................................................................... 43

Chapter 4 Numerical formulations .................................................................................. 47

4.1 Introduction ............................................................................................................... 47
4.2 Discretization of the load domain .............................................................................. 47
4.3 FEM discretizations ................................................................................................... 50
4.3.1 Discrete formulation of lower bound method..................................................... 50
4.3.2 Discrete formulation of upper bound method..................................................... 52

Chapter 5 Shakedown algorithms for elastic-plastic unbounded linearly kinematic

hardening structures ......................................................................................................... 57
5.1 Introduction ............................................................................................................... 57
5.2 Upper bound algorithm.............................................................................................. 57
5.3 Dual algorithm ........................................................................................................... 63

Chapter 6 Shakedown algorithm for elastic-plastic bounded linearly kinematic

hardening structures ......................................................................................................... 69
6.1 Introduction ............................................................................................................... 69
6.2 Upper bound algorithm.............................................................................................. 69

Chapter 7 ............................................................................................................................ 77
7.1 Structure 1: Thin plate under tension and temperature.............................................. 77

ii
 7.2 Structure 2: Square plate with a central circular hole ................................................ 80
7.3 Structure 3: Circular plate under surface pressure and uniformed bending .............. 83
7.4 Structure 4: Simple frame .......................................................................................... 84
7.5 Structure 5: Continuous beam ................................................................................... 88
7.6 Structure 6: Cylindrical pipe under complex loading ................................................ 91
7.7 Structure 7: Grooved rectangle plate under tension and bending .............................. 94
7.8 Structure 8: Tension-torsion experiment ................................................................... 96
7.9 Structure 9: Pressure vessel with nozzle.................................................................. 102
7.10 Structure 10: Short cylindrical shell under internal pressure p and ring load Q ... 106

Chapter 8 .......................................................................................................................... 109

Conclusions ................................................................................................................... 109
Further studies ............................................................................................................... 110

References: ....................................................................................................................... 111

iii
Nomenclature

B deformation matrix
c penalty parameter (big number to impose incompressibility condition)
D p ε p  total dissipation function
D p  εup  dissipation function corresponding to bounding surface
D p  εp  dissipation function corresponding to translated yield surface
E Young’s modulus
E tensor of elasticity
f , f0 body force
F yield function
fb back-stress surface
fu bounding surface
fy initial yield surface
f subsequent yield surface
HCF high cycle fatigue
i, NG Gaussian point and total number of Gaussian points in the structure
I1 , I 2 , I3 invariants of the stress tensor
J 2 , J3 second and third invariant of the stress deviator tensor
k, m load vertex, number of load vertices  m  2n 
kv yield stress in pure shear (von Mises yield criterion)
LCF low cycle fatigue
n number of time-dependent loads
ng number of Gaussian points in an element
ne number of elements
P loads
Pˆ k load at vertex k
pp, blkh, perfect plasticity, bounded linearly kinematic hardening,
ublkh unbounded linearly kinematic hardening
t time
t, t 0 traction, given traction
u, du actual displacement, incremental displacement
u, u0 , du velocity, given velocity, incremental velocity
V structure
V ¸ V , Vu boundary, traction boundary, displacement boundary
x coordinate vector
wi weighting factor at the Gaussian point i
Wex external power of loading
Win internally dissipated power
 load factor
 el elastic limit factor
 lim limit load factor

v
 sd lower bound shakedown factor
 
sd upper bound shakedown factor (kinematic theorem)
ε p , u plastic strain increment, displacement increment
 small number to avoid “zero strain”
εe elastic strain
εp, εp total plastic strain and rate
εup , εup plastic strain and rate corresponding to ratchetting
ε , ε
p p
plastic strain and rate corresponding to alternating plasticity
 gradient-operator
L load domain
k , ˆk , ˆs linesearch results
π, π back-stress, time-independent back-stress
ρ, ρ residual stress, time-independent residual stress
σ, σ D actual stress, deviatoric stress
σ0 plastic admissible stress
σE fictitious elastic stress
 y , u yield stress, ultimate strength

vi
 Chapter 0

Introduction

Overview
The main business of structural engineers is to design structures safely, economically and
efficiently. Elastic analysis does not fully exploit the capacity of structures made of ductile
materials. On the other hand limit analysis finds the ultimate strength capacity (static
collapse) with the assumption that applied loads on the structure are time-independent and
proportional. This analysis defines the limit state of the structure made of elastic perfectly
plastic material, and is specified in design codes (strength design codes) or standards.
Actually, applied loads on the structures are mostly neither monotonic nor proportional,
e.g. wind load on the buildings, traffic load on the bridges, waves onto the offshore oil-
rigs, cyclic load on the machine devices, internal pressure in pipes, varying thermal load
etc.. Then the structure may fail by fatigue or unserviceability before reaching its ultimate
strength capacity. Shakedown analysis defines the bounds of low cycle fatigue and
incremental plastic collapse. A structure which is designed based on the shakedown limit is
safer than a design based on the plastic limit. The concept of ratchetting (incremental
plastic collapse) and alternating (low cycle fatigue) are rarely mentioned in the existing
civil engineering design codes. However, all design codes of pressure vessels and piping
allow extended loading by local plastifications in the first load cycles. For this they employ
shakedown analysis implicitly (in the ASME Code) or also explicitly in the so called direct
route of the European standard EN 13445-3, see Staat et al. (2005), Zeman et al. (2006).

The first shakedown theorem was formulated by Bleich in 1932, the static theorem
was extended by Melan in 1936, the kinematic shakedown theorem was stated by Koiter in
1960. Since then there have been many studies on shakedown for elastic perfectly plastic
material. Among them, finite element solutions are introduced by Maier (1969),
Belytschko (1972), Polizzotto (1979), and then shakedown analysis has been extended in
many directions. The influence of geometrical nonlinearities has been considered by Maier
(1973), Weichert (1986, 1990), Groß-Weege (1990), Polizzotto (1996), it has been solved
for contact and friction problems by Anderson and Collins (1995), Polizzotto (1997), Li
and Yu (2006), Ponter and Chen (2006), Zhao et al. (2008), shell structures have been
modelled by Sawczuk (1969a, b), Groß-Weege (1989), Yan (1997), Bisbos and
Papaioannou (2006), Tran (2008), Tran et al. (2008), composite and multilayer structures
and materials have been treated in Tarn et al. (1975), Weichert et al. (1999), Ponter and
Leckie (1998a, b), Carvelli (2004), and the extension to probabilistic problems where
material data and loading are considered as uncertain quantities has been achieved by Staat
and Heitzer (2003), Tran et al. (2009, 2010).

1
 Among algorithms, there are also have many approaches such as: dual algorithms
where the upper and lower bounds are calculated simultaneously were developed by
Andersen and Christiansen (1995), Vu (2001). Makrodimopoulos and Bisbos (2003) with
Second Order Cone Programming, Stein and Zhang (1992, 1993), Staat and Heitzer (1997,
2002, 2003), with basic reduction technique, Ponter (2008) with Linear Matching or
Elastic Modulus methods, Weichert and Hachemi (2010) with Interior Point Difference of
Convex functions, etc.

For more realistic materials and more economic and efficient structures, the
hardening effect should be taken into account. Among hardening models, the isotropic
hardening law is generally not reasonable in situations where structures are subjected to
cyclic loading because it does not account for the Bauschinger effect, and rejects the
possibility of incremental plasticity. The unbounded kinematic hardening model has
already been used by Melan (1938) and later by Prager (1956). It cannot define the plastic
collapse and also incremental plasticity, but only low cycle fatigue. Introducing a bounding
surface in Melan-Prager’s model a two-surface model of plasticity is achieved which
appears to be most basic, suitable and simple for shakedown analysis. Many researchers
have investigated the benefit of hardening such as Maier (1973), Ponter (1975), König and
Siemaszko (1988). Their works are restricted to the unbounded kinematic hardening
material, and they have the similar conclusions for unbounded kinematic hardening model
as mentioned above.

Weichert and Gross-Weege (1988) use the Generalized Standard Material Model
(GSM) which was introduced by Halphen and Nguyen (1976). This model can be realized
by using a simple two-surface yield condition. They used Airy’s stress function to satisfy
the equilibrium conditions in the interior of the structures fulfilled. Recently smoothed
Finite Element Methods have been proposed as a more simple way to achieve solutions
which lie between the displacement FEM and the equilibrium FEM, Liu (2010). Smoothed
FEM have been used in shakedown analysis by Tran and Staat (2009), Tran et al. (2010).

Stein and Zhang (1992, 1993), Zhang (1991) extend the basic reduction technique
for perfectly plastic material to the more realistic bounded kinematic hardening materials
by using the overlay model (also called fraction or multiple subvolume model) which
preserve the characteristic structure of the perfectly plastic formulation. The overlay model
imposes that all the layers are discretized in the same way, i.e. the elements which lay on
top of each other have the same nodes.

Bodoville and de Saxcé (2001) introduce the bipotential approach for Armstrong-
Fredrick non-linear kinematic hardening material. The bipotential concept gives a
particular application for non-associative plasticity. This concept is confirmed by the good
agreement between the analytical solution and the numerical shakedown loads.

2
 Staat and Heitzer (2002, 2003) use a modified basic reduction method, which was
originally used for formulating for perfectly plastic material. The method is applicable with
arbitrary three dimensional finite elements. Their lower bound shakedown solution has
been validated by experiments and has been implemented in the general FE-code
PERMAS.

Nguyen Q.S. (2003), Pham D.C. (2003, 2005, 2007), extend the shakedown
theorems of Melan and Koiter for perfectly plastic materials to the theorems for isotropic
hardening, unbounded kinematic hardening and bounded kinematic hardening materials
with the two-surface model. Pham D.C. (2007) gives the very interesting simplified
theorems.

Shakedown analysis for structures made of hardening material in special cases has
been investigated. The special formulations include: second order geometry effect by
Maier (1972, 1973), cracked structures by Stein and Huang (1996), soil mechanical
problems by Boulbibane and Weichert (1997), Hamadouche and Weichert (1999), Nguyen
Q.S. (2003), Zhao et al. (2008).

Motivations and aims of the thesis

So far, most shakedown solutions for kinematic hardening are either lower bound solutions
or analytical solutions. The main aim of the thesis is to investigate the influence of
hardening on shakedown limit, specifically on the ratchetting by plastic direct method, and
starting from upper bound approach, which seems to be more effective than the lower
bound approach when dealing with realistic problems modelled with several 100 thousands
of unknowns and constraints.

The thesis develops a new FEM based upper bound algorithm for limit and
shakedown analysis of hardening structure by a direct plastic method with von Mises yield
criterion. The hardening model is a simple two-surface model of plasticity with a fixed
bounding surface. The initial yield surface can translate inside the bounding surface, so
that: (1) it always stays inside the bounding surface, or (2) its centre cannot move outside
the back-stress surface. Theoretically, two above ways of bounding are similar and this is
proven in chapter 1 of the thesis.

As the two-surface model is only based on yield stress  y and ultimate strength
 u , so it does not depend on the hardening curve, consequently it is a model of linearly
kinematic hardening.

There are three possibilities for the translated surface at shakedown limit. The first
situation is when the translated surface does not touch the bounding surface, this means
only low cycle fatigue occurs. The shakedown load multiplier of bounded kinematic

3
hardening material is equal to the load multiplier of perfectly plastic material. The second
situation occurs when the translated surface is fixed on the bounding surface, this means
only ratchetting occurs. The shakedown load multiplier of bounded kinematic hardening
material is equal to the load multiplier of perfectly plastic material times the ratio  u  y .
The last situation is the one which is between the two above mentioned situations, when
the translated surface is moving on the bounding surface.

Methods
Mathematically, the upper bound shakedown solution is a nonlinear programming
problem.

Firstly, numerical method is used to solve problem. By finite element method, the
structure V is decomposed into ne hexahedral 20-node elements with the Gaussian points
ng , and NG for one element and for whole structure respectively. Shakedown analysis is
checked only in the Gaussian points instead of whole structure. By the two convex-cycle
theorems, which were introduced by König and Kleiber (1978), any possible time-
dependent load domain could be described by a finite number of load vertices. In
shakedown analysis, it is sufficient to verify for all load vertices instead of whole load
domain. The upper bound shakedown load multiplier is the minimum of the plastic
dissipation function at all Gaussian points and all load vertices.

Secondly, to solve the nonlinear optimization problem with constraints, the penalty
function method is used to deal with compatibility and incompressibility conditions. The
normalized condition is treated by Lagrange multiplier method. In fact, normalized
condition can also be treated by penalty function method, however the application of
penalty function method requires more effort and computer memory to solve, see Vu
(2001). The Newton-Raphson method is cited for solving the Karush-Kuhn-Tucker
conditions.

The new algorithm is a useful tool for estimating the shakedown load multiplier of
structures made of bounded linearly kinematic hardening material with very large number
of elements.

Organization of the thesis

Besides the introduction chapter, the thesis is structured in eight chapters that can be read
more or less independently. Chapter 1 presents the fundamentals of the theory of plasticity,
with emphasis on models of the hardening effect. Chapter 2 presents the basic theories of
limit analysis and chapter 3 presents shakedown analysis theories. In these chapters, limit
and shakedown solutions based on Melan’s theorem as well as on Koiter’s theorem are

4
presented with emphasis on the shakedown problem for elastic-plastic bounded linearly
kinematic hardening materials. Limit and shakedown analysis states the design problem as
a nonlinear optimization problem for the minimum of a failure load and for its dual, the
maximum of a safe load. Chapter 4 transforms the “exact” solutions of shakedown limit
analysis into numerical solutions, based on discretization of the load domain by convexity
and spatial discretisation by the finite element method. Chapter 5 develops an upper bound
as well as primal-dual shakedown algorithms for unbounded kinematic hardening
structures, and chapter 6 develops an upper bound algorithm for elastic-plastic bounded
linearly kinematic hardening structures. In these chapters, the nonlinear programming
problems are solved by combination of the penalty function method and the Lagrange
multiplier method. Chapter 7 provides some numerical applications and validations to
verify the algorithms. Chapter 8 contains some conclusions and further study and
discussions.

Original contributions
According to the author’s knowledge, the original contributions of the thesis are:

 A demonstration that two ways of bounding are equivalent, presented in

chapter 1. In the bounded linearly kinematic hardening model, the translated
yield surface is bounded to stay inside the bounding surface, or equivalently its
centre cannot move outside the back-stress surface.

 A new dual algorithm for shakedown analysis for unbounded linearly kinematic
hardening structures, presented in chapters 4 and 5.

 A new upper bound algorithm for limit and shakedown analysis for bounded
linearly kinematic hardening structures, presented in chapters 4 and 6.

 Tests against analytical solutions, presented in chapter 7.

5
 Chapter 1

Fundamentals in plasticity

1.1 Inelastic behaviour of material

In the geometrically linear theory it is assumed that the total strain  ij can be decomposed
additively into an elastic or reversible part  ije and an irreversible part  ijp (see Fig. 1.1). If
some thermal effects occur, a thermal strain term  ij should be added and thus

 ij   ije   ijp   ij . (1.1)

Fig. 1.1c shows the idealized diagram of the uniaxial tensile test of mild steels,
where  y and  u are yield stress and ultimate strength respectively. The graph indicates
that there is an initial elastic range for the structural steels in which there is no permanent
deformation on removal of the load. When the stress exceeds the elastic limit, the curve
goes to inelastic range, consisting of two parts: initially, a perfectly plastic range occurs in
which the steels yield, that is strain increases without increase in stress. Then follows a
strain hardening range, where increase in strain is accompanied by a significant increase in
stress. If the hardening range is ignored, the curve reduces to the elastic-perfectly plastic
model as shown in Fig. 1.1b. Normally, the plastic strains are much larger than the elastic
strains, and the material behaviour may be further simplified as rigid plastic, as shown in
Fig. 1.1a.

  
Hardening range
Plastic range Plastic range Plastic range
u
y y y
Elastic range
Elastic range

  
   
p e p e

 
p
 

a) Rigid plastic b) Elastic-perfectly plastic c) Elastic plastic hardening

Figure 1.1: Simplified uniaxial stress-strain diagrams.

The elastic part of the strain obeys Hooke’s law of linear elasticity

 ije  Cijkl
1
 kl (1.2)

7
where the elastic constants Cijkl are components of a tensor of rank 4. For an isotropic
material, this tensor is expressed in the form below
E
Cijkl 
1   1  2 
 ij kl 
E
2 1   
 ik jl   il jk  (1.3)

where E denotes the Young’s modulus,  the Poisson ratio, and  ( ,   i, j, k , l ) is the
Kronecker delta (Cartesian unit tensor of second order). The inverse relationship of (1.2)
can be written as
2
 ij  2G ije  G ij kke (1.4)
1  2 
E
where G  is the shear modulus of elasticity.
21   

The plastic strain rate obeys an associated flow law

f
εp   (1.5)
σ

where  is a non-negative plastic multiplier and f  σ, , π  represents a time-independent

yield surface, which will be discussed in the section 1.3.

1.2 Yield function-yield surface

To develop a mathematical theory of plasticity, a basic assumption is made that there exists
a continuous scalar yield function f  σ, , ξ  , which has following properties:
 The equation f  σ, , ξ   0 represents a convex hypersurface, called yield or
loading surface, in the stress space σ for a given temperature  and an array of
internal variables ξ which are determined experimentally. The plastic strain
rate ε p can be nonzero in the region where f  σ, , ξ   0 .
 f  σ, , ξ   0 represents the elastic region, which occupies the interior of the
yield surface. In this region, both plastic strain rate ε p and all internal variables
ξ are zeros.
 f  σ, , ξ   0 corresponds to a region in stress space which is inaccessible for
the material.

For multi-axial stresses, it is necessary to define yield criteria which will cause
yielding. We present here after two most well-known yield criteria in plasticity. From
fig.1.1, we can see that the initial yielding is the same for the cases with or without
hardening.

8
 According to Bridgman (1923), the plastic deformation of metals essentially is
independent of hydrostatic stress. Therefore only the deviatoric stress σ D
1
σ D  σ  tr  σ  I (1.6)
3
causes yielding. Similarly we can express the strain deviator e as
1
e  ε  tr  ε  I . (1.7)
3

1.2.1 Von Mises yield criterion

The von Mises yield criterion is defined by the yield function

f  σ   J 2  kv2  0 (1.8)

where
kv is a material strength parameter. For a perfectly plastic material, kv is a
constant independent of strain history, for a hardening material kv will be
allowed to change with strain history.
J 2 is the second principal invariant of the stress deviator σ D , which has the form1:

1
J2  σD : σD . (1.9)
2
If the material is subjected to a pure shear  12 , while all other stress components

vanish, then J 2  122 , and yielding should occur when 12  kv . Hence the constant
kv   y 3 is the yield stress in pure shear.

1.2.2 Tresca yield criterion

Tresca (1868) concluded that the decisive factor for yielding is the maximum shear stress
in the material. He proposed the yield criterion stipulating that the maximum shear stress
has a constant value during plastic flow.
Using the principal stresses 1,  2 ,  3 is the simplest way to express Tresca’s idea.
If the principal axes of stress are so labelled that
1   2   3 , (1.10)

then Tresca’s yield condition is

f  1   3  2kT  0. (1.11)

σ : σ   tr σ 2  which the negative of the standard definition.

1
1
For a general tensor σ this form is I 2 
2

9
However, f in this form violates the rule that the manner in which the principal axes are
labelled 1, 2, 3 should not affect the form of the yield function. To obey this rule, we
observe that Tresca’s condition states that during plastic flow one of the differences
1   2 ,  2   3 ,  3   1 has the value 2kT . Hence we may write:

f  σ   Max  1   2 ,  2   3 ,  3  1   2kT  0 , (1.12)

y
where kT  ,  y is yield stress. The equation (1.12) is now symmetrical with respect to
2
principal stresses, and can be put into an invariant form

f  J 2 , J 3   4 J 23  27 J 32  36kT2 J 22  96kT4 J 2  64kT6  0 , (1.13)

where J 2 , J 3 are the second and third invariants of the stress deviation tensor σ D with J 2
given in (1.9), and

J 3  det  ijD    ijD Djk kiD

1
3 (1.14)
  11 22 33  2 12 23 31   11D 232   22D 312   33D 122 .
D D D

1.3 Perfect plasticity material and the initial yield surface

The yield function of perfectly plastic material, without hardening and temperature
dependence is

f  σ   F σ    y2  0 . (1.15)

For isotropic materials, the yield function depends only on the invariants of stress.

f  f  I1 , I 2 , I3  (1.16)

where I1 , I 2 , I3 are the three invariants of the stress tensor  ij . In terms of principal
stresses, the yield function is expressed as:

f  f 1, 2 , 3  . (1.17)

The yield surface f 1 , 2 , 3   0 can be plotted in a three-dimensional stress

space. The von Mises yield criterion is f  J 2  k 2 , with k being a constant. The surface
f  0 appears as an infinite circular cylinder whose axis is equally inclined to the
coordinate axes, where the principal stresses 1 , 2 , 3 are taken as the coordinate axes
(Fig. 1.2).

10
 Axis is the line
1  2  3
1

Yield surface
2
J2 - k = 0

3

2

Figure 1.2: Yield surface according to von Mises criterion

in the principal stress plane.

Since the effect of hydrostatic pressure on yield can be neglected, the yield function
will be independent of I1  1   2   3 . It means that the yield function could be
expressed advantageously as:

f  f  J 2 , J3  (1.18)

where J 2 , J 3 are the second and third invariants of the stress deviation tensor σ D .
When f depends on J 2 , J 3 alone, it can be written in the form

f  f 1   3 , 2   3  (1.19)

and the yield surface f  0 can be represented in two-dimensional plot with

1   3 ,  2   3 as coordinate axes, (Fig. 1.3).
Another way of representing a yield surface when it is unaffected by hydrostatic
pressure is to project the yield surface on the deviatoric plane, 1   2   3  0 , (Fig. 1.4).

2 3 1
von Mises, J2 = const.
von Mises

Tresca Tresca
Max. Shear = const.


1 3

2 3

Figure 1.3: Yield surfaces plotted on Figure 1.4: Projection of yield

the plane 1   3  ,  2   3  . surfaces on the deviatoric
plane.

1.4 Hardening materials and the subsequent yield surfaces

The yield function of hardening material depends not only on the invariants of stress, but
also invariants of strain and on the strain history.

11
 Obviously, we can model strain hardening by relating the size and shape of the
yield surface to plastic strain in some appropriate ways.
With the assumptions that the plastic deformation is independent of the hydrostatic
pressure and that the plastic flow is incompressible, the yield surface in the principal stress
space 1 , 2 , 3  is a cylinder of infinite length and the axis 1   2   3 (see Fig. 1.2, the
cylinder is not necessary with circular cross section). The deviatoric plane, which has
1   2   3  0 is perpendicular to the axis 1   2   3 (Fig. 1.2). The yield surface can
be represented by its cross section on the plane 1   2   3  0 . The cross sectional curve
is closed convex, and piecewise smooth. It may change in size and shape during plastic
deformation. Hereafter, we present some hardening models based on hypotheses and
assumptions that approximate real behaviour more or less exactly (Fung & Pin Tong, 2001,
Lubliner, 2005).

