Chapter 2

The Era of Elasticity

2.1 Fundamentals of Elasticity

2.1.1 Basic Field Equations
In order to establish a solution in continuum mechanics, three basic sets of relations have to
be fulfilled; (1) equilibrium conditions which guarantee that the body is always in
equilibrium; (2) compatibility conditions which guarantee that the body remains continuous;
and (3) constitutive relations which connect stresses and strains of a material behavior. These
relations can be expressed in tensor notations as follows.

Equilibrium: The equilibrium conditions for an arbitrary volume, V , Fig. 2.1, are

V
 ( ij , j + Fi )dV = 0 (2.1)

where  ij is the stress tensor and Fi is the body force. Eq. (2.1) may be written in the usual
(X, Y and Z) notation as

Fig. 2.1 Equilibrium of a material body.

 x  xy  xz
+ + + Fx = 0
x y z
 yx  y  yz
+ + + Fy = 0 (2.2)
x y z
 zx  zy  z
+ + + Fz = 0
x y z

1
where  x ,  y and  z stand for the normal stress components and  xy ,  yz , … stand for the
shear stress components.

Compatibility: The compatibility conditions (or conditions of body continuity) can be

expressed as

 ij ,kl +  kl ,ij −  ik , jl −  jl ,ik = 0 (2.3)

Upon expanding these expressions the following can obtained with respect to the usual (X, Y
and Z) notation

 2 x   y  2 xy
2

+ =2
y 2 x 2 xy
 2 y  2 z   yz 2

+ =2
z 2 y 2
yz
 2 z  2 x  2 zx
+ = 2
x 2 z 2 zx
   yz  zx  xy   2 x
− + +  = (2.4)
x  x y z
 yz
   zx  xy  yz    y
2
 − + +  =
y  y z x  zx
   xy  yz  zx   2 z
− + + =
z  z x y  xy

where  x ,  y and  z represent the normal strain components and  xy ,  yz , … represent the
shear strain components.

For a material element in volume, V, the three equations of equilibrium, Eq. (2.2) and
the six equations of compatibility, Eq. (2.4), give a sum of nine equations. On the other hand,
the total number of unknowns is fifteen (six stress components, six strain components and
three displacement components). The remaining six equations necessary to obtain a solution
are the material dependant equations (constitutive relations). The inter-relations between the
three sets of relations are schematically illustrated in Fig. 2.2.

Constitutive Relations: The constitutive relations can be expressed in tensor notation as

follows

 ij = C ijkl kl (2.5)

2
where C ijkl is the material elastic constant tensor. Equation (2.5) is a simple generalization of
Hooke’s law experiment in simple tension test; and therefore, it is referred to as the
generalized Hooke’s law. This equation can be written in matrix form for an isotropic linear
elastic material as follows

Fig. 2.2 Inter-relations between mechanics variables.

  = C   (2.6)

where the matrix C  is called the elastic constitutive or elastic moduli matrix and is given by

(1 − )   0 0 0 
 
  (1 − )  0 0 0 
   (1 − ) 0 0 0 
 
E  (1 − 2 ) 
C  =  0 0 0 0 0 
( )(
1 +  1 − 2 ) 2

 (1 − 2 ) 
 0 0 0 0
2
0 
 
 (1 − 2 ) 
 0 0 0 0 0
2 
(2.7)

where E is the Young’s modulus and  is the Poisson’s ratio.

2.1.2 Solution Process-An Illustration

Linear elasticity is based on two fundamental assumptions; the stress-strain relation is linear
and reversible. The first assumption allows for the application of the principle of
superposition. The second assumption means that the material is load path independent,
which permits for total load application; i.e., no need to go to incremental solution or to
follow load history. This leads to a simple and direct formulation of the equations of

3
mechanics at a point in a body. Nevertheless, the solution of the fifteen equations, except in
very special cases, represents a great mathematical challenge and in some cases is impossible
to achieve without exploring nontraditional thinking. Fig. 2.3 shows the solution of the
problem of half-space under concentrated load; one of the few problems that can be handled
from a direct solution of the fifteen equations.

Fig. 2.3 Half-space problem.

2.2 The Concept of Generalized Stress and Generalized Strain

2.2.1 Introduction
In the concept of continuum mechanics the field relations derived at a material point can be
used to establish solutions for limited applications. In order to widen the scope of
applications these field relations should be formulated in terms of elements from which the
structure is composed; e.g., bar, plate, shell and finite element. For example, for a bar
element instead of the six stress components,  ij , at a point one can deal with the bending
moment (generalized stress), M. Correspondingly, the six strain components,  ij , are
replaced with the angle of relative rotation or curvature (generalized strain),  . The same
concept applies for other structure elements such as plate, shell, finite element… etc.

One may ask how to decide on which variable is a generalized-stress or generalized

strain? Bending moment in a beam can be considered as a generalized-stress whereas other
variables, such as shear, cannot be considered as generalized stresses because the
deformations associated with their correspondent strains are negligible. In beam for example
the curvature signifies the dominant deformation (bending deformation) and therefore it is
considered as a generalized-strain and its correspondent stress variable (moment) is
considered as a generalized-stress.

4
 Another question may arise about the significance of this concept and its impact. The
answer is very simple; for example, in a beam problem with continuum mechanics approach
the solution searches for fifteen variables (15 equations in 15 unknowns). On the other hand,
in a beam analogy (bending theory) the solution searches for one variable. The reduction in
the effort is thus very obvious when using this concept since the effort is an exponential
function of the number of variables. In this regard, Einstein was asked once why he would
use one type of soap instead two as other people do. He replied: two soaps are too many
variables, one is enough and much simpler.

In the following sections equilibrium is utilized to develop the generalized stress in

terms of the stress tensor. Compatibility is utilized to derive the generalized strain in terms of
the strain tensor. The constitutive relations are employed to connect the generalized stresses
and the generalized strains. These developments are given for different structure elements;
e.g. bar, plate bending, shell element and finite element.
2.2.2 Bar Element as a Start
Axially Loaded Element: For a bar element subjected to an axial force (generalized stress),
P , causing a uniform axial stress,  , equilibrium leads to

P = A (2.8)

where A is the cross-sectional area of the bar. The stress,  , causes an axial strain,  , which
is associated with a change in bar length (generalized strain),  . From compatibility the
generalized strain is related to the strain as follows


= (2.9)
x

Upon integration of the previous equation

 = L (2.10)

where L is the bar length. The generalized stress, P , can be related to the generalized strain,
 , upon employing Hooke’s law.

 = E (2.11)

where E is the Young’s modulus. This leads to


P = A = EA = E A
L
or
 EA 
P=  (2.12)
 L 

5
The term (EA L ) connecting the two variables P and  is the bar axial stiffness.

Flexural Element: The same simple and systematic steps followed before for a bar under
axial force, can be followed for the development of the variables of a bar under pure bending.
Consider a bar element of length S subjected to end moments, M , such that the element
deforms as shown in Fig. 2.4. Upon applying equilibrium at the end section of the element

M =  ydA (2.13)

Fig. 2.4 Bar element under pure bending.

The generalized strain (or curvature),  , can be related to the strain in a beam,  , upon the
adoption of linear strain distribution within the section depth according to Bernoulli’s
hypothesis (plane sections perpendicular to the neutral axis before bending remain plane and
perpendicular to the neutral axis after bending). Thus,

 = y (2.14)

The generalized stress, M , is related to the generalized strain,  , through the application of
the stress-strain relation  = E .

M =  ydA =  EydA =  Ey 2 dA = E  y 2 dA

or
M = EI (2.15)

where I is the moment of inertia of the beam cross-section. It should be noted that only the
bending moment can considered as a generalized stress, whereas other parameters such as
shear or torsion cannot; since the deformations associated with their correspondent stresses
are negligible.

2.2.3 Plate Element as a Next Step

Compatibility: For a plate element in Cartesian coordinates the thickness is denoted as h and
the deflection is denoted as w . Since there is no normal force applied to the end sections of
the plate segment shown in Fig. 2.5 the neutral surface is assumed to coincide with the middle

6
surface of the plate. Based on Kirchhoff’s hypothesis for thin plates with small deflection
(plane sections perpendicular to the neutral surface before bending remain plane and
perpendicular to the neutral surface after bending) the elongation of a fiber parallel to the x-
or y-axis is proportional to its distance z from the middle surface. The generalized strains
(curvature of the deflection in the x- and y-direction,  x and  y , respectively) can be taken as
1 2w
x = =− 2 (2.16a)
rx x
1 2w
y = =− 2 (2.16b)
ry y
where rx and ry are the radii of curvatures in the x and y directions. From kinematics the
strains at a distance z from the middle surface in the x and y directions,  x and  y ,
respectively, are

u   w  2w
x = = − z  = −z 2 (2.17a)
x x  x  x
v   w  2w
y = = − z  = −z 2 (2.17b)
y y  y  y

where u and v are the displacement components in the x and y directions.

