Chapter 1

Nonlinear Mechanics

1.1 Introduction

Numerous different nonlinearities can occur in solid mechanics which are either of geometrical or of physical
nature. The treatment of associated problems demands a large bandwidth of methods and algorithms
which will be discussed in the following chapters. Based on introductory examples, different phenomena
of nonlinear behaviour will be described to introduce the reader the nature of the problems. Deliberately
simplified mechanical models are used, which are just complicated enough to represent the desired nonlinear
feature. All solutions can still be solved analytically which helps to understand the problem. However,
engineering problems cannot be formulated with such simplified models. Due to that numerical methods
have to be applied for real world applications.

1.2 Geometrical Nonlinearity

In structural analysis, it is usually sufficient to consider only small deformations and strains since many
parts of the structure can only undergo small strains to maintain their usability. With this restriction, a
linear constitutive equation can be introduced when elastic deformations are present. However, even under
this assumption, there are many problems which depict large displacements or rotations, such as cables,
beams or shells. Such problems require a nonlinear theory which includes the geometry in an exact way.
Some examples which represent different geometrically nonlinear behaviour are discussed in the following.

1.2.1 Large Displacements of a Rigid Beam

The first example for geometrically nonlinear behaviour is a rigid beam of length l, (Fig. 1.1) which is
supported by an elastic rotational spring with stiffness c at its left end.
Equilibrium at the deformed system yields directly, (see Fig. 1.1),

F l cos φ = cφ. (1.1)

Equation (1.1) relates the force F in a nonlinear way to the beam rotation φ. The nonlinearity stems from
the change of geometry in the equilibrium equation. Hence this type of behaviour is known as geometrical
nonlinearity.
For small rotations φ the approximation cos φ → 1 is obtained. With this the linear solution F = cφ/l
can be derived from (1.1). Figure 1.2 shows the increase of the force as a function of the rotation φ for both
cases. One observes clearly that the linear solution deviates from the exact geometrically nonlinear one for
large rotations.
Often the so-called second order theories are applied to include nonlinear effects in the mechanical model.
The idea is to describe the nonlinear terms using a Taylor series which is terminated after the second term.
Cristal Plasticity 2

Figure 1.1: System and loading

Figure 1.2: Force versus rotation

φ2
For the presented example, the relation of the second order theory is obtained with cos φ ≈ 1 − 2 .

cφ
F = φ2
(1.2)
l(1 − 2 )

This equation approximates the exact solution up to a rotation of φ ≈ π/3 very well (see Fig. 1.2). The
solution (1.2) deviates, however, for larger rotations from the exact solution.
In case that the flexibility of the beam has to be considered too, we would have to include also nonlinear
strain-displacement relations for the beam.

1.2.2 Large Displacements of an Elastic System

In this example, the influence of flexibility of a structure is investigated in the context of geometrically
nonlinear behaviour. Let us consider two horizontal elastic springs which have a linear force–displacement
relation. The structure is loaded by a point force F, (see Fig. 1.3).
To obtain the load–deflection curve of the force F with respect to the vertical displacement w, the
kinematical relation between the vertical displacement w and the elongation f of the spring has to be
formulated as well as the equilibrium and the constitutive law for the spring. The kinematical relation is
given by, (see Fig. 1.3), r 
2 2 2 w 2
w + l = (l + f ) → f = l 1+( ) −1 (1.3)
l
and
w
sin φ = . (1.4)
l+f
Equilibrium follows from Fig. 1.3 by considering symmetry as
F
SF sin φ = . (1.5)
2
Cristal Plasticity 3

Figure 1.3: Geometry and equilibrium

Figure 1.4: Load-deflection curve

Furthermore, the constitutive equation for the spring is assumed to be linear

SF = cf (1.6)

where c is the spring stiffness and SF denotes the force in the spring. Inserting (1.4) in (1.5) yields with
(1.6)
w F
cf = . (1.7)
l+f 2
This equation can be reformulated using (1.3) in terms of the vertical displacement w
" #
w 1 F
1− p = . (1.8)
l w 2 2cl
1+( ) l

The associated load-deflection curve is depicted in Fig. 1.4.

The load-deflection curve has a horizontal tangent at zero displacement. For that reason, this example
cannot be formulated in terms of a linearized theory. However, it is possible to derive a second order theory
for small values of w (w/l << 1) by a Taylor series expansion of the square root in (2.8):
1 w 1 w
+ ( )2 ≈ 1 − ( )2 .
1 l 2 l
This approach yields the cubical polynomial
 
