ADVANCED STRUCTURAL ENGINEERING

(PLATES AND SHELLS)

By

Dr. DJOMO MBONG Thierry Landry Michel

Mechanics, Materials & Structures Specialist

And

Dr.-Engr. PENKA Jules Bertrand

Polytechnic Civil Engineer

Department of Civil Engineering and Architecture

(CEA)
National Higher Polytechnic Institute
(NAHPI)
The University of Bamenda

1
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 Objective:

Evaluation of stresses and deformations of plates and shells for a better design (i.e. good choices of
dimensions, materials and eventual reinforcements), and to avoid therefore failures of structures.

Applications:

Plates and shells are represented in many civil engineering buildings, bridges and realizations or
constructions.

 For Plates (flat surface structures) :

- Side panels and slabs - Roofs of buildings

- Tank Bottoms
- Walls

2
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
  For Shells (curved surface structures) :

- Vaults (as roofs)

- Domes (as roofs)

- Cylindrical culverts

- Tank surfaces

3
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 SECTION A:
PLATES

I- INTRODUCTION

1- Definition:

A plate is a flat structural element for which the thickness is small compared with the surface
dimensions. The thickness is usually constant but may be variable and is measured normal to the middle surface
of the plate (see fig.1).

Figure 1 : A plate
The bending properties of a plate depend greatly on its thickness h as compared with its other dimensions.
In the following discussion, we shall distinguish between three kinds of plates:
- Thin plates with small deflections,
- Thin plates with large deflections,
- Thick plates.

2- Plate Theory

a. Some Considerations

Let the plate mid-surface lie in the x  y plane and the z  axis be along the thickness direction, forming
a right handed set (see fig.2).

Figure 2 : Cartesian Axes

4
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 The stress components  ij acting on a typical
element of the plate are shown in Fig.3. The
representation for the strain components  ij is the same
as for the stress components.

b. Thin Plates with Small Deflection

According to the criterion often applied to define

a thin plate (for purposes of technical calculations), the
ratio of the thickness to the smaller span length should be
less than 1/20. Figure 3 : Stresses acting on a material element.

If deflections w of a plate are small in comparison with its thickness h , a very satisfactory approximate
theory of bending of the plate by lateral loads can be developed (called the plate theory) by making the following
assumptions:
(1) There is no deformation in the middle plane of the plate. This plane remains neutral during bending.
(2) Points of the plate lying initially on a normal-to-the-middle plane of the plate remain on the normal-
to-the-middle surface of the plate after bending (see fig.4).

Figure 4 : Deformed line elements remain perpendicular to the mid-plane.

(3) The normal stresses (or normal strains) in the direction transverse to the plate can be disregarded;
i.e.  zz  0 (or  zz  0 ).
The last supposition becomes unreliable in the vicinity of highly concentrated transverse loads. These
three assumptions also called the Kirchhoff hypothesis are the basis of the Classical Plate Theory. When the
deflections are not small, the bending of plates is accompanied by strain in the midplane, and assumption (1) is
inapplicable (Thin plates with large deflections).

c. Notation and Stress Resultants

Knowing that for plates, the stress components are allowed to vary through the thickness of the plate, so
the stress resultants are obtained by integrating the stresses through the thickness of the plate.
In general there will be:

Type of stress resultants Descriptions

Moments M Two bending moments and One twisting moment
Out-of-plane forces V Two shearing forces
In-plane forces N Two normal forces and One shear force

5
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
They are defined as follows:
 In-plane normal forces and bending moments (see fig.5) :

In-plane normal forces

h 2 h 2

Nx    xx dz; Ny    yy dz 1
h 2 h 2

Bending moments
h 2 h 2

Mx    z xx dz; M y    z yy dz  2 
h 2 h 2 Figure 5: In-plane normal forces and bending moments.

 In-plane shear force and twisting moment (see fig.6):

In-plane shear force

h 2

N xy    xy dz  3
h 2

Twisting moment
h 2

M xy   z xy dz  4
h 2 Figure 6: In-plane shear force and twisting moment.