Fig. 2.5 Plate bending element.

Constitutive Relations: From the generalized Hooke’s law

x y
x = − (2.18a)
E E
x y
 y = − + (2.18b)
E E

7
where  is the Poisson’s ratio. This leads to

x =
E
( x +  y ) (2.19a)
1 − 2
y =
E
( x +  y ) (2.19b)
1 − 2

Upon substitution from Eq. (2.17) into Eq. (2.19)

− Ez  2w 2w 
x =  2 +  2  (2.20a)
1 − 2  x y 
− Ez   2 w  2 w 
y =  +  (2.20b)
1 −  2  x 2 y 2 

Equilibrium: From equilibrium the generalized stresses, M x and M y , are related to the
stresses,  x and  y , as follows
h/2
Mx = 
−h / 2
x zdz (2.21a)

h/2
My = 
−h / 2
y zdz (2.21b)

In this regard, the twisting moments, M xy and M yx , and shearing forces, Q x and Q y , cannot
be considered as generalized stresses since the deformations associated with their
correspondent stresses are negligible. Upon substitution from Eq. (2.20) into Eq. (2.21) and
carrying out the integration

 2w 2w 
M x = − D 2 +  2  (2.22a)
 x y 
 2w 2w 
M y = − D 2 + 2  (2.22b)
 x y 
where
Eh 3
D=
( )
(2.23)
12 1 −  2

The generalized stresses, M x and M y can be related to the generalized strains (curvatures),
 x and  y , by solving Eqs. (2.16) and (2.22).

8
 1  
M x = D ( x + y ) = D  +  (2.24a)
 rx ry 
 
 1 
M y = D ( x +  y ) = D  +  (2.24b)
 rx ry 
 

2.2.4 Shell Element as a Further Extension

The most common shell structures are thin cylindrical shells under axisymmetric loading, as
in water tanks, towers… etc. Therefore, they are taken as an example of shell structures to
show how generalized stresses and generalized strains are derived and related.

For this element, Fig. 2.6, there are three generalized strains and three corresponding
generalized stresses; (1) the radial strain (which is the tangential strain of the middle surface),
  , and its corresponding tangential force, N ; (2) the curvature of the cylinder wall in the x
direction,  x , and the corresponding moment, M x ; and (3) the curvature of the cylinder wall
in the tangential direction,  , and the corresponding moment, M  .

Fig. 2.6 Element of a circular cylindrical shell.

Equilibrium: Upon considering the equilibrium of the shell element in Fig. 2.6 in the x, y
directions and the moment about the y-axis (and neglecting higher order terms) the following
relations can be obtained.

 dN x 
 dx dx  rd = 0 (2.25)
 
dQx N
+ =p (2.26)
dx r
dM x
− Qx = 0 (2.27)
dx

9
Eq. (2.25) means that N x should be equal to either a constant value or zero; in this derivation
it is considered equal to zero. Upon differentiating Eq. (2.27) with respect to x and
substituting from Eq. (2.26) the following can be obtained.

d 2 M x N
+ =p (2.28)
dx 2 r

Kinematics for Membrane Action: The kinematical relations and constitutive relations are
discussed for membrane action first and later for bending behavior. Define the membrane
strains in the x and radial directions as  x and   , respectively, and the displacements in the x
and radial directions as u and w , respectively, Fig. 2.7. Then,

du
x = (2.29)
dx
w
 = (2.30)
r

Constitutive Relations: From the generalized Hooke’s law

x  Nx N 1
x = − = −  = ( N x − N ) (2.31)
E E Et Et Et

x  N x N 1
 = − + = − + = ( − N x + N ) (2.32)
E E Et Et Et

Fig. 2.7 Radial deformation of a

circular cylindrical shell element.

From Eqs. (2.31) and (2.32) the following values of N x and N in terms of  x and   can be
obtained.

Et
Nx = ( x + ) (2.33)
1 − 2
Et
N = ( x +  ) (2.34)
1 − 2

10
For an open cylinder N x = 0 ; i.e.,

 x = − (2.35)
Then,
N = Et (2.36)

Fig. 2.8 Bending deformation of a circular cylindrical shell element.

Kinematics for Bending Action: For bending strain define the strains in the x and radial
directions as  xb and  b , and their corresponding stresses as  xb and  b , respectively. Based
on Kirchhoff hypothesis the strains  xb and  b are linear functions of the distance z from the
neutral axis, Fig. 2.8; that is,

 xb = z  x (2.37)

The notation  x , Fig. 2.8, is used here for the curvature of the cylinder wall in the x direction;
however, other notations such as k x may be commonly used in textbooks.

2w
x = (2.38)
x 2

In order to assess  b refer to Fig. 2.8,

11
  + b =
( r − z +w ) d  − ( r − z ) d  w  z   z 
 1 +    1 +  (2.39)
( r − z )d  r  r   r
Since ( z r ) 1 it can be neglected in the above equation; hence,

 b  0 (2.40)

This will also lead to

 + b  w
 = = = 0 (2.41)
r r r2

Constitutive Relations: From the generalized Hooke’s law

 xb =
E
1 − 2 (
 xb +b =
E
1 − 2
)
( z x ) (2.42)

Equilibrium: Thus, the generalized stress M x is

t 2 t 2
E Et 3
M x =   zdz = b
 x z dz = 12 1 − 2 x = D x
2
(2.43)
−t 2
1 − 2
x
−t 2 ( )
The stress  b can be obtained from Hooke’s law as follows

E
 b =
E
1 − 2 (
 xb + b =
E
1 − 2) xb = ( )
E
1 − 2 (
 z x ) =
1 − 2
( z x ) (2.44)

Hence, the moment M  is

E  Et 3
t 2 t 2

M  =    zdz = b
 x z dz = 12 (1 − 2 ) x =  D x
2
(2.45)
−t 2
1 − 2 −t 2

In summary, the generalized stresses, N and M x , and generalized strain,   and  x , are
related as follows,

N = Et (2.46)
M x = D x (2.47)

The curvature   0 ; therefore, the moment M  =  M x cannot not be considered as a

generalized stress.

12
2.2.5 Finite Element as a Recent Progress
Kinematics: In the finite element method, Fig. 2.9, formulation starts with the kinematical
(compatibility) conditions. In this step the generic displacements of an element (internal
displacements within an element), u, are related to the generalized strains (the nodal
displacements of the element), q  , by means of assumed shape functions, N  . This
assumption is equivalent to Bernoulli assumption in beams under bending (plane sections
perpendicular to the neutral axis before bending remain plane and perpendicular to the neutral
axis after bending) or Kirchhoff’s hypothesis in plates under bending.

u  =  N q  (2.48)

With the displacement within the element, the strain vector,  , at any point within the
element can be obtained by differentiation of u in Eq. (2.48).

  =  B q  (2.49)

where B  is composed of the derivatives of the shape function, N  .

Fig. 2.9 Sample finite element.

Equilibrium: The next step is to impose equilibrium condition in order to obtain the relation
between the generalized stresses (nodal force vector), F  , and the internal stress vector at
any point,  . This step can be achieved upon the application of the principle of virtual
work. This demonstrates the fact that the principle of virtual work is nothing other than a
principle of equilibrium. Thus,

F  =  BT  dV (2.50)

Constitutive Relations: The third step is to impose the constitutive relations which can be
expressed in the following general form

  = C   (2.51)

13
where C  is called the elastic constitutive or elastic moduli matrix. Upon substitution from
(2.49) and (2.51) into (2.50)

F  =  k q  (2.52)

where

k  =  BT C BdV (2.53)

Thus, the formulation of generalized stresses and generalized strains and their relation
in the finite element is the same as other structure elements such as bars or plates, except in
the application of equilibrium and compatibility. Equilibrium is justified through the
principle of virtual displacement and compatibility is achieved through the assumed shape
function.

2.3 Theory of Structures

2.3.1 Generalized Stress-Generalized Strain Relations of Beam Member
In the preceding section it has been illustrated how the generalized stress (moment), M , and
generalized strain (curvature),  , of a beam element are related. This relation is utilized to
develop the relation between the generalized stress (end moment), M , and the generalized
strain (end rotation),  , of a beam member.

