FINITE ELEMENT METHOD (MAK336E)

Mechanical Engineering Department, ITU

Prof. Dr. Kenan Y. Şanlıtürk

3.0 FE THEORY: GENERAL CONTINUUM

3.1 Fundamentals
Concentrated
y Ri
forces
S
t

A : Surfaces on which
displacements are
dV
fixed or known.

b S : Surfaces on which
A tractions (forces) are
known.

Fig. 3.1 Arbitrarily shaped solid body.

Consider an arbitrarily shaped 3-dimensional solid body in Fig.3.1. The body is

supported in some ways and subjected to certain external loads: internal body forces per
→ →
unit volume b (such as gravity), external surface forces per unit area t (not necessarily
→
normal pressures) and, concentrated forces Ri applied at locations denoted by i. These
vectors can be expressed as

⎧bx ⎫ ⎧t x ⎫ ⎧ Rx ⎫
→
⎪ ⎪ →
⎪ ⎪ →
⎪ ⎪
b = {b} = ⎨b y ⎬ , t = {t} = ⎨t y ⎬ , R = {R} = ⎨ R y ⎬ (3.1)
⎪b ⎪ ⎪t ⎪ ⎪R ⎪
⎩ z⎭ ⎩ z⎭ ⎩ z⎭

KYŞ FEM: Chapter 3 1

The stresses (and strains) inside the volume V have two types of components, direct and
shear stresses. It is also very convenient to write these stresses in the form of column
vectors as:

⎧σ xx ⎫ ⎧ε xx ⎫
⎪σ ⎪ ⎪ε ⎪
⎪ yy ⎪ ⎪ yy ⎪
⎪σ ⎪ ⎪ε ⎪
{σ } = ⎪⎨ zz ⎪⎬ {ε } = ⎪⎨ zz ⎪⎬ (3.2)
⎪σ xy ⎪ ⎪ε xy ⎪
⎪σ yz ⎪ ⎪ε yz ⎪
⎪ ⎪ ⎪ ⎪
⎪⎩σ zx ⎪⎭ ⎪⎩ε xz ⎪⎭

The displacement at any point within the volume V has three components, which is
expressed by a vector as:

⎧ u ( x, y , z ) ⎫
{d } = ⎪⎨ v( x, y, z ) ⎪⎬ (3.3)
⎪w( x, y, z )⎪
⎩ ⎭

The equations of compatibility, relating strains to displacements, u, v, w are:

∂u ∂v ∂w
ε xx = , ε yy = , ε zz =
∂x ∂y ∂z
(3.4)
∂u ∂v ∂v ∂w ∂u ∂w
ε xy = + , ε yz = + , ε xz = +
∂y ∂x ∂z ∂y ∂z ∂x

and this relation can be written in matrix form as

⎡∂ ⎤
⎢ ∂x 0 0⎥
⎢ ∂ ⎥
ε
⎧ xx ⎫ ⎢0 0⎥
⎪ε ⎪ ⎢ ∂y ⎥
⎪ yy ⎪ ⎧ u ( x, y , z ) ⎫ ⎢ ∂⎥
⎪⎪ε zz ⎪⎪ ⎪ ⎪ ⎢0 0
∂z ⎥⎥
⎨ ⎬ = [S ]⎨ v( x, y, z ) ⎬ where [S ] = ⎢ ∂ ∂ (3.5)
⎪ε xy ⎪ ⎪w( x, y, z )⎪ ⎢ 0⎥
⎪ε zy ⎪ ⎩ ⎭ ⎢ ∂y ∂x ⎥
⎪ ⎪ ⎢ ∂ ∂⎥
⎪⎩ε zx ⎪⎭ ⎢0 ⎥
⎢ ∂z ∂y ⎥
⎢∂ 0
∂⎥
⎢⎣ ∂z ∂x ⎥⎦

KYŞ FEM: Chapter 3 2

The stress-strain relationship for an isotropic material can also be written in matrix form
as:

⎡1 − ν ν ν 0 0 0 ⎤
⎢ 1 −ν ν 0 0 0 ⎥⎥
⎢
⎢ 1 −ν 0 0 0 ⎥
{σ } = [D]{ε }, [D] = E
⎢ ⎥
(1 + ν )(1 − 2ν ) ⎢ 0.5 − ν 0 0 ⎥
⎢ Sym 0.5 − ν 0 ⎥
⎢ ⎥
⎣⎢ 0.5 − ν ⎦⎥
(3.6)

These are the basic equations from Theory of Elasticity expressed in matrix form. The
next step is to use Principle of Virtual Displacement (PVD) to establish the equilibrium
equations.

3.2 Principle of Virtual Displacements (or Work)

When deriving the stiffness matrix for the truss elements using the direct approach, it was
necessary to find a relationship between the stresses in the elements and the forces at the
nodes. It was possible to obtain such a relationship “directly” because of the simplicity
of the element geometry and the stress being constant over its area (and length).
However, for more complicated elements, e.g., stress varying over its area, this is not so
easy.

The Principle of Virtual Displacement (PVD) provides a completely general energy-

based approach for deriving element matrices for all kinds of elements. PVD simply
states that the equilibrium of the body requires that for any compatible, small virtual
displacements, "the work done within a body (the strain energy) is equal to the external
work done by the forces applied to the body".

Imagine that a virtual displacement {δd } = {δu ( x, y, z ), δv( x, y, z ), δw( x, y, z )} is imposed

T

on a body. The virtual strains corresponding to the virtual displacement are:

{δε } = {δε xx , δε yy , δε zz , δε xy , δε yz , δε zx }T (3.7)

According to the PVD:

δU = δW (3.8)

KYŞ FEM: Chapter 3 3

Where δU is the virtual strain energy and δW is the work done by the applied external
forces due to virtual displacements. The virtual strain energy per unit volume is the
virtual work done by the stresses, i.e.

(δε xx σ xx + δε yyσ yy + δε zzσ zz + δε xyσ xy + δε yzσ yz + δε zxσ zx ) = {δε }T {σ } (3.9)

Integrating Eq.(3.9) over the whole volume yields the total internal virtual work (strain
energy):

δU = ∫ {δε }T {σ }dV (3.10)

V

(It may be helpful to think of the virtual work done by a stress component on a small
piece of the material as analogous to the virtual work done by a uniaxial force on a
spring).