w 1 w 2 F
( ) = , (1.9)
l 2 l 2cl

which approximates the exact solution for w/l < 0.4 well.
Cristal Plasticity 4

Figure 1.5: System and loading

Figure 1.6: Load-deflection curve

1.2.3 Bifurcation Problem

The solution of a nonlinear problem is not always unique. This feature will be discussed by means of the
stability problem described in Fig. 1.5. This example is equivalent to the first one; only the load acts now
in horizontal direction. Formulating equilibrium at the deformed system yields:
φ Fl
F l sin φ = cφ → = . (1.10)
sin φ c
This equation has multiple solutions. The trivial solution is φ = 0, which is valid for all values of F. For
Fl
c > 1 (|φ| ≥ | sin φ|), there exist two more solutions which are depicted in Fig. 2.6.
In total, three solutions of (1.10 exist for Fcl > 1. Hence the solution is no longer unique. The point
Fl
( c = 1) at which the three different solutions start is known as bifurcation point. An essential question is
now which solution path will be followed by the system when the load is increased beyond the bifurcation
point. The answer for that is provided by the theory of stability. For the case at hand, it can be shown that
the trivial solution is instable. The physical meaning of this is that for a small disturbance of the trivial
solution φ = 0 equilibrium is lost and the system will change to a new stable equilibrium state. Usually large
deformations and even dynamical effects occur in such situation. In technical construction, such behaviour
leads most of the times to a total collapse. Hence the identification of such instable solution is of great
practical importance.
The other solutions of (1.10) are stable.
and insensitive against small disturbances. An approximation of (1.10) using second order theory yields
with sin φ ≈ φ − φ3 /6 and 1/(1 − x) ≈ 1 + x

φ2 Fl
1+ = . (1.11)
6 c
Fl
This equation reproduces the same behaviour as (1.10), since three solutions are also present for c > 1.
Cristal Plasticity 5

Figure 1.7: System and loading

Remark: Often, in practical applications, it is only of interest to compute the bifurcation point of a
structure under a given loading. In its vicinity, it can be assumed that φ is small and hence approximate
sin φ ≈ φ. From (1.10), the linear homogeneous equation

(F l − c)φ = 0 (1.12)

is obtained which is either fulfilled trivially for φ = 0 or non-trivially for F = Fc = cl . One calls Fc the
critical load (Fc is eigenvalue of the eigenvalue problems (1.12)). The critical load Fc is equivalent to the
load related to the bifurcation point in the exact equation (1.10).

1.2.4 Snap-Through Problem

In this example, the system depicted is considered which consists of two trusses – modelled as springs of
length L0 – under the action of a vertical point force F. The springs are supported at their left and right
end, respectively.
With the kinematical relations (h − w)2 + l2 = L2 and h2 + l2 = L20 , the length change of the spring is
as follows: s 
 2 s  2
h−w h 
f = L − L0 = l  1 + − 1+ . (1.13)
l l

Equilibrium with regard to the deformed system can be written as

F
N sin(α − φ) = − , (1.14)
2
see Fig. 1.7 with the normal force N. This relation leads with sin(α − φ) = (h − w)/L to
h−w F
N =− . (1.15)
L 2
The spring characteristic is linear, hence N = cf.
Inserting (1.13) into (1.15) yields the nonlinear relation between force F and displacement w
 
L − L0 F w−h L0  − F = 0.
c(h − w) =− ⇒ 1− q (1.16)
L 2 l l 1 + ( w−h )2 2cl
l

The associated load-deflection curve is plotted for L0 /l = 1.25.

The load increases until point D and decreases afterwards for increasing displacement w. The latter
situation cannot be reached in case of a static loading. To model this process in a physically correct way a
dynamic process should be introduced once point D is passed. The system cannot not be in static equilibrium
Pr. B. RADI 6

Figure 1.8: Load-deflection curve

until point E is reached for a load larger or equal than the load at D. The associated process, in which the
system changes from one equilibrium state to another instantaneous, is called snap-through.
Due to that, point D is called snap-through point. Since the solution is not unique at such point
(equilibrium can be found for load F in D or E which is associated with two different displacements w) a
snap-through problem is a
stability problem. Furthermore, the load is limited at point D which leads to the notion of a limit
point. Depending on the geometry and the loading, snap-through behaviour can be observed in many
technical structures like trusses, beams or shells. Usually snap-through is connected with a total failure
of the structure. However, there are also applications which rely on snap-through behaviour. One of such
problems is related to the opening of, e.g. a glass of jam. Here a loud noise of the lid proves that the glass
was not opened before and that the contents are untouched. The noise is related to a snap-through of a thin
shell (the lid) following from a change of the internal pressure. The deformations related to this process are
very small, thus the lid is not damaged.
From the previous examples, it can be concluded that geometrical nonlinearity leads to numerous different
phenomena, which can occur individually or in combined form. In principle, always nonlinear behaviour
of a new structural design has to be considered until it can be proven that a linearized treatment of the
problem is adequate.