 Out-of-plane shearing forces (see fig.7):

h 2

Vx   
h 2
 xz dz

 5
h 2

Vy   
h 2
zy dz

Figure 7: Out-of-plane shearing forces

Note that the above forces and moments are actually forces and moments per unit length. This allows
one to have moments varying across any section. If one considers an element with dimensions x and y ,
- The actual moments acting on the element are: M x  y , M y x , M xy x and M xy y
- The actual forces acting on the element are: N x  y , N y x , N xy x , N xy y , V x  y and V y x

II- THE MOMENT-CURVATURE EQUATIONS

1- From Beam Theory to Plate Theory

6
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 In the beam theory, based on the assumptions of plane sections remaining plane and that one can neglect
the transverse strain, the strain varies linearly through the thickness. In the notation of the beam, with y positive
up,  xx   y R , where R is the radius of curvature, R positive when the beam bends “up”. In terms of the
curvature  v x  1 R , where v is the deflection, one has:
2 2

 2v
 xx  y 2  6
x
The beam theory assumptions are essentially the same for the plate, leading to strains which are proportional to
distance from the neutral (mid-plane) surface, z , and expressions similar to Eq.(6). This leads again to linearly
varying stresses  xx and  yy (  zz is also taken to be zero, as in the beam theory).

2- Curvature and Twist

The plate is initially undeformed and flat with the mid-surface lying in the x  y plane. When deformed,
the mid-surface occupies the surface w  w  x , y  and w is the elevation above the x  y plane (see fig.8).
The slopes of the plate along the x and y directions are respectively  x   w  x and  y  w y .

Figure 8: Deformed Plate

a. Curvature

The curvature is defined as the reciprocal of the radius of

curvature of the elastic curve or as the rate of change of the slope
with respect to the arc of length of the elastic curve.
Therefore according to fig.9, the curvature  x in the x -
direction can be written as:
 x  1 Rx or  x  d ds
By replacing the slope  and the arc of length ds by their Figure 9: Angle and arc-length used in
expression, one finds that: the definition of curvature

 2 w x 2 1
x  3 2   7
1   w x   2 Rx
 
As with the beam, when the slope is small, one can take   tan  w x and  s   x and Eq.(7)
reduces to (and similarly for the curvature in the y direction):

7
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
  2w 1  2w 1
x  2  And y  2   8
x Rx y Ry

b. Twist

Not only does a plate curve up or down,

it can also twist. For example, shown in fig.10
is a plate undergoing a pure twisting (constant
applied twisting moments and no bending
moments).
If one takes a row of line elements lying
in the y direction, emanating from the x axis,
the further one moves along the x axis, the more
they twist (see fig.10).
Figure 10: A twisting plate
The twist can be defined analogously to
the curvature, and denoted by 1 Txy ; it is a measure of the “twistiness” of the plate:

1  2w
  9
Txy xy
The signs of the moments, radii of curvature and curvatures are illustrated in fig.11. Note that the
deflection w may or may not be of the same sign as the curvature. Note also that when M x  0 (respectively
M y  0 ),  2 w x 2  0 (respectively  2 w y 2  0 ).

Figure 11: Sign convention for curvatures and moments

On the other hand, for the twist, with the sign convention being used, when M xy  0 ,  w xy  0 ,
2

as depicted in fig.10.

8
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 3- Strains in a Plate

Consider a line element parallel to the x axis, of length x . Let the element displace as shown in fig.12.
Whereas w was used in the previous section on curvatures to denote displacement of the mid-surface, here, for
the moment, let w  x , y , z  be the general vertical displacement of any particle in the plate. Let u and v be the
corresponding displacements in the x and y directions. Denote the original and deformed length of the element
by dS and ds respectively.

Figure 12: Deformation of a material fibre in the x direction.

a. Normal strains

The unit change in length of the element (that is, the exact normal strain) is, using Pythagoras’ theorem,
pq  pq
2 2 2
ds  dS  u   v   w 
 xx    1   1 10
dS pq  x   x   x 
In the plate theory, it will be assumed that:
- The displacement gradients (i.e. u x , u y , u z , v x , v y , v z , w z ) are
small of order   1 say, so that squares and products of these terms may be neglected.
- However, for the moment, the squares and products of the slopes (i.e.  w x  ,  w y  ,
2 2

 w x  w y  ) will be retained, as they may be significant, i.e. of the same order as the strains,
under certain circumstances.
So, Eq.(10) now reduces to:

u  w 
2
u 1  w 
2

 xx  1 2    1     11
x  x  x 2  x 
where the Taylor series approximation  1  e   1  ne have been used twice. So using the similar approach
n

for all the directions, all the normal strains can therefore be defined as:
2
u 1  w  v 1  w  w
2

 xx     ;  yy     ;  zz  12
x 2  x  y 2  y  z

9
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 b. Shear strains

Consider next the angle change for line elements initially lying parallel to the axes (see fig.13). Let  be
the angle r pq , so that    2   is the change in the initial right angle rpq .

Figure13: The deformation of fig.12, showing shear strains.