The virtual work done by applied external forces (δW), i.e. by body forces, tractions and
concentrated loads etc, can also be written in a similar way:

δW = ∫ {δd }T {b}dV + ∫ {δd }T {t}dS + ∑ δd i Ri (3.11)

V S i

Using the statement of the virtual displacements, i.e., the work done within a body (the
strain energy) is equal to the external work done by the forces applied to the body" leads
to:

∫ {δε } {σ }dV = ∫ {δd } {b}dV + ∫ {δd } {t}dS + ∑ δd R

T T T
i i (3.12)
V V S i

This very important fundamental work equation is the basis of the finite element
displacement method. In the Finite Element Method, we approximate the body in Fig.3.1
as an assemblage of discrete finite elements with the elements being interconnected at
nodal points on the element boundaries. Therefore, we can equally apply the principle of
virtual work to individual elements.

Now, consider a single finite element where the displacement field is represented using
the shape function matrix [N]

{d } = [N ]{u} (3.13a)

and strains are

{ε } = [B ]{u} (3.13b)

KYŞ FEM: Chapter 3 4

Note that {d} refers to the analytical functions in Eq.(3.3) and {u} refers to a vector
which contains nodal (scalar) displacements - not in just one but in all directions - for an
element. ({u}vector should not be confused with the displacements in x-direction only!.)
In other words, the size of the {u} vector is equal to the number of degrees-of-freedom of
a finite element.

To apply the PVD, the virtual displacement vector {δd} is written in terms of nodal
displacements, i.e.,

{δd } = [N ]{δu} (3.14a)

Correspondingly, the virtual strains are

{δε } = [B ]{δu} (3.14b)

and recalling that

{σ } = [D ]{ε } = [D ][B ]{u} (3.14c)

Inserting Eq.(14) into Eq.(3.12), but noting that the integration is over the element
volume Ve and element surface Se , leads to

⎛ ⎞ ⎛ ⎞
{δu}T ⎜⎜ ∫ [B ]T [D][B]dVe ⎟⎟{u} = {δu}T ⎜⎜ ∫ [N ]T {b}dVe + ∫ [N ]T {t}dS e + {R}⎟⎟ (3.15)
⎝ Ve ⎠ ⎝ Ve Se ⎠

The term {δu} cancels in Eq.(3.15) and the result is

T

⎛ ⎞ ⎛ ⎞
⎜ [B ]T [D ][B ]dV ⎟{u} = ⎜ [N ]T {b}dV + [N ]T {t}dS + {R}⎟
⎜ V∫ e
⎟ ⎜ V∫ e ∫ e
⎟ (3.16)
⎝ e ⎠ ⎝ e Se ⎠

Equation (3.16) provides a well-defined relationship to obtain the element stiffness

matrix and the force vector due to body forces, tractions and concentrated loads.
Inspection of Eq.(3.16) reveals that it is exactly in the form of

[k ]{u} = { f } (3.17)

where

[k ] = ∫ [B]T [D][B]dVe (3.18)

Ve

KYŞ FEM: Chapter 3 5

and the external force vector is the sum of three forces: effect of element body forces
{fB}, effect of the surface tractions {fS}and the concentrated forces {fC }. That is

{f } = {f B }+ {f S }+ { fC } (3.19a)

where
{ f B } = ∫ [N ]T {b}dVe (3.19b)
Ve

{ f S } = ∫ [N ]T {t}dS e (3.19c)
Se

{ f C } = {R} (3.19d)

In dynamic analyses, inertia forces have to be included. Considering inertia forces will
lead to the derivation of the elemental mass matrix. This can be achieved, for example,
by considering the inertia forces as if they are another type of body force which is
proportional to accelerations, but in opposite direction. So, if the acceleration vector is
denoted by {a} and they are interpolated using shape function matrix as before we have:

⎧ .. ⎫
⎪ u.. ( x , y , z ) ⎪
{a} = ⎪⎨ v( x , y , z ) ⎪⎬ = [N ]⎧⎨u ⎫⎬
..
(3.20a)
⎪ .. ⎪ ⎩ ⎭
⎪ w ( x , y , z ) ⎪
⎩ ⎭

Then the inertia force (body force) per unit volume can be expressed as:

{binertia } = − ρ [N ]⎧⎨u ⎫⎬
..
(3.20b)
⎩ ⎭
Where ρ is the mass per unit volume, i,e., density. Then the elemental inertia forces are
written as:

⎧ ⎫
∫ [N ] {binertia }dVe = − ∫ ρ [N ] [N ]dVe ⎨u ⎬
..
T T
(3.20c)
Ve Ve ⎩ ⎭

⎧ .. ⎫
The equation above shows that the inertial body force is of the form: − [m]⎨u ⎬ , hence the
⎩ ⎭
elemental mass matrix, [m], is given by:

[m] = ∫ ρ[N ]T [N ]dVe (3.20d)

Ve

KYŞ FEM: Chapter 3 6

3.3 Example 1: Conical Rod Elements

Given the geometry and properties below. Find u(x=L).

A=1000 mm2

u(x) P=20 kN

x
E=3*10 4 N/mm2

A=100 mm2

L=1000 mm

Fig. 3.2 The geometry of a conical rod.

Analytically the displacement as a function of x can be found by following the following

simple steps. First:

20000 N
σ ( x) =
(1000 − 0.9 x)mm 2
σ ( x) 2
Then the axial strain is: ε ( x) = =
E 3(1000 − 0.9 x)

Note that
∂u du 2
ε ( x) = = =
∂x dx 3(1000 − 0.9 x)
Hence,

x
2
u ( x) = ∫ dx
0
3(1000 − 0.9 x)

u(x) = -0.741 ( ln(1000-0.9x) - 3ln10) mm

Hence
u(L) = 1.706 mm (True value)

and σ(L) = 200 N/mm2 (True value)

KYŞ FEM: Chapter 3 7

3.3.1 FE Solution: One Linear Element

Steps 1-4: Idealisation, Discretization, Element Stiffness

e
x Element e e
u1=u1 u2=u2

e 1 2 ef =F
f1=F1 2 2

Fig. 3.3 Conical Rod. FE model with one element

Assuming a linear variation for u(x)

x x
u ( x) = (1 − )u1 + u 2 (3.21a)
l l

which can be written as

⎧u ⎫ ⎧ x ⎫⎧ u1 ⎫
u ( x) = [N ]{u} = {N 1 N 2 }⎨ 1 ⎬ = ⎨(1 − )
x
⎬⎨ ⎬ (3.21b)
⎩u 2 ⎭ ⎩ l l ⎭⎩u 2 ⎭