No Sway: Fig. 2.10a illustrates a beam member AB with two end moments (generalized
stresses), M A and M B , associated with end rotations (generalized strains),  A and  B . The
moment at section located at a distance x from the origin of coordinates, M ( x) , can be
written as

 M + MB 
M ( x) = − M A +  A x (2.54)
 L 

From the generalized stress-generalized strain relation of a beam element, the curvature at
section x,  (x) , is

M ( x) 1  M A + M B  
 ( x) = =  x − MA (2.55)
EI EI  L  

14
Fig. 2.10 Beam member subjected to end moments and sway.

The curvature,  (x) , is related to the deflection, v(x) , and hence the moment as
follows

d 2 v( x) 1  M A + M B  
=  ( x) =   x− MA (2.56)
dx 2
EI  L  

Upon integrating Eq. (2.56) twice, and introducing the boundary conditions, v x =0 = 0 and
v x = L = 0 , the integration constants can determined. Then, the equation of deflection will be

1  M A + M B  3 M A 2  2M A − M B  
v( x) =  x − x +  Lx  (2.57)
EI  6L  2  6  

The equation of rotation,  (x) , will be

1  M A + M B  2  2M A − M B  
 ( x) =   x − M Ax +   L (2.58)
EI  2L   6  

Applying Eq. (2.58) at member ends A and B

L
A = ( 2M A − M B ) (2.59a)
6 EI
and
L
B = ( − M A + 2M B ) (2.59b)
6 EI

Solving Eqs. (2.59) in terms of M A and M B leads to

15
 EI
MA = ( 4 A + 2 B ) (2.60a)
L
EI
MB = ( 4 B + 2 A ) (2.60b)
L

Eqs. (2.60) represent the generalized stress-generalized strain relation of a beam member with
no sway.

Sway: If the beam member experiences sway only, Fig. 2.10b, the generalized stress will be
the shear, Q , and the corresponding generalized strain will be the vertical displacement
(sway),  . For this case it can be easily proven, as derived for the case of no sway, that

12 EI
Q=  (2.61)
L3

QL
The shear, Q , will be associated with two end moments, M A = M B = − ; hence,
2
−6 EI
MA = MB =  (2.62)
L2

The relation in Eq. (2.62) can give advantage in limiting the number of variables to one
generalized stress. Thus, Eqs. (2.60) and (2.62) can be combined to give the following
generalized stress-generalized strain relations for the case of a beam member with sway, Fig.
2.10c.

EI  
MA =  4 A + 2 B − 6  (2.63a)
L  L
EI  
MB =  4 B + 2 A − 6  (2.63b)
L  L

Eqs. (2.60) or (2.63) are commonly known as the slope-deflection equations.

The concept of the generalized stresses and generalized strains has been widely used
along with the principle of superposition to develop many successful methods of structural
analysis such as the slope-deflection method and the moment distribution method. In
addition, it has been used to solve many mechanics problems such as column buckling. Such
a contribution is illustrated in the following sections.
2.3.2 Single Span Beam Problems
The solution of single span beam problems can be obtained from direct application of the
slope deflection equations and with the enforcement of the appropriate boundary conditions.
Examples of these problems are given in Fig. 2.11 with their solution. For the problem in Fig.
2.11c the end moment at A is set equal to zero in Eq. (2.60a) leading to  A = − B / 2 , which
upon substitution in Eq. (2.60b) gives the illustrated result. The moment at the fixed end of

16
the beams in Figs. 2.11a and 2.11b is the carry over moment, which is half the moment
applied at the other end (the end free to rotate); i.e., the carry over factor in this case is equal
to 0.5. On the other hand, in the beam of Fig. 2.11c the carry over moment and hence the
carry over factor are equal to zero.

Fig. 2.11 Examples of single span beam problems.

For the simply supported beam subjected to uniform load in Fig. 2.12a the end rotation
can be obtained by following the same solution procedure in the previous section for
obtaining the slope-deflection equations. The solution of the beam subjected to two end
moments in Fig. 2.12b can be obtained from the slope-deflection equations. Upon utilizing
the principle of superposition, the solutions of the previous two beams can be utilized to
obtain the fixed end moments of the beam in Fig. 2.12c through the enforcement of the
boundary conditions. The same procedure can be followed to obtain the fixed end moments
for other cases.

Fig. 2.12 Fixed-end moments.

The concept of the generalized stresses and generalized strains has been widely used
along with the principle of superposition to develop many successful methods of structural
analysis such as the slope-deflection method and the moment distribution method. In

17
addition, it has been used to solve many mechanics problems such as column bucking. Such a
contribution is illustrated next.
2.3.3 Slope-Deflection Method for Simple Structures
The slope-deflection method is one of the successful methods used for the analysis of
statically indeterminate skeletal structures and for other applications. It is a form of the
stiffness method where the unknowns are the displacements of the joints; i.e., the size of the
problem depends on the number of unspecified degrees of freedom (degree of kinematic
indeterminacy). The application of the method is illustrated in the following example.

The continuous beam shown in Fig. 2.13 has two degrees of freedom (the rotations at
B and C), which means that the solution leads to the formation of two simultaneous linear
equations in  B and  c . The solution starts with writing the slope-deflection equations of
each span as an isolated beam. In these equations we write the moment at each joint of the
span, which is the summation of the fixed end moment due to applied loads plus the moment
associated with the end rotations or sway (or settlement). Then, we apply equilibrium at
joints B and C (summation of end moments at each joint is equal to the external moment
applied at the joint, which is zero in this case). Thus, two simultaneous linear equations in  B
and  c are derived; and hence, these joint rotations are obtained. Consequently, the end
moments can be calculated followed by the end shear.

Fig. 2.13 Example of a continuous beam.

The same procedure of slope-deflection method used in beams applies to frames

except in frames with sway where equilibrium equation(s) is written for the sway. For
example, for the frame in Fig. 2.14 the number of degrees of freedom is three (the rotations at
B and C and the lateral sway,  B or C ). The sway equation is obtained from the
equilibrium of the forces in the direction of the sway; in this case the external horizontal loads
and the shear at A and D. The shear is obtained from member equilibrium.

Fig. 2.14 Example of a frame with a sway.

18
 In this method it is obvious that the equations of the individual spans are developed
from the three basic conditions; equilibrium, compatibility and constitutive relations, as
illustrated earlier. Also, in the complementary steps of the solution, compatibility was the
initiation of the solution, the rotation at any joint is controlled, next the formation of the
slope-deflection equations are obtained by applying the same three sets of conditions. In
addition, equilibrium is enforced in order to obtain the simultaneous linear equations of the
unknown rotations. Nevertheless, as a result of adopting the generalized stress and
generalized strain concept the solution is much easier here in comparison with the continuum
mechanics approach.
2.3.4 Moment-Distribution Method for Frame Structures
The moment-distribution method is another successful method for the analysis of
statically indeterminate beams and frames. It is a form of the displacement method, in which
equilibrium, compatibility and constitutive conditions are utilized to obtain the properties of
individual members with the aid of the generalized stress and generalized strain concept.
These properties include the fixed end moments, stiffness and carryover factor, as illustrated
in the preceding section. The rigid joints are assumed locked and the corresponding moments
(fixed end moments) are calculated. Then, the out-of-balance moment at each joint is
calculated and distributed between the members connected to the joint according to their
stiffness. This step is followed by transferring a percentage of the correction moment
between the joints of each member according to the carryover factor. Afterwards, the joints
balance is checked and another adjustment of the moments connected to the joints is achieved
followed by the carryover moment. This last iterative procedure is repeated until convergence
takes place.
2.3.5 Structural Design
It has been illustrated how the three sets of conditions, equilibrium, compatibility and
constitutive conditions, are utilized to develop successful methods for structural analysis.
With the introduction of the concept of the generalized stress and generalized strain the effort
of analysis has been remarkably reduced. Two variables replaced the continuum mechanics
stress and strain tensors and the constitutive relations with the implementation of Bernoulli’s
hypothesis. The solution will always search for one unknown, M , and the other unknown
comes from the generalized stress-generalized strain relation.

After performing structural analysis and obtaining the internal forces the stress,  ,
can be calculated and checked against the allowable value,  all .

M
= y   all (2.64)
I

The allowable value is provided by the code and is based on real tests. This value should
guarantee safety and in order to do so it accounts for residual stresses and other environmental
factors. The code is meant by other issues such as how to avoid local buckling, bracing… etc.