1.4.1 Isotropic hardening and its subsequent yield surface

The isotropic hardening model assumes that the material remains isotropic during the
process of plastic loading and that the subsequent yield surface is a uniform expansion of
the initial yield surface. The initial and subsequent yield surfaces have the same centre,
(Fig. 1.5).
In a general case, the subsequent yield surfaces can be expressed in the form

f  f *  J 2 , J3     0 , (1.20)

where  is an internal variable that characterizes the hardening of the material. The strain-
hardening hypothesis assumes that  is a monotonically increasing function, which
depends only on the effective plastic strain but not the strain path.
1 
Subsequent yield surface
Initial yield surface
von Mises, Tresca von Mises u
y
Initial yield surface
Tresca 

b) Uniaxial curve
2 3

a) Projection of surfaces on
deviatoric stress plane.

Figure 1.5: A model for isotropic hardening.

12
 An isotropic hardening law is generally not useful in situations where the structure
is subjected to cyclic loading. It does not account for the Bauschinger effect, and so it
predicts that after a few cycles the structure will just harden until it responds elastically.
This model rejects the possibility of incremental plasticity.

1.4.2 Bauschinger effect

When a metal specimen is subjected to increasing tensile stress (“cold-worked”) the higher
tensile yield stress is accompanied by a lower compressive strength if the load direction is
reversed to compression. This is known as the Bauschinger effect (Fig. 1.6) which is
explained the microscopic stress distribution of the material caused by the dislocation
structure in the cold-worked metal.
.

Tension

y
y
(1) (2) y
y 
 y
y

Compression

Figure 1.6: An expression of the Bauschinger effect.

1.4.3 Kinematic hardening and its subsequent yield surface

The kinematic hardening leads to a translation of the loading surface corresponding to the
propagation and generation of dislocations in crystalline solids. The initial yield surface
can translate without changing its shape and size. The translation may be limited by
obstacles for the motions of the dislocations causing a back-stress on the dislocations so
that their motion becomes progressively more difficult. This is observed as work-hardening
effect i.e. increasing applied stress is necessary to produce additional plastic deformation
(Lubliner, 2005).

a. Unbounded kinematic hardening model

The original Melan-Prager model is characterized by unbounded translation of the loading

surface in the multi-axial stress space. Fig. 1.7a shows that the subsequent yield surface
can translate unboundedly, and as mentioned before, this hardening model rejects the
possibility of incremental plasticity. The subsequent yield surface translates with the
evolution of the back-stress π . This evolution for loading and unloading paths starts once
the structure begins to yield, see Fig 1.7b for linearly kinematic hardening material.

13
 Subsequent yield
1  
surface f


Initial yield y 

y
surface f y




2 y
3

a) Projection of surfaces on b) Uniaxial curve
deviatoric stress plane.

Figure 1.7: A model for unbounded linearly kinematic hardening.

The initial surface is defined in Eq. (1.15), and the subsequent yield surface according to
the von Mises criterion is defined as

f  σ, π   F σ  π   y2  0 , (1.21)

where π is called back-stress

2
π  Cε p , (1.22)
3
with the associated plastic flow, as defined in Eq. (1.5)
f
εp   , (1.23)
σ
and C is a material constant.

Unbounded kinematic hardening means that the ultimate strength  u of the

material is infinite, so this model is not realistic and not suitable for limit analysis.
Moreover, similarly as isotropic hardening model, a structure made of unbounded linearly
kinematic materials can fail only by alternating plasticity, i.e. it is impossible for such a
structure to be involved in incremental plastic collapse, while the alternating plasticity
limit (plastic fatigue) exists and is the same for materials with and without hardening, cf.
Gokhfeld & Cherniavsky (1980)
More realistic two surface models of bounded kinematic hardening have been
introduced by Weichert and Gross-Weege (1988), Stein and Zhang (1992, 1993), and
investigated by Staat and Heitzer (2002, 2003), Nguyen (2003) etc. We present these
models in the next section.

14
b. Bounded with two-surface plasticity model

In the two-surface model of plasticity the translation of f is constrained by a bounding

surface f u . This model means that one surface is the initial surface f y and the other one is
the bounding surface f u (see Fig. 1.8). The stress σ is bounded by the ultimate strength
 u . The initial yield surface can translate inside the bounding surface, without changing its
shape and size, such that the subsequent yield surface always stays inside a bounding
surface, consequently, the centre of the subsequent yield surface cannot move outside the
back-stress surface f b . Fig. 1.8b shows the case of nonlinear kinematic hardening of e.g.
the Armstrong-Frederick model in which the bounding surface is reached asymptotically.
Our two-surface shakedown analysis uses a linearly kinematic hardening model.
According to Staat and Heitzer (2002), both models give very close shakedown limits.

1 
Bounding surface f u u



y
y

y
Subsequent yield surface f

y


u
Initial yield surface f y 

2 3
Back-stress surface f b

a) Projection of surfaces on b) Uniaxial curve

deviatoric stress plane

Figure 1.8: A two-surface model for bounded kinematic hardening.

The initial surface is defined in Eq. (1.15), and the subsequent yield surface is
defined in Eq. (1.21). The subsequent yield surface may or may not touch the bounding
surface. It is bounded by one of the two following ways:

a) it always stays inside a bounding surface, expressed by the following condition

F σ    u2 . (1.24a)

b) its centre cannot move outside the back-stress surface f b , expressed by the
following condition

F  π   u   y  .
2
(1.24b)

15
In Eqs. (1.24a) and (1.24b), inequality presents the subsequent yield surface does not touch
the bounding surface.
Intuitively, from Fig. 1.8, we can see that conditions (1.24a) and (1.24b) are equivalent,
otherwise, it is easily proved.
Firstly, we prove: (1.24a)  (1.24b) . From the triangle inequality

F σ   F  σ  π   π   F σ  π  F  π   y   u   y    u (a)

 F σ  π   y , and

F σ  π  F  π   y   u   y    (b)
 F  π   u   y 

then

F σ     F  π   
  y  . (Q.E.D.)
2 2
u u (1.25)

Finally, we prove: (1.24b)  (1.24a) .

 u   y  y
If we choose π    σ so that σ  π  σ . Then for any σ satisfying F σ    u2
  u   u

we will find:

 u   y    u   y 
2

F  π  F   F σ    u   y 
2
σ   (c)
 u   u 
From (c) and π chosen before, we have

 y  y 
2

F σ  π   F  σ     F σ    y2  F σ    u2 (d)
 u   u 
then

F π   u  y 
2
   F σ    . 2
u (Q.E.D.) (1.26)

Some authors such as Zhang G. (1991), Pham D. C. (2005, 2007) use condition
(1.24a). Other authors like Weichert, Groß-Weege (1988), Heitzer, Staat (2003) use
(1.24b).2 Zhang G. (1991, pp. 43-44) argues that F  π   u   y 
2
is safer than
F σ   u2 . The above prove however shows that both conditions are equivalent.

Zhang (1991), Heitzer and Staat (2003) have proved that when  u  2 y then
shakedown limit of bounded kinematic hardening structure is equal to that of unbounded

2
Heitzer (1999) has proven that the optimization problems of lower bound shakedown analysis based on
either constraint (1.24a) or (1.24b) have the same solution. The above prove is independent of shakedown
analysis.

16
kinematic hardening structure. Gokhfeld & Cherniavsky (1980) also stated that the
problem of alternating plasticity can be solved directly by using the fictitious elastic field
σ E . It is due to the fact that the plastic fatigue limit is determined by the condition that
anywhere in the structure; the maximum magnitude of fictitious equivalent (von Mises)
elastic stress cannot exceed two times the yield limit of the material. Moreover, intuitively,
from Fig. 1.6, the evolution of the initial yield surface takes place within the range of 2 y .

1.5 Drucker’s postulate

Drucker’s postulates were developed in the 1950s in an attempt to provide the missing link
between material behaviour and mathematics, (see Bower, 2010).
The process of application and removal of the additional stress is called stress
cycle. Removing the additional stress enables the stress of the structure to return to the
original stress state, but the strain state can be different if plastic deformation occurred
during the stress cycle. Note that dσ : dεe is always positive, where dεe is the elastic strain
increment which is recoverable. Drucker (Fung & Pin Tong, 2001) accordingly defines a
work-hardening (or “stable”) plastic material if the following two conditions hold true.

 the work done during incremental loading is positive

dσ : dε  0; and (1.27)

 the work done in the loading-unloading cycle is nonnegative

dσ :  dε  dεe   dσ : dε p  0 (1.28)

where dε p is the plastic strain increment, which is not recovered by the process of stress
cycle. Eq. (1.28) sometimes known simply as Drucker’s inequality, is valid for both
work-hardening and perfectly plastic materials, where for perfect plasticity, it assumes
equality: dσ : dε p  σ : ε p  0 .
Drucker (Fung & Pin Tong, 2001) extended the definition of work-hardening to
allow for a finite dσ produced by the external agency. In fact, the initial stress, σ 0 , may
not at any point be outside the yield surface such that σ 0  σ . The work per unit volume
done by the external agency is  σ  σ 0  : dε p . Eq. (1.28) is replaced by

σ  σ  : ε
0 p
 0. (1.29)

Equation (1.29) is also called the principle of maximum plastic dissipation. It can be
written in the form

σ : ε p  D ε p , ξ   σ0 : ε p , (1.30)

17
where the plastic dissipation D  ε p , ξ  depends on the plastic strain rate ε p and the
internal variable ξ only. The back-stress is a suitable internal variable, i.e. ξ  π .

Consequences of Drucker’s postulate

 The yield surface and all subsequent loading surfaces must be convex.
The convexity of the yield surface plays a very important role in plasticity. It
permits the use of convex programming in limit and shakedown analysis. It should be
noted that Drucker’s postulate is quite independent of the basic laws of thermodynamics. It
does not hold if internal structural changes occur or for temperature dependent behaviour
(Kalisky, 1985). Furthermore, the yield surface fails to be convex if there is an interaction
between elastic and plastic deformations, i.e. if the elastic properties depend on the plastic
deformation (Panagiotopoulos, 1985).

 The plastic strain increment vector must be normal to the loading surface at a
regular point, and it must lie between the adjacent normals to the loading
surface at a corner of the surface.

Yield surface

p
 0
p

  0 Elastic
region

Inadmissible region

Figure 1.9: Normality rule.

Two above consequences appear to ensure that  σ  σ 0  : ε p  0 , and can be represented as

f
εp   . (1.31)
σ
In the case of having n intersected differentiable yield surfaces at a singular point, (1.31)
should be replaced by:
n
f k
ε   k
p
. (1.32)
k 1 σ

18
  The rate of change of plastic strain must be a linear function of the change of
the stress.

1.6 Plastic dissipation functions

The plastic dissipation function is defined in Eq. (1.30)

D ε p , π  σ : ε p (1.33)

σ is the existing stress tensor satisfying the yield condition f  σ, π   0 . Hereafter, we will
express the dissipation function of the initial yield surface, of the subsequent surface, and
of the bounding surface according to the von Mises criterion. As mentioned before, the
effect hydrostatic stress can be neglected in metal plasticity, (see Eq. (1.6) and Eq. (1.7)).

1.6.1 Initial surface

The initial yield surface is defined in Eq. (1.15)

f  σ   f y  F σ    y2  0 (1.34)

For the von Mises criterion it can be described as follows

fy  3
2
σD : σD   y  0 . (1.35)

The dissipation function corresponding to (1.35) is defined as

D p ε p   σ : ε p   y 2
3
εp :εp . (1.36)

It is quite clear that if there is no evolution of the initial surface, then the initial
yield surface is exactly the yield surface of a perfectly plastic material.

The total plastic strain rate ε p can be separated into two components, König (1987)

ε p  x, t   εup  x, t   εp  x, t  . (1.37)

The first term of Eq. (1.37) represents a perfectly incremental collapse process over
a time interval  0,T  , in which a kinematically admissible plastic strain increment εu  x 
is attained in a proportional and monotonic way

εu  x, t     x, t  ε u  x  (a)

εu  x   εu  x, T   εu  x,0  is kinematically admissible (b)
 (1.38)

T

  
 x , t  0 ,    x, t  dt  1 (c)
 0

and the second term of Eq. (1.37) represents an alternating plasticity process

19
 T
εp  x    εp  x, t  dt  0 . (1.39)
0

Obviously this alternating plasticity process should correspond to an alternating stress

process for an isotropic material.

1.6.2 Subsequent yield surface

The subsequent yield surface (translated surface) is defined in Eq. (1.21)

f  σ, π   f  F σ  π   y2  0 . (1.40)

For the von Mises criterion it can be described as follows

f  3
2 σ D
 π : σ D  π   y  0 . (1.41)

The dissipation function corresponding to (1.41) is defined as

D p  εp    σ D  π  : εp   y ε : εp .

2 p
3 
(1.42)

As mentioned in Eq. (1.37), εp involves only alternating plasticity (or LCF).

1.6.3 Bounding surface

The bounding surface is defined in Eq. (1.24a)

f  σ   fu  F σ    u2  0 . (1.43)

For the von Mises criterion it can be described as follows

fu  3
2
σD : σD  u  0 . (1.44)

The dissipation function corresponding to (1.44) is defined as

D p  εup   σ : εup   u 2
3
εup : εup . (1.45)

As mentioned in Eq. (1.37), ε up involves to ratchetting or incremental plastic collapse.

when f is bounded by f u , correspondingly, its centre cannot move outside the back-
stress surface f b which is defined in Eq. (1.24b)

f  π   fb  F  π   u   y  .
2
(1.46)

For the von Mises criterion it can be described as follow

fb  3
2
π : π   u   y   0 . (1.47)

20
The dissipation function corresponding to (1.47) is defined as

 
V
2
3 u   y  ε p dV , (1.48)

where
T
ε   ε p dt .
p
(1.49)
0

As proved in the last section, bounding conditions (1.43) and (1.46) are equivalent.

1.6.4 Total internal dissipation energy of bounded linearly kinematic hardening

Since the initial yield surface translates inside the bounding surface during the load cycles,
so the dissipation of the subsequent yield surface depends not only on the stress but also on
the strain history- Then it is more convenient to define the dissipation function for any case
of shakedown (see Eq. (1.50)) than for a certain case such as either alternating (see Eq.
(1.42)) or ratchetting (see Eq. (1.45)), (see Nguyen Q.S (2003)).
T T

  D  ε  dVdt     y ε p dVdt      y  ε p dV .
p p 2 2
3 3 u (1.50)
0V 0V V

On the right hand side of Eq. (1.50), the first term is the internal plastic energy of a
structure made of perfectly plastic material, and the second term is the effect of hardening.

1.7 Fundamental principles in plasticity

Consider a structure subjected to volume loads (body force) f and surface loads t 0 . The
stresses σ are said to be in equilibrium if they satisfy the equations of internal equilibrium
divσ  f  0 in V (1.51)
and the conditions of equilibrium on the traction boundary V of the body

σn  t 0 on V . (1.52)

Any stress field satisfying conditions (1.51) and (1.52) is called a statically admissible
field. Furthermore, if this stress field nowhere violates the yield criterion, f  σ   0 , it is
called a plastically admissible or licit stress field.
The actual flow mechanism is composed of the velocities u and strain rate ε in the
body which satisfy the compatibility condition and kinematical boundary conditions

ε
1
2

u   u 
T
 in V , (1.53)

u  u0 on Vu . (1.54)

Any mechanism  u, ε  satisfying conditions (1.53) and (1.54) is called kinematically

21
admissible. Furthermore, if this mechanism furnishes a non negative external power3

Wex   f T  udV  t
T
0  udS  0 (1.55)
V V

then it is called a licit mechanism. A kinematically admissible strain and displacement field
can be defined in a similar manner.

Principle of virtual power

One of the main tools in the mechanics of continua is the principle of virtual power,
which states that for an arbitrary set of virtual velocity variations  u that are
kinematically admissible, the necessary and sufficient condition to make the stress field σ
equilibrium is to satisfy the following variational equation

 σ :  εdV   f   udV  t   udS .

T T
0 (1.56)
V V V

Principle of complimentary virtual power

For an arbitrary set of virtual variations of the stress tensor  σ that are statically
admissible, the necessary and sufficient condition to make the strain rate tensor ε and
velocity vector u compatible is to satisfy the following variational equation

 ε :  σdV   n σ   udS .

T
(1.57)
V Vu

Equation of virtual power

From the two above principles, one can easily deduce that for all strain rate tensors
ε and velocity vectors u that are kinematically admissible, and for all stresses σ that are
statically admissible we have the following virtual power equation

 σ : εdV   f  udV  t  udS   nσ   u0dS .

T T T
0 (1.58)
V V V Vu

In the above variational principles no constitutive equation of the material is

assumed. Variational Eqs. (1.56) and (1.57) are applicable even if the body is not elastic,
for which the energy functional cannot be defined. The finite element analysis of structures
is based on these variational equations.

3
The notation of the product f T  u assumes matrix notation in Cartesian coordinates. In tensor notation the
transposition has no effect and we could simply write f  u  fi ui .

22
 Chapter 2

Limit analysis

2.1 Introduction
In this chapter we only present the basic theory of limit analysis but not all details because
we will state that limit analysis is a special case of shakedown analysis in next chapters.
Let us consider a structure of volume V made of ideal plastic or bounded
hardening material and subjected to external loading P . The external loading P consists
of general body force f in V and surface traction t on V . The bar (horizontal line)
placed over the loads f and t implies that these loads are time-independent. We assume
that:

 The material is ductile so that the structure can undergo large deformations
beyond elastic limit without fracture.

 The deflections of the structure under loading are small so that second-order
effects can be ignored. Therefore buckling cannot be considered in limit analysis.

 All loads are applied in a monotonic and proportional way, i.e.

P   P0 (2.1)

where P0   f0 , t0  denotes the nominal or initial load. If the value of the load factor  is
sufficiently small, the structure behaves elastically. As  increases and reaches a special
value, the first point in the structure reaches the plastic state. This state of stress is called
elastic limit. The corresponding load factor is denoted  el . Further increase of  will lead
to the expansion of plastic regions in the structure. The structure gradually reduces the
statically indeterminacy (redundancy). At limit state, the structure forms a collapse
mechanism and therefore fails to carry the applied load. If P0 represents the applied load,
the value  lim corresponding to the plastic collapse state is called the safety factor of the
structure or the limit load multiplier.
The theory of limit analysis offers a way to solve directly the problem of estimating
the plastic collapse load Plim , bypassing the spreading process of the plastic flow. The limit
value of the load is estimated and at the same time the limit state of stress in the whole
structure can be evaluated. The limit load and stresses so obtained are of great interest in
practical engineering whenever the model of perfect plasticity or bounded hardening and
the small deformation and small displacement assumption constitute a good approximation
of the material.

23
2.2 Definition of licit fields

2.2.1 Licit stress field: is the field that satisfies the two following conditions:

1) Equilibrium conditions (static conditions)

 σ
  f  0 in V (a)
 x (2.2)
σn  t on V
 0 (b)
2) Nowhere in the structure the yield condition is violated

f  σ, π   0 . (2.3)

2.2.2 Licit velocity field: is the field that satisfies the two following conditions:

1) Compatibility conditions (kinematic conditions)



1
2

ε  u   u 
T
 in V (a)
(2.4)

u  u0 on Vu (b)
and for incompressible material, it requires

tr  ε p   0 . (2.5)

2) External power is non-negative

f  udV   t0T  udS  0 .

T
(2.6)
V V

2.3 General theorems of limit analysis

Let us note that subscript and superscript of the load factor  indicate the following:  


and  

stand for lower bound and upper bound approach respectively,  
pp
and  
blkh

stand for perfect plasticity and bounded linear kinematic hardening material respectively,
and  lim  stands for limit analysis.

2.3.1 Lower bound theorem

The lower bound theorem of limit analysis states that: The true limit load factor  lim of
a perfectly plastic structure is the largest one among all possible static factors  lim


corresponding to the set of all licit stress fields σ , that is

 lim

  lim
pp
. (2.7)

24
 Then, the objective of the lower bound approach of limit analysis is to find the
maximum load factor  lim
pp
for which the structure is safe. The task of computing becomes
a nonlinear optimization problem

 lim
pp
 max  lim

(a)
s.t.: (2.8)
 F  σ    y2 in V (b)

divσ   lim f0

in V (c)

σn   lim t0

on V (d)

for the structure V , traction boundary V (with outer normal n ), yield function F , yield
stress  y , body forces  lim

f0 , and surface traction  lim

t0 .

For bounded linear kinematic hardening bodies, Eq. (2.7b) becomes

F  σ    u2 in V . (2.9)

Consequently, we have
 u pp
 lim
blkh
  lim . (2.10)
y

2.3.2 Upper bound theorem

The upper bound theorem states that: The true limit load factor  lim
pp
of a perfectly
plastic structure is the smallest one of the set of all factors  lim

corresponding to the set of
all licit velocity fields u

 lim
pp
  lim

(2.11)

where

 lim

 Win Wex (a)
Win   D p  ε p  dV (b)
V
(2.12)
Wex   f0T  udV   t0T  udS  0 (c)
V V

u  u 0 on Vu (d)

with Win and Wex are the total power of the internal deformation and the power of the
external loads of the structure.
According to Eq. (1.36), the dissipation function D p  ε p  in Eq. (2.12b) is

25
 D p ε p   σ : ε p   y 2
3 ε p : ε p for perfectly plastic structures,
(2.13)
D p ε p   σ : ε p   u 2
3 ε p : ε p for bounded kinematic hardening structures.

Then, the objective of the upper bound approach of limit analysis is to find the
minimum load factor  lim for which the structure fails by plastic collapse. The task of
computing becomes a nonlinear optimization problem (written in a normalized form)

 lim
pp
 min  D p  ε p  (a)
V

s.t. : (2.14)
 1

ε  2 u   u 
T

in V (b)

u  u 0 on Vu (c)

Wex   f0  udV   t0  udS  1
T T
(d)
 V V

and the limit load factor for bounded kinematic hardening as well as isotropic hardening
bodies is calculated as in Eq. (2.10).

26
Summary
From equations (2.7) and (2.11), we can summarize that the objective of limit analysis is to
find the lower bound limit load  lim

P0 , which is less than, or at best equal to the true load
capacity Plim   lim P0 , or is to find the upper bound limit load  lim

P0 , which is greater than,
or at best equal to the true load capacity Plim   lim P0 .

Safe load  True load capacity  Overload

or (2.15)
 
lim   lim   
lim

An “exact” limit analysis solution must satisfy both the licit stress field and the licit
velocity field. Otherwise, lower bound and upper bound solutions are approximate ones.
Duality shows that
Maximum safe load  True load capacity  Minimum overload
or (2.16)
max lim

  lim  min  lim


Limit load factor

Upper bound solution

+lim

Exact solution
Iteration

Lower bound solution

-lim

Figure 2.1: Exact and approximate solutions.

27
 Chapter 3

Shakedown theories

3.1 Behaviour of a structure

For structures subjected to varying loads, shakedown is more relevant than static collapse,
which has been investigated by limit analysis in the last chapter. A structure made of
perfectly plastic materials as well as linear or nonlinear kinematic hardening materials,
subjected to cyclic loads may behave in one of the following ways, depending on the
intensity and character of the applied loads.

