Introductory Course

of

Finite Elements Method

Ayman Aboel-Fetoh Embaby

Prof. Department of Structural Engineering
Ain-Shams University
2018
(Draft)
 PREFACE

The finite elements method has become the most powerful and reliable analytical
approach in simulating the physical response of simple as well complex structures
and industrial components. Its privilege extends in dealing with dynamic and
nonlinear responses, either from both geometrical and material points of views.

The objective of the current introductory course is to acquire the students the
necessary background to enable in deriving stiffness matrix of different finite
elements from the energy point of view. More comprehensive reading are required
to cover different aspects.
1. Element characteristic matrix
In the field of structures mechanics, the element characteristic matrix is called stiffness matrix. It
relates the internal forces produced at the nodes with the corresponding nodal displacements.
The term nodes is referred to the element connectivity with other adjacent elements, supports, or
any other adjacent media. Three methods are normally used in deriving the element stiffness
matrix, namely Direct method, Energy method, and Weighted residual method.

The direct method is limited to simple types of structures having limited number of nodes, relying
on basic theories of structure mechanics. The Energy approach is considered as the most easier
approach in deriving stiffness matrix. It relies on the total system potential energy (weak form) of
the structures. The Weighted residual method is used in lieu of the Energy approach in the
absence of the energy function, while having the differential equations (strong form).

The present course is dedicated to the Energy approach along with the simple applications of the
direct method.

2. Direct Method
In this method the coefficient kij of the stiffness matrix (ith row, and jth column), which relates the
force corresponds to ith degree of freedom with unit deformation corresponds to jth degree of
freedom, is derived by assigning a unit deformation along jth degree of freedom, while assuming
zero deformations for the rest of degrees of freedom. Using structural compatibility along with
the equilibrium equations the forces related to the degrees of freedom which are required to
maintain the aforementioned deformed configuration can be calculated. These forces form the j th
column of the stiffness matrix, since they are all related to the unit deformation of the jth degree
of freedom. The term degrees of freedom is the number of independent deformation
components by which the deformed system can be described.

Repeating the above procedures, the whole columns of the stiffness matrix cab be derived. It is
worth noting that stiffness matrix is symmetrical, therefore, the above procedures results in
kij=kji.
Example 1

Using direct method, derive the stiffness matrix

for the shown spring system, noting that degrees
of freedom are horizontal displacement and
rotation of node 1, and 2.

Answer

The shown four degrees of freedom are (u1, θ1, u2, and θ2).

Step 1:

Assuming u1=1, and the rest (θ1,u2,θ2) are zero.

As can be noticed that the above assumed deformed configuration, the two springs will be
compressed. Using the free body diagram of the system, the forces produced assigning the first
degree of freedom a unit value, while keeping the rest with zero values are called F11, M11m F21,
and M21, where the second index (1) refers to first degree of freedom.

From horizontal equilibrium of each side of

the free body diagram the following can be
found:

F11=(K1+k2)

F21=-(K1+k2)

From sum of the moment about point 1=0

results in the following:

M11=(K2-K1)*a

From sum of the moment about point 2=0 results in the following:

M21=(K1-K2)*a

Therefore, the first column of the stiffness matrix Ki1 contains the following rows:

(K1+k2), (K2-K1)*a, -(k1+k2), and (k2-K1)*a

Step 2:

Assuming θ1=1, and the rest (u1,u2,θ2) are zero.

As can be noticed that according to the

above assumed deformed
configuration, the top spring will be
extended while bottom will be
compressed. Using the free body
diagram of the system, the forces
produced assigning the second degree
of freedom a unit value, while keeping
the rest with zero values are called F12,
M12, F22, and M22, where the second index (2) refers to second degree of freedom.

From horizontal equilibrium of each side of the free body diagram the following can be found:

F11=(K2-k1)*a

F21=-(K1-k2)*a

From sum of the moment about point 1=0 results in the following:

M11=(K1+K2)*a2

From sum of the moment about point 2=0 results in the following:

M21=-(K1+K2)*a2

Therefore, the second column of the stiffness matrix Ki2 contains the following rows:

(K1+K2)*a2,(K1+K2)*a2, -(K1-k2)*a, -(K1+K2)*a2

Repeating the above procedures for u2=1, and θ2=1, the overall stiffness matrix can be found as
following:

𝐾1 + 𝐾2 (𝐾2 − 𝐾1)𝑎 −(𝐾1 + 𝑘2) (𝐾2 − 𝐾1)𝑎

⎡ ⎤
(𝐾2 − 𝐾1)𝑎 (𝐾1 + 𝐾2)𝑎 (𝐾1 − 𝐾2)𝑎 −(𝐾1 + 𝐾2)𝑎
[𝐾] = ⎢ ⎥
⎢−(𝐾1 + 𝐾2) (𝐾1 − 𝐾2)𝑎 (𝐾1 + 𝐾2) (𝐾2 − 𝐾1)𝑎 ⎥
⎣ (𝐾1 − 𝐾2)𝑎 −(𝐾1 + 𝐾2)𝑎 (𝐾2 − 𝐾1)𝑎 (𝐾1 + 𝐾2)𝑎 ⎦
Problems

Derive stiffness matrices for the shown spring systems.