Taking the dot product of the of the vector elements pq and pr  :
pq  r r   qq  r r   qq  pr 
cos 
pq  pr 
 u   u   v   v   w   w 
 x  x    z    x    z    x    z  z   13 
     
           
x z x z x z
u   v   w 
2 2 2
 u   v  
2 2
w 
2

x  1        z        1  
 x   x   x   z   z   z 
Again, with the displacement gradients u x , u y , u z , v x , v y , v z , w z of order

  1 (and the squares  w x  at most of order   1 ). Therefore Eq.(13) becomes:

2

 u   v   v   w 
x  z    x     z    z  x 
 z   x   z   x   u  w   v  v  14
cos      
x  z z x  x  z 
For small  ,   sin   cos , and using the similar approach for all the directions, all the shear strains
can therefore be defined as:
1  u v w w  1  u w  1  v w 
 xy     ;  xz    ;  yz     15
2  y x x y  2  z x  2  z y 

The normal strains (Eq.(12)) and the shear strains (Eq.(15)) are non-linear. They are the starting point for
the various different plate theories.

10
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 c. Von Karman Strains

Introduce now the assumptions of the classical plate theory.

- The assumption that line elements normal to the mid-plane remain inextensible implies that:
w
 zz  0
z
This implies that w  w  x , y  ; so that all particles at a given  x, y  through the thickness of the plate
experience the same vertical displacement.
- The assumption that line elements perpendicular to the mid-plane remain normal to the mid-plane after
deformation then implies that :
 xz   yz  0
The strains now read
2
u 1  w  v 1  w 
2

 xx     ;  yy     ;  zz  0
x 2  x  y 2  y 
1  u v w w 
 xy     ;  xz  0 ;  yz  0 16
2  y x x y 
These are known as the Von Kármán strains.

d. Membrane strains and Flexural strains

Since  xz  0 and w  w  x, y  , one has from Eq.(15b):

u w w
  u  x, y, z    z  u0  x , y 
z x x
It can be seen that the function u0  x , y  is the displacement in the mid-plane. In terms of the mid-surface
displacements u0 , v 0 , w 0 , then,
w w
u  u0  z ; v  v0  z ; w  w0 17 
x y
And the Von Karman strains becomes therefore:
2
u0 1  w0  v0 1  w0   2 w0
2
 2 w0
 xx   z  yy     z
x 2  x 
; ;
x 2 y 2  y  y 2
1  u v  1  w w0   2 w0
 xy   0  0   0  18
x  2  x z 
z
2  y xy
The first terms are the usual small-strains, for the mid-surface. The second terms, involving squares of
displacement gradients, are non-linear, and need to be considered when the plate bending is fairly large (when
the rotations are about 10 – 15 degrees). These first two terms together are called the membrane strains. The
last terms, involving second derivatives, are the flexural (bending) strains. They involve the curvatures.
When the bending is not too large (when the rotations are below about 10 degrees), one has (dropping the
subscript “0” from w ):

11
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 u0  2w v  2w 1  u v   2w
 xx   z 2 ;  yy  0  z 2 ;  xy   0  0   z 19
x x y y 2  y x  xy
Finally, when the mid-surface strains are neglected, according to the final assumption of the classical plate theory,
one has:
 2w  2w  2w
 xx   z 2 ;  yy   z 2 ;  xy   z  20 
x y xy

e. Compatibility

Note that the strain fields arising in the plate satisfy the 2D compatibility relation (see Continuum
Mechanics notes - Chapter 3 – pp.34):
 2 xx   yy  2 xy
2

 2  21
y 2 x 2 xy
This can be seen by substituting Eq.(20) into Eq.(21).

4- The Moment-Curvature equations

Now that the strains have been related to the curvatures, the moment-curvature relations, which play a
central role in plate theory, can be derived.

a. Stresses and the Curvatures/Twist in a Linear Elastic Plate

From Hooke’s law (see Continuum Mechanics notes – Chapter 4 – Eq.(26)), taking  zz  0 ,
1  1  1
 xx   xx   yy ;  yy   yy   xx ;  xy   xy  22 
E E E E E
So, from Eq.(20), and solving Eq.(22), one obtains:
E   2w  2w  E   2w  2w  E  2w
 xx   z  2   2  ;  yy   z  2   2  ;  xy   z  23 
1  2  x y  1  2  y x  1   xy

b. Moment-Curvature Equations

Substituting Eq.(23) into the definitions of the moments, Eqs.(2) and (4), and integrating, one has:

  2w  2w    2w  2w   2w
M x  D  2   2  ; M y  D  2   2  ; M xy   D  1     24 
 x y   y x  xy
Where
Eh3
D  25 
12  1   2 

12
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
Equations (24) are the moment-curvature equations for a plate. The moment-curvature equations are analogous
to the beam moment-deflection equation  v x  M EI . The factor D is called the plate stiffness or
2 2

flexural rigidity and plays the same role in the plate theory as does the flexural rigidity term EI in the beam
theory.
While looking on Eq.(23) and Eq.(24), one can establish relations between the moments and the
stresses defined as:
12 zM x 12 zM y 12 zM xy
 xx   ;  yy   ;  xy   25a 
h3 h3 h3

5- Principal Curvatures

Consider the two Cartesian coordinate systems shown in

Fig.14, the second  t  n  obtained from the first  x  y  by
a positive rotation  . The partial derivatives arising in the
curvature expressions can be expressed in terms of derivatives
with respect to t and n as follows: with w  w  x , y  , an
increment in w is:
w w
w  x  y  26 
x y
Figure 14: Two different Cartesian coordinate
Also, referring to Fig.14, systems.

With n  0 With t  0
x  t cos   ; y  t sin   x  n sin   ; y  n cos  

From the above table, one has:

- For the first order derivatives (corresponding to the relations of the slope in the two Cartesian
coordinate systems)
w w w w w w
 cos    sin   ;   sin    cos    27 
t x y n x y

- For the second order derivatives (corresponding to the relations of the curvatures and twist in the
two Cartesian coordinate systems)
 2w  2w  2w  2w
 cos   2  sin   2  sin  2 
2 2
;
t 2 x y xy
 2w  2w  2w  2w
 sin   2  cos   2  sin  2 
2 2
 28 
n2 x y xy
 2w   2w  2w   2w
 cos   sin    2  2   cos  2 
t n  y x  xy

13
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
Or in terms of radii of curvature:
1 1 1 1
 cos 2    sin 2    sin  2  ;
Rt Rx Ry Txy
1 1 1 1
 sin 2    cos 2    sin  2   29 
Rn Rx Ry Txy

1  1 1  1
 cos   sin       cos  2 
Ttn  Ry Rx  Txy
These equations transform between curvatures in different coordinate systems. So:
There will be some angle  for which the twist is zero; at this angle, one of the curvatures will be the
minimum and one will be the maximum at that point in the plate. These are called the principal curvatures.

III- PLATES SUBJECTED TO PURE BENDING AND TWISTING

1- Pure Bending of an Elastic Plate

Consider a plate subjected to bending moments M x  M 1  0 and M y  M 2  0 , with no other

loading, as shown in Fig.15.

Figure 15: A plate under Pure Bending.

.
From equilibrium considerations, these moments act at all points within the plate – they are constant throughout
the plate. Thus, from the moment-curvature equations (24), one has the set of coupled partial differential
equations
M1  2 w  2w M2  2w  2w  2w
 2  2 ;  2  2 ; 0  30 
D x y D y x x y
Solving for the derivatives,
 2 w M1   M 2  2 w M 2   M1  2w
  0  31
x 2 D  1   2  D 1   2 
; ;
y 2 x y

Integrating all these equations allow us to obtain:

14
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 M1   M 2 2 M 2   M1 2
w  x, y   x  y  Ax  By  C  32 
2D 1   2  2D 1   2 

Where A , B and C are integration constants. Their values will depend on certain conditions at special locations
of the plate (either the edges (i.e. boundary conditions) or at some paticular points). For example, if one supposes
that the deflection w and slopes w x , w y are zero at the origin x  y  0 (so the origin of the axes are
at the plate-centre), then A  B  C  0 ; all deformation will be measured relative to this reference. It follows
that:
M2  M1  2  M1  2 
w  x, y       x   1   y   33 
2 D  1   2   M 2   M 2  
Once the deflection w is known, all other quantities in the plate can be evaluated – the strain from Eq.(20),
the stress from Hooke’s law or directly from Eq.(23), and moments and forces from Eqs.(1)-(5).
The character of the deformed plate is plotted in Fig.16 for various ratios M 2 M 1 (for   0.3 ).
- When the curvatures  w x and  w y are of the same sign, the deformation is called synclastic
2 2 2 2

(Upper plots of Fig.16).

- When the curvatures  w x and  w y
2 2 2 2
are of opposite sign, the deformation is said to be
anticlastic (Lower plots of Fig.16).

Figure 16: Bending of a Plate.