1. If the applied loads are sufficiently low, the behaviour of the structure is purely
elastic  ε  ε e  . When the first point of the body reaches the yield stress, it
reaches the elastic limit, (see Fig. 3.1). The behaviour of the structure according
to this state does not influence its integrity, since macroscopic plastic
deformation and damage do not occur at all (except of high cycle fatigue,
HCF). However, the load carrying potential of the structure is not fully
exploited.

2. If the load intensities is higher than the elastic limit, but do not exceed a certain
limit (shakedown limit load), then plastic deformation occurs initially  ε p  0  .
After some cycles, the structure reaches a steady state in which the total plastic
energy dissipated in the structure becomes bounded, no further plastic
deformation occurs  ε p  0  . The structure behaves as if it was elastic. We can
say that the structure shakes down or the structure adapts itself, (see Fig. 3.2).

 

max max

 
min min

Figure 3.1: Purely elastic. Figure 3.2: Shakedown.

29
 3. If the load intensity is higher than the shakedown limit, then no stabilization
situation is established, plastic flow continues to develop. The response of the
structure may be one of the two following modes (or both simultaneously):

 If the plastic strain increments in each load cycle,  ε p  0  , are of the

same sign, the plastic deformations in each cycle accumulates,  ε p  0  ,

so that after enough cycles the total strains (and therefore displacements)
become so large that the structure departs from its original form and
becomes unserviceable. This phenomenon is called incremental collapse or
ratchetting (see Fig. 3.3).

 If the plastic strain increments change sign in every cycle  ε p  0  , they

tend to cancel each other  ε p 0  , and the net deformation remains small.
However, after a sufficient number of cycles, the structure may fail by
fracture due to Low Cycle Fatigue (LCF). This phenomenon is called
alternating plasticity or low cycle fatigue, (see Fig. 3.4). Hardening does not
affect this state.

 

max max

 
min min

Figure 3.3: Ratchetting. Figure 3.4: Alternating plasticity.

Note that: (1) incremental collapse and alternating plasticity may appear
simultaneously, and (2) in the alternating plasticity mode, there is no essential difference
between elastic-perfectly plastic and kinematic hardening materials.

4. If the load intensity is higher than the ultimate strength capacity of the structure,
 P  Plim  , then the structure collapses instantaneously. This collapse is relevant
to static collapse (see Fig. 3.5). The ultimate strength capacity of the structure
made of unbounded kinematic hardening material is infinite.

30
 

max


min

Figure 3.5: Instantaneous plastic collapse.

3.2 Definition of load domain and fictitious elastic stress field

In shakedown analysis, the applied loads may vary independently, so it is necessary to
define the load domain L . This load domain contains all possible load histories.

3.2.1 Load domain

We study here the shakedown problem of a structure subjected to n time-dependent

(thermal and mechanical) loads Pk0 t  with time denoted by t , each of them can vary
independently within a given range

Pk0  t   I k0   Pk , Pk    k , k  Pk0 , k  1, n . (3.1)

These loads form a convex polyhedral domain L of n dimensions with m  2 n

vertices in the load space, as shown in Fig. 3.6a and 3.6b for two variable loads and three
variable loads respectively. These load domains can be represented in the following linear
form
n
Pt     k t Pk0 (3.2)
k 1

where

k  k  t   k , k  1, n . (3.3)

31
 L
P2 +
(a): Convex load domain with
L two variable loads
^ ^
P1 +2 P2
P (t)  2(t) 2
P  t    k  t  Pk0
k 1
L
-
-1 +1 P1
four special instants in L:
1(t)
Pˆ  Pˆ    P 0 ,   P 0 
1 1 1 1 2 2
^ ^
P4
-2 P3 ...
Pˆ 4  Pˆ 4  1 P10 , 2 P20 
  : upper bound load factor
  : lower bound load factor
Figure 3.6a: Load domain with two variable loads.

Moment (M)

^ (b): Convex load domain

P2 ^
P3
with three variable loads

+M 3
P  t    k  t  Pk0
^ ^
P4
P1
-p k 1
-N +N
Axial Force eight special instants in L:
Pˆ  Pˆ    p 0 ,   N 0 ,   M 0 
^
P6 (N)
+p ^
P7 1 1 p N M

 -
M ...
^
P5 ^
P8 Pˆ 8  Pˆ 8   p p 0 ,  N N 0 ,  M M 0 

Pressure (p)
Figure 3.6b: Load domain with three variable loads.

Intuitively, from these load domains, we can see that for limit analysis, the
monotone load is a singleton with only one point belonging to the load domain. In this way
limit analysis is a special case of shakedown analysis.

3.2.2 Fictitious elastic stress field

In the case of limit and shakedown analysis, it is useful to describe this load domain L in
the stress space. To this end, we use here the notion of a fictitious infinitely elastic structure
which has the same geometry as the actual one. The fictitious elastic response σ E  x, t  is
defined as the response which would appear in the fictitious structure if this structure was

32
subjected to the same loads as the actual one. Applying the superposition principle, this
fictitious elastic stress tensor field may be written in a form similar to (3.2)
n
σ E  x, t    k  t  σ Ek  x  . (3.4)
k 1

and the concept can be expressed in Fig. 3.7.

Actual applied loads P Fictitious elastic stress E

Actual applied loads

E
Fictitious elastic stress
P

A=B
A=P.u B= E
Displacement u Strain 
u 
Figure 3.7: An illustration of fictitious elastic stress field.

With the concept of the fictitious elastic stress field, we can write:
T  T  T

0  V
   A    0 V σ  x, t  : ε dVdt  0 .
T E p
f udV t udA dt (3.5)

  

3.3 Fundamental theorems of shakedown

In this section, we will present two basic theorems of shakedown, where the static or lower
bound theorem investigates the shakedown or safe region, while the kinematic or upper
bound theorem investigates the non-shakedown or failure region.

3.3.1 Static shakedown theorem (Melan)

Let L denote the load domain (3.2) and let σ  x, t  denote the fictitious elastic stress
E

response for all possible load combinations in L . If after some load cycles the
structure has already shaken down, everywhere in the structure plastic deformation
ceased to develop: ε p  0 .
If there exist a time-independent back-stress field π  x  satisfying

F  π  x    u  x    y  x  
2
(3.6)

there must exist a constant residual stress field ρ  x  such that the actual stress field σ  x,t 

σ  x, t   σ E  x, t   ρ  x   π  x  (3.7)

33
does not anywhere violate the yield criterion:

F  σ E  x, t   ρ  x   π  x    y2  x  (3.8)

The following theorem shows that this is the necessary and sufficient condition for a
structure to shake down (for perfect plasticity see Koiter 1960):

Theorem 3.1:

Shakedown occurs if there exists a permanent residual stress field ρ  x  , statically

admissible, such that:

F  σ E  x, t   ρ  x   π  x    y2  x  (3.9)

Shakedown will not occur if no ρ  x  exists such that

F  σ E  x, t   ρ  x   π  x    y2  x  . (3.10)

Based on the above static theorem, we can find a permanent statically admissible
residual stress field in order to obtain a maximum load domain that guarantees (3.10).
The obtained shakedown load multiplier  sd is generally a lower bound. The shakedown
problem can be seen as a maximization issue in nonlinear programming

 sd  max  (a)

s.t.: (3.11)
 F  σ E  x, t   ρ  x   π  x     y2  x  x  V , t (b)
  
 F  π  x     u  x    y  x  
 2
x  V (c)

div ρ  x   0 x  V (d)

ρ  x  n  0
 x  V (e)

For the perfectly plastic behaviour  u   y  , the back-stresses π are identical

zero due to the second inequality (3.11c). Melan’s original theorem for unbounded
kinematic hardening can be also deduced from the previous formulation if  u   . Then
the second inequality is not relevant anymore and the back-stresses π are free variables.

3.3.2 Kinematic shakedown theorem (Koiter)

Using plastic strain fields to formulate shakedown criterion, the kinematic shakedown
theorem is the counterpart of the static one. The theorem was given by Koiter (1960) and
some of its applications in analysis of incremental collapse were derived by Gokhfeld
(1980), Sawczuk (1969a, b). The same as proposed by Koiter for plastic strain field, we
introduce here an admissible cycle of the plastic strain field ε p , corresponding to a cycle
of the displacement field u . The plastic strain rate ε p may not necessarily be compatible

34
at each instant during the time cycle T but the plastic strain accumulation over the cycle
defined below
T
ε p   ε p dt (3.12)
0

must be compatible, such that


2 
ε p  1   u      u  T
  in V (a)
(3.13)

u  0 on Vu (b)

Theorem 3.2:

The structure will shake down if

T T

 σ  x, t  : ε
E p
dVdt   D p  ε p  dVdt . (3.14)
0V 0V

The structure will not shake down if

T T

 σ  x, t  : ε
E p
dVdt   D p  ε p  dVdt . (3.15)
0V 0V

For the bounded kinematic hardening model, (see section 1.5.4), we have:
T T

  D ε  dVdt     y ε p dVdt      y  ε p dV .
p p 2 2
3 3 u (3.16)
0V 0V V

In equation (3.16),  u   y  is the penalty parameter associated with the constraint

ε p  0 , where ε p is defined in Eq. (3.12).

The first term on the right hand side of Eq. (3.16) is exactly the total dissipation
energy of perfectly plastic structures, while the second term takes the hardening effect into
account. It is quite clear that:

 If  u   y , it is the perfectly plastic behavior

 If alternating plasticity occurs, which means ε p  0 , then the right hand side
of Eq. (3.16) retains only the first term. Again like the perfectly plastic
behavior, then

 alternating
blkh
  alternating
pp
(3.17)

The conclusion (3.17) has been stated in literature, e.g. Gokhfeld &
Cherniavsky (1980), Stein & Huang (1995), Pham D.C. (2007).
 For limit analysis, Eq. (3.16) can be rewritten as

35
  D  ε  dV    y ε p dV      y  ε p dV
p p 2 2
3 3 u
V V V
(3.18)
 2
3  u ε dV p

Then
 u pp
 lim
blkh
  lim . (3.19)
y

Based on the kinematic theorem, an upper bound of the shakedown limit load
multiplier  sd can be computed. The shakedown problem can be seen as a mathematical
minimization problem of nonlinear programming
T

  D  ε  dVdt
p p

 sd  min T 0V
(a)
  σ  x, t  : ε
E p
dVdt
0V

s.t.: (3.20)
 p T p
ε   ε dt (b)
 0
 p 1


ε  2   u      u  
T
 in V (c)

u  0 on Vu (d)

As presented before, there are two modes of shakedown: alternating plasticity or

low cycle fatigue (LCF), and incremental plasticity or ratchetting. Equation (3.20b) is the
definition of accumulated plastic strain. If ε p  0 then LCF occurs, else if ε p  0 , then
ratchetting occurs. We can calculate LCF and ratchetting limits separately or
simultaneously by separated shakedown method or unified shakedown method
respectively. The next section presents these methods.

3.4 Methods of calculating LCF limit and ratchetting limit

3.4.1 Separated shakedown method

This method defines the LCF limit and the ratchetting limit separately. As presented in
Chapter 1, Eqs (1.37) to (1.39), any plastic strain history ε p  x, t  can be separated into two
components

ε p  x, t   εup  x, t   εp  x, t  , (3.21)

where εup  x, t  involves ratchetting, and εp  x, t  involves LCF. There are three
possibilities (cf. Pham D. C., (2005))

36
 1) εup  x, t   0 and εp  x, t   0 : ratchetting appears
2) εup  x, t   0 and εp  x, t   0 : LCF appears
3) εup  x, t   0 and εp  x, t   0 : ratchetting and LCF appear simultaneously

These possible modes can be defined precisely in the following way

1. A perfect incremental collapse process (over a certain time interval 0, T  ) is a
process of plastic deformation εup  x, t  in which a kinematically admissible
plastic strain increment εup  x, t   εup  x, T   εup  x,0  is attained in a
proportional and monotonic way, namely

ε up  1
2   u     u   T
in V (a)
u  0 on Vu (b)
ε up    x, t  εup  x  (c) (3.22)
  x, t   0 (d)
  x, 0   0 (e)
  x, T   1 (g)

2. An alternating plasticity process is any process of plastic deformation εp  x, t 

within a certain time interval 0, T  such that the total increment of the plastic
strain εp  x  over this period is zero,
T
εp  x    εp  x, t  dt  0 . (3.23)
0

Based on the upper bound theorem, the criteria of safety with respect to incremental
collapse may be obtained by substituting the plastic strain history (3.22) into the
shakedown condition (3.14) with the dissipation (1.45), then it leads to (cf. Pham D. C.
(2005, 2007)).

Theorem 3.3a:

The incremental collapse will not happen if

T T

 σ  x, t  : εu dVdt   D εu  dVdt .

E p p p
(3.24)
0V 0V

The incremental collapse will happen if

37
 T T

 σ  x, t  : εu dVdt   D εu  dVdt

E p p p
(3.25)
0V 0V

with

D p ε p   σ : ε p   2 ε p : ε p (a)
 u u u 3 u u

 p T

εu  x    ε u  x, t  dt  0
p
(b)
 0 (3.26)
 p 1

εu  2   u     u  T
in V (c)

u  0 on Vu (d)

and the criteria of safety with respect to LCF may be obtained by substituting the plastic
strain history (3.23) into the shakedown condition (3.14) with the dissipation (1.42), then it
leads to (cf. Pham D.C. (2005, 2007))

Theorem 3.3b:

The alternating plasticity will not happen if

T T

 σ  x, t  : ε dVdt   D ε  dVdt

E p p p
(3.27)
0V 0V

The alternating plasticity will happen if

T T

 σ  x, t  : ε dVdt   D ε  dVdt

E p p p
(3.28)
0V 0V

with

D p ε p   σ : ε p   2 ε p : ε p (a)
  3  


y

 p T (3.29)
ε  x       x, t  dt  0
p
(b)

 0

3.4.1.1 Incremental collapse criterion

If the safety condition against any form of perfectly incremental collapse is considered, the
plastic strain field is assumed by (3.22). Substituting (3.22) into (3.24), one obtains
T T
Wex   σ E
 x, t    x, t  ε  x  dVdt  Win   D p  εup  dVdt .
p
u (3.30)
0V 0V

By taking into account the properties of the dissipation function and the plastic
strain history (3.22), we can write

38
 T T
Win   D p  εup  dVdt   dt  D p  εup  dV   D p  εup  dV . (3.31)
0V 0 V V

From the shakedown condition (3.30), the smallest upper bound of incremental
limit could be attained when the external power Wex assumes its maximum and the internal
 (x, t ) is selected in such a
dissipation Win takes its minimum. To this end, the function 
 (x, t )  0 only when the product σ E  x, t  εu  x  takes its maximum possible
way that 
value for a given load domain L . In this case, the external power Wex can be written as
T
Wex   σ E  x, t    x, t  : εup  x  dVdt   σ E  x  : εup  x  dV (3.32)
0V V

in which

σ E  x  : εup  x   max  σ E  x, t  : εup  x   . (3.33)

By this way, the safety condition against any form of perfectly incremental collapse
thus has the form

  σ E  x  : εup  x  dV   D p  εup dV . (3.34)

V V

If the load variation domain is prescribed by (3.2), (3.3) and (3.4), namely n
independently varying loads, the formulation (3.34) becomes

   k σ Ek  x  : εup  x dV   D p  εup dV

n
(3.35)
V k 1 V

in which

 k
 
if σ Ek  x  : εup  x   0
k    (3.36)
 k
 if σ Ek  x  : εup  x   0

From condition (2.34), the shakedown load multiplier against incremental collapse
 can be formulated as a non-linear minimization problem

sd

 Win 
 sd  min   (3.37)
 Wex 
or in normalized form

 sd  min Win

(3.38)
s.t.: Wex  1

39
3.4.1.2 Alternating plasticity criterion
If the safety condition against alternating plasticity is considered, the plastic strain field
must be satisfied (3.23). The shakedown condition (3.27) in this case has the form
T T
  σ E  x, t  : εp  x, t  dVdt   D p  εp  dVdt (3.39)
0V 0V

with
T
  1,  ε  x, t  dt  0 for all x  V .
p
(3.40)
0

Starting from the kinematic theorem and the last constraint in (3.40), the
optimization problem leading to the most stringent limit condition can be established at
each point x separately. The global safety factor against alternating plasticity limit will be
the minimum of local  (x ) defined as
T
1
 max  σ E  x, t  : εp  x, t  dt
 x ε
0

s.t.: (3.41)
T p p
  D  ε  dt  1
0
T
 ε p x, t dt  0
   
0
By solving this problem, the static shakedown condition against any form of
alternating plasticity can be obtained.
A given structure is safe against alternating plasticity if there exists a time-
independent stress field ρ which, if superimposed on the envelope of elastic stresses, does
not violate the yield condition

f  σ E  x, t   ρ  x   π  x    0 . (3.42)

It should be noted that the stress field ρ in (3.42) is an arbitrary time-independent

stress field and not necessarily self-equilibrated as it is required in Melan’s theorem (3.9)
and (3.10). We define a general stress response
n
σ*  x     k  k  σ Ek  x   π  x   (3.43)
k 1

where σ Ek  x  and π  x  are the elastic stress field and back-stress field respectively in the
reference structure when subjected to the k  th load and

k  k k  k

k  , k  . (3.44)
2 2

40
In view of (3.41), the plastic shakedown load multiplier (lower bound) may be calculated
as
1
  min (3.45)
x F σ  x   ρ  x   π  x 
*

where
f  F 1 . (3.46)

The sign of  k must be chosen so that the value of function F is maximal. By

considering the alternating characteristic of the stress corresponding to an alternating strain
rate, the optimal time-independent stress field ρ can be defined by

ρ   k  σ Ek  π  .
n
(3.47)
k 1

Then the plastic shakedown limit load multiplier can be finally represented as
1
  min . (3.48)
 
F   k  σ Ek  x   π  x   
x n

 k 1 

3.4.2 Unified shakedown method

This method defines the LCF limit and the ratchetting limit simultaneously. It is not
necessary to separate the plastic strain history ε p  x, t  into two components εup  x, t  and
εp  x, t  as presented in Eq. (3.17). Then theorems 3.3a and 3.3b for the upper bound
approach can be re-stated as, (see Nguyen Q-S. (2003)):

Theorem 3.4

The necessary and sufficient condition for shakedown to occur is that there exists a
plastic accumulation mechanism  p such that

T E T

  σ  x , t  : ε p
dVdt    3  y ε dVdt  
2 p 2
3  u   y  ε p dV
0 V 0V V
 T
(3.49)
ε p  ε p dt


0 compatible

In computations, in most cases it is not practicable to apply lower and upper bound
theorems in their above form to find directly the shakedown limit defined by the minimum
of the incremental plasticity limit and alternating plasticity limit. The difficulty here, for
the lower bound approach is the presence of the time-dependent stress field σ E  x, t  in
(3.9) and (3.10) and the theorems are stated only for time-independent residual stress ρ  x 

41
and time-independent back-stress π  x  , and for upper bound approach the difficulty is
appearance of the time integration in equation (3.20). These obstacles can be overcome
with the discretization of the load domains, which will be presented in next chapter.

42
Summary

S.1: Problem definitions

S.1.1: Elastic perfectly plastic structures

Upper bound solution

T
 sd  min   D p  ε p  dVdt ; (a)
εp
0V

s.t.:
 p T p
ε   ε dt (b)
 0

  
 tr ε p  0 (c)
 p 1


ε  2   u      u  
T
 in V (d)

u  0 on Vu (e)

T
 σ E  x, t  : ε p dVdt  1 (3.50)
 0 V
(g)

where

D p ε p   2
3  y ε p  2kv J 2  ε p  . (3.51)

Lower bound solution

 sd  max  (a)
 F  σ E  x, t   ρ  x     y2  x  x  V , t (b)
  
 (3.52)
s.t.: div ρ  x   0 x  V (d)

 ρ xn  0 x  V (e)


43
S.1.2: Unbounded kinematic hardening structures

Upper bound solution

T
 sd  min   D p  ε p  dVdt ; (a)
εp
0V

s.t.:
 p T p
ε   ε dt  0 (b)
 0

  
 tr ε  0
p
(c)
 p 1


ε  2   u      u  
T
 in V (d)

u  0 on Vu (e)

T
 σ E  x, t  : ε p dVdt  1 (3.53)
 0 V
(g)


where D p  ε p  is calculated in Eq. (3.51), as for the case of perfectly plastic material

D p ε p   2
3  y ε p  2kv J 2  ε p  .

Lower bound solution

 sd  max  (a)
s.t.: (3.54)
 F  σ E  x, t   ρ  x   π  x     y2  x  x  V , t (b)
  

div ρ  x   0 x  V (d)

 ρ xn  0 x  V (e)


44
S.1.3: Bounded kinematic hardening structures
Upper bound solution
T
  min   D p  ε p  dVdt

sd (a)
εp
0V

s.t.:
 p T p
ε   ε dt (b)
 0

  
 tr ε p  0 (c)
 p 1


ε  2   u      u  
T
 in V (d)

u  0 on Vu (e)

T
 σ E  x, t  : ε p dVdt  1 (3.55)
 0 V
(g)

where
T T

  D  ε  dVdt     y ε p dVdt      y  ε p dV
p p 2 2
3 3 u (3.56)
0V 0V V

Lower bound solution

 sd  max  (a)
s.t.: (3.57)
 F  σ E  x, t   ρ  x   π  x     y2  x  x  V , t (b)
  
 F  π  x      x     x  2 x  V (c)
   u y

div ρ  x   0 x  V (d)

ρ  x  n  0 x  V (e)

45
S.2 Perfectly plastic and kinematic hardening shakedown limit factors

If elastic perfectly plastic and kinematic hardening materials have the same yield stress  y ,
then
if  u  2 y then the bounded kinematic hardening becomes an unbounded model.
else

 The alternating limit factor is the same for elastic perfectly plastic and kinematic
hardening materials

 alternating
pp
  alternating
blkh
. (3.58)

 The shakedown limit factor of kinematic hardening material is bounded by

 u pp
 sdpp   sdblkh    2 el . (3.59)
 y sd

 The hardening curve does not affect the shakedown limit factor. Therefore only the
initial yield surface and the fixed bounding surface have to be considered.

S.3 Lower bound and upper bound approaches

Compared to each other, lower bound and upper bound have advantages and disadvantages,
where the advantages of the one approach are the disadvantages of the other one.

Advantages of lower bound approach

 Avoids the non-differentiability of the objective function, which must be
regularized via internal dissipation energy.
 There is no incompressibility constraint in the nonlinear programming problem.

Disadvantages of lower bound approach

 Suffers from nonlinear inequality constraints.
 Finite element methods based on stress method are more difficult.
 It is difficult to present alternating limit and ratchetting limit separately.