19
2.4 Theory of Structural Stability
2.4.1 Generalized Stress-Generalized Strain Relations of Beam-Column
Member
Similar to the case of beam member, the relation between generalized stress (end moment),
M , and generalized strain (end rotation),  , of a beam-column member can be developed.
For the framed member in Fig. 2.15a, upon writing the equilibrium relations in the deformed
geometry the following relations can be obtained for the case of no sway.

MA =
EI
L
( sii A + sij B ) (2.65a)

MB =
EI
L
( s ji A + s jj B ) (2.65b)

where the stability functions, sii , sij , s ji and s jj , are as given in Table 2.1. In case there is a
sway of value  , Fig. 2.15b, the generalized stress-generalized strain relations will be

EI  
MA =  sii A + sij B − ( sii + sij )  (2.66a)
L  L

EI  
MB =  s ji A + s jj B − ( s ji + s jj )  (2.66b)
L  L

Eqs. (2.65) or (2.66) are commonly referred to as the slope-deflection equations.

Fig. 2.15 Beam-column member subjected to end moments and sway.

20
 Table 2.1 Stability functions for a beam-column
Function Condition of axial load, P
Compression Tension Zero
sii = s jj kL sin kL − ( kL ) cos kL ( kL ) cosh kL − kL sinh kL
2 2 4
2 − 2cos kL − kL sin kL 2 − 2 cosh kL + kL sinh kL

sij = s ji ( kL )
2
− kL sin kL kL sinh kL − ( kL )
2 2
2 − 2 cos kL − kL sin kL 2 − 2cosh kL + kL sinh kL

P
k=
EI

2.4.2 Buckling Analysis of Structural Members

The generalized stress-generalized strain relations of a beam-column member (or slope-
deflection equations) are the analytical tool of beam-column. This is illustrated by the
following three examples.

Example (2.4.1): This example, Fig. 2.16a, is a beam-column with one fixed end and one
hinged end. In this member the rotation at end A is zero; and therefore, the slope-deflection
equations at ends A and B are

MA =
EI
L
( sij B ) (2.67a)

MB =
EI
L
( s jj B ) (2.67b)

The moment at joint B, M B = 0 ; hence, from Eq. (2.67b) and since  B  0

s jj = 0
or
kL sin kL − ( kL ) cos kL
2

=0 (2.68)
2 − 2 cos kL − kL sin kL

Since kL  0 , Eq. (2.68) is simplified to the following characteristic equation of the beam-
column

tan kL − kL = 0 (2.69)

The value of kL which satisfies Eq. (2.69) is

kL = 4.4934
Thus,

21
 PL2
k 2 L2 = 20.19 = (2.70)
EI

i.e., the critical load of the beam-column is

20.19 EI  2 EI
Pcr = = (2.71)
( 0.7 L )
2
L2

Fig. 2.16 Beam-column with one fixed end and one hinged end.

The critical load of this member could have been obtained by solving the differential
equation of equilibrium. The equilibrium equation of the segment AC, Fig. 2.16c, of this
member is

 x 
−M int + Py − M F 1 − =0 (2.72)
 L 
Since
M int = − EIy '' (2.73)

Thus, the equilibrium equation can be written as

MF  x 
y '' + k 2 y − 1 − =0 (2.74)
EI  L 

22
which has a general solution in the form

MF  x 
y (x ) = C 1 sin kx + C 2 cos kx + 1 −  (2.75)
P  L

After imposing the boundary conditions, Eq. (2.75) would lead to the same characteristic
equation of the member, Eq. (2.69).

Example (2.4.2): This example, Fig. 2.17, is a beam-column with one fixed end and one
guided end. In this member the rotation at both ends A and B is zero; and therefore, the
slope-deflection equations at ends A and B are

 −  EI 
 − ( sii + sij ) L  = L2 ( sii + sij )
EI
MA = MB = (2.76)
L  

From the member equilibrium

M A + M B = P (2.77)

Substituting from Eq. (2.76) into Eq. (2.77) and simplifying

PL2
= sii + sij
2 EI
or
k 2 L2
= sii + sij (2.78)
2

Upon substitution for the stability functions into Eq. (2.78) and simplification, the following
is obtained.

kL sin kL = 0 (2.79)

The lowest value of kL which satisfies Eq. (2.79) is

kL = 
Then,
PL2
k 2 L2 =  2 = (2.80)
EI

i.e., the critical load of the beam-column is

 2 EI
Pcr = (2.81)
L2

23
Fig. 2.17 Beam-column with one fixed end and one guided end.

The critical load of this member can be obtained by solving the differential equation of
equilibrium. The equilibrium equation of one segment of this member is

P
M int − + Py = 0 (2.82)
2
Since
M int = EIy '' (2.83)

thus, the equilibrium equation can be written as

P
y '' + k 2 y = (2.84)
2 EI

which has a general solution in the form


y ( x) = C1 sin kx + C2 cos kx + (2.85)
2

After imposing the boundary conditions, Eq. (2.85) would lead to the same characteristic
equation of the member, Eq. (2.79).

Example (2.4.3): This example, Fig. 2.18, is a continuous beam-column with two outer ends
fixed and an interior joint that prevents lateral displacement. Both column spans have the
same flexural rigidity, EI , but they are of different lengths, 0.5L and L . The rotations at
ends A and C are zero; therefore, the slope-deflection equations of the moments at B are
M BA =
EI
0.5L
( )
siiBA B =
L
(
2 EI BA
sii  B) (2.86a)

M BC =
L
( sii  B )
EI BC
(2.86b)

From equilibrium at joint B

M BA + M BC = 0 (2.87)

Equations (2.86) and (2.87) yield the following relation

24
 2 siiBA + siiBC = 0 (2.88)

From Table 2.1

0.5kL sin 0.5kL − ( 0.5kL ) cos 0.5kL

2

s BA
= (2.89a)
2 − 2 cos 0.5kL − 0.5kL sin 0.5kL
ii

kL sin kL − ( kL ) cos kL
2

s BC
= (2.89b)
2 − 2 cos kL − kL sin kL
ii

Upon substitution for siiBA and siiBC into Eq. (2.88) and simplification the characteristic
equation can be obtained, which upon solution yields

kL = 5.412

This gives the critical load of this continuous beam-column

2.97 2 EI
Pcr = (2.90)
L2

Fig. 2.18 Continuous beam-column example.

Euler Load: The critical load of any beam-column of different boundary conditions can be
determined by following the previous procedure using the slope-deflection equations.
Alternatively, a solution can be obtained by forming the differential equation of equilibrium
then solving the equation with imposing the appropriate boundary conditions. One of the
interesting cases is a column of two hinged ends, Fig. 2.19, for which the critical load is

 2 EI
Pcr = (2.91)
L2

The critical load of a two hinged ends column is called Euler load, PE ; i.e.,

 2 EI
PE = (2.92)
L2

25
Fig. 2.19 Pin ended column.

2.4.3 The K Factor and the Alignment Charts

K Factor: The critical load of any beam-column of any boundary conditions, Pcr , can be
written in a unified form as follows

Pcr =  2 EI L2b (2.93)

where Lb is the buckling length of the column, which is the distance between the inflection
points of the beam-column in its deformed shape as a pin-pin column. From Eqs. (2.92) and
(2.93), the column critical load, Pcr , can be related to Euler load, PE , as follows,

PE P
Pcr = = E2 (2.94)
(Lb L ) K
where the factor K is the ratio between the column buckling length, Lb , and its length, L .
( 0.7 )
2
The value of K of the beam-column in Fig. 2.16a is 0.7 ( Pcr = 2.041PE = PE or
Lb = 0.7 L ), while K is equal to 1.0 for the beam-column in Fig. 2.17 and off course for the
pin ended column. The different values of K , Table 2.2, is given for beam-columns of
different boundary conditions, and it ranges from 0.5 for a braced fixed ended column to 2.0
for a column of one fixed end and one free end. The value of K of the continuous beam-
column in Fig. 2.18 is 0.58 ( Pcr = 2.97 PE = PE ( 0.58 ) or Lb = 0.58L ), which is
2

approximately the average value of cases (a) and (b) in Table 2.2.

26
Table 2.2 Values of K of Idealized Columns

In the design of a structural system, where columns are not isolated members, the
assessment of the critical load should follow different approach. For instance, the critical load
of a system can be determined by the slope-deflection equations or a second-order analysis.
On the other hand, the value of the K factor can be assessed with the aid of the alignment
charts which are derived from the slope-deflection method.