.

a. Case one: Equal Bending moments

15
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 In the special case of equal bending moments, with M1  M2  M0 say, one has:

w  x, y  
M0
2 D 1   
 x2  y2   34 
This is the equation of a sphere. In fact, from the relationship between the curvatures and the radius of curvature
R (Eq.(8)),
D 1   
R  Cte  35 
M0
and so the mid-surface of the plate in this case deforms into the surface of a sphere with radius given by Eq.(35),
as illustrated in Fig.17.

Figure 17: Deformed plate under Pure Bending with equal moments.
.

b. Case two: Pure cylindrical deformation

In order to get a pure cylindrical deformation, w  f  x  say, one needs to apply moments M x and
M y   M x , in which case, from Eq.(33),
Mx 2
w x  x  36 
2D
Note that when there is only one moment, M y  0 say, there is still curvature in both directions. In this case,
one can solve the moment-curvature equations to get:
 2w  2w  2w
2 
x2   y2 
Mx Mx
w  x, y     ;   2  37 
2D 1    x 2 D  1   2  y 2 x
which is an anticlastic deformation.

2- Pure Torsion of an Elastic Plate

In pure torsion, one has the twisting moment M xy  M  0 with no other loading, Fig.18. From the
moment-curvature equations (Eq.(24)),
 2w  2w  2w  2w M  2w
0  2 ; 0  2  2 ;    34 
x 2 y y x D  1    xy

16
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
Solving for the derivatives,
 2w  2w  2w M
 0;  0;   35 
x 2 y 2 xy D 1   

Figure 18: Twisting of a Plate.

Integrating all these equations allow us to obtain:
M
w  x, y    xy  Ax  By  C  36 
D 1   
Where A , B and C are integration constants. As previously,
if one supposes that the deflection w and slopes w x ,
w y are zero at the origin, then A  B  C  0 . It follows
that:
M
w  x, y    xy  37 
D 1   
The middle surface is deformed as shown in Fig.19, for a
negative M xy . Note that there is no deflection along the lines
x  0 or y  0 . The principal curvatures will occur at 45°to
the axes (see Eqs.(29)):
1 M 1 M
 ;   38 
R1 D 1    R2 D 1    Figure 19: Deformation for a (negative) twisting
moment.

IV- EQUILIBRIUM AND LATERAL LOADING

In this section, lateral loads are considered and these lead to shearing forces V x , V y , in the plate.
In general, a plate will at any location be subjected to a lateral pressure q , bending and twisting moments
M x , M y , M xy and out-of-plane shear forces V x and V y ; q is the normal pressure on the upper surface of
the plate:

 0, z  h 2
 zz  x , y     39 
q  x, y  ,
 z  h 2

17
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
These quantities are related to each other through force equilibrium.

1- Force Equilibrium

Consider a differential plate element with one corner at  x, y    0,0 , Fig.20, subjected to moments,
pressure and shear force. Taking force equilibrium in the vertical direction (neglecting a possible small variation
in q , since this will only introduce higher order terms):
 V y   V 
 z y  y y  x  Vx y   Vx  xx x  y  qxy  0  40 
F  V  x  V   y
 
Which gives the vertical equilibrium equation:
Vx V y
  q  41
x y
Next, taking moments about the x axis:
 M xy   M y 
 x xy  xy x 
M  M  y  M   x  y  M y  x   y
M 
y
 y  x
   
 V y   y 
 yV y x   y  y   V y  y  x   y  V  y  42 
 y   2  x
 y   V x   y 
 y    Vx  x  y   y   qx y  0
 2  x   2 
Using Eq.(41), this reduces to (and similarly for moments about the y -axis),

M x M xy M xy M y
Vx    ; Vy     43 
x y x y

Figure 20: A plate element subjected to moments, pressure and shear forces.

2- Relations directly from the Equations of Equilibrium

18
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 The equilibrium relations (41), (43) can also be derived directly from the equations of equilibrium using
the stress tensor (see continuum mechanics notes):
 xx  xy  xz  yx  yy  yz  zx  zy  zz
  0 ;    0;   0  44 
x y z x y z x y z
Taking the first of these (which ensures equilibrium of forces in the x direction), multiplying by z and integrating
over the plate thickness, gives:
;
h 2
 xx
h 2
 xy h 2
 xz

h 2
z
 x
dz  
h 2
z
 y
dz   z
h 2
 z
dz  0
 45 
    h 2 
h 2 h 2
h 2

x   h 2 xx  y   h 2 xy   xz   h 2  h 2 xz
  z dz    z dz    z    dz  0

and, since the shear stress  xz must be zero over the top and bottom surfaces, one has the first equation of (43).
Applying a similar procedure to the second equilibrium equation gives the second equation of (43). Finally,
integrating directly the third equilibrium equation without multiplying across by z , one arrives at Eq.(41).