46
 Chapter 4

Numerical formulations

4.1 Introduction
In general, the problem of limit and shakedown analysis can be transformed into an issue
of mathematical nonlinear programming. In this part, we will transform the integrations
(3.55) and (3.56) into summation forms, where 1) by load domain discretization, the
integration over a certain time interval t  0, T  is transformed into a summation form for
k  1, m , where k is a loading vertex, m  2n , n is the number of variable loads, and 2) by
finite element discretization, the integration over the entire structure V is transformed into
the numerical integration for i  1, NG , where i is a Gaussian point and NG is the number
of Gaussian points in the structure. With these discretizations, the limit and shakedown
analysis is reduced to checking the restrictions only at all load vertices m and all Gaussian
points NG instead of checking for the entire load domain L and the whole structure V .
Discrete formulations of lower bound methods for perfectly plastic structures have
been presented by e.g. Hachemi, Hamadouche, Weichert (2003) and for unbounded and
bounded kinematic hardening structures are proposed by e.g. Hachemi and Weichert
(1996), Heitzer, Staat (2003). In Vu D.K. (2001) a primal dual approach to the
discretization of the upper and lower bound method has been presented for perfectly plastic
structures. In this chapter, we will mainly extend Vu’s discrete formulation of the upper
bound method to bounded and unbounded kinematic hardening structures.

4.2 Discretization of the load domain

Let L be a load domain containing any possible load which acts on the structure V . Any
load P  x,t   L could be specified by a variable t . For a variable cyclic loading the load
domain contains infinitely many loads (for a monotonic load in limit analysis it is
presented by one single load.) In shakedown analysis the sufficient conditions must be
verified for all the non-countable loads P  x,t  . This situation leads to difficulties due to
appearances of time integration in Eq. (3.55) and time-dependent stress field σ E  x, t  in
Eq. (3.56). Fortunately, these difficulties can be overcome with the help of the following
two convex-cycle theorems, introduced by König and Kleiber (1978).

Theorem 4.1:

“Shakedown will happen over a given load domain L if and only if it happens over the
convex hull of L ”.

47
Theorem 4.2:

“Shakedown will happen over any load path within a given load domain L if it happens
over a cyclic load path containing all vertices of L ”.

These theorems can be demonstrated via the load domain of two variable loads P1
and P2 as in Fig. 4.1a and Fig. 4.1b respectively.
P2 P2

^
P ^
P ^
P ^
P
2 1 2 1

P1 P1

^ ^
P ^
P ^
P
P 3 4 3 4

a) Shakedown happens over b) Shakedown happens over a cyclic load path

the convex hull of L containing all vertices of L

Figure 4.1: Critical cycles of load for shakedown analysis.

The theorems hold for convex load domains and convex yield surfaces, which
permits us to consider one cyclic load path instead of all loading history. This means that
the cyclic loading could be described by a finite number of load cases Pˆ k , k  1, m where
m  2n is the total number of vertices of L , n is the total number of varying loads Pk .
These n loads Pk vary in a given interval  Pkmin , Pkmax  , e.g. for a cyclic pressure load, the
pressure is bounded by minimum and maximum pressure (see König and Kleiber (1978)).
We restrict ourselves to problems where the traction boundary V remains constant (see
e.g. König (1987)) for moving loads on plates, and Kapoor, Johnson (1994) for structures
with contact). By defining the load cases Pˆ  x  ,..., Pˆ  x  via the load limits in each case,
1 m

any load P  x,t   L in a convex load space L is given as unique convex combination of
the Pˆ k  x  , as follows

48
 m
P  x, t      tk  Pˆ k  x ,
k 1

s.t.:
m
(4.1)
   t   1, and   t   0, k  1, m
k 1
k k

1 if t  tk
  tk   
0 if t  tk
n
where Pˆ k  x    k Pk0 . (4.2)
k 1


Let σ E x, Pˆ k  be the fictitious elastic stress in the body corresponding to the k-th load
vertex. From the principle of superposition for the elastic stresses we derive the convex
combination of the stresses σ E  x, t  by

 
m m
σ E  x, t     tk σ E x, Pˆ k    tk σ kE  x  . (4.3)
k 1 k 1

Then the external energy is computed as

T m

 σ  x, t  : ε
E p
dVdt    σ kE  x  : ε k dV . (4.4)
0V k 1 V

The strain accumulated over a complete load cycle is represented by

T m
ε p   ε p dt   ε k (4.5)
0 k 1

and the total internal dissipation energy Eqn. (3.56) in the structure is computed as
T

  D  ε  dVdt =  
m
p p
y
2
3 ε k : ε k dV
0V k 1 V
. (4.6)
2 k   k 
   u   y 
m m

3   ε  :   ε  dV
V  k 1   k 1 
The inequality (3.54b) becomes

  
F  σ E x, Pˆ k  ρ  x   π  x    y2  x 

x V , k  1, m , (4.7)

where ρ  x  is a time-independent residual stress and π  x  is a time-independent back-

stress.

49
4.3 FEM discretizations

From the consequence of load domain discretization, in the next section we only consider
all load vertices Pˆ k  x  of the load domain L instead of P  x,t  . And from now on, all
quantities are time-independent ones.

4.3.1 Discrete formulation of lower bound method

To calculate the fictitious elastic stresses σ kE in the reference body V E which is the same
as the original body V , subjected to the same loading Pˆ k in the load domain L , we use
the virtual work principle combined with the finite element discretization with shape
functions for the displacement fields. Then, the fictitious elastic stresses σ kE are in
equilibrium with body forces f k and surface tractions t k if the following equality holds

ε : σ kE dV    uTk  fk dV   u  t k dS
T T
k k (4.8)
V V Vu

for any virtual displacement  u and any virtual strains  ε satisfying the compatibility
condition.

 ε  12   u      u   
T
in V (a)
  (4.9)
u  0 on Vu (b)

The virtual displacement field  u within each element e is approximated by interpolation

of nodal values:
ND
 u   Nl ule (4.10)
l 1

where N l and  ule denote respectively the l -th shape function matrix and the vector of
virtual displacements of the l -th node of the finite element e which has ND nodes. The
virtual strain field is derived as follows

 ε    u    N ue   B  x  ue . (4.11)

Here B  x     N is the deformation matrix. Eq. (4.11) is called strain-displacement

relation or relation of lower bound and upper bound method.
By means of the finite element method, the structure V can be discretized into ne elements
Ve , the integration (4.8) is subdivided into the integrations over each element e .
ne ng ne

 wiBieT σikE   fke

e 1 i 1 e 1
(4.12)

where

50
 σikE  Eeεek  EeBieuek (4.13)

ng is the total number of Gaussian points in each element e , wi is the weight factor at
Gaussian point i , B ie is the deformation matrix at Gaussian point i of element e , σ ikE is
fictitious elastic stress vector at Gaussian point i corresponding to load vertex Pˆ k , f ke is
the vector of nodal loads of element e corresponding to vertex Pˆ k , Ee is the matrix of
elastic moduli of element e . Here B ie and Ee are the two constant matrices for any load
vertex.
Eq. (4.12) can be written in the following form
NG ne

 wiBTi σikE   fke

i 1 e 1
(4.14)

where NG  ng  ne is total number of Gaussian points in the structure.

Simultaneously, we can write the equilibrium equation (4.14) for residual stress as
NG

wB ρ
i 1
i
T
i i  BT ρ  0 . (4.15)

Equation (4.15) is numerical form replaced for derivative form of Eq. (3.57d & 3.57e),
where B and ρ are the global deformation matrix and the global residual stress vector,
respectively. They have the following form

B   w1B1 , w2B 2 ,..., wi Bi ,..., wNG B NG  (a)

(4.16)
ρT  ρ1T , ρ2T ,..., ρiT ,..., ρTNG  (b)

With the discretized residual stresses ρ calculated at Gaussian points only, the discretized
necessary limit and shakedown conditions are derived with the yield stress  y , i at every
Gaussian point i . Eq. (4.7) becomes

F  σikE  ρi  πi    y2,i . (4.17)

The restrictions of the optimization problems are checked only at the Gaussian points.
Finally, the shakedown limit load factor for bounded kinematic hardening
structures based on the lower bound approach, which is expressed in Eq. (3.57), becomes

51
  sd  max  (a)
s.t.: (4.18)

 F  σ ik  ρi  πi    y ,i i  1, NG, k  1, m
E 2
(b)


 F  π i    u , i   y , i 
2
i  1, NG (c)
 NG
 w BT ρ  BT ρ  0
 i i i i i  1, NG (d)
i 1

Here σ ikE is calculated in Eq. (4.14). The problem (4.18) has  NSC  NG  2   1 unknowns:
ρi , πi , and  , NG inequality constraints for yield condition, NG inequality constraints for
bounding condition, and NSC  NG equality constraints for equilibrium condition, where
NG is number of Gaussian points of structure, NSC is number of stress components of each
Gaussian point.

4.3.2 Discrete formulation of upper bound method

As in the lower bound discretization, the structure V is subdivided into ne elements Ve ,

the integration (4.6) of the internal dissipation energy can be written as

 m k  m k
  u   y 
m

  y
k 1 V
2
3 ε k : ε k dV + 2
3   ε  :   ε dV
 k 1   k 1 
V
(4.19)
2 k   k 
ε k : ε k dVe     u   y 
m ne ne m m
    y 2
3 3  ε    ε dVe
:
k 1 e 1 Ve e 1 Ve  k 1   k 1 

Using the Gauss-Legendre integration technique, the integration over the e-th element Ve
is approximated by numerical integration, Eq. (4.19) can be transformed into

 m k  m k
ε k : ε k dVe +    u   y 
ne m ne

  
e 1 k 1 Ve
y
2
3
e 1 Ve
2
3   ε  :   ε dVe
 k 1   k 1 
 m ik   m ik 
εik : εik   wi  u   y 
m ne ng ne ng
  wi y 2
3
2
3 ε  :ε  (4.20)
k 1 e 1 i 1 e 1 i 1  k 1   k 1 
 m ik   m ik 
εik : εik   wi  u   y 
NG m NG
  wi y 2
3
2
3 ε  :ε 
i 1 k 1 i 1  k 1   k 1 

Similarly, the integration (4.7) of the external energy can be written as

m m ne m ne ng

  σ kE  x  : ε k dV    σ kE  x  : ε k dVe   wiσikE : εik

k 1 V k 1 e 1 Ve k 1 e 1 i 1
(4.21)
m NG
  wi σ ikE : εik
k 1 i 1

52
in the above equations, ε ik is the strain tensor and σ ikE is the fictitious elastic stress tensor
at Gaussian point i and load vertex k .
In order to avoid the singularity of the dissipation function (4.20), we introduce
here a very small value  02 , where 0   02 1 , such that
T

  D  ε  dVdt   w 
NG m
p p
i y
2
3 εik : ε ik   02
0V i 1 k 1
(4.22)
 ik   ik 
  wi  u   y 
NG m m

 ε  :  ε   0 .
2 2
3
i 1  k 1   k 1 

From now onwards we work with matrix notations instead of tensor notations but
without changing the symbols. The functions of the strain tensor ε ik , Eq. (4.22) can be
written as,
T

  D  ε  dVdt   wi y
NG m
p p 2
3 εik Dεik   02
0V i 1 k 1
(4.23)
    2
  wi  u   y 
NG m m

  εik  D   εik    0
2
3
i 1  k 1   k 1 

where ε ik is the strain vector corresponding to load vertex Pˆ k , at Gaussian point i

T
εik  11ik  22ik  33ik 212ik 2 23ik 213ik 
T
(4.24)
 11ik  22ik  33ik  12ik  23ik  13ik 

and D is a diagonal square matrix. In a three-dimensional model, D has the form as

 1 1 1
D  Diag 1 1 1 . (4.25)
 2 2 2 

The external energy in Eq.(4.21) can be rewritten as function of vectors of strain ε ik as:
T m NG

  σ  x, t  : ε dVdt   wiεik σik .

E p T E
(4.26)
0V k 1 i 1

The strain accumulation, which has been defined in Eq. (3.55b), must be compatible at
each Gaussian point i . Eqs. (3.55 b, d, e) then becomes
m

ε
k 1
ik  Bi u . (4.27)

And the incompressibility condition (3.55c) can be expressed in vector form

DM εik  0 (4.28)

where DM is a square matrix. In a three-dimensional problem, it has the following form

53
 1 1 1 0 0 0
1 1 1 0 0 0 

1 1 1 0 0 0
DM    . (4.29)
0 0 0 0 0 0
0 0 0 0 0 0
 
0 0 0 0 0 0

Finally, with the load domain and finite element discretization, the optimization
problem for the shakedown limit load factor for bounded kinematic hardening structures
based on the upper bound approach presented in Eq. (3.55) becomes

 m NG
 blkh

 min  2
3  y  wi2εTik Dε ik  wi2 02

p
ε
k 1 i 1


 u y   i  ik  εik  wi2 02 
NG m m
 2
3    w 2
ε T
D (a)
i 1 k 1 k 1 
s.t.:
(4.30)
m
 ε ik  Bi u i  1, NG (b)
 k 1

DM εik  0 i  1, NG, k  1, m (c)
 m NG
 wi εTik σ ikE  1 (d)
 k 1 i 1

Once again, as mention with respect to Eq. (1.50) in chapter 1, the first term in the
objective function in Eq. (4.30) corresponds to the perfectly plastic material, and the
second term represents the hardening effect. If we omit this second term, then Eq. (4.30)
yields exactly the upper bound shakedown limit of a perfectly plastic structure.
It is quite clear that, if there is only one load vertex, k  1 , then it leads to the limit
analysis problem, consequently, limit analysis is a special case of shakedown analysis.
Then Eq. (4.30) is reduced to

 NG

 lim
 blkh
 min  2
3  u  wi2εTi Dεi  wi2 02  (a)
 
p
ε
i 1

s.t.: (4.31)
εi  Bi u i  1, NG (b)

DM εi  0 i  1, NG (c)

 NG
 wi εi σ i  1
T E
(d)
 i 1

54
Equation (4.31) leads to the limit load factor for perfect plasticity material, if in the Eq.
(4.31a),  u is replaced by  y , and we obtain

 u  pp
 lim
 blkh
  . (4.32)
 y lim

If the hardening is unbounded, that means  u is infinite, and shakedown limit bases only
on alternating mode, it depends only on yield stress  y , then Eq. (3.53) becomes

 m NG

 blkh

 min  2
3  y  wi2εTik Dε ik  wi2 02  (a)
 
p
ε
k 1 i 1

s.t.:
m
 εik  Bi u i  1, NG (b)
 k 1
m
 εik  0 i  1, NG (c)
 k 1
D ε  0 i  1, NG, k  1, m (d)
 M ik
 m NG
 wi εik σ ik  1
T E
(e)
 k 1 i 1 (4.33)

If the constraint (4.33c), which does ensure the LCF condition, is removed, then Eq. (4.33)
provides the solution for perfectly plastic structures. In the next chapter, shakedown
algorithms for unbounded linearly kinematic hardening structures will be presented. The
upper bound problem has the same number of linear equality constraints as the number of
inequality constraints (most of them nonlinear) of the lower bound problem.

55
 Chapter 5

Shakedown algorithms for elastic-plastic unbounded

linearly kinematic hardening structures

5.1 Introduction

In this chapter, we solve problem (4.33) which has been established in chapter 4. Solution
 sdublkh for alternating plastic will be found by an upper bound algorithm in section 5.2, and
by the primal-dual algorithm in section 5.3. Details of the dual relationship between upper
bound and lower bound were proved in Vu D.K. (2001) for perfect plasticity. Since  u is
infinite, the inequality constraint (3.57c) is always satisfied.

5.2 Upper bound algorithm

For sake of simplicity, let us define some new quantities as follows:

New strain vector

eik  wi D 2 εik
1
(5.1)

New fictitious elastic stress vector

1
t ik  D 2σikE (5.2)

New deformation matrix

ˆ  w D 12 B
B (5.3)
i i i

where D is a diagonal square matrix, presented in Eq. (4.25), and

 12  1 1 1 
D  Diag 1 1 1
  2 2 2 
D12  Diag 1 1 1 2 2 2 
 
Then the nonlinear programming problem (4.33) of the upper bound shakedown limit
becomes

57
 m NG
 ublkh

 min  2kv eTik eik   2 (a)
k 1 i 1

s.t.:
m
 eik  Bi u
ˆ i = 1,NG (b)
 k 1

m
 eik  0 i = 1,NG (c)
k 1

1 D e  0 i = 1, NG k = 1,m (d)
 3 M ik
 m NG
 eT t  1  0 (e) (5.4)
 k 1 i 1 ik ik

To solve the non-linear constrained optimization problem (5.4) we use the penalty
method to deal with the compatibility constraint (5.4b), the alternating constraint (5.4c)
and the incompressibility constraint (5.4d). The penalty function is written as


NG  m


T

FP   

i 1 
eTik eik   2 
c m
2 k 1

 eik  Bˆ iu  
eik  Bˆ iu 
c m T

2 k 1
e e
ik ik 
c m T

2 k 1

eik DM eik  (5.5a)

 k 1 
or

 m
 
T

  
ˆ u  c  eT  D  I  e  (5.5b)
NG
c m m
FP    e e 
T
ik ik
2
  eik  B
ˆu
i eik  B i ik M ik ik
i 1  k 1 2 k 1 2 k 1 

where:
c is penalty parameter such that c  1 . This parameter c may be dependent on
integration points or load vertices and c should be adjusted to fit different compatibility
criteria. However, at this stage, for the sake of simplicity, c is let to be constant
everywhere. Theoretically, when c goes to infinity we will recover related conditions.
 2 is a positive number and its value is reduced to zero as the procedure converges
to solution. When  2 goes to zero, the accurate solution may be expected provided that c
is sufficiently large. These parameters will be studied in chapter 7.
I ik is an identity matrix.

Then the modified upper bound formulation (5.4) becomes:

 ublkh

 2kv  min FP  (a)
s.t.: (5.6)
NG m

 e
i 1 k 1
T
t 1  0
ik ik (b)

using Lagrange multiplier method to solve problem (5.6), leads to the Lagrange function
FPL ,

58
  NG m T 
FPL  FP     eik t ik  1 . (5.7)
 i 1 k 1 
For finding a minimum of a function, its first derivative must be equal to zero. The
stationary condition called Karush-Kuhn-Tucker condition for the Lagrange function FPL
states that

 FPL  m
 
eik
 c   eik  B ˆ iu   c  DM  I ik  eik   t ik  0, i,k (a)
 eik eik eik    k 1 
T 2


 FPL NG
T 
m
ˆ iu   0
  c Bi   eik  B
ˆ (b) (5.8)
 u i 1  k 1 
 F  NG m 
 PL    eTik t ik  1  0 (c)
   i 1 k 1 

Eq. (5.8a) can be rewritten as:

 m ˆ iu 
eik  c e e     eik  B
T
ik ik
2

 k 1  (5.9)
 c eTik eik   2  DM  I ik  eik   eTik eik   2 t ik  0, i,k

For the sake of simplicity, let us define some functions as follow

 2 
m

      eik  Bˆ iu   c eik eik    D M  I ik  eik
T T 2
 ik
f e ik c e e
ik ik
  k 1 
   eTik eik   2 t ik  0 i,k (a)

 T 
m
 (5.10)
hi  Bi   eik  Bˆ iu   0
ˆ (b)
  k 1 
  NG m 
nik    eTik t ik  1  0 (c)
  i 1 k 1 
From (5.10b) it follows
T  ˆ 
NG
NG m


i 1
h i  
i 1
ˆ
B i   eik  Bi u   0
 k 1 
so that using the Newton-Raphson method to solve the system (5.10) we obtain

 2 
m

     deik  Bˆ i du   d eik eik   t ik  fik
T T 2
 ik ik
H d e c e e
ik ik (a)
  k 1 
 NG T  m
ˆ i   deik  Bˆ i du    hi
NG

 B (b) (5.11)
 i 1  k 1  i 1
 NG m T NG m
  t ik deik  1   t ik eik
T
(c)
 i 1 k 1 i 1 k 1

where

59
 Hik  I ik  c eTik eik   2  DM  I ik  
 
1   m   (5.12)
 c  eik  Bˆ iu   c  DM  I ik  eik   t ik  eTik
2  
eik eik     k 1
T
 

The algorithm reaches convergence more stably if we keep only the first term of H ik in Eq.
(5.12), as

Hik  Iik  c eTik eik   2  DM  Iik  . (5.13)

 
this approximation will be tested in chapter 7.

From (5.11a), we can calculate the increment deik of the strain vector

 m ˆ i du   d eTik eik   2 Hik1t ik  H ik1fik .

deik  c eTik eik   2 Hik1   deik  B (5.14)
 k 1 

Writing (5.14) for k  1, m and then sum them up, we obtain

m m m m

 deik  cK i1  eTik eik   2 Hik1Bˆ i du  d K i1  eTikeik   2 Hik1t ik  K i1  Hik1fik (5.15)
k 1 k 1 k 1 k 1

where
m
K i  I i  c eTik eik   2 Hik1 . (5.16)
k 1

Computation of du

To calculate the increment du of the displacement vector, we substitute (5.15) into (5.11b)
and obtain

T   NG
NG NG m m


i 1
Bi Ci Bi du   Bi  d K i  eik eik   Hik t ik  K i  Hik fik    hi
ˆ T ˆ
i 1
ˆ

1

k 1
T 2 1 1

k 1
1

 i 1
(5.17)

where
 m

Ci   Ii  cK i1  eTik eik   2 Hik1  . (5.18)
 k 1 
From (5.16) we have
m
c eTik eik   2 Hik1  K i  Ii . (5.19)
k 1

Substituting (5.19) into (5.18) leads to

60
 Ci  K i1 . (5.20)

Then (5.17) can be rewritten as

Sdu  d f1  f2 (5.21)

where

 NG

S   Bˆ Ti K i1Bˆ i (a)
 i 1

 NG m

1
f    B ˆ T
i K i  eik eik   H ik t ik
1 T 2 1
(b) (5.22)
 i 1 k 1
 NG m NG
f 2   B ˆ Ti K i1  H ik1fik   hi (c)
 i 1 k 1 i 1

NG
Solving (5.21) with unknown is du , and substituting fik from (5.10a), h
i 1
i from (5.10b)

into (5.22c), after some manipulations, we obtain

du    d  S1f1  u . (5.23)

Computation of deik
m
Substituting fik from (5.10a),  de
k 1
ik from (5.15) into (5.14), after some manipulations,

we obtain

deik   deik 1    d  deik 2 (5.24)

where

 deik 1 =  eik (a)

 (5.25)
 deik 2  c eik eik   Hik K i Bˆ iS f1  eik eik   H ik K i t ik
T 2 1 1 1 T 2 1 1
(b)

Computation of   d 

From (5.11c) we have

NG m

 t  e
i 1 k 1
T
ik ik  deik   1 . (5.26)

Substituting (5.24) into (5.26), then solving for unknown   d  , we obtain

1   tTik  eik   deik 1 

NG m

  d  i 1 k 1
NG m
. (5.27)
 t  de 
i 1 k 1
T
ik ik 2

If the existing value of eik is already normalized, i.e.