Alignment Charts: The alignment charts have been developed by Julian and Lawrence
(1959) for a simplified assessment of the K factor. In the development of the charts the two
models in Fig. 2.20 are used for a system that is braced against side sway and an unbraced
system. In both models the following assumptions are adopted:
1. All columns are prismatic and behave elastically.
2. The axial forces in the beams are negligible.
3. All columns in a story buckle simultaneously.
4. At a joint, the restraining moment provided by the beams is distributed among the
columns in proportion to their stiffnesses.
5. At buckling the girders are bent as shown in the models.
In both models in Fig. 2.20 column c2 (member AB) is the column in question. After
writing the slope-deflection equations of either model and applying equilibrium at joints A
and B a relation of K is obtained in terms of two end stiffness parameters of the column, GA
and GB , where

(I L )c
 of column stiffnesses meeting at joint A
GA = A
= (2.95a)
(I
A
L )b  of beam stiffnesses meeting at joint A
(I L )c
 of column stiffnesses meeting at joint B
GB = B
= (2.95b)
(I
B
L )b  of beam stiffnesses meeting at joint B

27
 (a) Sub-assemblage model for braced frame (b) Sub-assemblage model for unbraced frame
Fig. 2.20 Sub-assemblage models.

The indices b and c stand for beam and column, respectively, and the indices A and B stand
for joints A and B, respectively. The obtained relation of K is expressed in a monograph
form (alignment chart) for both cases of bracing conditions, Fig. 2.21.

(a) Braced System (b) Unbraced System

28
Fig. 2.21 Alignment Charts.

2.4.4 Stability Analysis of Framed Structures

The application of the slope-deflection method for stability analysis of framed system is
illustrated by the following two examples, Fig. 2.22. The first example is a frame where sway
is not allowed and in the second example is a frame where sway is allowed.

Example (2.4.4): In this example it is required to determine the collapse load of the portal
frame shown in Fig. 2.22a, where sway is assumed to be prevented. The deflected shape of
the frame is assumed as illustrated by the dashed lines. Upon writing the slope-deflection
equations for the column joints A and B and eliminating  A since M A = 0 the following
equation is obtained,
EI  
2
sijc
M BA = c  siic −  B (2.96)
Lc  siic 

The index c stands for column. Writing the slope-deflection equations for the beam joints B
and C and eliminating  C since  C = − B the following equation is obtained,

M BC = b ( siib − sijb ) B
EI
(2.97)
Lb
The index b stands for beam. Assuming that the axial force in the beam is so small that it can
be neglected; then, siib = 4 and sijb = 2 ; i.e.,
2 EI b
M BC = B (2.98)
Lb

Fig. 2.22 Examples of framed systems.

Upon applying equilibrium at joint B,

EI c  
2
sijc 2 EI b
M BA + M BC =  iic
s −   B + B = 0
Lc  siic  Lb

29
Leading to
EI c   2 EI b
2
sijc
 siic −  + =0 (2.99)
Lc  siic  Lb

This equation is the characteristic equation of the frame. For the special case I c = I b and
Lc = Lb , Eq. (2.99) becomes
2
sijc
siic − +2=0 (2.100)
siic
By trial and error the value of kL that satisfies Eq. (2.100) is
P
kL = L = 3.59
EI
which gives a value of the critical load

EI
Pcr = 12.9 (2.101)
L2

Example (2.4.5): In this example it is required to determine the collapse load of the portal
frame shown in Fig. 2.22b, where sway is allowed. The deflected shape of the frame is
assumed as illustrated by the dashed lines. Upon writing the slope-deflection equations for
the column joints A and B and eliminating  A since M A = 0 the following equation is
obtained,
EI    
2 2
sijc sijc
M BA = c  siic −   −  s −   (2.102)
Lc  siic  
B iic
 siic L
 
c 
Writing the slope-deflection equations for the beam joints B and C and after the elimination of
 C ; since the beam is in double curvature  C =  B , the following equation is obtained,

M BC = b ( siib + sijb ) B
EI
(2.103)
Lb
Assuming that the axial force in the beam is so small that it can be neglected; then, siib = 4
and sijb = 2 ; i.e.,
6 EI b
M BC = B (2.104)
Lb
Upon applying equilibrium at joint B,

EI  2
sijc   2
sijc    6 EI b
M BA + M BC = c  siic −   B −  siic −   + B = 0
Lc  siic   siic  Lc  Lb
Leading to
 s ijc2 I L   2
s ijc 
 s iic − + 6 b c B  B −  s iic −  = 0 (2.105)
 s iic I c Lb   s iic  Lc

30
From the story shear equilibrium of the frame

M AB + M BA + P M CD + M DC + P
+ =0 (2.106)
Lc Lc
It should be noted that

M AB = M DC = 0 (hinged) (2.107a)
M CD = M BA (anti-symmetry) (2.107b)

Upon substitution from the last equation in the shear equation in terms of M BA and
substitution for M BA from Eq. (2.102) and simplification, the following can be obtained,
 2
sijc   2
sijc P 2 
 siic −   B −  siic − − Lc  = 0
L
(2.108)
 siic   siic EI c  c

Eqs. (2.105) and (2.108) are the frame equilibrium equations and they can be written in matrix
form as follows

 I b Lc 
S + 6 I L −S   B 
 c b
   = 0
 2 
PL   0 (2.109)
−S S − c   Lc   
 
 EI c 
where
2
sijc
S = siic − (2.110)
siic
For the special case I c = I b = I and Lc = Lb = L , Eq. (2.109) becomes
 
S + 6 −S   B   0 
 − S S − k 2 L2     =  0  (2.111)
    
L
where k = P EI . At bifurcation both  B and  increase without bound and therefore in
2

order for Eq. (2.111) to be valid

S +6 −S
=0 (2.112)
−S S − k 2 L2

Eq. (2.112) is the characteristic equation of the frame. By trial and error the value of kL that
satisfies this equation is

p
kL = L = 1.35
EI

31
which gives a value of the critical load

EI
Pcr = 1.82 (2.113)
L2

In the preceding examples the slope-deflection method has been used to determine the
critical load of a framed system. In case this calculation regime is difficult to perform as in
case of large structures codes of practice adopt approximate methods in order to account for
the second-order stability effects. These methods are commonly known as the amplification
(or magnification) methods, which are presented next.
2.4.5 Amplification Factors for Second-Order Effects
With reference to Fig. 2.23 structural stability has two fundamental effects; the first is the P-
effect, and the second is the P- effect. The first effect may have a softening or stiffening
effect on structural members; nevertheless, such effect is generally detrimental for columns.
With reference to Fig. 2.23a, the P- effect is associated with additional loads on columns due
to the interaction between the load and nodal displacements (the shift of the load from its
original position because of structural deformations). The lateral force, H , acting on the
frame alone cause a lateral drift,  ; however, with presence of vertical forces,  P , this drift
and the overturning moment will increase further as a result of the interaction between the
vertical forces and the lateral drift.

The member stability (P- effect), Fig. 2.23b, will have a softening effect on a slender
member if P is compression and a stiffening effect if P is tension. The softening effect arises
from the additional moments due to the interaction between the axial compressive force and
the lateral deflection of the member, . If the axial force is tension the interaction between
this force and the member lateral deflection will reduce the moments along the column span.
Usually columns are subjected to compression rather than tension and therefore the P- effect
would be critical.

Fig. 2.23 P- effect and P- effect.

32
 In order to approximately assess the stability effects without having to carry out a
slope-deflection analysis or a second-order analysis, the moment amplification (or
magnification) has been developed. In this method the P- effect and P- effect are estimated
independently as illustrated in the following.