1.3 Physical Nonlinearity

Within the treatment of geometrically nonlinear problems, which have been considered so far, only linear
elastic stress–strain relations, such as the linear characteristic for the elongation of a spring, were used.
This is a good approximation for many materials but holds only under certain restrictive assumptions like
small strains. Simple examples show that this is not always true. An elongated rubber band depicts, e.g.
with increasing elongation a greater stiffness; furthermore a metal wire suffers permanent deformation under
bending which is due to elasto-plastic material behaviour. Permanent deformations occur in the last case
once a limiting stress is exceeded. This behaviour is also called plastic flow and the limiting stress is known
as yield stress.
As an example for elasto-plastic deformations, the system depicted in Fig. 1.9 is considered. It consists
of two bars with same cross section but different materials. Generally, elasto-plastic material behaviour is
assumed as shown in Fig. 1.9 where the stress is limited by the yield stress σy . The material data for Young
modulus are chosen as E1 = 2E2 = 2E; furthermore the yield stresses of the two materials are different
σy1 = 3σy2 = 3σy where the subscript is related to the bars.
Under the assumption that the loading is so small that the yield stress will not be reached in one of the
bars, a purely elastic response occurs. In that case, equilibrium can be formulated
F
N1 + N2 = F → σ1 + σ2 = , (1.17)
A
Pr. B. RADI 7

Figure 1.9: System and loading

where the normal force Ni in bar i is related to the stress by Ni = Aσi . The kinematical relations yield
u
u1 = u2 = u,  = (1.18)
l
for the displacements in normal direction ui and the associated strain .
Finally, the linear elastic material law of Hooke relates stresses to strains
u
σi = Ei  = Ei . (1.19)
l
A combination of all equations results in a linear relation between force F and displacement u:
u u Fl
E1 A + E2 A = F → u = . (1.20)
l l (E1 + E2 )A

The stresses in both bars follow as

F F
σi = Ei → σi = Ei . (1.21)
(E1 + E2 )A 3EA

Note that bar 2 starts to yield for F = 3Aσy while bar 1 is still elastic. The associated displacement is
u = σy l/E. In case of a further increase of the load, the constant normal force N2 = Aσy in bar 2 has now
to be considered.
This results, analogous to (1.20), in a displacement of
 
F l
u= − σy . (1.22)
A 2E

Finally, bar 1 yields with N1 min = 3Aσy at an applied force of F = 4Aσy .

After that, it is not possible to increase the load further and the load- displacement diagram in Fig. 1.10
is obtained.
Again a nonlinear relation between load and displacement is observed as in the previous examples.
However, here this nonlinear behaviour stems from the elasto-plastic material law defined in Fig. 1.9.
In many technical applications, nonlinear behaviour is observed which results from a combination of
geometrical and physical nonlinearities. Examples are deep drawing of metal sheets, bulk forming processes
or car crashes.

1.4 Nonlinearity Due to Boundary Conditions

Another but different source for nonlinearities is related to special boundary constraints. One major cause
of such nonlinear behaviour are boundary constraints which change with the deformation state of a system
Pr. B. RADI 8

Figure 1.10: Load-displacement diagram

Figure 1.11: System and loading

(e.g. during the increase of a prescribed load). These occur when one body comes into contact with another
one during a deformation process. Here a penetration of one body into the other is ruled out and the contact
zone between the two bodies changes depending on the load level.
The essential characteristics of contact problems are discussed using the following simple model. We
consider to bars with stiffness EA, see Fig. 1.11. The system is subjected to a point force F. Both bars are
separated by a gap δ. We look for the displacement and stress state in the bars on the condition that one bar
cannot penetrate the other one. This condition leads with Fig. 1.11 to an inequality for the displacements

u1 − u2 ≤ δ. (1.23)

Here the ”less” sign is correct in the case that bar 1 does not touch bar 2. The ”equal” sign is true for
contact of the bars.
In case of u1 − u2 < δ displacement u2 is zero. Thus the displacement of bar 1 is

Fl
u1 = . (1.24)
EA
An increase of the force F such that F > EA δl is fulfilled leads to contact of bars 1 and 2, then equation
u1 − u2 = δ is valid. With the displacements of both bars at x = 2l
Fl N1 2l N2 l
u1 = + and u2 = − (1.25)
EA EA EA
and the condition that the normal force at the contact point has to be equal for both bars (N1 (2l) =
N2 (2l) = N ) follows from u1 − u2 = δ:
 
Fl Nl 1 δ
+3 =δ→N = EA − F . (1.26)
EA EA 3 l
Pr. B. RADI 9

Using (1.25) and (1.26) yields, in case of contact, a relation between force F and displacement u1
 
u1 δ
F = EA 3 − 2 . (1.27)
l l

This equation includes, for u1 = δ, the limiting case of the beginning contact in which N = 0 is obtained
from (1.26). Hence the displacement u1 of bar 1 can also be computed from (1.24).
Figure 1.11 depicts the resulting nonlinear load-deflection curve where the nonlinear behaviour stems
only from the contact mechanism.
Basically, geometrical constraints and equilibrium equations have to be considered in contact formula-
tions which are not differentiable since the system can assume two different states of being in contact or
being not in contact. This is reflected by the kink in the load deflection curve.
Since contact nonlinearities are linked in many technical applications to further nonlinearities, like finite
or inelastic deformations, it is especially complicated to construct robust end efficient algorithms for contact
problems.
All discussed examples show that the source of nonlinear behaviour is quite different. Besides geo-
metrical effects, material properties or changing boundary conditions yield nonlinear behaviour. In the
following chapter, the underlying theoretical basis is generalized for two- and three-dimensional solids and
the necessary numerical solution schemes based on the finite element method will be developed.