61
 NG m

 t
i 1 k 1
T
e 1
ik ik

the normalized form of (5.27) is

NG m

 t  de  T
ik ik 1
  d   i 1 k 1
NG m
. (5.28)
 t  de 
i 1 k 1
T
ik ik 2

Algorithm

1. Initialize displacement and strain vectors: u 0 and e 0 such that the normalized
condition (5.4e) is satisfied:

NG m

 t
i 1 k 1
e  1.
T 0
ik ik (5.29)

Normally the fictitious elastic solution must be computed first in order to define the load
domain L in terms of the fictitious elastic generalized stress σ ikE . Hence u 0 and e 0 may
assume fictitious values (after being normalized) for their initialization. Set up initial
values for the penalty parameter c and for  . Set up convergence criteria and maximum
number of iterations.

2. Calculate S , f1 , f 2 from (5.22) at the current values of u and e .

3. Calculate  deik 1 ,  deik 2 from (5.25),   d  from (5.28), then calculate deik
from (5.24), and du from (5.23).
4. Perform a line search to find k such that:

k  min FP  u  du, e  de  . (5.30)

Update displacement, strain and  as:

u  u  k du (a)
eik  eik  k deik (b) (5.31)
    d (c)
5. Check convergence criteria: if they are all satisfied, then go to step 6, otherwise go
to steps 2.
6. Stop.

62
5.3 Dual algorithm

The primal issue is chosen from upper bound approach, then penalty and Lagrange
functions as well as Karush-Kuhn-Tucker conditions are expressed respectively in Eqs.
(5.5b), (5.7) and (5.8). By the means of dual algorithm, we can find the upper bound and
lower bound of shakedown limit simultaneously, and when these bounds are closed
together, we gain the result confidently. Besides, the dual algorithm makes the process to
converge faster than the upper bound or the lower bound individually.

If we define:
ˆ e ,
γ ik  cD ˆ  D  I 
D (a)
M ik M M ik

 m  (5.32)
βi  c   eik  Bˆ iu  (b)
 k 1 
Then (5.8a) and (5.8b) are rewritten as follows

e  eT e   2  β  γ   t   0 i,k (a)
 ik ik ik i ik ik

 NG (5.33)
 Bˆ i βi  0
T
(b)
 i 1
For sake of simplicity, we define some functions as:

f  e  eT e   2  γ  β   t   0 i,k (a)
 ik ik ik ik ik i ik

 NG

m i   B ˆ Ti βi  0 (b)
 i 1

  NG m T 
 ik   eik t ik  1  0
n  (c) (5.34)
  i 1 k 1 
g  γ  cD ˆ M eik  0 i,k (d)
 ik ik

  m
 i
h  β i  c   eik  B ˆ iu   0 i (e)
  k 1 

Using the Newton Raphson method to solve system (5.34) we obtain

M de  eT e   2  dγ  dβ  d t   f (a)
 ik ik ik ik ik i ik ik

 NG NG

 Bˆ i dβi   Bˆ i βi
T T
(b)
 i 1 i 1


 t ik  eik  deik   1
NG m
T
(c) (5.35)
 i 1 k 1
dγ  cD ˆ M deik  g ik (d)
 ik
  m 
 i
d β  c   deik  Bˆ i du   hi (e)
  k 1 

63
where
 eTik 
Mik  I ik   γ ik  βi   t ik   . (5.36)
 eTik eik   2 

Substituting (5.35d) into (5.35a) we have:

ˆ ik deik  eTik eik   2  gik  dβi  d t ik   fik

M (5.37)

where
ˆ ik  Mik  eTik eik   2 cD
M ˆM. (5.38)
 

Now we can compute the incremental vector deik of the strain from (5.37)

ˆ ik1 eTik eik   2  gik  dβi  d t ik   M

deik  M ˆ ik1fik (5.39)

Writing (5.39) for k  1, m and then sum of them, we obtain

m m m

 de
k 1
ik   M
ˆ ik1 eTik eik   2  g ik  dβi  d t ik    M
k 1
ˆ ik1fik .
k 1
(5.40)

Substituting fik from (5.34a), g ik from (5.34d), dβi from (5.35e) into (5.40), after some
manipulations, we obtain

Ei  M ik eik eik   2  cD
ˆ M eik 
m
1 1 m ˆ 1 T

k 1
de ik  
c k 1
m
 Ei1  M
ˆ ik1 eTik eik   2 Bˆ i du
k 1

ˆ ik1 eTik eik   2   eik  Bˆ iu 

m m
 Ei1  M (5.41)
k 1  k 1 
m
1
 Ei1  M ˆ ik1eik
c k 1
m
1
 Ei1  M ˆ ik1 eTik eik   2   d  t ik
c k 1

where

I m
ˆ ik1 eTik eik   2  .
Ei   i   M  (5.42)
 c k 1 
Computation of du

To calculate the increment du of the displacement vector, we substitute (5.34e) into

(5.35e) then (5.35e) and (5.41) into (5.35b). After some manipulations, and noting that

64
 m m
Ii ˆ ik1 eTik eik   2  Ei  I i
Ei   Mˆ ik1 eTik eik   2   M
c k 1 k 1 c
then (5.43)
 m
ˆ 1 eTik eik   2   1 Ei1
i  M ik
1
I
 i  E  c
 k 1 
we obtain
NG NG

 Bˆ E
i 1
T
i iBˆ i du   Bˆ Ti Ei1Bˆ iu
1

i 1

  e
NG m
+  Bˆ Ti Ei1  M
ˆ ik1 M
ˆ ik  I ik  eTik eik   2 cD
ˆM ik (5.44)
i 1 k 1
NG m
  Bˆ Ti Ei1  M
ˆ ik1 eTik eik   2   d  t ik
i 1 k 1

From (5.36) and (5.38) we can express

ˆ ik  I ik   γ ik  βi   t ik  eTik ˆM.
M  c eTik eik   2 D (5.45)
e e 
T
ik ik
2

Then (5.44) could be rewritten as

NG NG

 Bˆ Ti Ei1Bˆ i du   Bˆ Ti Ei1Bˆ iu
i 1 i 1
NG m
eTik eik
 B
ˆ Ti Ei1  M
ˆ ik1  γ ik  βi   t ik  (5.46)
i 1 k 1 eTik eik   2
NG m
 B
ˆ Ti Ei1  M
ˆ ik1 eTik eik   2   d  t ik
i 1 k 1

Solve (5.46) with unknown is du , we obtain

du  du1    d  du2 (5.47)

where

du1  u  Sˆ 1fˆ1
 (a)
 ; (5.48)
du 2  Sˆ fˆ2
1
 (b)

 NG
 Sˆ   Bˆ Ti Ei1Bˆ i (a)
 i 1
 NG m
eTik eik
 f1   Bi Ei  M ik  γ ik  βi   t ik  T
ˆ ˆ ˆ
T 1 1
(b) (5.49)
 i 1 k 1 eik eik   2
 NG m
 fˆ2   Bˆ Ti Ei1  M ˆ ik1 eTik eik   2 t ik (c)
 i 1 k 1

65
Computation of dβi for the preparation of computing deik
m
To substitute h i from (5.31),  de
k 1
ik from (5.41), and du from (5.47) into (5.35e), after

some manipulations, we obtain

dβi   dβi 1    d  dβi 2 (5.50)

where

 dβi 1  Ei1  Mˆ ik1  Ii  c 

m
eTik eik   2 DM eik
k 1

  m 
 Ei1 Bˆ i du1    eik  Bˆ iu    βi (a) (5.51)
  k 1 
m
 dβi 2  Ei1Bˆ i du 2  Ei1  Mˆ ik1 eTik eik   2 t ik (b)
k 1

Computation of deik

To substitute fik from (5.34a), g ik from (5.34d), and dβi from (5.50) into (5.39), after
some manipulations, we obtain

deik   deik 1    d  deik 2 (5.52)

where
 de   M
 ik 1

ˆ ik1 eTik eik   2  cD

ˆ M eik   dβi    M
1 
ˆ ik1 eik  eTik eik   2 βi  (a)
(5.53)
 deik   Mˆ ik1 eTik eik   2  dβi   M ˆ ik1 eTik eik   2 t ik (b)
 2 2

Computation of   d 
To substitute (5.52) into (5.35c), then solving for unknown is   d  , we obtain

1   tTik  eik   deik 1 

NG m

  d  i 1 k 1
NG m
. (5.54)
 t  de 
i 1 k 1
T
ik ik 2

If the existing value of eik is already normalized, that is

NG m

 t
i 1 k 1
T
ik ike 1

66
then the normalized form of (5.54) is
NG m

 t  de  T
ik ik 1
  d   i 1 k 1
NG m
. (5.55)
 t  de 
i 1 k 1
T
ik ik 2

Algorithm

1. Initialize strain vectors e 0 such that the normalization condition (5.4e) is satisfied:

NG m

 t
i 1 k 1
e  1.
T 0
ik ik

Set the stress vectors equal to zero vectors


 γ ik  0
0

 0 i  1, NG; k  1, m
βi  0


Set up initial values for the penalty parameter c and for  . Set up convergence
criteria, maximum number of iterations.
2. Calculate  deik 1 ,  deik 2 at current value of e , using equation (5.53).
3. From (5.52) calculate   d  . From (5.52) calculate deik
4. Perform a line search to find ˆk such that:

ˆk  min FP  e   de  (5.56)

where Fp is the penalty function (5.5a)

Update strain eik and  as

eik  eik  ˆk deik (a)

(5.57)
    d (b)

5. Calculate the vector of incremental stress dγ ik , from (5.35d)

ˆ M deik  gik
dγ ik  cD (5.58)

6. Perform a line search to find ˆs such that:

ˆs  max 
s.t.: (5.59)
 γ ik  t ik     dγ ik  t ik d  1

67
 Update vector stress γ ik and scalar  with a chosen parameter  : 0    1

 γ ik  γ ik  ˆs dγ ik (a)
 (5.60)
    ˆs d (b)

7. Check convergence criteria: if they are all satisfied then go to step 7, otherwise go
to steps 2.
8. Stop.

The algorithm is also presented in Fig. 5.1

NG m

 t
i 1 k 1
e 1
T 0
ik ik u 0 , e0

 γ ik0  0
 ˆ , fˆ fˆ
 0 i  1, NG; k  1, m S 1, 2  deik 1 ,  deik 2
βi  0


  d  du, deik

ˆk  FP  u   du, eik   deik   min ˆ M deik  gik

dγ ik  cD
 m ˆ i du   hi
dβi  c   deik  B
 k 1 

ˆs  max 
s.t.:  γ ik  βi  t ik     dγ ik  dβi  t ik d   1

 γ ik  γ ik  ˆs dγ ik
u  u  ˆk du 
βi  βi  ˆs dβi
e  e  ˆ de  ˆ
    s d
ik ik k ik

convergence
criteria
satisfied ?

Stop

Figure 5.1: Flow chart for the primal-dual algorithm for shakedown analysis of unbounded
linearly kinematic hardening structures.

68
 Chapter 6

Shakedown algorithm for elastic-plastic bounded

linearly kinematic hardening structures

6.1 Introduction

In this chapter, we solve problem (4.30) which has been established in chapter 4. Once
again, the penalty function method combined with the Lagrange multiplier method is used
to solve the constraint nonlinear optimization problem. Solution  sdblkh will be found by
upper the bound algorithm in section 6.2. This algorithm gives solutions for unbounded
kinematic hardening as presented in chapter 5 if  u  2 y , and it gives solutions for
structures made of perfectly plastic material if  u   y .

6.2 Upper bound algorithm

For sake of simplicity, we recall the new strain vector eik , the new fictitious elastic stress
vector t ik , and the new deformation matrix Bˆ i as presented in Eqs. (5.1), (5.2) and (5.3) in
chapter 5. Then the nonlinear programming problem (4.30) of the shakedown limit load
factor becomes

 m NG
 blkh

 2
3
 y min  eTik eik   2
eik
 k 1 i 1
  y  NG m m 
2
   ik  ik  
u T
e e  (a)
y i 1 k 1 k 1 
s.t.:
m (6.1)
 eik  Biu
ˆ i  1, NG (b)
 k 1
1
 DM eik  0 i  1, NG k  1, m (c)
3
 m NG T
 eik t ik  1 (d)
 k 1 i 1
By writing the penalty function for the compatibility condition (6.1b) and the
incompressibility condition (6.1c), we obtain

69
  m
NG m m
FP    eTik eik   2  a e e T
ik ik 2
i 1  k 1 k 1 k 1

 
T
c m T c m   m
  eik DM eik    eik  Biu    eik  Bˆ iu  
ˆ
2 k 1 2  k 1   k 1   (6.2)
s.t.:
m NG

 e
k 1 i 1
T
t 1
ik ik

where
u  y
a . (6.3)
y

Following (6.2) the modified kinematic formulation (6.1) becomes

   2kv  min FP  (a)

s.t.: (6.4)
m NG

 e
k 1 i 1
T
t 1
ik ik (b)

where kv   y 3 is the yield stress in pure shear according to the von Mises criterion.
The corresponding Lagrange function of (6.4) is
NG 
 m m m
FPL   a e e T
ik ik   2   eTik eik   2
i 1 
 k 1 k 1 k 1
(6.5)

T
c m T c m   m   m NG 
  eik DM eik    eik  Biu    eik  Bˆ iu       eTik t ik  1
ˆ
2 k 1 2  k 1   k 1   k 1 i 1 

Writing the Karush-Kuhn-Tucker conditions for the Lagrange function FPL we obtain

 m

 F e ik
eik
 PL a k 1

 eik m m
eTik eik   2
  eTik  eik   2
 k 1 k 1

  m 
  cDM eik  c   eik  Bˆ iu    t ik  0 i,k (a) (6.6)
  k 1 
 FPL
ˆ Ti   eik  B ˆ iu   0
NG m
  c  B (b)
 u i 1  k 1 
 F  NG m

 PL    eTik t ik  1  0 (c)
   i 1 k 1 

70
Eq. (6.6a) can be rewritten as

 m  m m
a   eik  eTik eik   2  eik  eTik  eik   2
 k 1  k 1 k 1

m m
  eTik  eik   2 eTik eik   2  cDM eik 
k 1 k 1
(6.7)
m m
 m 
 e e
k 1
T
ik
k 1
ik  2
e e   c   eik  Bˆ iu 
T
ik ik
2

 k 1 
m m
  eTik  eik   2 eTik eik   2  t ik   0
k 1 k 1
i,k

Let us define some functions for shorter notations

  m  T m m
 ik    ik  ik ik    ik  ik  e ik  
2 T 2
f a e e e e e
  k 1  k 1 k 1
 m m
  cDM eik  eTik  eik   2 eTik eik   2
 k 1 k 1

  m  m T m


 c  
 k 1
e ik  ˆ
B i   eik  e ik  
u
 k 1 k 1
2
eTik eik   2
 (6.8)
 m m

   t ik  eTik  eik   2 eTik eik   2  0 i,k (a)

 k  1 k  1


ˆ Ti   eik  B ˆ iu   0   h i   Bˆ Ti   eik  Bˆ iu   0
m NG NG m
hi  B (b)
  k 1  i 1 i 1  k 1 
  NG m

nik    eTik t ik  1  0 (c)

  i 1 k 1 

Using the Newton-Raphson method to solve the system (6.8) we have

 m

ik  deik  cbik B i du  bik t ik d  fik

M d e  N ˆ (a)
 ik ik
 k 1

 NG T  m  NG
T 
m

 Bi   deik  Bi du    Bi   eik  Bˆ iu 
ˆ ˆ ˆ (b) (6.9)
 i 1  k 1  i 1  k 1 
 NG m T NG m
 t ik deik  1   eik t ik
T
(c)
 i 1 k 1 i 1 k 1

71
where

 m m 
M ik  
 e e T
ik ik   2 I ik  cbik DM 

 k 1 k 1 
m m   m  

k 1
eT
ik 
k 1
e ik   2


a   eik  eTik
 k 1  T 

   aik eik  (6.10)
eTik eik   2 
m m

 
k 1
eik  eik  
T

k 1
2



Nik  a eTik eik   2  cbik I ik 
  m 
T

eTik eik   2
 e ik  
  k 1 
e ik   m
T
 
 (6.11)
  T  aik   eik  
2  eik eik    k 1  
m m 2

 eik  eik  
T
 
k 1 k 1  

 m ˆ iu   cDM eik   t ik
aik  c   eik  B  (a)
 k 1 
(6.12)
m m
bik  e e
k 1
T
ik
k 1
ik  2
e e 
T
ik ik
2
(b)

The algorithm reaches convergence more stably if we keep only the first term of M ik as
well as of Nik in the equations (6.10) and (6.11), i.e.

 m m 
M ik  
  eT
ik  eik   I ik  cbik D M 
2
 (6.13)
 k 1 k 1 


Nik  a eTik eik   2  cbik Iik  (6.14)

The tests in chapter 7 demonstrate that the algorithm converges to the correct solutions
with this approximation.

From (6.9a), we can calculate the incremental strain vector deik

m
deik  Mik1Nik  deik  cbik Mik1B
ˆ du  b M 1t d  M 1f .
i ik ik ik ik ik
(6.15)
k 1

Writing (6.15) for k  1, m and then sum them up, we have

m m m m

 de
k 1
ik  cQi1  bik Mik1B
k 1
ˆ du  d Q1  b M 1t  Q1  M 1f
i i ik ik ik i ik ik
k 1 k 1
(6.16)

72
where
m
Qi  I i   Mik1Nik . (6.17)
k 1

Substituting fik from (6.8a) into (6.16), after some manipulations, we have
m m

 deik  cQi1  bik Mik1Bˆ i du

k 1 k 1

m  m m m 
 Qi1  M ik1  a  eik eTik eik   2  eik e e T
2 
 k 1 ik ik

k 1  k 1 k 1  (6.18)
 m
m
 m
 Q  M cbik   eik  Bˆ iu   Qi1  M ik1cbik D M eik
1
i
1
ik
k 1  k 1  k 1
m
   d  Qi1  M ik1t ik bik
k 1

Computation of du

m
To calculate the incremental displacement vector du , we substitute  dek 1
ik from (6.18)

into (6.9b), after some manipulations, we obtain

NG NG NG m

 Bˆ Ti EiBˆ i du   Bˆ Ti EiBˆ iu   Bˆ Ti Ei  eik

i 1 i 1 i 1 k 1

NG m  m m m 
  Bˆ Ti Qi1  M ik1  a  eik eTik eik   2  eik  e ik  e ik   
T 2
 k 1 
i 1 k 1  k 1 k 1  (6.19)
NG m
  Bˆ Ti Qi1  M ik1cbik DM eik
i 1 k 1
NG m
   d   Bˆ Ti Qi1  M ik1t ik bik
i 1 k 1

where
 m

Ei   I i  cQi1  bik Mik1  . (6.20)
 k 1 
Then

du  u  S1f1    d  S1f2 (6.21)

73
where

 NG

S   Bˆ Ti Ei Bˆ i (a)
 i 1

 NG m

 1  i i  eik
 ˆ T
f B E
 i 1 k 1
 NG m  m m m  (6.22)
  Bˆ T Q 1 M 1  a e eT e   2  e
  i i  ik
  ik ik ik ik  eT
ik  ik e   2
 cbik M ik 
D e

(b)

i 1 k 1  k 1 k 1 k 1 
 NG m

2

f   i 1
Bˆ T
i Q i  M ik t ik bik
1

k 1
1
(c)

Computation of deik

m
Substituting  de
k 1
ik from (6.18) and fik from (6.8a) into (6.15), after some manipulations,

we obtain

deik   deik 1    d  deik 2 (6.23)

where

  1 m
1 
 ik 1  ik ik i  M ik  M ik  β1
  1 1
de  M N Q (a)
  k 1 
 (6.24)
 de    M 1N Q 1 M 1  M 1  β
m

 ik 2  ik ik i  ik  2
ik (b)
k 1 

 m m m
 1β  a  e ik e T
e
ik ik   2
 e ik  ik  e ik  
e T 2

 k 1 k 1 k 1
  m

  cDM eik bik  c   eik  Bˆ iu  bik  cbik Bˆ i du1 (a) (6.25)
  k 1 

 2  ik 2
β  t  cBˆ du (b)
i



74
The algorithm is described in the following flow chart.

NG m

 t
i 1 k 1
e 1
T 0
ik ik
u 0 , e0

S, f1 , f 2  deik 1 ,  deik 2

  d  du, deik

k  FP  u  du, e  de   min

u  u  k du
eik  eik  k deik
    d

convergence
criteria
satisfied?

Stop

Figure 6.1: Flow chart for the upper bound algorithm for shakedown analysis of bounded
linearly kinematic hardening structures.

75
 Chapter 7
Validations and parametric studies

In this chapter, firstly we validate our theory and algorithms by analysing many examples,
which are available in literature. For limit analysis, the results are verified by
 lim
blkh
  u  y  lim
pp
, where  u   y . For shakedown analysis, the results are verified by

 sdblkh   sdpp , for perfectly plastic structures


 pp  u pp
 sd   sd    sd  2 el ,
blkh
for bounded and unbounded kinematic hardening
 y

Secondly, we do the parametric study, where the penalty parameter c and the value
of  are varied to see their influence on the solution, and to confirm their interval value for
stable results.
Test cases cover 2D and 3D finite elements, thermal load to mechanical loads, and
from simple to complicated structures and loadings.

7.1 Structure 1: Thin plate under tension and temperature

Problem definitions

This problem has investigated in Heitzer, Staat (2003), using a numerical approach
based on the lower bound shakedown theorem.
Geometry: The square plate has the dimensions: B  L  400  400 mm2 , supported at
the top and bottom sides, see Fig. 7.1.1.
Material: Young’s modulus: E  2.1105 N/mm2 , yield stress:  y  160 N/mm2 ,
hardening effect (ultimate strength):  u  y  1.5 for bounded hardening,
and  u  y  2 for unbounded hardening, Poisson’s ratio:   0.3 ,
coefficient of thermal expansion: t  105 K 1 .
Loads: uniform distributed tension p  N/mm  applied on the lateral sides, and
uniform distributed temperature T  °C  on the plate, these loads vary in the
domain: p  0, p0  , T  0, T0  , where p0 and T0 is the elastic analytical
solution for pure tension and pure temperature, respectively, see details in
Heitzer, Staat (2003)
1 1
p0   y and T0  y . (7.1.1)
1    2 E t

77
FEM model: a quarter of the structure is modelled with sixteen 8-node-2D elements with
2  2 Gaussian points . The structure is considered as a plain stress
problem.

Figure 7.1.1: Thin plate under tension and temperature.

Numerical results

Tests are run with c  1010 and   1024 . The results are presented in Fig. 7.1.2, together
with the results of Heitzer, Staat (2003).

Figure 7.1.2: Interaction diagram for elastic and shakedown limits of thin plate under
tension and temperature, normalized by p0 and T0 in formula (7.1.1).

Parametric studies

To study the influence of the penalty parameter c , we keep the value   1024 fixed, and
let c vary from 104 to 1017 and do shakedown analysis for the load combination
 p0 ,0.3T0  . For  u  y  1.0 and  u  y  1.5 the algorithm gives stable solutions if

78
c  106 ,1014  . For  u  y  2.0 the algorithm gives stable solutions if c  104 ,1015  ,
Fig. 7.1.3.