P- Effect: When the lateral forces H act on a frame and cause a primary lateral
deflection,  i , which can be obtained from a first-order analysis, the vertical forces will
interact with this deflection. As a result, additional lateral deflection and additional moments
take place, which is called the P- effect. This effect can be estimated by different methods
such as the story magnifier method and the multiple-column magnifier method (Chen and Lui,
1991). In the former method, it is assumed that each story behaves independently of other
stories and that the additional moment in the columns caused by the P- effect is equivalent to
that caused by a lateral force of P / h ( h is the story height). Hence, the sway stiffness of
the story can be determined as

H H + P / h
SF = = (2.114)
i 
where H is the sum of all story horizontal forces producing  i which is the first order
translational deflection of the story under consideration, P is the axial loads on all columns
in that story. Solving Eq. (2.114) for the total deflection 

 1 
=  i (2.115)
 1 − Pi / Hh 

Since every story is assumed to behave independently of other stories, the sway moment as a
result of the story swaying is proportional to the lateral deflection of the story. Therefore, the
total moment due to sway and second order effect (in terms of the primary story moment,
M isway ) is

 1 
M =  M isway (2. 116)
 1 − Pi / Hh 

P- Effect: In order to estimate the P- effect the beam-column in Fig. 2.24 is considered.
The vertical loads and end moments produce primary moment, M i , and primary deflection,
vi . As a result of the interaction between the axial force, P , and the primary deflection,
additional moment, M ii , and additional deflection, vii , which are called the P- effect, take
place. The total moment, M , and total deflection, v , are thus,

M = M i + M ii (2.117a)
v = vi + vii (2.117b)

33
For an approximate assessment of this effect (Chen and Lui, 1991), assume that M ii has the
shape of half sine wave and that vmax (=  =  i +  ii ) occurs at midspan,
x
M ii = P sin (2.118)
L
The moment M ii is related to the second derivative of vii
M
vii'' = − ii (2.119)
EI
The negative sign in Eq. (2.119) is because the moment increases while the slope decreases
and vice versa. From Eqs. (2.118) and (2.119)
P x
vii'' = − sin (2.120)
EI L
Upon double integration of Eq. (2.120) and introducing the boundary conditions ( vii = 0 at
x = 0, L ) and calculating  ii = vii at midspan,
P
 ii = − (2.121)
PE
Since
 =  i +  ii (2.122)
Substituting from Eq. (2.121) into Eq. (2.122) and solving for 

1
= i (2.123)
1 − P PE

Fig. 2.24 P- effect.

If the maximum primary moment is assumed to take place in the midspan the following
equation for the maximum moment can be obtained.

34
 1 + P PE
M max = M i max (2.124)
1 − P PE
where
 i PE
= (2.125)
M i max

Equation (2.124) can be written in the following form

Cm
M max = M i max (2.126)
1 − P PE

where
Cm = 1 + P PE (2.127)

It should be mentioned that the value of C m in Eq. (2.127) is applicable only when the
maximum primary moment, M i max , takes place at or near midspan.

Example 2.4.6: For demonstration, the amplification factor methods are applied to the frame
shown in Fig. 2.25. The axial rigidity and flexural rigidity of all frame members have the
same values, EA = 2.619 106 kN and EI = 3.852 104 kN .m 4 , respectively. From the first-
order analysis, the lateral displacements of joints B and C are  B = 26.41mm and
C = 26.22mm respectively, and the maximum deflection of member AB is
 iAB = 7.33mm and of member DC is  iDC = 10.373mm .

Fig. 2.25 Frame example 2.4.6.

For the P- effect, P  i / Hh = 0.05263 . From Eq. (2.116) M = 1.056M isway ; i.e.,
the end moment of column AB at B is, M = 1.056 115.3 = 121.76kN .m , and the end
moment of column DC at C is, M = 1.056  249.8 = 263.8kN .m .

35
 For the P- effect, PE = 152.2 102 kN for either column AB or column DC. For
Column AB, P PE = 0.0123 and M i max 90.4kN .m . From Eq. (2.125),  AB = 1.234 ; thus
from Eq. (2.124) M max = 1.028M i max . For Column DC, P PE = 0.014 and
M i max 125.0kN .m ,  DC = 1.263 and M max = 1.0282M i max .

2.5 Theory of Plates

The generalized stress-generalized strain relations of thin plate element under bending have
been derived from the three sets of conditions; equilibrium, compatibility and constitutive
law. By this, a major simplification has been introduced to the solution of the bending
problem; i.e., dealing with two variables instead of fifteen variables and treating a whole
element instead of a field point. Expanding the discussion of this topic by considering the
equilibrium of the element shown in Fig. 2.26, the following differential equation of
equilibrium can be derived.

Fig. 2.26 Plate element and stress resultants.

4w 4w 4w p

+ 2 + = (2.128)
x 4 x 2y 2 y 4 D

where p is the load intensity.

36
 In this section the concept of generalized stress and generalized strain with the three
sets of conditions are employed to derive different solutions of thin plate bending problems.

The simply supported plate shown in Fig. 2.27 is subjected to a sinusoidal load given
by
x y
p ( x, y ) = po sin sin (2.129)
a b

where a and b are the plate dimensions in the x and y directions, respectively. Upon
substitution of the load formula in the differential equation of equilibrium

Fig. 2.27 Simply supported plate.

4w 4w  4 w po x y
+ 2 + = sin sin (2.130)
x 4
x y
2 2
y 4
D a b

From kinematics,

2w
w = 0 and = 0 at x = 0 and x = a ` (2.131a)
x 2
2w
w = 0 and = 0 at y = 0 and y = b (2.131b)
y 2

In order to solve this plate problem the deflection expression must satisfy the aforementioned
kinematic relations and it should have the same form as the load expression. Hence, the
deflection should have the following form

x y
w = C sin sin (2.132)
a b

where C is a constant that can be obtained upon substitution from Eq. (2.132) into Eq.
(2.130).

po
C = 2
(2.133)
 1 1 
D  2 + 2 
4

a b 

37
or
po x y
w = 2
sin sin (2.134)
 1 1  a b
D  2 + 2 
4

a b 

The stress resultants, Q x , Qy , M x , M y , M xy and M yx can be obtained from

  2w 2w 
Q x = −D  + 
x  x 2 y 2 
  2w 2w 
Q y = −D  + 
y  x 2 y2 
 2w 2w 
M x = −D  2 + 2  (2.135)
 x y 
 2w 2w 
M y = −D  + 2 
 x y 
2

 2w 
M xy = (1 − )D   = −M yx
 xy 

Of course the strains and stresses at any point can be obtained if needed.

2w
 x = −z 2
x
2w
 y = −z 2
y
2w
 xy = −2 z
xy
x =
E
( x +  y ) (2.136)
1 − 2
y =
E
( x +  y )
1 − 2
E xy
 xy =
2(1 +  )

Let’s examine the same simply supported plate in Fig. 2.27 if subjected to a general
distributed load p ( x, y ) . The load can be represented by double Fourier series (Navier) as
follows

38
  
m x n y
p ( x, y ) =  pmn sin sin (2.137)
m =1n =1 a b

m / x n / y
In order to obtain pmn both sides of Eq. (2.137) are multiplied by sin sin ; then,
a b
both sides of the equation are integrated, which leads to

m x n y
a b
4
p mn =
ab   p ( x , y ) sin
0 0
a
sin
b
dxdy (2.138)

From the previous solution of the plate under sinusoidal load and upon applying the principle
of superposition, a solution of a simply supported plate under general loading condition can
be obtained
1   p mn m x n y
4 
w = sin sin (2.139)
(
D  m =1 n =1 ( m a )2 + ( n b )2
)
2
a b

If p ( x, y ) = po then

m x n y 16 po
a b
4 po
pmn =
ab   sin
0 0
a
sin
b
= 2
 mn
(2.140)

and
m x n y
  sin sin
16 p
w = 6 o  a b (2.141)
 D m =1 n =1
(
mn ( m a ) + ( n b )
2
)
2 2

m, n = 1,3,5,... since the terms of equations vanish for even values of m or n .

2.6 Theory of Shells

The generalized stress-generalized strain relations of thin cylindrical shell under axisymmetric
loading have been derived from the three sets of conditions; equilibrium, compatibility and
constitutive law. Thus, the number of variables is reduced from fifteen to three, which is a
remarkable simplification. In addition, a solution of a whole structural element or structure
can be obtained at once. The discussion is extended in this section in order to show how the
concept of generalized stress-generalized strain along with the three basic sets of conditions
are employed to obtain solutions of different problems of thin cylindrical shell under
axisymmetric loading. The discussion starts with the derivation of the general solution of this
class of problems.

The following equilibrium equations have been derived earlier in Section 2.2.4,

39
 dQx N
+ =p (2.142)
dx r
dM x
− Qx = 0 (2.143)
dx
Upon differentiating Eq. (2.143) with respect to x , and substitution for dQx dx from (2.142)
the following equation can be obtained
d 2 M x N
+ =p (2.144)
dx 2 r
Upon substitution from Eq. (2.46) for N , from Eq. (2.47) for M x and for  x =  2w x 2
into Eq. (2.144) the following differential equation of equilibrium can be obtained.