3- The Sophie Germain’s equation

Eliminating the shear forces from (41) and (43) leads to the differential equation:
2Mx  2 M xy  2 M y
2   q  46 
x 2 xy y 2
Finally, substituting in the moment-curvature equations (24) leads to:
 4w  4w  4w q
 2     47 
x 4 x 2y 2 y 4 D
This is called the governing differential equation for deflection of thin plates or sometimes called the equation
of Sophie Germain after the French investigator who first obtained it in 1815. This partial differential equation
is solved subject to the boundary conditions of the problem, i.e. the fixing conditions of the plate.
Eq.(47) can be succinctly expressed as:
q
4w    48 
D
Where the del operator (or Laplacian operator)  is defined as:
2

2 2
2    49 
x 2 y 2

V- BOUNDARY CONDITIONS

For a plate, solution of Equations (47) or (48) requires that two boundary conditions be satisfied at each
edge. These may be a given deflection and slope, or force and moment, or some combination.

19
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 We now treat the boundary conditions or a rectangular plate with edges a and b parallel to the x and
y axes, as shown in Fig.21.

Figure 21: A plate structure.

We can now formulate a variety of commonly encountered situations. The boundary conditions that
apply along the edge x  a of the rectangular plate with edges parallel to the x and y axes (Fig.22) are as
follows:
- Clamped, fixed, or built-in edge (Fig.22a).
In this case, both the deflection and the slope must vanish. That is,
w  x , y 
w  x  a, y   0 ; 0  50 
x x a

- Simply supported edge (Fig.22b).

At the simple support considered, both the deflection and the bending moment are zero. Therefore implying the
following conditions:
 2w  x, y 
w  x  a, y   0 ; 0  51
x 2 x a

- Free edge
Such an edge at x  a is free of moment and vertical shear force. Therefore implying the following conditions:
  2w  x, y   2w  x, y     3w  x, y   3w  x, y  
   0 
;  2    0  52 
 x 2 y 2  x a  x 3 x y 2 
x a

- Sliding edge (Fig.22c).

In this case, the edge is free to move vertically, but the rotation is prevented. The support is not capable of resisting
any shear force. Thus,
w  x , y    3w  x, y   3w  x, y  
 0;   2    0  53 
x x a   x 3
 x  y 2
 x a

Some other types of boundary conditions may be treated similarly. It is observed that the boundary
conditions are of two basic kinds:
 A geometric or kinematic boundary condition describes the end constraint pertaining to deflection
or slope;

20
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 Figure 22: Various boundary conditions: (a) clamped edge; (b) simply supported
edge; (c) sliding edge.
 A static boundary condition equates the internal forces (and moments) at the edges of the plate to
the given external forces (and moments).
Accordingly, in Equations (50), both conditions are kinematic; in Equation (52), both are static; and in Equations
(51) and (53), the conditions are mixed.

21
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 SECTION B:
SHELLS
CHAPTER I:
MEMBRANE STRESSES IN SHELLS

I- INTRODUCTION

Until now, our concern was with the analysis of thin flat plates. We now extend the discussion to curved
surface structures termed thin shells. Examples of shells include pipes, culverts, roof domes, tunnels,
nonrectangular tanks etc. Each of these has walls that are curved. Inasmuch as a curved plate can be viewed as
a portion of a shell, the general equations for thin shells are also applicable to curved plates.
We shall limit our treatment to shells of constant thickness, small in comparison with the other two
dimensions. As in the treatment of plates, the plane bisecting the shell thickness is called the midsurface. To
describe the shape of a shell, we need only know the geometry of the mid-surface and the thickness of the shell
at each point.
Shells of technical significance are often defined as thin when the ratio of thickness t to radius curvature
r is equal to or less than 1/20.
The major thin-shell structures include concrete shell structures, often cast as a dome, bridge, or saddle
roof; lattice shell structures, so-called grid-shell structures, usually in the form of a dome or hyperboloid
structure; membrane or fabric structures and other tensile structures, cable domes, and pneumatic structures. The
thin-walled shell structures have more strength with respect to their self-weight and high stiffness, they are
gaining in popularity in areas where there is risk of earthquake and hurricane. Thus, it is necessary to analyze the
shell structures to ensure the best and safe design method.
Membrane stresses in typical shell structures in the forms of sphere, cylinder, cone, ellipsoid, toroid,
hyperbolic paraboloid, and multisphere are discussed in the examples.