Figure 7.1.3: Common logarithm lg c for stable solution.

To study the influence of  , we keep the value c  1010 fixed, and let  vary from
1040 to 109 and do shakedown analyses for the load combination  p0 ,0.3T0  . The
algorithm gives stable solutions for all considered material models if   1014 , Fig. 7.1.4.

Figure 7.1.4: Common logarithm lg  for stable solution.

79
Subconclusions 7.1

 For  u  y  1 , the present results coincide with the results for perfectly plastic
material as implemented in Vu D.K. (2001).
 The unbounded kinematic hardening curve is the alternating plasticity limit.
Most of the shakedown limits for this structure with and without hardening
effect are ratchetting limits. The hardening effect is quite obvious.
 The present results compare well to those of Heitzer, Staat (2003), which are
based on the lower bound approach.
 The algorithm gives the steady results for penalty parameters c  106 ,1014  and
  1014 .

7.2 Structure 2: Square plate with a central circular hole

Problem definitions

This problem has been treated very frequently in literature. For limit analysis, Gaydon
(1954), Gaydon and McCrum (1954) presented the analytical solution of the plastic
collapse limit for both lower and upper bound, for the Tresca yield criterion and potential
function in Gaydon (1954), and the von Mises yield criterion and potential function in
Gaydon, McCrum (1954). For limit and shakedown analysis using the von Mises yield
criterion, Staat, Heitzer (2003) presented the FEM based lower bound solutions. Tran T.N.
(2008) presented the FEM based upper bound solutions using the exact Ilyushin yield
surface and shell elements. Vu D.K. (2001) presented the FEM based primal-dual
solutions. etc.. Further perfectly plastic analyses of the problem have been compared in
Schwabe (2000).
As stated in Tran T.N. (2008), Vu D.K. (2001) this structure fails with alternating
plasticity. Therefore we can state that there is no benefit of hardening (refer to equation
(3.17) in chapter 3). We do limit and shakedown analysis for this structure for two aims:
one is the validation for  u  y  1.0 , and the other one is to confirm the benefit of the
hardening effect.

Geometry: By the symmetry of both geometry and applied loads, a quarter of the
structure is described in Fig. 7.2.1. The dimension of the central circular hole
varies from R L  0.1 to R L  0.9 , L  1000 mm .
Material: Young’s modulus: E  2.1105 N/mm2 , yield stress:  y  116.2 N/mm2 ,
hardening effect (ultimate strength):  u  y  1.4 for bounded hardening,
and  u  y  2 for unbounded hardening, Poisson’s ratio:   0.3 .

80
Loads: uniformly distributed loads: vertical tension load p1 and lateral tension load
p2 . These loads vary independently in the domain: p1  0,  y  ,

p2  0,  y  .
FEM model: using about 800 8-node-2D elements (depending on the size of the circular
hole), with 2  2 Gaussian points for numerical integration to model a quarter
of the structure, Fig. 7.2.2. The structure is considered as a plain stress
problem.

Figure 7.2.1: Square plate with central Figure 7.2.2: FEM mesh
circular hole

Numerical results

Firstly, to validate our algorithm, we choose  u  y  1.0 for perfectly plastic material.
Then the results are compared to those in Gaydon, McCrum (1954) and Vu D.K. (2001)
and are shown in Table 7.2.1. The analytical limit load is exact for R L  0.5,1 because
lower bound and upper bound assume the same value. These tests are run with c  1010 and
  1022 .

81
Table 7.2.1: Comparison of load factors
Limit analysis Shakedown analysis
R Analytical Vu Present
L Present u u u
lower upper perfectly  1.0  1.4  2.0
bound bound plastic y y y
0.1 0.97063 0.99215 0.97143 0.48661 0.48656 0.48651 0.48652
0.2 0.89425 0.92376 0.89591 0.42768 0.42766 0.42765 0.42766
0.3 0.79122 0.80829 0.79226 0.35308 0.35304 0.35302 0.35303
0.4 0.67602 0.69048 0.67677 0.27067 0.27064 0.27064 0.27065
0.5 0.55682 0.55682 0.55759 0.19021 0.19020 0.19021 0.19021
0.6 0.43801 0.43801 0.43851 0.12126 0.12125 0.12125 0.12125
0.7 0.32195 0.32195 0.32229 0.06652 0.06652 0.06652 0.06652
0.8 0.20991 0.20991 0.21023 0.02855 0.02855 0.02855 0.02855
0.9 0.10249 0.10249 0.10268 0.00693 0.00693 0.00694 0.00694

Secondly, to discover the influence of hardening, here we analyse the plate with
R L  0.2 made of hardening material with  u  y  1.4 . The interaction diagram of load
factors is plotted in Fig. 7.2.3, where the shakedown limits with and without hardening
effect are coincident.

Figure 7.2.3: Interaction diagram for load factors of square plate with central
circular hole under biaxial tension.

Subconclusions 7.2

 When  u  y  1 , the present results are quite close to the results for perfectly
plastic material in literature.

82
  For this structure, the shakedown bound of kinematic hardening coincides with
the shakedown bound of perfectly plastic material, since inadaptation is
alternating plasticity.
 The algorithm gives the stable solutions when the penalty parameter
c  106 ,1012  , and   1015 .

7.3 Structure 3: Circular plate under surface pressure and uniformed bending

Problem definitions

This problem has been investigated in Zhang G. (1992), using a numerical approach based
on the static shakedown theory.

Geometry: The circular plate described in Fig. 7.3.1 has the thickness d  0.01 m ,
radius R  1.0 m .
p

M M

Figure 7.3.1: Circular plate under surface pressure and uniformed bending.

Material: Young’s modulus: E  2.1104 kN/cm2 , yield stress:  y  36.0 kN/cm2 , for
bounded hardening:  u  y  1.2 , and for unbounded hardening  u  y  2 ,
Poisson’s ratio:   1 3 .
Loads: uniformly distributed pressure p on surface and uniformed bending
moment M . These loads vary independently in the domain: p  0, 1 ,
M  0, 1 .
FEM model: using 600 20-node-3D elements with 2  2  2 Gaussian points each one to
model a quarter of the structure.

Results

Analytical plastic collapse limits of pure bending M 0 and pure pressure p0 introduced by
Sawczuk, Sokól-Supel (1993), see these results also in Zhang G. (1992):
d2
M0   y (a)
4
(7.3.1)
 0.5d 
2

p0  6.517 y (von Mises) (b)

R2

83
 Fig. 7.3.2 shows the interaction diagram of surface pressure vs. bending moment.
The results are normalized by the analytical plastic collapse limits.

Figure 7.3.2: Interaction diagram for limit and shakedown analysis of

a circular plate under surface pressure and bending.

Subconclusions 7.3
 Numerical solutions of limit analysis are quite close to analytical solutions.
 When the plate is subjected to pure bending or pure surface pressure, then
plastic collapse limit is equal to the shakedown limit, and hardening effect is
quite clear. For the case of unbounded kinematic hardening, plastic collapse
limit is infinite, but shakedown limit is equal to that of  u  y  2.0 .
 For the case of combined load, ratchetting will occurs only if either bending
predominates or pressure predominates.

7.4 Structure 4: Simple frame

Problem definitions

This problem is investigated for perfectly plastic material in Tran T.N. et al. (2010), using
edge-based smoothed finite element method (ES-FEM) by primal-dual shakedown
algorithm, and in Garcea et al. (2005) using a mixed finite element method in which both
stress and displacement fields are interpolated. The structure is recalled here to validate the
present algorithm by choosing  u  y  1.0 , and to further investigate the effect of
hardening by giving  u  y  1.35 .

84
Geometry: The simple frame is described in Fig. 7.4.1. It is investigated for two cases:
a) sway frame where support A can move free in ox direction, Fig 7.4.1a
and b) nonsway frame where support A and C are fixed, Fig. 7.4.1b.

Figure 7.4.1: Simple frame under vertical and horizontal loads.

Material: Young’s modulus: E  2.0 105 N/mm2 , yield stress:  y  100 N/mm2 ,
hardening effect (ultimate strength):  u  y  1.35 for bounded hardening,
and  u  y  2 for unbounded hardening, Poisson’s ratio:   0.3 .
Loads: uniform distributed loads: p1 applied on AB and p2 applied on BC. These
loads vary independently in the domain: p1  1.2, 3 , p2  0.4, 1 . The
load domain is described in Fig 7.4.2.

Figure 7.4.2: Load domain.

FEM model: using 636 8-node-2D elements with 2  2 Gaussian points each one to
model whole structure, Fig. 7.4.3. The structure is considered as a plain
stress problem.

85
 R=20

Figure 7.4.3: FEM mesh.

Numerical results

Firstly, to validate our algorithm, giving  u  y  1.0 for perfectly plastic material, the
attained results are compared to those in Garcea et al. (2005), Tran T.N. et al. (2010).
Table 7.4.1 and table 7.4.2 show comparisons for limit analysis and shakedown analysis
respectively. These tests are run with c  1010 and   1025 .

Table 7.4.1 Comparison of plastic collapse limit (perfectly plastic material)

sway frame nonsway frame
Author P̂4 P̂2 P̂3 P̂4 P̂2 P̂3
Garcea et al. 2.975 2.831 2.465 7.804 4.207 3.949
Tran et al. 2.970 2.792 2.659 7.901 4.241 4.008
Present 2.889 3.000 2.692 8.022 4.317 4.039

Table 7.4.2 Comparison of shakedown limit

sway frame nonsway frame
Author [p1, p2] = [3.0, 1.0] [p1, p2] = [3.0, 1.0]
Garcea et al. (perfectly plastic material) 2.473 3.925
Tran et al. (perfectly plastic material) 2.487 4.006
Present (perfectly plastic:  u  y  1.0 ) 2.517 4.015
Present (kin. hardening:  u  y  1.35 ) 3.206 5.217

86
 Secondly, to investigate the influence of hardening, choosing  u  y  1.35 for
bounded kinematic hardening, and  u  y  2.0 for unbounded kinematic hardening. The
interaction diagram is plotted in Fig. 7.4.4 for sway frame and Fig 7.4.5 for nonsway
frame.

Figure 7.4.4: Interaction diagram for limit and shakedown analysis of

a simple sway frame.

Figure 7.4.5: Interaction diagram for limit and shakedown analysis of

a simple nonsway frame.

87
Parametric studies

To study the influence of penalty parameter c , keeping the value   1025 , and let c
varies from 105 to 1016 and do limit and shakedown analysis for the load combination
 p1, p2   3, 1.0 ,  u  y  1.5 . The algorithm gives stable solutions if c  106 ,1013  for
limit analysis, and c  106 ,1012  for shakedown analysis, see Fig. 7.4.6.

To study the influence of  , keeping the value c  1010 , and let  varies from 1030
to 1010 . The algorithm gives stable solutions for every case when   1014 , see Fig. 7.4.7.

Figure 7.4.6 : Safety region of c for Figure 7.4.7: Safety region of  for
stable solution stable solution.

Subconclusions 7.4

 When  u  y  1 , the present results compare well to those in Garcea et al.

(2005), and Tran T.N. et al. (2010).
 for  u  y  1 all of shakedown limits are ratchetting limit.
 for  u  y  1.35 all of shakedown limits are either ratchetting limit, or
ratchetting and alternating simultaneously, so the benefit of hardening effect is
quite clear.
 Algorithm gives the stable results when penalty parameter c  106 ,1012  , and
  1015 .
 For sway as well as nonsway frames, shakedown limit is either equal to or little
bit smaller than plastic collapse limit, so hardening effect to shakedown limit is
obvious.

7.5 Structure 5: Continuous beam

Problem definitions

Together with structure 4, this structure is also investigated in Garcea et al. (2005) and
Tran T.N. et al. (2010).
Geometry: The continuous steel beam is described in Fig. 7.5.1.

88
 Figure 7.5.1: Continuous beam.

Material: Young’s modulus: E  1.8 105 N/mm2 , yield stress:  y  100 N/mm2 ,
ultimate strength:  u  1.35 y , Poisson’s ratio:   0.3 .
Loads: uniform distributed loads: p1 and p2 vary independently in the
domain: p1  1.2, 2 , p2  0, 1 . The load domain is described in Fig 7.5.2.

Figure 7.5.2: Load domain for structure 5.

FEM model: using 589 8-node-2D elements with 2  2 Gaussian points each one to model
half structure, Fig. 7.5.3. The structure is considered as a plain stress
problem.

Figure 7.5.3: FEM mesh.

Numerical results

Table 7.5.1 shows the present results of limit and shakedown analysis, compared to those
of different analysis methods in literature.

89
Table 7.5.1 Comparison of plastic limit collapse and shakedown results
limit shakedown
[p1,p2]= [p1,p2]= [p1,p2]= [p1,p2]= [p1,p2]=
Author
[2.0, 0.0] [0.0, 1.0] [1.2, 1.0] [2.0, 1.0] [1.2, 2.0]
Garcea et al. 3.280 8.718 5.467 3.280 3.244
Tran et al. 3.402 9.192 5.720 3.388 3.377
Present (perfectly plastic) 3.300 8.744 5.500 3.300 3.264
Present (kin. hardening) 4.455 11.804 7.425 4.455 4.406

Table 7.5.2 shows the inadaptation modes for the certain combined load:
p1  0.6 p1 combines with p2  p2 . It can be seen that, for this load combination, the
structure is failure by ratchetting if hardening is ignored, but it will failure by alternating if
hardening is considered.

Table 7.5.2 Comparison between

alternating and ratchetting for certain
case of loading
perfectly kinematic
plastic hardening
alternating 6.700 6.700
ratchetting 5.399 6.704

Interaction diagram of shakedown load multiplier is plotted in Fig. 7.5.4.

Figure 7.5.4: Interaction diagram for shakedown bounds of continuous beam.

The results are not normalized.

90
Subconclusions 7.5

 When  u  y  1 , the present results compare well to those in literature.

 For the certain structure subjected to certain load, it will be inadaptation either
u
by alternating or by ratchetting, depends on ratio .
y
 Benefit of hardening is quite clear for this continuous beam.

7.6 Structure 6: Cylindrical pipe under complex loading

Problem definitions
This closed-end pipe is investigated for perfectly plastic material in Vu D. K. (2001), using
primal-dual shakedown algorithm, in Yan A.M. (1997), using shell element, and in Tran
T.N. (2008) using shell elements and Illyushin plastic function, etc. The structure is
subjected to bending M b and torsion M t moments, internal pressure p and axial tension
T . It is recalled here to validate present algorithm by choosing  u  y  1.0 , and to further
investigate the effect of hardening by choosing  u  y  1.25 .

Material: Young’s modulus: E  2.1105 N/mm2 , yield stress:  y  160 N/mm2 ,

ultimate strength:  u  1.25 y , Poisson’s ratio:   0.3 .
FEM model: using 60 20-node-3D elements with 2  2  2 Gaussian points each one to
model whole structure with the dimensions: length L  2700 mm , mean
radius r  300 mm , and thickness h  60 mm , see Fig. 7.6.1.

Figure 7.6.1: FEM mesh of cylindrical pipe.

Results

a) Analytical solutions: the analytical solutions of plastic collapse limit for cylindrical pipe
under complex loading can be cited from [133], etc. and presented hereafter:

Pure bending capacity:

 2 h2 
M b lim  4 y h  r    3647.52 106 Nmm . (7.6.1)
 12 

91
 Pure torsion capacity:
2
M t lim   r 2 h y  3134.24 106 Nmm . (7.6.2)
3
Pure tension capacity:
Tlim  2 rh y  18095573.6 N . (7.6.3)
Pure internal pressure capacity:
h
plim   y  32 N/mm2 . (7.6.4)
r
and the normalized load multiplier when bending, internal pressure and tension are
combined is:

4  3n2  n  2n  
m cos   x
, (7.6.5)
2  4  3n2 2 
 
where

m  M M b lim

n  p plim (7.6.6)

nx  T Tlim
If the axial tension force comes from only internal pressure on closed ends, then
nx  n 2 , and formula (7.6.5) becomes:

4  3n2
m . (7.6.7)
2
b) Numerical solutions

FE analysis is done for structure subjected to combined internal pressure p and bending
M b . Results are presented in table 7.6.1, normalized by pure bending capacity in formula
(7.6.1) and pure internal pressure in formula (7.6.4).
Limit analysis is implemented for  u  y  1.0 to be compared to formula (7.6.7),
and interaction diagram is plotted in Fig. 7.6.2.
Shakedown analysis is implemented for the load domain: p  [0,1]; M b  [1, 1] .
Interaction diagram is plotted in Fig. 7.6.3 shows the effect of hardening when
 u  y  1.25 for bounded kinematic hardening and  u  y  2 for unbounded kinematic
hardening.

92
Table 7.6.1: Limit and shakedown load multipliers of cylindrical pipe
subjected to internal pressure and bending
shakedown shakedown shakedown
limit factor
load elastic factor factor factor
perfectly
combination factor (perfectly (bounded (unbounded
plastic
plastic) hardening) hardening)
0.0p_1.0M 0.7338 1.0012 0.7338 0.7338 0.7338
0.2p_1.0M 0.7228 0.9870 0.7297 0.7310 0.7304
0.4p_1.0M 0.7011 0.9478 0.7228 0.7236 0.7231
0.6p_1.0M 0.6570 0.8914 0.7131 0.7132 0.7134
0.8p_1.0M 0.6023 0.8267 0.7011 0.7013 0.7014
1.0p_1.0M 0.5509 0.7608 0.6667 0.6855 0.6853
1.0p_0.8M 0.6168 0.8540 0.7696 0.8128 0.8127
1.0p_0.6M 0.6921 0.9556 0.8906 0.9894 0.9894
1.0p_0.4M 0.7727 1.0546 1.0179 1.2318 1.2346
1.0p_0.2M 0.8486 1.1306 1.1204 1.3874 1.5506
1.0p_0.0M 0.9019 1.1589 1.1586 1.4482 1.8091

Figure 7.6.2: Interaction diagram for limit bounds. Comparison

between analytical and numerical solutions.

93
 Figure 7.6.3: Interaction diagram for elastic and shakedown bounds,
normalized by pure plastic collapse limits, M b lim and plim .

Subconclusions 7.6

 The present results of limit analysis for  u  y  1 compare well to analytical

solutions.
 The hardening effect increases the pressure capacity clearly if the moment is less
than 0.5M b lim .
 The shakedown bound of unbounded hardening structure cannot exceed two
times of elastic bound.

7.7 Structure 7: Grooved rectangle plate under tension and bending

Problem definitions:

This structure made of perfectly plastic material has been analysed by many authors. Limit
analysis for pure tension load  pN  0, pM  0  has been done by Prager and Hodge
(1951), Casiaro and Cascini (1982) and Yan A.M. (1997). For combined tension and
bending  pN  0, pM  0  , limit and shakedown analyses have been done by Heitzer
(1999), Vu D.K. (2001) and Tran T.N. (2008). Their results are close together, so it is
sufficient to compare the present results to one out of those.

94
Geometry: The structure is described in Fig. 7.7.1 has dimensions in mm: R  250 ,
L  2R , and h  4R .
Material: Young’s modulus: E  2.1105 N/mm2 , yield stress:  y  116.2 N/mm2 ,
ultimate strength:  u  1.25 y , Poisson’s ratio:   0.3 .
Loads: The load domain is defined by: pM  0,  y  , and pN  0,  y  .
FEM model: Since the symmetry of geometry and applied load, only half of structure is
modeled by using 500 8-node-2D elements. The problem is considered as
plain stress. FEM mesh is showed in Fig. 7.7.2.

pM

pN

pN

pM

Figure 7.7.1: Structure 7. Figure 7.7.2: FEM mesh.

Numerical solutions

The interaction diagram for limit states including elastic, shakedown and plastic collapse
limits are plotted together with the results from Tran T.N. (2008), and displayed in Fig.
7.7.3.

95
Figure 7.7.3: Interaction diagram for limit states of grooved rectangular plate in
tension and bending.

Subconclusions 7.7

 The present results for  u  y  1 compare well to those in Tran T.N. (2008).
 The structure failures only in the alternating plasticity mode. The LCF bound is
two times the elastic bound, so that there is no hardening effect on the
shakedown limit.

7.8 Structure 8: Tension-torsion experiment

Problem definitions

The hollow steel bar is subjected to tension force N and torsion moment M . This
problem is investigated by Heitzer et al. (2003), Staat, Heitzer (2002) by both experiment
and FE analysis,. Their FE results compare well to analytical solutions for linear and
nonlinear kinematic hardening obtained for bounded linear kinematic hardening and linear
kinematic hardening with the Amstrong-Frederick model respectively (see Staat, Heitzer
(2002)).

Geometry: the structure of the test specimen is described in Fig. 7.8.1.

96
 Figure 7.8.1: Tension torsion specimen.

Material: Young’s modulus: E  2.07 105 N/mm2 , yield stress:  y  485 N/mm2 ,
ultimate strength:  u  631 N/mm2 ,  u  y  1.3 , Poisson’s ratio:   0.3 .
Loads: there are four cases of load domains considered: (a) constant torsion and
cyclic tension with nonzero mean stress, M : deadload, N  0, 1 , (b)
constant torsion and cyclic tension with zero mean stress,
M : deadload, N   1, 1 , (c) constant tension and cyclic torsion with zero
mean stress, N : deadload, M   1, 1 , and (d) both tension and torsion are
cyclic loads, with fully reversed, N   1, 1, M   1, 1 .
FEM model: To reduce the computing time, we use 360 20-node-3D elements to model
only the critical part of the structure, Fig. 7.2.2. This simplification does not
affect the results.

Figure 7.8.2: FEM mesh of tension – torsion experiment.

Results

a) Analytical solutions: the analytical solutions for this structure can be cited from Heitzer
et al. (2003), and presented hereafter:

Elastic analysis:

For pure torsion and axial symmetry, the normal stress vanishes and only shear stresses
  r     z  r  occur. The stress for constant moment M at the radius r is calculated as:

2M
 z  r     r   r. (7.8.1)
  ra4  ri 4 

97
The maximum shear stress is at the outer radius. For ra  4 mm and ri  2.4 mm then:

2M M
 max    ra   ra  1/mm3 . (7.8.2)
  ra  ri 
4 4
87.5

with the equivalent von Mises stress  eq  3 max and the structure starts yielding at:

y 485
M el  87.5 mm3  87.5 Nmm  24501.3 Nmm . (7.8.4)
3 3
Limit analysis:

pure tension capacity is easily calculated:

- for perfectly plastic material:
pp
Nlim  N pl    ra2  ri 2   y    42  2.42  485 N  15602 N . (7.8.5)

- for bounded linearly kinematic hardening material:

blkh
Nlim   u  y  Nlim
pp
 1.3 15602  20282.6 N . (7.8.5)

pure torsion capacity:

pp
- for perfectly plastic material: M lim is obtained from elastic moment M el with
plastic limit load factor  pl , (see details in Issler, Ruoß, Häfele (1995))
pp
M lim  M pl   pl M el  1.2  24501.3 Nmm  29401.56 Nmm , (7.8.6)

where M el is calculated in formula (7.8.4), and

4 1 d D 4 1   4.8 8 
3 3

 pl      1.20098 . (7.8.7)
3 1   d D 4 3 1   4.8 84

- for bounded linearly kinematic hardening material:

blkh
M lim   u  y  M lim
pp
 1.3  29401.56 Nmm  38222.0 Nmm . (7.8.5)

b) Numerical solutions:

The interaction diagrams are plotted in Fig. 7.8.3a to 7.8.3d, corresponding to load
domains (a) to (d). In the Fig. 7.8.3a, for the reason of comparison, the shakedown bounds
are normalized by N z 0  15.602 kN and M z 0  27.98 Nm , taken from Heitzer et al.
(2003). In the Fig. 7.8.3b to 7.8.3d, the shakedown bounds are normalized by analytical
solutions of pure tension Nlimpp
 15.602 kN and pure torsion M limpp
 29401.56 Nmm . All
tests are run with   1020 , and c  1010 .