 4 w Etw p
+ = (2.145)
x 4 Dr 2 D
Define
Et
4 4 = (2.146)
Dr 2

Then, Eq. (2.145) can be written as follows,

4w p ( x)
+ 4 4 w = (2.147)
x 4
D

Eq. (2.147) is the differential equation of equilibrium of thin cylindrical shells subjected to
axisymmetric loading. The general solution of this equation has the form

w = e  x ( C1 cos  x + C2 sin  x ) + e −  x ( C3 cos  x + C4 sin  x ) + f ( x ) (2.148)

The term f ( x ) is the particular solution and the constants C1 to C 4 can be obtained from the
boundary conditions.

In the following the concept of generalized and stress-generalized strain along with the
three basic sets of conditions are utilized to obtain solutions for long thin cylindrical shell
under different loading conditions. In the following example a long cylinder is solved for a
shear force at the edge, Fig. 2.28a, and for a bending moment at the edge too, Fig. 2.28b. For
both problems the function f ( x ) is set equal to zero since there is no applied pressure. In
addition, the deflection associated with the term e  x tends to approach infinity as x becomes
a large value. This contradicts the real behavior of the cylinder under the shown edge load, in
which the deflection and moments dissipate as x increases. Therefore, the constants C1 and
C 2 of Eq. (2.148) should vanish; thus, the equation reduces to

40
Fig. 2.28 Example of long circular cylindrical shell.

w = e −  x ( C3 cos  x + C4 sin  x ) (2.149)

In order to obtain the constants C3 and C 4 the boundary conditions for each problem are
introduced. For the problem in Fig. 2.28a the conditions are:

M x ( x = 0) = 0 (2.150a)
dM x
Qx ( x = 0 ) = H o = (2.150b)
dx x =0

For the problem in Fig. 2.28b the conditions are:

M x ( x = 0) = M o (2.151a)
dM x
Qx ( x = 0 ) = 0 = (2.151b)
dx x =0

The solution of the problem in Fig. 2.28a is:

Ho
Mx = e−  x sin  x

Q x = 2H oe −  x cos (  x + 4 )
N = 2 H o  re −  x cos  x
M =  M x (2.152)
Ho − x
w= e cos  x
2 D 3
dw −Ho − x
= e sin (  x + 4 )
dx 2 D 2

The solution of the problem in Fig. 2.20b is:

41
 M x = 2M o e −  x cos (  x − 4 )
Qx = −2M o e −  x sin  x
N = −2 2M o  2 re −  x sin (  x − 4 )
M =  M x (2.153)
−M o −  x
w= e sin (  x − 4 )
2 D 2

dw M o −  x
= e sin (  x − 2 )
dx D

2.7 Finite Element

The finite element (FE) method is based on three concepts; (1) the concept of
generalized stresses and generalized strains; (2) equilibrium is justified through the principle
of virtual displacement; and (3) compatibility is achieved through the assumed shape function.
As a consequence, it has become possible to analyze any structure of any geometry and of
different materials.

In this section the solution of a few examples are presented in order to show (1) the
level of accuracy which can be reached using the FE method; and (2) how structural behavior
can be reflected in the FE solution and thus to verify some mechanics assumptions. In all
examples, the elasticity modulus, E = 25000MPa , and Poisson’s ratio,  = 0.2 .

The first example is a cantilever beam subjected to an end moment, Fig. 2.29, where
the exact solution can be obtained from mechanics (Section 2.3.1). The rotation,  ( x ) , is

M
 ( x) = ( x) (2.154)
EI
and the deflection, v(x) , is
M  x2 
v( x) =   (2.155)
EI  2 
The moment M = 1.44 108 N .mm and the flexural rigidity EI = 9  1012 N .mm 2 ; thus,
 ( x = L) = 0.0384rad and v( x = L) = 46.08mm . Hence, the vertical deflection at 1 and 2 is,
v1 = v2 = 46.08mm upward, and the horizontal displacement at 1, u1 = zero , while the
horizontal displacement at 2 is calculated from the rotation, u2 = −11.52mm . From the FE
solution of this problem it is found that, u1 = zero , u 2 = −11.51mm , v 1 = 46.12mm and
v 2 = 45.98mm . When these FE results are compared with the exact values it is evident how
much accurate the FE results are when using proper meshing and modeling. Also, from the
FE results the strain distribution at section 3 is almost linear, which conforms to the
simplification of Bernoulli hypothesis of beams.

42
Fig. 2.29 Example of a cantilever beam.

The second example, Fig. 2.30, is a beam with different values of shear span to depth
ratio, a d . In this example, the division line between a deep beam and a short beam can be
noticed from the strain distribution. From the FE results in Table 2.3, for a shear span to
depth ratio, a d  2.0 , the beam behaves as a deep beam, which agrees with St. Venant’s
principle. Also, the stress trajectories show the arch action in deep beams and the truss model
in ordinary beam.

Fig. 2.30 Example of a beam of different shear span to depth ratios.

Table 2.3 FE results of the example beam of different shear span to depth ratios, Fig. 2.30

43
ad Strain Distribution at Strain Distribution at Stress Trajectories
Section 1 Section 2

1.0
-150

-50

50

150

-450

-150

150

450
2.0 Strain (microstrain) Strain (microstrain)
-150

-50

50

150

-450

-150

150

450
3.0 Strain (microstrain) Strain (microstrain)
-150

-50

50

150

-450

-150

150

450

4.0 Strain (microstrain) Strain (microstrain)

-150

-50

50

150

-450

-150

150

450

The third
Strainexample
(microstrain) is a simply supported rectangular plate, Fig. 2.27, with dimensions,
Strain (microstrain)
a = 4.00m , b = 3.00m , and t = 100mm . The plate is subjected to a uniform load
p = 10.0kN / m 2 . The exact value of maximum deflection can be obtained from mechanics
(section 2.5) to be w = 2.54mm and from FE it is equal to 2.50mm .

It should be mentioned that the advantage of the FE is not only to analyze any problem
of any shape but also it can handle discontinuity problems; e.g., openings, bends and stress
concentration, and of irregular boundary conditions. The fourth example shown in Fig. 2.31
is a deep beam with a central opening. From the FE results illustrated in the figure the stress
distribution at the central section (section 1) is nonlinear and the stress trajectories agree with
the logic flow of forces. The same observations are valid for the deep beam with an eccentric

44
opening, Fig. 2.32. Another example shown in Fig. 2.33 is a cantilever with a dapped end,
where the illustrated FE results reflect the behavior of such discontinuity region.

Stress Distribution at Section 1 (kPa)

0

5000
-10000

-5000

10000
Stress Trajectories
Stress (kPa)

Fig. 2.31 Example of a deep beam with a central opening.

45
 Stress Distribution at Section 1 (kPa)

5000
-10000

-5000

10000
Stress Trajectories
Stress (kPa)

Fig. 2.32 Example of a deep beam with an eccentric opening.

200

Stress Trajectories
-400

-200

400
0

200
-400

-200

400

Strain (microstrain) Strain (microstrain)

Strain Distribution at Strain Distribution at
Section 1 (microstrain) Section 2 (microstrain)

Fig. 2.33 Example of a cantilever with a dapped end.

In conclusion, the FE can analyze any structure of any shape and of any boundary
conditions. If the FE can be considered “exact” which means the best possible solution it can
be used for calibration and verification; for instance, verification of kinematic assumptions
such as Bernoulli assumption. It can illustrate the limits of simplified models; for example,
beam model, and the reliability of simplification; e.g. Bernoulli hypothesis. The FE can
handle those problems where simplified models are not valid; for example, openings in
beams, abrupt change of thickness, discontinuity in general.

46
2.8 The Allowable Stress as a Basis for Design
There is a fundamental two-stage process in structural design operation: firstly, the forces
acting on each structural member must be defined, and secondly, the load-carrying capacity of
each member must be determined. The first stage involves an analysis of the stresses acting
within the structural members; the second involves knowledge of the load-carrying capacity
of the structural members. In allowable stress design, the first stage is based on a linear
elastic analysis while the second stage is based on full scale tests of structural members
reduced by an acceptable safety factor to the allowable stress level as specified by
specifications.

This was the era of great advancement in the use of the mathematical theory of
elasticity for structural design; but one placed too much emphasis on the elastic response of an
idealized version of the real structure to normal working loads. It turned the engineer so far
toward the analysis of the working load range that much more important task of insuring the
safety of a design against collapse or failure was submerged till the emerging development of
the theory of plasticity in the 1950’s.