From strain-displacement relationship

∂N 2 ⎫⎧ u1 ⎫ ⎧ u1 ⎫
εx =
∂
[N ]{u} = ⎧⎨ ∂N1 ⎬⎨ ⎬ = [B ]⎨ ⎬ (3.22a)
∂x ⎩ ∂x ∂x ⎭⎩u 2 ⎭ ⎩u 2 ⎭

which gives

⎡ 1 1⎤ ⎧ u1 ⎫
ε x = ⎢− ⎨ ⎬ = [B ]{u} (3.22b)
⎣ l l ⎥⎦ ⎩u 2 ⎭

Note that length of the element l=L as we are using a single element for modeling the
conical rod. The stress-strain relationship is very simple since a one-dimensional
formulation is being used, hence the elasticity matrix is

[D] = E (3.23)

KYŞ FEM: Chapter 3 8

We can now insert the identified matrices into the general stiffness matrix formulation:

[k ] = ∫ [B]T [D][B]dVe (3.24a)

Ve

⎧ 1⎫
⎪− ⎪ 1 − 1⎤
[k ] = ∫ E ⎨ 1l ⎬⎧⎨− 1 1⎫⎬dVe = E2 ⎡⎢
l ⎣− 1 1 ⎦V∫e
⎥ dVe (3.24b)
Ve ⎪ ⎪⎩ l l ⎭
⎩ l ⎭

But
l

∫ dVe = ∫ A( x)dx
Ve 0
(3.25a)

Also, the cross sectional area as a function of x can be interpolated using the cross
sectional areas at nodes 1 and 2:

⎧A ⎫
A( x) = [N ]⎨ 1 ⎬ (3.25b)
⎩ A2 ⎭

Hence

l l
⎧ A1 ⎫ l
A + A2
∫ A( x)dx = ∫ [N N 2 ]⎨ ⎬dx = ∫ ((1 − ) A1 + A2 )dx = l ( 1
x x
1 ) = l Am (3.25c)
0 0 ⎩ A2 ⎭ 0
l l 2

A1 + A2
where the mean area is Am = (3.25d)
2

Therefore, the elemental stiffness matrix reduces to

1 − 1⎤
[k ] = EAm ⎡⎢ ⎥ (3.26)
l ⎣− 1 1 ⎦

Since there is only one element in the model, element stiffness matrix is equal to the
overall stiffness matrix, i.e.

1 − 1⎤
[k ] = [K ] = EAm ⎡⎢ ⎥ (3.27)
l ⎣− 1 1 ⎦

KYŞ FEM: Chapter 3 9

Hence the global equilibrium equation is

EAm ⎡ 1 − 1⎤ ⎧ u1 ⎫ ⎧ F1 ⎫
⎢− 1 1 ⎥ ⎨u ⎬ = ⎨ F ⎬ (3.28)
l ⎣ ⎦⎩ 2 ⎭ ⎩ 2 ⎭

Steps 5-6: Boundary Conditions and Solution

Force boundary condition : F2=20000 N

Displacement boundary condition : u1=0

Therefore, from Eq.(3.28) we obtain:

u2 =1.21 mm

Steps 7: Supplementary Computation

From
⎧ 1 1⎫⎧ u1 ⎫ E
σ x = [D ]ε x = [D ][B ]{u} = E ⎨− ⎬⎨ ⎬ = u 2 ,
⎩ l l ⎭⎩u 2 ⎭ L
we obtain
σx=36.36 N/mm2

(Note that [B] matrix is not a function of x for this particular case, hence constant stress
element). From the first row of Eq.(3.28) we find the reaction force at Node 1

F1= - 20000 N as it should be.

1.8

1.6

1.4
displacement u - mm-

1.2

1 Single linear element

0.8

0.6

0.4
Exact solution
0.2

0
0 200 400 600 800 1000
x -c oordinate -m m -
Fig.3.4 Conical Rod. Comparison of exact and FE solution with 1 linear element.

KYŞ FEM: Chapter 3 10

3.3.2 FE Solution: Two Linear Elements

Step1-2: Idealisation and Discretization

Element 1 Element 2

1 2 3

500 500

Fig.3.5 Conical Rod. 2 linear elements.

Step 3: Elemental stiffness matrices

For a general linear element, the stiffness matrix is given

1 − 1⎤
[k ] = EAm ⎡⎢ ⎥
l ⎣− 1 1 ⎦
In this case,

l=L/2=500 mm

and the mean elemental area

1
Am=775 mm2
2
Am=325 mm2

Then the blown-up element stiffness equations are:

⎡ 1 Am −1Am 0⎤ ⎧u1 ⎫ ⎧ 1 f 1 ⎫
E ⎢ 1 ⎥⎪ ⎪ ⎪ ⎪
For element 1: − Am 1
Am 0⎥ ⎨u 2 ⎬ = ⎨ 1 f 2 ⎬ (3.29a)
L/2 ⎢
⎢ 0
⎣ 0 0⎥⎦ ⎪⎩u 3 ⎪⎭ ⎪⎩ 1 f 3 ⎪⎭

KYŞ FEM: Chapter 3 11

 ⎡0 0 0 ⎤ ⎧ u1 ⎫ ⎧ 2 f1 ⎫
E ⎢ ⎪ ⎪ ⎪ ⎪
For element 2: 0 2
Am − 2 Am ⎥⎥ ⎨u 2 ⎬ = ⎨ 2 f 2 ⎬ (3.29b)
L/2 ⎢
⎢⎣0 − Am
2 2
Am ⎥⎦ ⎪⎩u 3 ⎪⎭ ⎪⎩ 2 f 3 ⎪⎭

Step 4: Assembly

[K ]=1 [k ]+ 2 [k ] (3.30a)

Therefore, the overall equation is:

⎡ 1 Am −1Am 0 ⎤ ⎧u1 ⎫ ⎧ F1 ⎫
E ⎢ 1 ⎥⎪ ⎪ ⎪ ⎪
− Am 1
Am + 2Am − Am ⎥ ⎨u 2 ⎬ = ⎨ F2 ⎬
2
(3.30b)
L/2 ⎢
⎢ 0
⎣ − 2 Am 2
Am ⎥⎦ ⎪⎩u 3 ⎪⎭ ⎪⎩ F3 ⎪⎭

Step 5-6: Boundary Conditions and Solution

Force boundary conditions : F2=0, F3=20000N

Displacement boundary conditions: u1=0

Hence, deleting the first row and column of the above matrix equation and inserting the
numerical values, we can find that

u2=0.430 mm
u3=1.456 mm

Step 7: Supplementary Calculations

For element 1: compute F1 using the first row of Eq.(3.31) which gives
F1=19995 N. (This should be -20000…the difference is due rounding error!!)