98
Figure 7.8.3a: Shakedown interaction diagram for load domain (a), M is a dead load, and
N  0,1 , normalized by N z 0  15.602 kN and M z 0  27.98 Nm .

Figure 7.8.3b: Limit and Shakedown interaction diagram for load domain (b), N is a dead
load, and M   1, 1 , normalized by Nlim
pp
 15.6 kN and M lim
pp
 29.4 Nm .

99
Figure 7.8.3c: Limit and Shakedown interaction diagram for load domain (c), M is a
dead load, and N   1, 1 , normalized by Nlim  15.6 kN and M lim  29.4 Nm .

Figure 7.8.3d: Limit and Shakedown interaction diagram for load domain (d),
M   1, 1 , and N   1, 1 , normalized by Nlim  15.6 kN and M lim  29.4 Nm .

Keep   1020 and let c varies from 105 to 1015 . Tests are run for load
combination: M  M lim
pp
and N  Nlim
pp
in the load domain (a). The influence of the penalty
parameter c on the shakedown load factor is displayed in Fig. 7.8.4.

100
Figure 7.8.4: Influence of common logarithm lg c of the penalty parameter on the results.

Subconclusions 7.8

 Present results compare well to those in literature, which are obtained by lower
bound approach.
 For the load domain (a), shakedown bound coincides with limit bound. The
structure failures in ratchetting mode. Hardening effect is easily to be seen as:
 u pp
 lim
pp
  sdpp ;  lim
blkh
  sdblkh   lim .
y

 For the load domain (b): N is the dead load and M   M max , M max  , if tension
is large enough, then ratchetting appears. The benefit of hardening is obvious.
Contrary, for the load domain (c), M is the dead load and N    Nmax , Nmax  , if
torsion is large enough, then ratchetting appears. The benefit of hardening is
obvious. From Fig. 7.8.3b and 7.8.3c, we can see that when  sdpp   lim
pp
, it is
possible for  sdblkh   lim
blkh
.
 For the load domain (d): M   M max , M max  and N    Nmax , Nmax  ,
shakedown bound coincides with elastic bound. There is no hardening effect.
 The algorithm gives stable results if c  108 , 1013  .

101
7.9 Structure 9: Pressure vessel with nozzle

Problem definitions

This pressure vessel with nozzle made of perfectly plastic material has been investigated in
the literature. Tran T.N. (2008) has performed limit and shakedown analysis, using shell
elements and the upper bound shakedown algorithm, and Hsieh et al. (2000) has done limit
analysis using 20-node volume elements.
Based on the results in Tran T.N. (2008), for any combination of loading, the
shakedown limits are always equal to two times the corresponding elastic limits. The
structure is recalled here to validate the present algorithm by choosing  u  y  1.0 , and to
further investigate the effect of hardening by choosing  u  y  1.4 .

Geometry: The structure with dimensions in mm is described in Fig. 7.9.1. It is fixed at

the base. Top end of nozzle is closed. In Tran T.N. (2008), height of nozzle
is chosen as three times of its diameter, 1200 mm , and height of vessel is
chosen as its diameter, 2000 mm . To reduce computing time, we choose
the height of the nozzle is 800 mm , (two times of its diameter), and vessel’s
height is 1000 mm (half of its diameter). We have tested and conclude that,
with these new heights, the results do not change.

M tor
M in

M out

Internal pressure p

Figure 7.9.1: Pressure vessel with nozzzle under complex load.

102
Material: Young’s modulus: E  2.1105 N/mm2 , yield stress:  y  340 N/mm2 ,
ultimate strength is chosen as:  u  1.4 y , Poisson’s ratio:   0.3 .
Loads: The structure is subjected to internal pressure p applied on both the vessel
and the nozzle combined with tension force N , in-plane bending moment
M in , out of plane bending moment M out , and torsion moment M tor . These
loads vary independently in the normalized domain:  0, 1 . Besides the
tension force N , pressure on the closed end of nozzle also contributes
tension force.

FEM model: using 5920 20-node-3D elements to model whole structure. In thickness
direction, it is discretised by two elements. FEM mesh and local refinement
detail are described in Fig. 7.9.2.

Figure 7.9.2: FEM mesh and local refinement detail.

Numerical results

Firstly, to validate our algorithm, choosing  u  y  1.0 for perfectly plastic material, the
attained results are compared to those in literature. The comparison is presented in table
7.9.1 for limit states consist of: elastic, shakedown and plastic collapse. These tests are run
with c  1010 and   1020 .

103
Table 7.9.1: Comparison of limit states for individual loads
Out of
Internal Axial In-plane
plane Torsion
Limit states Author & element type pressure force bending
bending (kNm)
(MPa) (MN) (kNm)
(kNm)
Hsieh et al., 3D-20 node 1.37 0.45 64.7 66.5 193.8
Elastic Tran, Shell- 4 node 1.44 0.39 64.4 68.2 193.3
Present, 3D-20 node 1.46 0.42 60.5 65.5 164.8
Hsieh et al. 3.54 1.63 282.6 265.8 625
Plastic Tran 3.63 1.67 270.8 270.8 635
collapse
Present,  u  y  1.0 3.19 1.49 281.5 273.6 555.1
Tran 2.87 0.78 128.8 136.4 386.6
Shakedown Present,  u  y  1.0 2.76 0.56 121.0 131.1 330.2
Present,  u  y  1.4 2.76 0.56 121.0 131.1 330.2

Secondly, we do limit and shakedown analysis for structures under four load
combinations: (1) internal pressure plus in-plain bending, (2) internal pressure plus out of
plain bending, (3) internal pressure plus torsion, and (4) internal pressure plus tension. The
interaction diagrams are plotted in Fig. 7.9.3 to 7.9.6.

Figure 7.9.3: Interaction diagram for limit states of nozzle-vessel in combined internal
pressure and in-plain bending, normalized by plim  3.27 MPa and M in-lim  281.5 kNm .

104
Figure 7.9.4: Interaction diagram for limit states of nozzle-vessel in combined internal
pressure and out of plain bending, normalized by plim  3.27 MPa and
M out-lim  273.06 kNm .

Figure 7.9.5: Interaction diagram for limit states of nozzle-vessel in combined internal
pressure and torsion, normalized by plim  3.27 MPa and M tor-lim  555.1 kNm .

105
Figure 7.9.6: Interaction diagram for limit states of nozzle-vessel in combined internal
pressure and tension, normalized by plim  3.27 MPa and Nlim  1.49 MN .

Subconclusions 7.9

 The present results compare well to those in literature, except the torsion
capacity is 11% different from Hsieh et al. solution and 12.6% different from
Tran’s solution.
 Inadaptation occurs for this structure by alternating plasticity, consequently
there this no benefit of hardening.

7.10 Structure 10: Short cylindrical shell under internal pressure p and ring load Q

Problem definitions

The incremental collapse mode of this structure has been investigated by Groß-Weege and
Weichert (1992).

Geometry: The short cylindrical shell is described in fig. 7.10.1.

106
 Q

p h
 R  0.1

 L  0.3
 R

Figure 7.10.1: Short cylindrical shell under internal pressure and ring load.

Material: Young’s modulus: E  2.1105 N/mm2 , yield stress:  y  360 N/mm2 ,

ultimate strength:  u  1.5 y , Poisson’s ratio:   0.0 .
Loads: both internal pressure p and ring load Q vary independently in the domain:
p  0, 1 , Q  0, 1 .

FEM model: Fig. 7.10.2 shows the quarter of structure which is modelled by 225 20-node-
3D elements.

Figure 7.10.2: FEM mesh.

Numerical results

The interaction diagram of internal pressure p and ring load Q in fig. 7.10.3 is normalized
by the analytical limit loads plim and Qlim
h
plim   y (a)
R (7.10.1)
h

Qlim   y L 2  1
R
(b)

Fig. 7.10.3 shows that: For perfectly plastic material, incremental collapse occurs for all
load domains, except of the domain  p  0.8 plim , Q  Qlim  , where incremental collapse
and LCF occur simultaneously. Our results are in good agreement with results of Groß-
Weege and Weichert (1992). For kinematic hardening material, if the ring load Q is small,

107
only ratchetting occurs, hence the shakedown load multiplier of hardening material,  sdblkh ,
is  u  y times higher than the shakedown load multiplier of perfectly plastic material,
 sdpp . When Q is high enough, LCF occurs, hence  sdblkh is less than  u  y  sdpp , but of
course, greater than  sdpp .

Figure 7.10.3: Interaction diagram of internal pressure p and ring load Q for structure 10,
normalized by plim and Qlim .

108
 Chapter 8

Conclusions and further studies

Conclusions
We have developed a new upper bound algorithm for shakedown analysis of elastic plastic
bounded linearly kinematic hardening structures. The results are very close to lower bound
solutions or to analytical solutions in the literature. The advantage of the upper bound
algorithm is that it can deal with large-scale problems with a very high number of degrees
of freedom, meanwhile the lower bound algorithm suffers from a large number of
nonlinear inequality constraints which makes the size of the optimization problem become
very large.

For unbounded linearly kinematic hardening structures a primal-dual algorithm has

been developed which converges faster than the upper bound algorithm.

The algorithm gives stable results when   1020 , and the penalty parameter
c  108 ,1012  . If c is outside this interval, it strongly influences the results.

For  u   y the bounding surface coincides with the initial yield surface and the
elastic perfectly plastic model is obtained. If  u  2 y we have the unbounded kinematic
hardening model.

For limit analysis, we always obtain

 u lim
 blkh
lim
  pp . (c1)
y
Let  el ,  sdpp , and  sdblkh denote respectively elastic limit, shakedown limit for elastic
perfectly plastic, and shakedown limit for bounded kinematic hardening material, then:

 u pp
 sdpp   sdblkh   sd  2 el . (c2)
y

The left equality occurs if the subsequent yield surface translates inside the bounding
surface, the middle equality occurs if the subsequent yield surface is fixed on the bounding
surface, the last equality occurs when yield surface translates unboundedly. If a structure
shakes down in the alternating plasticity mode, then there is no benefit of the hardening
effect.

109
 For the sake of simplification, as stated in Pham D.C. (2007), we can use the
shakedown algorithm of elastic perfectly plastic material to do a shakedown analysis of
bounded kinematic hardening structures with replacing  y by  u to obtain the ratchetting
limit. However, this simplification ignores the inequalities in relation (c2).

The shakedown limit does not depend on the hardening curve, but only on the
initial yield stress  y and the ultimate strength  u .
Elastic analysis is sensitive to stress concentration such as sharp reentrant corners
of the geometry. As a consequence shakedown analysis shows the same sensitivity if the
failure is by alternating plasticity.

Further studies
Shakedown analysis for hardening structures can be developed for structural reliability
analysis, where geometry, material properties and applied loads are considered as random
variables.

The influence of temperature on material properties should be investigated, since

the hardening effect plays an important role for structures under thermal load. This is an
unsolved problem for temperature dependent elastic data (Oueslati, de Saxcé, 2009).

For more realistic modelling of hardening material, nonlinear kinematic hardening

should be considered although Staat, Heitzer (2002) showed that the bounded linearly
kinematic hardening model may give results which are quite close to the nonlinear
Frederick-Armstrong model.

Mixed models which combine isotropic and kinematic hardening model should be
investigated. The subsequent yield surface may also change its shape. In most cases no
detailed information is available for such complicated material. It is one of the great
advantages of shakedown analysis that it can give robust results with incomplete
information on material models and load history. Therefore is may be recommended to
first study if complex material models give shakedown limits which differ from the limits
obtained with more simple models. This would help make a decision about the necessary
effort in theory and material testing.

110
References:
[1] Abdel-Karim M., Ohno N. Kinematic hardening model suitable for ratchetting
with steady-state. International Journal of Plasticity 16 (2000) 225-240.
[2] Andersen K.D., Christiansen E. Limit analysis with the dual affine scaling
algorithm. Journal of Computational and Applied Mathematics 59 (1995) 233-
243.
[3] Anderson I.A., Collins I.F. Plane strain stress distributions in discrete and blended
coasted solids under normal and sliding contact. Wear 185 (1995) 23-33.
[4] Artioli E., Auricchio F., Beirão da Veiga L. Second-order accurate integration
algorithms for von-Mises plasticity with a nonlinear kinematic hardening
mechanism. Comput. Methods Appl. Mech. Engrg. 196 (2007) 1827–1846.
[5] Bari Shafiqul, Hassan Tasnim. An advancement in cyclic plasticity modelling for
multiaxial ratchetting simulation. International Journal of Plasticity 18 (2002)
873–894.
[6] Bari Shafiqul, Hassan Tasnim. Kinematic hardening rules in uncoupled modelling
for multiaxial ratchetting simulation. International Journal of Plasticity 17 (2001)
885–905.
[7] Bazaraa M.S., Sherali H.D., Shetty C.M. Nonlinear programming Theory and
Algorithms. John Wiley & Sons, Inc., New York (2006).
[8] Belytschko T. Plane stress shakedown analysis by finite elements. Int. J. Mech.
Sci. 14, (1972) 619-625.
[9] Bisbos C.D., Pardalos P.M. Second-order cone and semidefinite representations of
material failure criteria. Journal of Optimization Theory and Applications, 134 (2),
(2007) 275-301. http://dx.doi.org/10.1007/s10957-007-9243-8
[10] Bisbos C.D., Papaioannou G. Shakedown analysis of FEM-discretized steel shell
structures under the Ilyushin yield criterion. In Baniotopoulos C.C. (ed.): II. Int.
Conference on Nonsmooth Nonconvex Mechanics with Applications in
Engineering (II. NNMAE), Thessaloniki, July 7-8, (2006) 197-204.
[11] Bodovillé G., de Saxcé G. Plasticity with non-linear kinematic hardening-
modelling and shakedown analysis by the bipotential approach. Eur. J. Mech.
A/Solids 20 (2001) 99–112.
[12] Boulbibane M., Weichert D. Application of shakedown theory to soils with non-
associated flow rules. Mechanics Research Communications, 24 (5), (1997) 513-
519.
[13] Bousshine L., Chaaba A., de Saxcé G. Plastic limit load of plane frames with
frictional contact supports. International Journal of Mechanical Sciences 44
(2002) 2189–2216.
[14] Bouby C., de Saxcé G., Tritsch J.-B. A comparison between analytical
calculations of the shakedown load by the bipotential approach and step-by-step
computations for elastoplastic materials with nonlinear kinematic hardening.
International Journal of Solids and Structures 43 (2006) 2670–2692.

111
[15] Bower A.F. Mechanic of solids. CRC Press, Boca Raton (2010).
[16] Capsoni A., Corradi L. A plane strain formulation of the elastic-plastic
constitutive law for hardening von Mises materials. Int. J. Solids Structures 32
(23), (1995) 3515-3531.
[17] Carvelli V. Shakedown analysis of unidirectional fiber reinforced metal matrix
composites. Computational Materials Science 31 (2004) 24-32.
[18] Casciaro R., Cascini L. A mixed formulation and mixed elements for limit
analysis. Int. J. Num. Meth. In Eng., 18 (1982) 211-243.
[19] Chaboche J.L. A review of some plasticity and viscoplasticity constitutive
theories. International Journal of Plasticity 24 (2008) 1642–1693.
[20] Chen Haofeng, Ponter A.R.S. A method for the evaluation of a ratchet limit and
the amplitude. Eur. J. Mech. A/Solids 20 (2001) 555–571.
[21] Chen Xu, Jiao R., Kim K.S. Simulation of ratchetting strain to a high number of
cycles under biaxial loading. International Journal of Solids and Structures 40
(2003) 7449–7461.
[22] Chen Xu, Jiao R. Modified kinematic hardening rule for multiaxial ratchetting
prediction. International Journal of Plasticity 20 (2004) 871–898.
[23] Christensen Peter W. A nonsmooth Newton method for elastoplastic problems.
Comput. Methods Appl. Mech. Engrg. 191 (2002) 1189-2019.
[24] Cocchetti G., Maier G. Elastic–plastic and limit-state analyses of frames with
softening plastic-hinge models by mathematical programming. International
Journal of Solids and Structures 40 (2003) 7219–7244.
[25] Code_Aster 7.3. User’s manual. A product of EDF R&D, Clamart, France (2003).
http://www.code-aster.org
[26] Colak Ozgen U. Kinematic hardening rules for modelling uniaxial and multiaxial
ratchetting. Materials and Design 29 (2008) 1575–1581.
[27] Corona E., Hassan T. and Kyriakides S. On the performance of kinematic
hardening rules in predicting a class of biaxial ratchetting histories. International
Journal of Plasticity, 12 (1), (1996) 117-145.
[28] Chang T.Y., Chu S.C. Elastic-plastic deformation of cylindrical pressure vessels
under cyclic loading. Nuclear Engineering and Design 27 (1974) 228-237.
[29] De Saxcé G., Tritsch J.B., Hjiaj M. Shakedown of elastic-plastic structure with
nonlinear kinematic hardening by the bipotential approach, in D. Weichert, G.
Maier eds., Inelastic Analysis of Structures under Variable Loads, Kluwer,
Dordrecht, (2000), 167-182.
[30] Druyanov B., Roman I. Concept of the limit yield condition in shakedown theory.
Int. J. Solids Structures 34 (13) (1997) 1547-1556.
[31] El Abdi R., Samrout H. A non-linear kinematic hardening model for a steel under
complex loadings. Computers and Structures 76 (2000) 675-681.

112
[32] Fedele R., Filippini M., Maier G. Constitutive model calibration for railway wheel
steel through tension–torsion tests. Computers and Structures 83 (2005) 1005–
1020.
[33] Feng Xi-Qiao, Liu Xin-Sheng. On shakedown of three-dimensional elastoplastic
strain-hardening structures. International Journal of Plasticity, 12 (10) (1997)
1241-1256.
[34] Fortin J., Hjiaj M., de Saxcé G. An improved discrete element method based on a
variational formulation of the frictional contact law. Computers and Geotechnics
29 (2002) 609–640.
[35] Fung Y.C., Pin Tong. Classical and computational solid mechanics, Advanced
Series in Engineering Science- Vol. 1, World Scientific Publishing, Singapore
2001.
[36] Gao Bingjun, Chen Xu, Chen Gang. Ratchetting and ratchetting boundary study of
pressurized straight low carbon steel pipe under reversed bending. International
Journal of Pressure Vessels and Piping 83 (2006) 96–106.
[37] Garcea G., Armentano G., Petrolo S., Casciaro R. Finite element shakedown of
two-dimensional structures. Int. J. Numer. Mech. Engng 63 (2005) 1174-1202.
[38] Gaydon F.A. On the yield-point loading of a square plate with concentric circular
hole. Journal of the Mechanics and Physics of Solids, 2, (1954) 170 – 175.
[39] Gaydon F.A., McCrum A.W. A theoretical investigation of the yield point loading
of a square plate with central circular hole. Journal of the Mechanics and Physics
of Solids, 2, (1954) 156 -169.
[40] GiD 7.2. User’s manual. A product of Centre for Numerical Methods in
Engineering (CIMNE), Barcelona (2002). http://www.gid.cimne.upc.es
[41] Gokhfeld D.A., Cherniavsky O.F. Limit analysis of structures at thermal cycling,
Sijthoff & Noordhoff, The Netherlands (1980).
[42] Groß-Weege J. On the numerical investigation of the shakedown behaviour of
shells (in German). ZAMM 69, (1989) 480
[43] Groß-Weege J. A unified formulation of statical shakedown criteria for
geometrically nonlinear problems. Int. J. Plasticity 6 (1990) 433.
[44] Groß-Weege J., Weichert D. Elastic-Plastic shells under variable mechanical and
thermal loads. Int. J. Mech. Sci, 34 (11), (1992) 863-880.
[45] Hachemi A., Weichert D. Application of shakedown theory to damaging inelastic
material under mechanical and thermal load. Int. J. Mech. Sci, Vol. 39, No. 9,
(1997) 1067-1076.
[46] Hahm J.H., Kim K.H. Anisotropic work hardening of steel sheets under plane
stress. International Journal of Plasticity 24 (2008) 1097-1127.
[47] Halphen B., Nguyen Q.S. Sur les matériaux standards généralisés. J. Méc. 14,
(1975) 39-63.

113
[48] Hamadouche M.A., Weichert D. Application of shakedown theory to soil
dynamics. Mechanics Research Communications, 26 (5), (1999) 565-574.
[49] Hassan Tasnim, Zhu Yimin, Matzen V.C. Improved ratchetting analysis of piping
components. International Journal of Pressure Vessels and Piping 75 (1998) 643–
652.
[50] Hassan Tasnim, Taleb Lakhdar, Krishna Shree. Influence of non-proportional
loading on ratchetting responses and simulations by two recent cyclic plasticity
models. International Journal of Plasticity 24 (2008) 1863–1889.
[51] Heitzer M. Traglast- und Einspielanalyse zur Bewertung der Sicherheit passiver
Komponenten. Jül-3704 Report of Forschungszentrum Jülich, Ph.D. Thesis
RWTH Aachen (1999).
[52] Heitzer M., Pop G., Staat M. Basic reduction for the shakedown problem for
bounded kinematic hardening material. J. Global optimization, 17 (2000), 185-
200.
[53] Heitzer M., Staat M. FEM-computation of load carrying capacity of highly loaded
passive components by direct methods. Nuclear Engineering and Design, 193 (3),
(1999) 349-358. http://dx.doi.org/10.1016/S0029-5493(99)00190-9
[54] Heitzer M., Staat M. Basis reduction technique for limit and shakedown problems.
In [Staat, Heitzer (2003), 1-55].
[55] Heitzer M., Staat M., Reiners H., Schubert F. Shakedown and ratchetting under
tension–torsion loadings_analysis and experiments. Nuclear Engineering and
Design 225 (2003) 11–26.
[56] Hsieh M.F., Moffat D.G., Mistry J. Nozzles in the knuckle region of a
torispherical head: limit load interaction under combined pressure and piping
loads. Int. J. Pres. Vessels and Piping, 77, (2000) 807-815.
[57] Huang Y., Stein E. Shakedown of a cracked body consisting kinematic hardening
material. Engineering Fracture Mechanics 54 (1), (1996) 107-112.
[58] Issler L., Ruoß H., Häfele P. Festigkeitslehre Grundlagen, Springer, Berlin
Heidelberg (1995).
[59] Jospin J.R. Displacement estimates of pipe elbows prior to plastic collapse loads.
Nuclear Engineering and Design 178 (1997) 165–178.
[60] Kang Guozheng, Ohno Nobutada, Nebu Akira. Constitutive modelling of strain
range dependent cyclic hardening. International Journal of Plasticity 19 (2003)
1801–1819.
[61] Kapoor A., Johnson K.L. Plastic ratchetting as a mechanism of metallic wear.
Proceedings of the Royal Society of London Series A 445 (1994) 367–381.
[62] Kobayashi M. and Ohno N. Thermal ratchetting of a cylinder subjected to a
moving temperature front Effects of kinematic hardening rules on the analysis.
International Journal of Plasticity, 12 (2), (1996) 255-271.