In order to arrive to a reasonable estimate of the safety factor, SF , the two diagrams
in Fig. 2.34 are examined; one diagram is the idealized stress-stress relation based on linear
elasticity theory and the other is the actual behavior of the material. However, at the time of
the design method initiation, the material curve could be measured up to the first yield point,
while the strength beyond this point was not possible to predict and therefore it was neglected.
Hence, safety would have been related to this first yield point; i.e., the allowable stress,  all ,
would have been a fraction of this yield stress,  y .

y
 all = (2.156)
SF

The value of the safety factor would be based on the type of material (steel or
concrete, …), type of straining action (flexure, axial, shear, …), type of load and load
combination, type of structure, … etc. Values of safety factor were defined through time with
the accumulation of practical experience.

47
Fig. 2.34 Idealized and actual material stress-strain relation.

The allowable stress design method had been employed reasonably well in traditional
structures as a result of the accumulative experience particularly under normal loading
conditions. However, safety rules based on experience work well only for designs lying
within the scope of that experience. They cannot be relied on outside of that range, for
example, for the case of non-traditional structures; e.g., offshore structures, poles, nuclear
facilities… etc. In other words, structures susceptible to abnormal loads; e.g., extreme wind,
major earthquakes, explosive loads; …, etc, could not be properly handled with this method.
Therefore, it was a logical future development to establish more reliable and more rational
inelastic second-order analysis methods for design which can avoid the drawbacks of the
allowable stress design method. This further development of analysis and design methods
will be described in the era of plasticity chapter that follows.

2.9 Historical Sketch

Modern theory of elasticity began when the French mathematician Jacob Bernoulli combined
equilibrium equations with Hooke’s law (1678) to obtain the differential equation of the
elastica, in 1705, that is, the curve assumed by the deformed axis of the beam. 116 years after
Bernoulli, the general equations of equilibrium and vibration of elastic solids were made by
Navier in 1821, which led to the formulation of the linear theory of elasticity by A. Cauchy in
1822. This linear theory of elasticity remains virtually unchanged to the present day. During
this period following the discovery of Hooke’s law, the growth of the science of elasticity
proceeded from a synthesis of solutions of special problems. This began the development of
theory of flexural of beams, theory of torsion, theory of stability of columns, and some results
on bending and vibration of plates in early nineteenth century.

This chapter represents an attempt to present concisely several aspects of the theory of
elasticity from a unified point of view and to indicate those familiar methods of solution of
the field equations of elasticity that are familiar to structural engineers in particular. To this
end, we must single out and stress the contributions to the engineering theory by the Russian
elastician, a great teacher, in particular, the widely known work of great elegance and
importance by S. P. Timoshenko.

48
 It was during his years in American that Timoshenko made the major part of his
contributions and writings to the theory of elasticity and to its application to the design of
engineering structures and components. Timoshenko and his colleagues and students became
internationally known for their pioneering work in elasticity, especially those classical
textbooks used widely in engineering practice and university teaching in particular.
Timoshenko’s unparalleled monographs in the applications of elasticity to engineering
structures include:

1. Applied Elasticity, (with J. M. Lessells), D. Van Nostrand Company, Inc., New York,
1925.
2. Vibration Problems in Engineering, D. Van Nostrand Company, Inc., New York, 1928.
1937 and 1955 (with D. H. Young).
3. Strength of Materials, Part I, Elementary Theory and Problems, D. Van Nostrand
Company, Inc., Princeton, New Jersey, 1930, 1940 and 1955.
4. Strength of Materials, Part II, Advanced Theory and Problems, D. Van Nostrand
Company, Inc., Princeton, New Jersey, 1930, 1941 and 1956.
5. Theory of Elasticity, McGraw-Hill Book Company, New York, 1934 and 1951 (with J.
N. Goodier).
6. Elements of Strength of Materials, D. Van Nostrand Company, Inc., Princeton, New
Jersey, 1935, 1940 and 1949 (with G. H. McCullough), 1962 (with D. H. Young).
7. Theory of Elastic Stability, McGraw-Hill Book Company, New York, 1936 and 1961
(with J. M. Gere).
8. Engineering Mechanics, (with D. H. Young), McGraw-Hill Book Company, New York,
1937, 1940, 1951 and 1956.
9. Theory of Plates and Shells, McGraw-Hill Book Company, New York, 1940, 1959 (with
S. Woinowsky-Krieger).
10. Theory of Structures, (with D. H. Young), McGraw-Hill Book Company, New York,
1945 and 1965.
11. Advanced Dynamics, (with D. H. Young), McGraw-Hill Book Company, New York,
1948.
12. History of Strength of Materials, McGraw-Hill Book Company, New York, 1953.
13. Engineering Education in Russia, McGraw-Hill Book Company, New York, 1959.
14. The Collected Papers of Stephen P. Timoshenko, 1953, McGraw-Hill Book Company,
New York, USA.
15. As I Remember, D. Van Nostrand Company, Inc., Princeton, New Jersey, 1968.
16. Stephen Timoshenko 60th Anniversary Volume, Contributions to the Mechanics of
Solids dedicated to Stephen Timoshenko by his Friends, The MacMillan Company,
1938.

49
 S. P. Timoshenko was born on Dec. 22, 1878 in Ukrainia for parents who took a great
care of their children. He was self-motivated and since his childhood he had a dream to
become a structural engineer. After his graduation from real-gymnasium at the age of 18,
Timoshenko obtained most of his early education and practical experience in Russia. In his
early life he had two trips to Western Europe, which had a great influence on his life and
career. In his 1900 trip he visited Germany, Belgium, France and Switzerland, where he
realized how these countries were ahead of Russia in both culture and industrial development.
In his 1904 trip he was very much impressed with August Föppl at Munich Polytechnical
Institute.
In the academic year 1903-04 he read in English Love’s “Theory of Elasticity” and
Lord Rayleigh’s “Theory of Sound”. Those two books produced a great impression on him
and had a very considerable influence on his scientific work. Timoshenko went twice to the
University of Göttingen in Germany for a few months in 1905 and 1906, where he took some
courses in mathematics and mechanics and investigated the problem of lateral buckling of I-
beams. Timoshenko had his first teaching opportunity at Kiev Polytechnicum, where he
commenced lecturing in January 1907. During his whole teaching career, and different from
what was followed at his time, he used to start with the simplest problems and moves
gradually to the more complicated problems, which was much appreciated by students.

In 1920 Timoshenko escaped to Zagreb because of the long suffering for many years
under the Bolshevik regime and the 1st world war and because he was politically in danger. In
Zagreb he was satisfied with his scientific achievements; however, he decided to approach
academic life in the USA in 1922. Nevertheless, he had only a chance to work in industry,
first in a vibration company in Philadelphia for a few months and afterwards in Westinghouse
Company in Chicago. With many research accomplishments he stayed in Westinghouse for
five years until he had the opportunity to return to academic work as he always hoped. In
1927 he joined the University of Michigan where he achieved teaching and research work of
exceptional level such that his ideas and attitude towards mechanics became widespread
throughout the US. In 1936 he joined Stanford University until his formal retirement in 1944;
however, he continued his academic activities at Stanford but with a reduced scale and with
concentration on writing and revision of text books.

References
Chen, W. F., and Han, D. J., 1988, Plasticity for Structural Engineers, Springer-Verlag, New
York, USA.
Chen, W. F., and Lui, E. M., 1991, Stability Design of Steel Frames, CRC Press, Boca Raton,
Ann Arbor, USA.
Chen, W. F., and Lui, E. M., 1987, Structural Stability-Theory and Implementation, Elsevier,
New York, USA.
Jawad, M. H., 1994, Theory and Design of Plate and Shell Structures, Chapman & Hall, New
York, USA.
Kassimali, A., 2005, Structural Analysis, 3rd Ed., Thomson, Ontario, Canada.

50
Kelkar, V. S., and Sewell, R. T., 1987, Fundamentals of the Analysis and Design of Shell
Structures, Prentice-Hall, Inc., Englewood Cliffs, New Jersey, USA.
Livesly, R. K., 1975, Matrix Methods of Structural Analysis, 2nd Ed., Pergamon Press, New
York, USA.
Timoshenko, S. P., 1968, As I Remember, D. Van Nostrand Company, Inc., Princeton, New
Jersey, USA, Translated from Russian by R. Addis.
Timoshenko, S. P., 1953, History of Strength of Materials, McGraw-Hill Book Company,
New York, USA.
The Collected Papers of Stephen P. Timoshenko, 1953, McGraw-Hill Book Company, New
York, USA.
Timoshenko, S. P., and Goodier, J. N., 1970, Theory of Elasticity, 3rd Ed., McGraw-Hill
Book Company, New York, USA.
Timoshenko, S. P., and Woinowsky-Krieger, S., 1959, Theory of Plates and Shells, McGraw-
Hill Book Company, New York, USA.

51