Compute the strain and stress as before:

1
σx=25.8 N/mm2, 1εx=0.00086

For element 2:
2
σx=61.5 N/mm2, 2εx=0.00205

KYŞ FEM: Chapter 3 12

 1.8

1.6

1.4
2 linear elements
displacement u - mm-

1.2

0.8 1 linear element

0.6

0.4
Exact sol.
0.2

0
0 100 200 300 400 500 600 700 800 900 1000
x-coordinate -mm-

Fig.3.6 Conical Rod. Comparison of various solutions

3.3.3 FE Solution: One Quadratic Element

Steps 1 and 2: Idealisation and Discretization

l=L

Element e
x ui uj uk
i fi j fj k fk

u(x) Assumed quadratic form

for u(x)
uk

ui uj
x
Fig.3.7 Conical Rod. Single element with quadratic shape function.
(Nodes i, j and k can also be referred as Nodes 1, 2 and 3)

Assume a quadratic variation form for u(x) within the element:

u(x) = ax2 + bx + c (3.31)

KYŞ FEM: Chapter 3 13

In order to construct the shape functions, the constants, (a, b, c) are to be determined by
making use of the nodal conditions, i.e.,

u i = axi2 + bxi + c
u j = ax 2j + bx j + c (3.32)
u k = ax k2 + bx k + c

We have here 3 equations and 3 unknowns (a, b, c), hence we may solve it using
Cramer's rule as

u i ( x j − x k ) + u j ( x k − xi ) + u k ( x i − x j )
a=
∆
u i ( x − x ) + u j ( xi2 − x k2 ) + u k ( x 2j − xi2 )
2 2

b=
k j
(3.33)
∆
u i ( x j x k − x k x j ) + u j ( x k2 xi − xi2 x k ) + u k ( xi2 x j − x 2j xi )
2 2

c=
∆
where
∆ = xi2 ( x j − x k ) + x 2j ( x k − xi ) + x k2 ( xi − x j )

Hence, from Eqns. (3.32) and (3.33), we obtain:

⎧ui ⎫ ⎧ui ⎫
u ( x) = [N ]⎨u j ⎬ = {N i N k }⎨u j ⎬
⎪ ⎪ ⎪ ⎪
Nj (3.34)
⎪u ⎪ ⎪u ⎪
⎩ k⎭ ⎩ k⎭

where

Ni =
1
∆
(
( x j − x k ) x 2 + ( x k2 − x 2j ) x + ( x 2j x k − x k2 x j ) )
1
(
N j = ( x k − xi ) x 2 + ( xi2 − x k2 ) x + ( x k2 xi − xi2 x k )
∆
) (3.35)
1
(
N k = ( xi − x j ) x 2 + ( x 2j − xi2 ) x + ( xi2 x j − x 2j xi )
∆
)
Note that if xi=0 (that is the origin of the co-ordinate system is at Node 1), then xj=L/2
and xk=L. In this case we obtain:

KYŞ FEM: Chapter 3 14

 ⎛ 2 3 ⎞
N i = ⎜ 2 x 2 − x + 1⎟
⎝L L ⎠
⎛ 4 4 ⎞
N j = ⎜ − 2 x2 + x⎟ (3.36)
⎝ L L ⎠
⎛ 2 1 ⎞
Nk = ⎜ 2 x2 − x⎟
⎝L L ⎠
3
L
and ∆=−
4

Hence the assumed displacement function or interpolation function becomes

⎛ 2 3 ⎞ ⎛ 4 4 ⎞ ⎛ 2 1 ⎞
u ( x) = ⎜ 2 x 2 − x + 1⎟u i + ⎜ − 2 x 2 + x ⎟u j + ⎜ 2 x 2 − x ⎟u k (3.37)
⎝L L ⎠ ⎝ L L ⎠ ⎝L L ⎠

Remark: The procedure to obtain the shape functions as described here (Eq.(3.35) is
rather cumbersome. In Chapters 4 and 5, an elegant method (using natural co-ordinates)
will be discussed, which will supply the shape functions [N] directly.

The displacement-to-strain operator matrix [B] in the following equation

⎧ui ⎫
∂u ( x) ∂[N ] ⎧ ∂N j ∂N k ⎫⎪ ⎪
εx = = {u} = [B]{u} = ⎨ ∂N i ⎬⎨u j ⎬ (3.38)
∂x ∂x ⎩ ∂x ∂x ∂x ⎭⎪ ⎪
⎩u k ⎭
can be evaluated as:

[B]1x3 = 1 [2( x j − xk ) x + ( xk2 − x 2j ) 2( xk − xi ) x + ( xi2 − xk2 ) 2( xi − x j ) x + ( x 2j − xi2 ) ]

∆

(3.39)
Step 3: Elemental Stiffness Matrix:

[k ]3 x3 = ∫ [B]T3 x1 [D]1x1 [B]1x3 dVe (3.40)

Ve

Noting that [D]=E, and

⎛ x x⎞ ⎛ x⎞ ⎛ ∧ x
⎞
dVe = ⎜ Ai (1 − ) + Ak ⎟dx = ⎜ Ai − ( Ai − Ak ) ⎟dx = ⎜ Ai − A ⎟dx (3.41)
⎝ l l⎠ ⎝ l⎠ ⎝ l⎠
∧
where A = ( Ai − Ak )

KYŞ FEM: Chapter 3 15

The integral in Eq.(3.40) can now be written as:

(3.42a)
xk
∧
[k ]3 x 3 = ∫ [B]T3 x1 [D]1x1 [B]1x3 ( Ai − A x )dx
xi
l

The constant terms, E in [D] matrix and ∆ in [B], can be taken out of integration sign and
the above expression can be written in matrix form as:

⎡ k11* k12*
xk k13* ⎤
E ⎢ * ⎥
[k ]3 x3 = 2 ∫⎢ *
k 22 k 23 ⎥ dx (3.42b)
∆ xi
⎢ Sym * ⎥
k 33 ⎦
⎣