114
[63] Kan Q.H., Kang G.Z., Zhang J. Uniaxial time-dependent ratchetting visco-plastic
model and finite element application. Theoretical and Applied Fracture
Mechanics 47 (2007) 133–144.
[64] König J.A. Shakedown of Elastic-Plastic Structures, PWN-Polish Scientific
Publishers, Warsaw, and Elsevier, Amsterdam (1987)
[65] König J.A., Kleiber M. On a new method of shakedown analysis. Bulletin
del’Academie Polonaise des Sciences, Serie des sciences techniques 26 (1978)
165-171.
[66] Kunc R., Prebil I. Low-cycle fatigue properties of steel 42CrMo4. Materials
Science and Engineering A345 (2003) 278-285.
[67] Krieg R.D., Xu Shiqiang. Plane stress linear hardening plasticity theory. Finite
Elements in Analysis and Design 27 (1997) 41-67.
[68] Kulkarni S.C., Desai Y.M., Kant T., Reddy G.R., Prasad P., Vaze K.K., Gupta C.
Uniaxial and biaxial ratchetting in piping materials–experiments and analysis.
International Journal of Pressure Vessels and Piping 81 (2004) 609-617.
[69] Khoei A.R., Jamali N. On the implementation of a multi-surface kinematic
hardening plasticity and its applications. International Journal of Plasticity 21
(2005) 1741-1770.
[70] Krämer D., Krolop S., Scheffold A., Stegmeyer R. Investigations into the
ratchetting behaviour of austenitic pipes. Nuclear Engineering and Design 171
(1997) 161-172.
[71] Lakhdar Taleb, Cailletaud G., Liliana Blaj. Numerical simulation of complex
ratchetting tests with a multi-mechanism model type. International Journal of
Plasticity 22 (2006) 724–753.
[72] Lemaitre J., Chaboche J.L. Mechanics of solid materials. University press,
Cambridge (2000).
[73] Li H.X., Yu H.S. A nonlinear programming approach to kinematic shakedown
analysis of frictional materials. International Journal of Solids and Structures 43
(2006) 6594-6614.
[74] Liu G.R., Nguyen T.T. Smoothed Finite Element Methods. CRC Press, Boca
Raton (2010)
[75] Lubliner J. Plasticity Theory. Revised edition, University of California at Berkeley
(2005).
[76] Melan E. Theorie statisch unbestimmter Systeme aus ideal plastischem Baustoff.
Sitzber. Akad. Wiss. Wien IIa, 145 (1936) 195-218.
[77] Melan E. Zur Plastizität des räumlichen Kontinuums. Ing.-Arch. 8 (1938) 116-
126.
[78] Makrodimopoulos A., Bisbos C. Shakedown analysis of plane stress problems via
SOCP. (In Staat M., Heitzer M. (eds.) Numerical Methods for Limit and
Shakedown analysis – Deterministic and Probabilistic Problems. NIC Series Vol.

115
 15, John von Neumann Institute for Computing, Jülich (2003). http://www.fz-
juelich.de/nic-series/volume15/nic-series-band15.pdf ).
[79] Mahbadi H., Eslami M.R. Cyclic loading of beams based on the Prager and
Frederick–Armstrong kinematic hardening models. International Journal of
Mechanical Sciences 44 (2002) 859–879.
[80] Mahbadi H., Eslami M.R. Cyclic loading of thick vessels based on the Prager and
Armstrong–Frederick kinematic hardening models. International Journal of
Pressure Vessels and Piping 83 (2006) 409–419.
[81] Maier G. Shakedown theory in perfect elastoplasticity with associated and
nonassociated flow-laws: A finite element, linear programming approach.
Meccanica 4, (1969) 250-260.
[82] Maier G.A shakedown matrix theory allowing for work hardening and second-
order geometric effects. In: Foundations of plasticity (Sawczuk A., ed.), North-
Holland, Amsterdam (1973) 417-433.
[83] Mohammad Abdel-Karim. Numerical integration method for kinematic hardening
rules with partial activation of dynamic recovery term. International Journal of
Plasticity 21 (2005) 1303–1321.
[84] Mohammad Abdel-Karim. Shakedown of complex structures according to various
hardening rules. International Journal of Pressure Vessels and Piping 82 (2005)
427–458.
[85] Nadarajah C., Mackenzie D., Boyle J.T. Limit and shakedown analysis of nozzle
cylinder intersections under internal pressure and in-plane moment loading. Int. J.
Pres. Ves. & Piping 68 (1996) 261-272.
[86] Navarro A., Giráldez J.M., Vallellano C. A constitutive model for elastoplastic
deformation under variable amplitude multiaxial cyclic loading. International
Journal of Fatigue 27 (2005) 838–846.
[87] Nayebi A., El Abd R. Shakedown analysis of beams using nonlinear kinematic
hardening materials coupled with continuum damage mechanics. International
Journal of Mechanical Sciences 50 (2008) 1247-1254.
[88] Nguyen Q.S. On shakedown analysis in hardening. Journal of the Mechanics and
Physics of Solids 51 (2003) 101-125.
[89] Nguyen Q.S., Pham D.C. On shakedown theorems in hardening plasticity. C. R.
Acad. Sci. Paris, t. 329, Série II b, (2001) 307–314.
[90] Ohno N. and Wang J. Nonisothermal constitutive modelling of inelasticity based
on bounding surface. Nuclear Engineering and Design 133 (1992) 369-381.
[91] Ohno N., Wang J. On modelling of kinematic hardening for ratchetting behaviour.
Nuclear Engineering and Design 153 (1995) 205-212.
[92] Ohno N. Constitutive modelling of cyclic plasticity with emphasis on ratchetting.
Int. J. Mech. Sci. 40 (2-3), (1998) 251-261.

116
[93] Oueslati A., de Saxcé G. Static shakedown theorem for solids with btemperature
dependent elastic modulus. In: A.R.S. Ponter, D. Weichert, (eds.) Limit States of
Materials and Structures: Direct Methods. Springer, Netherlands (2009).
[94] Pham D.C. Shakedown theory for elastic-perfectly plastic bodies revisited.
International Journal of Mechanical Sciences 45 (2003) 1011-1027.
[95] Pham D.C. Shakedown static and kinematic theorems for elastic-plastic limited
linear kinematic-hardening solids. European Journal of Mechanics A/Solids 24
(2005) 35–45.
[96] Pham D.C. Shakedown theory for elastic plastic kinematic hardening bodies.
International Journal of Plasticity 23 (2007) 1240–1259.
[97] Pham D.C., Weichert D. Shakedown analysis for elastic plastic bodies with
limited kinematic hardening. Proc. Roy. Soc. London A 457 (2001) 1097-1110.
[98] Phạm P.T., Staat M. An upper bound algorithm for limit and shakedown analysis
of bounded linear kinematic hardening bodies. In: G. de Saxcé, et al. (eds.) Direct
Methods. Springer Netherlands, accepted.
[99] Phạm P.T., Vũ Đ.K., Trần T.N., Staat M. A primal-dual algorithm for shakedown
analysis of elastic-plastic bounded linearly kinematic hardening bodies. IV
European Conference on Computational Mechanics (ECCM 2010), Paris, France,
May 16-21, 2010.
[100] Polizzotto C. Upper bounds on plastic strains for elstic-perfectly plastic solids
subjected to variable loads. Int. J. Mech. Sci., 21 (1979) 317-327.
[101] Polizzotto C. Dynamic shakedown of elastic-plastic solids for a set of alternative
loading histories. Int. J. Non-Linear Mechanics. 19 (4) (1984) 363-371.
[102] Polizzotto C. A bounding technique for dynamic plastic deformations of damped
structures. Nuclear Engineering and Design 79 (1984) 363-376.
[103] Polizzotto C., Borino G., Caddemi S., Fuschi P. Shakedown problem for material
with internal variables. Eur. J. Mech. A./Solids, 10 (6) (1991) 621-639.
[104] Polizzotto C., Borino G. Shakedown and steady-state responses of elastic-plastic
solids in large displacements. Int. J. Solids Structures 33 (23) (1996) 3415-3437.
[105] Polizzotto C. Shakedown of elastic - plastic solids with frictionless unilateral
contact boundary conditions. Int. J. Mech. Sci. 39 (7) (1997) 819-828.
[106] Polizzotto C. Variational methods for the steady state response of elastic–plastic
solids subjected to cyclic loads. International Journal of Solids and Structures 40
(2003) 2673–2697.
[107] Polizzotto C. Shakedown theorems for elastic–plastic solids in the framework of
gradient plasticity. International Journal of Plasticity 24 (2008) 218–241.
[108] Ponter A.R.S., Chen Haofeng. A minimum theorem for cyclic load in excess of
shakedown, with application to the evaluation of a ratchet limit. Eur. J. Mech.
A/Solids 20 (2001) 539–553.

117
[109] Ponter A.R.S., Chen H.F., Ciavarella M., Specchia G. Shakedown analysis for
rolling and sliding contact problems. International Journal of Solids and
Structures 43 (2006) 4201–4219.
[110] Ponter A.R.S., Engelhardt M. Shakedown limits for a general yield condition-
implementation and application for a Von Mises yield condition. Eur. J. Mech.
A/Solids 19 (2000) 423–445.
[111] Ponter A.R.S., Fuschi P., Engelhardt M. Limit analysis for a general class of yield
conditions. Eur. J. Mech. A/Solids 19 (2000) 401–421.
[112] Ponter A. R. S., Leckie F. A. Bounding properties of metal-matrix composites
subjected to cyclic thermal loading. J. Mech. Phys. Solids, 46 (4), (1998a) 697-
717.
[113] Ponter A. R. S., Leckie F. A. On the behaviour of metal matrix composites
subjected to cyclic thermal loading. J. Mech. Phys. Solids, 46 (11), (1998b) 2183-
2199.
[114] Portier L., Calloch S., Marquis D., Geyer P. Ratchetting under tension–torsion
loadings experiments and modelling. International Journal of Plasticity 16 (2000)
303-335.
[115] Prager W. A new method of analysing Stress and Strain in Work Hardening
Plastic Solids. J. Applied Mechanics, 23 (1956), 493-496.
[116] Prager W., Hodge P.G. Jr. Theory of perfectly plastic solids. Wiley, New York
(1951).
[117] Prager W. (1956) A new method of analyzing stress and strain in work hardening
plastic solids. J. Appl. Mech. ASME, 23 (1956) 493-496.
[118] Puzrin A.M., Houlsby G.T. Fundamentals of kinematic hardening hyperplasticity.
International Journal of Solids and Structures 38 (2001) 3771-3794.
[119] Queiroz F.J.R., Oden J.T., Ponter A.R.S., Barrosd Felicio B. A posteriori error
estimator and adaptive procedures for computation of shakedown and limit loads
on pressure vessels. Comput. Methods Appl. Mech. Engrg. 150 (1997) 155-171.
[120] Queiroz F.J.R., Ponter A.R.S., Barros Felício Bruzzi. Adaptive F.E. method for
the shakedown and limit analysis of pressure vessels. European Journal of
Mechanics A/Solids 22 (2003) 525–533.
[121] Rahman Syed M., Hassan Tasnim, Corona E. Evaluation of cyclic plasticity
models in ratchetting simulation of straight pipes under cyclic bending and steady
internal pressure. International Journal of Plasticity 24 (2008) 1756–1791.
[122] Savaidis A., Savaidis G., Zhang Ch. Elastic-plastic FE analysis of a notched shaft
under multiaxial nonproportional synchronous cyclic loading. Theoretical and
Applied Fracture Mechanics 36 (2001) 87-97.
[123] Savaidis A., Savaidis G., Zhang Ch. FE fatigue analysis of notched elastic–plastic
shaft under multiaxial loading consisting of constant and cyclic components.
International Journal of Fatigue 23 (2001) 303–315.

118
[124] Savaidis A., Savaidis G., Zhang Ch. Elastic-plastic FE analysis of a notched
cylinder under multiaxial nonproportional fatigue loading with variable
amplitudes. Computers and Structures 80 (2002) 1907–1918.
[125] Sarbayev B.S. On the theory of plasticity of anisotropic solids with isotropic and
kinematic hardening. Computational Materials Science 6 (1996) 21 l-224.
[126] Sawczuk A. On incremental collapse of shells and under cyclic loading. In:
Theory of the thin shells (Niordson F.I., ed.), Springer, Berlin (1969a) 328-340.
[127] Sawczuk A. Evaluation of upper bounds to shakedown loads for shells. J. Mech.
Phys. Solids 17(1969b) 291-231.
[128] Sawczuk A. Shakedown analysis of elastic-plastic structures. Nuclear engineering
and design 28 (1974) 121-136.
[129] Sai K., Cailletaud G. Multi-mechanism models for the description of ratchetting
Effect of the scale transition rule and of the coupling between hardening variables.
International Journal of Plasticity 23 (2007) 1589–1617.
[130] Schiffner K. Overlay models for structural analysis under cyclic loading.
Computers & Strucrures 56 (2/3) (1995) 321-328.
[131] Schwabe F. Einspieluntersuchungen von Verbundwerkstoffen mit periodischer
Mikrostruktur. Ph.D. Thesis, RWTH Aachen (2000).
[132] Shahram Sarkani, Michaelov G., Kihl D.P. Stochastic fatigue damage
accumulation in a T-welded joint accounting for the residual stress fields.
International Journal of Fatigue 23 (2001) S71–S78.
[133] Skodeli M.-A.A, Bisbos C.D. Limit and shakedown analysis of 3D frames via
approximate ellipoidal yiels surfaces. Engineering Structures 32 (2010) 1556-
1557.
[134] Staat M., Heitzer M. Limit and shakedown analysis for plastic safety of complex
structures. Transactions of the 14th International Conference on Structural
Mechanics in Reactor Technology (SMiRT 14), Lyon, France, August 17-22
(1997). http://www.iasmirt.org/SMiRT14/B02-2
[135] Staat M., Heitzer M. Limit and shakedown analysis for design. Proceeding of 7th
German-Japanese Joint Seminar on Research in Structural Strength and NDE-
Problems in Nuclear Engineering, Stuttgart (1997) 4.3.1-4.3.19.
[136] Staat M., Heitzer M. LISA – a European project for FEM-based limit and
shakedown analysis. Nuclear Engineering and Design, 206 (2-3), (2001) 151-166.
http://dx.doi.org/10.1016/S0029-5493(00)00415-5
[137] Staat M. Some achievements of the European project LISA for FEM based limit
and shakedown analysis. In N. Badie (ed.) Computational Mechanics:
Developments and Applications - 2002. ASME PVP Vol. 441, Paper PVP2002-
1300, (2002) 177-185.
[138] Staat M., Heitzer M. The restricted influence of kinematic hardening on
shakedown loads. Proceedings of WCCM V, 5th World Congress on

119
 Computational Mechanics, Vienna, Austria, July 7-12 (2002).
http://opus.bibliothek.fh-aachen.de/opus/volltexte/2005/79/
[139] Staat M., Heitzer M. (eds.) Numerical Methods for Limit and Shakedown analysis
– Deterministic and Probabilistic Problems. NIC Series Vol. 15, John von
Neumann Institute for Computing, Jülich (2003). http://www.fz-juelich.de/nic-
series/volume15/nic-series-band15.pdf
[140] Staat M., Heitzer M., Lang H., Wirtz K. Direct finite element route for design-by-
analysis of pressure components. Int. J. Pressure Vessels Piping, 82 (1), (2005)
61-67. http://dx.doi.org/10.1016/j.ijpvp.2004.04.006
[141] Stein E., Huang Y.J. Shakedown for systems of kinematic hardening materials. In:
Inelastic behaviour of structures under variables loads (Mroz et al., eds.). Kluwer
Academic Publishers, (1995) 33-50.
[142] Stein E., Zhang G. Theoretical and numerical shakedown analysis for kinematic
hardening materials. Computational plasticity Fundamentals and applications. In:
prceeding of the international conference in Barcelona, Spain, 6-10th April (Owen
D.R.J, Oñate E. and Hinton E., eds.), Pineridge Press, (1992) 1-26.
[143] Stein E., Zhang G., Huang Y. Modelling and computation of shakedown problems
for nonlinear hardening materials. Computer Methods in Applied Mechanics and
Engineering, 103 (1993) 247-272.
[144] Stein E., Zhang G., König J.A. Shakedown with nonlinear strain hardening
including structural computation using finite element method. Int. J. Plasticity, 8
(1992) 1-31.
[145] Stein E., Zhang G., Mahnken R. Shakedown analysis for perfectly plastic and
kinematic hardening materials. In: Progress in Computational Analysis of
Inelastic Structures (Stein E., ed.), Springer, Wien, New York (1993), 175-244.
[146] Tarn Q., Dvorak G. J., Rao M. S. M. Shakedown of unidirectional composites.
International Journal of Solids and Structures, 11 (6), (1975) 751-764.
[147] Tran T.N. Limit and shakedown analysis of plates and shells including
uncertainties. Dissertation, Technische Universität Chemnitz, Germany (2008).
[148] Trần T.N., Kreißig R., Vu D.K., Staat M. Upper bound limit and shakedown
analysis of shells using the exact Ilyushin yield surface. Computers & Structures,
86 (17-18) (2008) 1683-1695. http://dx.doi.org/10.1016/j.compstruc.2008.02.005
[149] Trần T.N., Phạm P.T., Staat M. Reliability analysis of shells based on direct
plasticity methods. 8th World Congress on Computational Mechanics (WCCM8),
5th European Congress on Computational Methods in Applied Sciences and
Engineering (ECCOMAS 2008), Venice, Italy, June 30 – July 5 (2008).
[150] Tran T.N., Pham P.T., Vu D.K., Staat M. Reliability analysis of inelastic shell
structures under variable loads. In: A.R.S. Ponter, D. Weichert, (eds.) Limit States
of Materials and Structures: Direct Methods. Springer, Netherlands (2009).
[151] Tran T.N., Liu G.R., Nguyen X.H., Nguyen T.T. An edge-based smoothed finite
element method for primal-dual shakedown analysis of structures. Int. J. Numer.
Engng 82 (2010) 917-938.

120
[152] Trần T.N., Staat M. An edge-based smoothed finite element method for primal-
dual shakedown analysis of structures under uncertainty. In: G. de Saxcé, et al.
(eds.) Direct Methods. Springer, Netherlands, accepted.
[153] Vu D.K. Dual Limit and Shakedown analysis of structures. PhD Thesis,
Université de Liège, Belgium (2001).
[154] Vu D.K., Yan A.M., Nguyen Dang H. A primal-dual algorithm for shakedown
analysis of structure. Comp. Meth. Appl. Mech. Engrg., 193 (2004) 4663-4674.
[155] Vu D.K., Staat M., Tran I.T. Analysis of pressure equipment by application of the
primal-dual theory of shakedown. Communications in Num. Meth. in Engng., 23
(3), (2007) 213-225. http://dx.doi.org/10.1002/cnm.891
[156] Weichert D. On the influence of geometrical nonlinearities on the shakedown of
elastic-plastic structures. Int. J. Plasticity 2, (1986) 135-148.
[157] Weichert D. Advances in the geometrically nonlinear shakedown theory. In:
Kleiber M., König J. A. Inelastic solids and structures. Pineridge Press, Swansea,
U.K. (1990)
[158] Weichert D., Gross-Weege J. The numerical assessment of elastic-plastic sheets
under variable mechanical and thermal loads using a simplified tow-surface yield
condition. Int. J. Mech. Sci., 30 (1988) 757-767.
[159] Weichert D., Hachemi A., Schwabe F. Shakedown analysis of composites.
Mechanics Research Communications, 26 (3), (1999) 309-318.
[160] Weichert D., Maier G. Inelastic Behaviour of Structures under Variable Repeated
Loads: Direct analysis methods. CISM Courses and Lectures No. 432,
International Center for Mechanical Sciences, Springer Wien-New York (2002).
[161] Weichert D., Ponter A.R.S. (eds.). Limit States of Materials and Structures, Direct
Methods. Springer (2008).
[162] Wiechmann K., Stein E. Shape optimization for elasto-plastic deformation under
shakedown conditions. International Journal of Solids and Structures 43 (2006)
7145–7165.
[163] Weiß E., Postberg B., Nicak T., Rudolph J. Simulation of ratchetting and low
cycle fatigue. International Journal of Pressure Vessels and Piping 81 (2004)
235–242.
[164] Yaguchi Masatsugu, Takahashi Yukio. Ratchetting of viscoplastic material with
cyclic softening, part 2 application of constitutive models. International Journal of
Plasticity 21 (2005) 835–860.
[165] Yan A.M. Contribution to the direct limit state analysis of plastified and cracked
structures. PhD Thesis, Université de Liège, Belgium (1997).
[166] Yan A.M., Nguyen Dang H. Direct finite element kinematical approaches in limit
and shakedown analysis of shells and elbows. In: Inelastic Analysis of Structures
under Variable Loads: Theory and Engineering Applications (Weichert D., Maier
G., eds.) Kluwer, Academic Press, Dordrecht, (2000) 233-254.

121
 [167] Yan A.M., Vu D.K., Nguyen D.H. Kinematical formulation of limit and
shakedown analysis (2003). In: Staat M., Heitzer M. (eds) (2003).
[168] Yang Xianjie, Nassar Sayed. Constitutive modelling of time-dependent cyclic
straining for solder alloy 63Sn–37Pb. Mechanics of Materials 37 (2005) 801–814.
[169] Yun S., Palazotto A. Two back-stress kinematic hardening constitutive models.
Engineering Fracture Mechanics 74 (2007) 2844–2863.
[170] Yun Su-Jin, Palazotto A. Plastic deformation under cyclic loading using two-back-
stress hardening models. International Journal of Fatigue 30 (2008) 473–482.
[171] Yoshida Fusahito. A constitutive model of cyclic plasticity. International Journal
of Plasticity 16 (2000) 359-380.
[172] Zeman J., Rauscher F., Schindler S. Pressure Vessel Design: The Direct Route.
Elsevier, Oxford (2006).
[173] Zhang G. Einspielen und dessen numerische Behandlung von Flächentragwerken
aus ideal plastischem bzw. kinematisch verfestigendem Material, Ph.D. Thesis,
University Hannover, Germany (1991).
[174] Zhao J. et al. Bound for shakedown of cohesive-frictional materials under moving
surface loads. International Journal of Solids and Structures 45 (2008) 3290-
3312.

122

View publication stats