II- THEORIES AND GENERAL BEHAVIOR OF SHELLS

The analysis of shell-like structures often embraces two distinct, commonly applied theories.
- The first of these, the membrane theory, usually applies to a rather large part of the entire shell. A
membrane, either flat or curved, is identified as a body of the same shape as a plate or shell but incapable
of conveying moments or shear forces.
- The second, the bending theory or general theory, includes the effects of bending. However,
information relative to shell membrane stresses is usually of much greater practical significance than the
knowledge of the bending stresses. The former are also far simpler to calculate.

The basic kinematic assumptions associated with the deformation of a thin shell as used in small-
deflection analysis are as follows:
1. The ratio of shell thickness to radius of curvature of the midsurface is small in comparison with unity.
2. Deflections are small compared with shell thickness.

22
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 3. Plane sections through a shell, taken normal to the midsurface, remain plane and become normal to
the deformed midsurface after the shell is subjected to bending. This hypothesis implies that strains
 xz and  yz are negligible. Normal strain  z due to transverse loading may also be omitted.
4. The z -directed normal stress  z is negligible.

III- LOAD RESISTANCE ACTION OF A SHELL

The load-resisting action of a shell differs from that of other structural forms (beams, plates) is
underscored by noting the extraordinary capacity of an eggshell to withstand normal forces, this despite their
thinness and fragility. A shell, being curved, can develop in-plane forces (thrusts) to form the primary resistance
action in addition to those forces and moments existing in a plate or beam.
To describe the phenomenon, consider a part of a spherical shell of radius r and thickness t , subjected
to a uniform pressure of intensity p (Fig.1). The condition that the sum of vertical forces be zero is expressed as
2 r0 N sin   p r02  0 or
pr0 pr
N  1
2sin  2
in which N is the in-plane force per unit of circumference. This relationship is valid anywhere in the shell, as
N is seen not to vary with  . We note that, in contrast to the case of plates, load is sustained by the midsurface.
It is next demonstrated that the bending stresses play an insignificant role in the load-resisting action.
Based on the symmetry of the shell and the loading, the state of stress at a point, represented by an infinitesimal
element, is as shown in Fig.1. The compressive direct stress has the form :
N pr
   2
t 2t
The stress normal to the midsurface is negligible, and thus the direct strain, from Hooke’s law, is :
1 pr
       1     3
E 2 Et
Associated with this strain, the reduced circumference is 2 r   2  r  r  , or r   r  1    . The
variation in curvature  is thus:
1 1 1 1   1  
     1        1      ... 
2

r r r  1    r 1  r
Neglecting higher-order terms because of their small magnitude
and introducing Eq.(3), the above expression becomes
 p 1   
    4
r 2 Et
A relationship for the shell-bending moment is derived
from the plate formulas. For the spherical shell under
consideration, we have    x   y . Thus, using the plate

 
formulas (i.e. M   D  x   y ) we obtain:
Figure 1: Truncated spherical shell subjected
to uniform pressure.

23
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
  1    p   pt
2

 
2
M   D  x   y  D
2 Et 24
We have replace D  Et
3
12  1   2   from plate theory. Hence, the bending stress is given by:
 
6M p
b  2    5
t 4
The ratio of the direct stress to the bending stress is:
 2r
  6
b t
It is observed that the direct or membrane stress is very much larger than the bending stress as t 2r  1 . We
are led to conclude therefore that the applied load is resisted predominantly by the in-plane stressing of the
shell.

IV- GEOMETRY OF SHELLS OF REVOLUTION

The geometry of a shell is completely defined

if we know the shape of its midsurface and its
thickness at all points. Consider a particular type of
shell described by a surface of revolution (Fig.2).
Examples include the sphere, cylinder, and cone. The
midsurface of a shell of revolution is generated by
rotation of a so-called meridian curve about an axis
lying in the plane of the curve. Fig.2 shows that a
point on the shell is conveniently located by
coordinates  ,  , r0 and that elemental surface
ABCD is defined by two meridians and two
parallel circles or parallels. Figure 2: A shell of revolution.
The planes associated with the principal radii
of curvature r1 and r2 at any point on the midsurface of the shell are the meridian plane and the parallel plane
at the point in question, respectively. The corresponding principal curvatures of the meridian and parallel planes
at any point may be denoted by  1 and  2 , so that 1  1 r1 and  2  1 r2 .
Observe from the figure that the radii of curvature r1 and r2 are thus related to sides CD and AC . The
principal radius r2 generates the shell surface in the direction perpendicular to the direction of the tangent to the
meridian curve. The two radii r0 and r2 are related inasmuch as r0  r2 sin (Fig.2). It follows that lengths of
the curvilinear shell element are
LAC  r0d  r2 sind ; LCD  r1d  7
It is assumed in the above description that the principal radii of curvature of the shell r1 and r2 are known
constants. In the case of a radius of curvature that varies from point to point, the radii are computed by applying
the equation that defines the shell shape, along with various relationships of differential geometry of a surface.