To demonstrate the procedure, one of the individual elements of the [k] matrix, k11, is
obtained explicitly below. First let's specify the position of the nodal points:

l3 L3
xi=0, xj-xi=l/2, xk-xj=l/2, xj=l/2, xk=l, which makes ∆ = − =−
4 4

Then
xk
E
k11 = 2 ∫k
*
11 dx
∆ 0

∫ (2( x )
xk
E 2⎛ ∧ x
⎞
k11 = 2 − x k ) x + ( x k2 − x 2j ) ⎜ Ai − A ⎟dx
∆
j
0 ⎝ l⎠
2
⎛ l2 ⎞ ⎛
xk
E l ∧ x
⎞
= 2 ∫0 ⎜⎝ 2
⎜ − + − ⎟ −
2
2( ) x (l ) ⎟ ⎜ A A ⎟dx
∆
i
4 ⎠ ⎝ l⎠
2
⎛ 3l 2 ⎞ ⎛
xk
E ∧ x
⎞
= 2
∆ ∫0 ⎜⎝
⎜ − lx + ⎟
⎟
4 ⎠ ⎝
⎜ Ai − A
l⎠
⎟dx
x 2
E k 2⎛ 3l ⎞ ⎛ ∧ x
⎞
2 ∫
= l ⎜ − x + ⎟ ⎜ Ai − A ⎟dx
∆ 0 ⎝ 4⎠ ⎝ l⎠
⎛ ⎞
El 2 ⎜ l 3 3l 3 9l 3 ∧ l3 l 3 9l 3 ⎟
= 2 ⎜ Ai ( − + ) − A( − + )⎟
∆ ⎜ 3 4 16 4 2 32 ⎟
⎝ ⎠
Inserting the value of ∆ and simplifying the above expression yields:

E⎛ ∧
⎞
k11 = ⎜14 Ai − 3 A ⎟
6l ⎝ ⎠

KYŞ FEM: Chapter 3 16

The other components are also calculated in a similar fashion which gives:

⎛ ⎡ 14 − 16 2 ⎤ ⎡ 3 − 4 1 ⎤⎞
E⎜ ⎢ ⎥ ∧
⎢ ⎟
[k ]3 x3 = ⎜ Ai ⎢ 32 − 26⎥ − A⎢ 16 − 12⎥⎥ ⎟ (3.42c)
6l ⎜
⎝ ⎢⎣ Sym 14 ⎥⎦ ⎢⎣ Sym 11 ⎥⎦ ⎟⎠

Steps 4-6: Assembly, Boundary Conditions and Solution

Since we used one element in the model this time, elemental stiffness matrix is the same
as the global stiffness matrix. Inserting the numerical values
∧
Ai = 1000 mm 2 A = 900 mm 2
l=L=1000 mm, E=30 000 N/mm2

into Eq.(3.42c), one can obtain

⎡56500 − 62000 5500 ⎤ ⎧ u1 ⎫ ⎧ F1 ⎫

⎢ ⎥ N ⎪ ⎪ ⎪ ⎪
⎢ 88000 − 26000⎥ ⎨u 2 ⎬ = ⎨ F2 ⎬ N (3.43)
mm ⎪ ⎪ ⎪ ⎪
⎢⎣ Sym 20500 ⎥⎦ ⎩u 3 ⎭ ⎩ F3 ⎭

The boundary conditions are:

Force boundary conditions : F2=0, F3=20000 N
Displacement boundary conditions: u1=0
Introducing these boundary conditions into Eq.(3.43) and solving for the unknown
displacements we have:

u2=0.46 mm
u3=1.56 mm

Steps 7: Supplementary Calculations

The stresses are given by the relationship: σ x ( x) = E [B ]{u} . Thus:
σ(x=0) =8.4 N/mm2 and σ(x=L) =85.2 N/mm2

Note that the matrix [B] is a function of x this time, hence the stresses are not constant:
they vary x!

KYŞ FEM: Chapter 3 17

 1.8

1.6 2 linear elements

1.4
displacement u - mm-

1.2

0.8 1 linear element

0.6
Exact sol.
0.4

0.2
1 Quadratic element
0
0 100 200 300 400 500 600 700 800 900 1000
x-coordinate -mm-
Fig.3.8 Conical Rod. Comparison of various solutions

200

180

160

140
1 Quadratic element
Axial Stress - N/mm-

120 Exact sol.

100

80 1 linear element

60

40

20 2 linear elements
0
0 100 200 300 400 500 600 700 800 900 1000
x-coordinate -mm-
Fig.3.9 Conical Rod. Comparison of axial stress for various solutions.

KYŞ FEM: Chapter 3 18

3.4 2D Analysis: Plane Stress and Plane Strain

The basic principles of theory of elasticity applicable to 3-dimensional elastic problems

with isotropic material are summarized in Section 3.1. Under certain conditions, it is not
necessary to model structures in 3 dimensional space as 2-dimentional models can give
results with acceptable accuracy.

In structural analysis, these two dimensional cases are named as plane strain, plane
stress and axisymmetric conditions. In plane stress and strain cases, the displacement
field is uniquely described by the u and v displacements in the x-y plane. Similarly, 2D
analysis is also applicable for axisymmetric analysis.

Plane Strain:
Plane strain is defined as a deformation state in which w=0 everywhere and u and v are
functions of x and y but not of z. Thus strains in the z-direction are all zero, i.e., εzz=εzx=
εzy=0. A slice of straight-pressurized pipe and a slice of a underground tunnel which lie
along the z-axis are typical examples. The Elasticity Matrix for this condition can be
readily obtained by starting with the 3-dimensional case (Eq.3.6) and deleting the rows
and columns corresponding to εzz , εzx and εzy. This yields:

⎧σ xx ⎫ ⎧ε xx ⎫ ⎡1 − ν v 0 ⎤
⎪ ⎪ ⎪ ⎪ ⎢ ⎥
⎨σ yy ⎬ = [D ]⎨ε yy ⎬, [D ] =
E
⎢ v 1 −ν 0 ⎥ (3.44a)
⎪σ ⎪ ⎪ε ⎪ (1 + ν )(1 − 2ν )
⎩ xy ⎭ ⎩ xy ⎭ ⎢⎣ 0 0 0.5 − ν ⎥⎦

It should be noted that in plane strain, the stress in a direction perpendicular to the xy
plane (i.e., σzz ) is not zero. If needed, σzz can be calculated from the relation

σ zz − νσ yy − νσ xx
ε zz = 0 = (3.44b)
E

Note that plane strain situation is applicable when modelling a section through a ‘thick’
body.

Plane Stress
Unlike plain strain case, plane stress is a condition that prevails in a 'thin’ sheet where the
stresses in the perpendicular direction are zero, i.e., σzz=σzx= σzy=0. However, the strain,
εzz is not zero (but εzx= εzy=0). Again, starting with the general 3D case, one can delete
the rows and columns of the [D] matrix corresponding to εzx and εzy, then express the εzz
in terms of εxx and εyy. The resulting stress-strain relationship becomes:

KYŞ FEM: Chapter 3 19

 ⎧σ xx ⎫ ⎧ε xx ⎫ ⎡ ⎤
⎢ 1 ν 0 ⎥
⎪ ⎪ ⎪ ⎪
⎨σ yy ⎬ = [D ]⎨ε yy ⎬, [D ] =
E
⎢ν 1 0 ⎥ (3.45)
⎪σ ⎪ ⎪ε ⎪ (1 − ν 2 ) ⎢ 1 −ν ⎥
⎩ xy ⎭ ⎩ xy ⎭ ⎢0 0 ⎥
⎣ 2 ⎦

3.5 Example 2: Triangular Plane Stress Element

Consider the following plane stress problem

Given:
y
C

50 N/cm2

30 cm Thickness h=2 cm
E=20x106 N/cm2
ν=0.25

A B x

10 cm 10 cm

Find the displacements at points B and C and the reaction forces.

Fig.3.10 Plane stress problem with a triangular element.

Step 1: Idealisation
Use a single 3-noded plane stress element.

Step 2: Discretisation

Consider a general element e as shown in Fig.3.11. Note that this element is

- 3-noded,
- 2 degrees-of-freedom per node
- 6 degrees-of-freedom per element.

KYŞ FEM: Chapter 3 20

 v3
y
3
u3

v1
Typical element e

u1 v2
1

u2
2
x
Fig.3.11 A typical 3-noded triangular element

As for other elements, let us assume an interpolation function (that is the 'form' of
distribution within the element) for the displacements within the element:

u ( x, y ) = α u x + β u y + γ u
v ( x, y ) = α v x + β v y + γ v (3.46)

Apply the nodal conditions in order to find the shape function matrix [N]:

u1 = α u x1 + β u y1 + γ u
u 2 = α u x2 + β u y2 + γ u
u 3 = α u x3 + β u y 3 + γ u (3.47a)

In matrix form:

⎡ x1 y1 1⎤ ⎧α u ⎫ ⎧ u1 ⎫
⎢x ⎪ ⎪ ⎪ ⎪
⎢ 2 y 2 1⎥⎥ ⎨β u ⎬ = ⎨u 2 ⎬ (3.47b)
⎢⎣ x3 y 3 1⎥⎦ ⎪⎩γ u ⎪⎭ ⎪⎩u 3 ⎪⎭

Using the three equations, we can solve for the unknowns α, β and γ as a function of
nodal displacements and nodal co-ordinates. We may solve for the unknowns using
Cramer's rule as:

u1 y1 1
u2 y2 1
u3 y3 1
αu = (3.48a)
∆
where

KYŞ FEM: Chapter 3 21

 x1 y1 1
∆ = 2 A = x2 y2 1 (3.48b)
x3 y3 1

Note that the determinant in Eq.(3.48b) is equal to twice the area of the triangle. In a
similar fashion, one can obtain the other unknowns. The result is:
1
αu = (( y 2 − y3 )u1 + ( y3 − y1 )u 2 + ( y1 − y 2 )u 3 )
2A
1
βu = (( x3 − x2 )u1 + ( x1 − x3 )u 2 + ( x2 − x1 )u 3 ) (3.49)
2A
1
γu = (( x2 y3 − x3 y 2 )u1 + (x3 y1 − x1 y3 )u 2 + ( x1 y 2 − x2 y1 )u 3 )
2A

Inserting Eq.(3.49) into Eq.(3.46) and rearranging terms, we can write:

u ( x, y ) = N 1u1 + N 2 u 2 + N 3u 3 (3.50)

where

a = y 2 − y3
1
N1 = (ax + by + c ) where b = x3 − x 2 (3.51a)
2A
c = x 2 y 3 − x3 y 2

a' = y 3 − y1
1
N2 = (a' x + b' y + c') where b' = x1 − x3 (3.51b)
2A
c' = x3 y1 − x1 y 3

a ' ' = y1 − y 2
1
N3 = (a' ' x + b' ' y + c' ') where b' ' = x 2 − x1 (3.51c)
2A
c' ' = x1 y 2 − x 2 y1

Following the same procedure, it can be shown that

v( x, y ) = N 1v1 + N 2 v 2 + N 3 v3 (3.52)

Hence the displacement field can be written in matrix form as:

KYŞ FEM: Chapter 3 22

 ⎧ u1 ⎫
⎪v ⎪
⎪ 1⎪
⎧u ( x, y )⎫ ⎡ N 1 0 N2 0 N3 0 ⎤ ⎪⎪u 2 ⎪⎪
⎨ ⎬=⎢ ⎨ ⎬ (3.53)
⎩ v ( x, y ) ⎭ ⎣ 0 N1 0 N2 0 N 3 ⎥⎦ ⎪v 2 ⎪
⎪u 3 ⎪
⎪ ⎪
⎪⎩ v3 ⎪⎭
[N] {u}

KYŞ FEM: Chapter 3 23

Step 3: Computation of Element Stiffness

According to the "principle of virtual work", the elemental stiffness matrix is given by

[k ] = ∫ [B]T [D][B]dVe (3.54)

Ve

The [D] matrix in above expression is already defined before for plane stress situation
and will be reproduced here:

⎧σ xx ⎫ ⎧ε xx ⎫ ⎡ ⎤
⎢ 1 ν 0 ⎥
⎪ ⎪ ⎪ ⎪
⎨σ yy ⎬ = [D ]⎨ε yy ⎬, [D ] =
E
⎢ν 1 0 ⎥ (3.55)
⎪σ ⎪ ⎪ε ⎪ (1 − ν 2
) ⎢ 1 − ν⎥
⎩ xy ⎭ ⎩ xy ⎭ ⎢0 0 ⎥
⎣ 2 ⎦

The [B] matrix which is an operator to obtain strains from displacements. The strains on
x-y plane are:

∂u ∂v ∂u ∂v
ε xx = , ε yy = , ε xy = + (3.56)
∂x ∂y ∂y ∂x

Now, using the displacement distribution in Eq.(3.53), the strains can be calculated as:

∂N 1 ∂N 2 ∂N
ε xx = u1 + u 2 + 3 u3
∂x ∂x ∂x
∂N ∂N 2 ∂N
ε yy = 1 v1 + v 2 + 3 v3 (3.57a)
∂y ∂y ∂y
∂N ∂N 2 ∂N ∂N ∂N 2 ∂N
ε xy = 1 u1 + u 2 + 3 u 3 + 1 v1 + v 2 + 3 v3
∂y ∂y ∂y ∂x ∂x ∂x

or in matrix form:
⎡ ∂N 1 ∂N 2 ∂N 3 ⎤ ⎧u1 ⎫
⎢ 0 0 0 ⎥ ⎪ v1 ⎪
⎧ε xx ⎫ ⎢ ∂x ∂x ∂x ⎥⎪ ⎪
⎪ ⎪ ∂N 1 ∂N 2 ∂N 3 ⎥ ⎪⎪u 2 ⎪⎪
⎨ε yy ⎬ = [B ]{u} = ⎢ 0
⎢ 0 0 (3.57b)
⎨ ⎬
⎪ε ⎪ ∂y ∂y ∂y ⎥ ⎪v 2 ⎪
⎩ xy ⎭ ⎢ ∂N ∂N 1 ∂N 2 ∂N 2 ∂N 3 ∂N 3 ⎥ ⎪u ⎪
⎢ 1 ⎥ 3
⎣⎢ ∂y ∂x ∂y ∂x ∂y ∂x ⎦⎥ ⎪⎪ v ⎪⎪
⎩ 3⎭

Using the shape functions in Eq.(3.51), the matrix [B] in above expression can be
evaluated explicitly. In doing so we obtain:

KYŞ FEM: Chapter 3 24

 ⎡a 0 a ' 0 a ' ' 0 ⎤
1 ⎢
[B] = ⎢0 b 0 b' 0 b' '⎥⎥ (3.57c)
2A
⎢⎣b a b' a ' b' ' a ' '⎥⎦

Note that the operator matrix [B] is independent of x and y in this case. Therefore, the
strains (and stresses) are constant over the whole element, hence this element is usually
called 'Constant Strain Triangle' element. As a result, the matrices involved in Eq.(3.54)
can be taken out of the integral as:

[k ] = [B]T [D][B]∫ dVe (3.58)

Ve

Furthermore, the volume integral above is simply the area multiplied by the thickness, h,
hence the element stiffness matrix for the Constant Strain Triangular element can be
written as:

[k ] = hA [B]T6 x3 [D]3 x3 [B]3 x 6 (3.59)

Unfortunately, writing the above expression explicitly is quite cumbersome. It will be

computed using specific numerical values in Step 4.

Step 4: Assembly of Equations for the Whole System

Let's recall our basic equation for an element obtained using the Principle of Virtual
Displacements:

[k ]{u} = { f } (3.60)

Note that there is only one element in our example, hence the elemental stiffness matrix
is equal to that of the whole 'assembly', i.e.

[K]=[k] (3.61)

First, let's compute the stiffness matrix in Eq.(3.59) using numerical values. The area of
the element is A=20*30/2=300 cm2 and the numerical values of the parameters in the [B]
matrix in Eq. (3.57c) are:

a = y 2 − y 3 = −30 cm a' = y 3 − y1 = 30 cm a' ' = y1 − y2 = 0 cm

b = x3 − x 2 = −10 cm b' = x1 − x3 = −10 cm b' ' = x2 − x1 = 20 cm (3.62)
c = x 2 y 3 − x3 y 2 = 600 cm 2 c' = x3 y1 − x1 y 3 = 0 cm 2 c' ' = x1 y2 − x2 y1 = 0 cm 2

Then, inserting the numerical values into the shape function matrix, [N], expressed in
Eq.(3.50) to Eq.(3.53) gives:

KYŞ FEM: Chapter 3 25

 1 ⎡(−30x −10y + 600) 0 (30x −10y) 0 (20y) 0 ⎤
[N ] = ⎢
600⎣ 0 (−30x −10y + 600) 0 (30x −10y) 0 (20y)⎥⎦
(3.62)

and the 'Displacement to Strain Matrix' [B] and the 'Elasticity Matrix' become:

⎡− 30 0 30 0 0 0⎤
1 ⎢
[B] = 0 − 10 0 − 10 0 20⎥⎥ (3.63)
600 cm ⎢
⎢⎣ − 10 − 30 − 10 30 20 0 ⎥⎦

⎡8 2 0 ⎤
21.333E 6 N ⎢
[D] = 2 ⎢
2 8 0⎥⎥ (3.64)
8 cm
⎢⎣0 0 3⎥⎦

The stiffness matrix can then be calculated numerically using Eq.(3.59):

[k ] = [K ] = hA [B]T6 x3 [D]3 x3 [B]3 x 6 (3.65)

which yields:

⎡7500 1500 − 6900 − 300 − 600 − 1200⎤

⎢ 3500 300 − 1900 − 1800 − 1600⎥⎥
⎢
⎢ 7500 − 1500 − 600 1200 ⎥
[K ] = 4444.44⎢ ⎥ N / cm (3.66)
⎢ 3500 1800 − 1600⎥
⎢ Sym 1200 0 ⎥
⎢ ⎥
⎢⎣ 3200 ⎥⎦

The elemental force vector {f} is composed of:

{f } = {f B }+ {f S }+ { fC } (3.67)

For our example, only the traction forces are specified, hence the elemental force vector
due to traction becomes:

{ f } = { f S } = ∫ [N ]T {t}dS e (3.68)
Se

The traction forces are zero everywhere except along the BC edge. Furthermore, the
traction forces along the BC edge are decomposed into the x- and y-directions and the
above integral can be written as:

KYŞ FEM: Chapter 3 26

 ⎡ N1 0 ⎤
⎢0 N 1 ⎥⎥
⎢
⎢N 0 ⎥ ⎧t x ⎫
{f } = ∫ ⎢ 2 ⎥ ⎨ ⎬dS e (3.69a)
B −C ⎢
0 N 2 ⎥ ⎩t y ⎭
⎢N3 0 ⎥
⎢ ⎥
⎣⎢ 0 N 3 ⎦⎥

where tx=cos(198.43)*50 N/cm2= -47.43 N/cm2, ty=sin(198.43)*50 N/cm2= -15.81

N/cm2. Moreover, the incremental area dSe along the edge can be written as the thickness
times the incremental length along the same edge illustrated below:
C

l
B
A

This leads to expressing the integral in Eq.(3.69a) as:

⎡ N1 0 ⎤
⎢0 N 1 ⎥⎥
⎢
⎢N 0 ⎥ ⎧− 47.43⎫
{f } = ∫ ⎢ 2 ⎥⎨ ⎬h dl (3.69b)
B −C ⎢
0 N 2 ⎥ ⎩ − 15.81⎭
⎢N3 0 ⎥
⎢ ⎥
⎣⎢ 0 N 3 ⎦⎥

It can be shown that the shape functions along the edge BC can be written with respect to
the local axis l as:

l l
N 1 = 0, N 2 = 1 − , N3 =
L L

Therefore, the integration is expressed in a simple form as shown below:

KYŞ FEM: Chapter 3 27

 ⎡ 0 0 ⎤
⎢ 0 0 ⎥⎥
⎢ l
⎢1 − 0 ⎥
⎢ L ⎥
l ⎥ ⎧− 47.43⎫
L
⎢
{ f } = ∫ ⎢ 0 1 − ⎥⎨ (3.69c)
⎬h dl
L ⎩ − 15.81⎭
0
⎢ l ⎥
⎢ 0 ⎥
⎢ L ⎥
⎢ 0 l ⎥
⎢⎣ L ⎥⎦

The result of the above integration yields the elemental force vector:

⎡0 0⎤
⎢0 0 ⎥⎥
⎢L ⎡ 0 0 ⎤ ⎧ 0 ⎫
⎢ 0⎥ ⎢ 0 0 ⎥ ⎥ ⎪ 0 ⎪
⎢2 ⎥ ⎢ ⎪ ⎪
L ⎥ ⎧− 47.43⎫ ⎢15.81 0 ⎥ ⎧− 47.43⎫ ⎪⎪− 1500⎪⎪
{ f } = ⎢⎢ 0 ⎨ ⎬h = ⎢ ⎥⎨ ⎬2 = ⎨ ⎬N (3.69d)
2 ⎥ ⎩ − 15.81⎭ ⎢ 0 15.81⎥ ⎩ − 15.81⎭ ⎪ − 500 ⎪
⎢L ⎥
⎢ 0⎥ ⎢15.81 0 ⎥ ⎪− 1500⎪
⎢2 ⎥ ⎢ ⎥ ⎪ ⎪
⎢⎣ 0 15.81⎥⎦ ⎪⎩ − 500 ⎪⎭
⎢0 L⎥
⎣⎢ 2 ⎦⎥

It should be noted that the external force vector for the whole system, {F}, includes
elemental forces as well as the -unknown- reaction forces, if any. It is therefore quite
appropriate to consider the external force vector as a sum (assembly) of elemental force
vectors plus the reaction forces as:

{F } = ∑ { f } + {R} (3.70)
e

It is also possible to include the external point forces in {R} if they are not already
included in the elemental force vectors as {fC}. We can therefore write the final stiffness
equation for our system as [K]{u}={F}:

KYŞ FEM: Chapter 3 28

 ⎡7500 1500 − 6900 − 300 − 600 − 1200⎤ ⎧ u1 ⎫ ⎧ 0 ⎫ ⎧ R1x ⎫
⎢ ⎥ ⎪ ⎪ ⎪ ⎪ ⎪R ⎪
⎢ 3500 300 − 1900 − 1800 − 1600⎥ ⎪ v1 ⎪ ⎪ 0 ⎪ ⎪ 1y ⎪
N ⎢ 7500 − 1500 − 600 1200 ⎥ ⎪⎪u 2 ⎪⎪ ⎪⎪− 1500⎪⎪ ⎪⎪ R2 x ⎪⎪
4444.4 ⎢ ⎥⎨ ⎬ ⎨ = ⎬ N + ⎨ ⎬
cm ⎢ 3500 1800 − 1600⎥ ⎪v 2 ⎪ ⎪ − 500 ⎪ ⎪ R2 y ⎪
⎢ Sym 1200 0 ⎥ ⎪u 3 ⎪ ⎪− 1500⎪ ⎪ R3 x ⎪
⎢ ⎥⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎢⎣ 3200 ⎥⎦ ⎪⎩ v3 ⎪⎭ ⎪⎩ − 500 ⎪⎭ ⎪⎩ R3 y ⎪⎭
(3.71)

Step 5-6: Boundary Conditions and Solution

The boundary conditions for the system are:

Node Natural B.C Essential B.C

Rx Ry u v
1 ? ? 0.0 0.0
2 0.0 ? ? 0.0
3 0.0 0.0 ? ?

Imposing these boundary conditions into Eq.(3.71) and deleting the rows and columns
corresponding to the degrees-of-freedom with zero displacements gives:

⎡ 7500 − 600 1200 ⎤ ⎧u 2 ⎫ ⎧− 1500⎫

⎪ ⎪ ⎪ ⎪
4444.44 ⎢⎢− 600 1200 0 ⎥⎥ ⎨u 3 ⎬ = ⎨− 1500⎬ (3.72)
⎢⎣ 1200 0 3200⎥⎦ ⎪⎩ v3 ⎪⎭ ⎪⎩ − 500 ⎪⎭

Then the solution of the above equation yields the unknown displacements:

⎧u 2 ⎫ ⎧− 0.0687 ⎫
⎪ ⎪ −3 ⎪ ⎪
⎨u 3 ⎬ = 10 ⎨− 0.3156⎬cm (3.73)
⎪v ⎪ ⎪− 0.0094⎪
⎩ 3⎭ ⎩ ⎭

Step 7: Supplementary Calculations

Once the displacements are known, the unknown reaction forces are solved using
equation (3.71). They are:

R1x=3000 N R1y= 2500 N R2y= -1500 N (3.74a)

Strains are calculated via

KYŞ FEM: Chapter 3 29

 ⎧ε xx ⎫ ⎧− 0.3438⎫
⎪ ⎪ −5 ⎪ ⎪
⎨ε yy ⎬ = [B ]{u} = 10 ⎨− 0.0312⎬ (3.74b)
⎪ε ⎪ ⎪− 0.9374⎪
⎩ xy ⎭ ⎩ ⎭

and the stresses are:

⎧σ xx ⎫ ⎧ε xx ⎫ ⎧− 75.0⎫
⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎨σ yy ⎬ = [D ]⎨ε yy ⎬ = ⎨− 25.0⎬ N / cm
2
(3.74c)
⎪σ ⎪ ⎪ε ⎪ ⎪− 75.0⎪
⎩ xy ⎭ ⎩ xy ⎭ ⎩ ⎭

KYŞ FEM: Chapter 3 30