24
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
 V- SYMMETRICALLY LOADED SHELLS OF REVOLUTION
In axisymmetrical problems involving shells of revolution, no shear forces exist, and there are only
two unknown membrane forces, or membrane stress resultants, per unit length, N  and N  . The governing
equations for the so-called hoop force N  and the meridional force N  are derived from two equilibrium
conditions. Fig.3(a) and (b) shows two different views of the element ABCD cut from the shell of Fig.2.
Prescribed by the condition of symmetry, the membrane forces and the loading display no variation with  . The
externally applied forces per unit surface area are represented by the components p y and pz in the y and z
directions, respectively.
Description of the z equilibrium requires that the z  components of the loading as well as of the forces
acting on each edge of the element be considered. The z  directed distributed load carried on the surface area
of the element is
pz r0r1d d
The force acting on the top edge of the element equals N  r0d . Neglecting higher terms, the force on

the bottom edge is also N  r0d . The z  directed component at each edge is then N r0d sin  d 2  . This

force is nearly equal to N r0d  d 2  , yielding the following magnitude of the resultant for both edges:
N  r0d d
As the cross-sectional area along each of the two sides of the element is r1d  , the force on these areas is
N  r1d . The resultant in the direction of the radius of the parallel plane for both such forces is N  r1d d
producing the following component in the z  direction
N r1dd sin
For the forces considered above, from F z  0 , we have:
N  r0  N r1 sin   pz r0 r1  0  8

Figure 3: Diagrams for analysis of symmetrically loaded shells of revolution: (a and b) Membrane
forces and surface loads on a shell element.

25
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
This expression may be converted to simpler form by dividing by r0 r1 and replacing r0 by r2 sin  . By so doing,
one of the basic relations for the axisymmetrically loaded shell is found as follows:
N N
   pz  9a 
r1 r2
The equilibrium of forces in the direction of the meridional tangent, that is, in the y  direction, is
expressed as
d
d
 
N r0 d d  N r1d d cos  p y r1d r0d  0 10
The first term represents the sum of normal forces acting on edges AC and BD , while the third term is the
loading component. The second term is the component in the y  direction of the radial resultant force
N  r1d d acting on faces AB and CD . Dividing Eq.(10) by d d , the equation of equilibrium of the
y  directed forces is now:
d
d
 
N  r0  N r1 cos    p y r1r0 11
It is noted that an equation of equilibrium that can be used instead of Eq.(11) follows readily by isolating part of
the shell intercepted by angle  (Fig.4a). Here force F represents the resultant of all external loading applied
to this free body. Recall that from conditions of symmetry, forces N  are constant around the edge. Equilibrium
of the vertical forces (see Fig.4b) is therefore described by 2 r0 N  sin   F  0 , and it follows that
F
N    9b 
2 r0 sin 
We show that Eq.(9b) is equivalent to Eq.(11). From Equations (9a) and (9b) we are able to determine the
membrane forces N  and N  from which the stresses are readily obtained. Negative algebraic results indicate
compressive stresses.

Figure 4: (a) Meridian forces and resultant loading (F) acting on a truncated shell defined by the angle
; (b) components the meridional force .

VI- SOME TYPICAL CASES OF SHELLS OF REVOLUTION

The membrane stresses in any particular axisymmetrically loaded shell in the form of a surface of
revolution may be determined from the governing expressions of equilibrium developed in Section 5. Treated in

26
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa
the following paragraphs are several common structural members. As pointed out previously, the principal radii
of curvature r1 and r2 at any point on the midsurface of the shell denote the meridian plane and parallel plane,
respectively.

1- Spherical Shell

For spherical shells we can set the mean radius a  r1  r2 . Then Equations (9a)-(9b) appear in the form :
N   N    pz a
F 12
N  
2 a sin 2 
The simplest case is that of a spherical shell subjected to constant internal pressure p.

27
Advanced Structural Engineering - Department Civil Engineering & Architecture - NAHPI - UBa