Derive stiffness matrix for the shown truss

system, having constant cross section.
3. System potential energy
Stable static structural system include potential energy (π). The potential energy consists of strain
energy (U) stored in the elements, depending to its stress and strain state, and work done (W) by
external load as a results of the system deformations. This can be expressed as follows:

π=U-W (1)

The strain energy is defined as follows:

𝑈= ∫ (𝜎) (𝜖)𝑑𝑣 (2)

Where (σ) and (ε) are stress and stress vectors at any infinitesimal volume (dv). It is worth noting
that the half appears in above integral is due to integration of any infinitesimal strain energy (dU)
stored in infinitesimal volume with stress state (σ) subjected to infinitesimal change of strain (dε)
as follows (for convenience one dimensional state (i.e. one stress and one strain will be used):

𝑈=∫ ∫ 𝑑𝑈 𝑑𝑣 = ∫ ∫ 𝜎 𝜀𝑑𝜖 𝑑𝑣 = ∫ ∫ 𝐸𝜖 (𝑑𝜖)𝑑𝑣 = ∫ 𝜀 𝐸𝑑𝑣 (3)

In general six components of stress and strain are included in structural elements. However, one or
two dimensional state may exist in case of 1-D structural elements (beams, frames, and link/rode
element) and 2-D (plane stress or plane strain) problems.

Strain energy of beam element

For a beam element subjected to deflection w , and having length L, and area A, the infinitesimal
volume dv=dAdx.

𝑈 = ∫ ∫ 𝜎𝜖 𝑑𝐴 𝑑𝑥 = ∫ ∫ 𝜎 𝑑𝐴𝑑𝑥 (4)

Substituting 𝜎 = in the above results in the following:

𝑈= ∫ ∫ ( ) 𝑑𝐴 𝑑𝑥 (5)

Since 𝐼 = ∫ 𝑦 𝑑𝐴, and EI w”=M, where w is the deflection, the strain energy is expressed as
follows:
𝑈= ∫ (𝑤 " ) 𝑑𝑥 (6)

Where w” is the curvature of (second derivative of the deflection) any infinitesimal length dx of the
element.

Strain energy of link (rode) element

For a link member of length L undergoing to axial displacement u, the strain energy U is expressed
as follows:

𝑈= ∫ 𝜎𝜀 𝑑𝑣 = ∫ 𝐸𝜀 𝑑𝑥 = ∫ ( ) 𝑑𝑥, 𝑤ℎ𝑒𝑟𝑒 𝜀 = (7)

3.1 Principle of Stationary Potential Energy

“Among all admissible configuration of conservative system, those satisfy the equation of
equilibrium make the potential energy (π) stationary with respect to small admissible
configuration”.

The admissible configuration has to satisfy the essential boundary conditions. The essential
boundary conditions ranges from the zeroth derivative of the displacement function up to (m-1) th
derivative, where m is the order of derivatives appears in the energy function. The conservative
system is that system whose potential energy is load-path independent (no plastic strain, nor
damping are exist.)

Assuming that a stable structural system is under equilibrium under the action of external loads
and state of deformations described by n independent degrees of freedom (Δ1, Δ2, Δ3, ……. Δn) .
The variation in the potential Energy dπ due to admissible variation of the degrees of freedom
(dΔ1, dΔ2, dΔ3, ……. dΔn) is described as follows:

𝑑𝜋 = 𝑑∇1 + 𝑑∇2 + 𝑑∇3 + ⋯ + (8)

∇ ∇ ∇ ∇

Since dΔ1, dΔ2, dΔ3, ……. dΔn are independent, the above variation dπ = 0 only if each term in the
above variation is equal to zero. Therefore, for equilibrium condition under the action of the
external load, and state of deformation described by (Δ1, Δ2, Δ3, ……. Δn ), the following
independent equations has to be satisfied:

=0 (9)
∇

3.2 Rayleigh-Ritz Method

For a continuous solid body subjected to the action of external loads, the task of choosing
admissible function that satisfies the governing differential equation (strong form) is almost
impossible, rather admissible field defined in terms of specific generalized coordinates or degrees
of freedom can be chosen to satisfy the above stationary principle of the potential energy (weak
form). Accordingly, describing the deformed body in terms of compatible and admissible
displacement field in terms of n independent degrees of freedom (or generalized coordinates),
along with using strain-displacement relationship as well as stress-strain relationship, n
independent equilibrium equations can be achieved using the stationary principle.

Rayleigh-Ritz approach may yield close results for deformation, but less accurate results may be
expected for higher derivatives (stress for example). This is attributed to the fact that the
admissible field satisfies only the essential boundary conditions, while the non-essential boundary
conditions are not satisfied.

A necessary condition for convergence to the exact solution using trial admissible fields is the
completeness. The completeness is that the polynomial describing the displacement field must
include all lower order of the polynomial including the constant. This to ensure the capability of
capturing all lower order deformation, even rigid body mode, under arbitrary loading action. In
other words, increasing the order of polynomial may fail to converge to exact solution of linear
deformation if such term was missing in the displacement field.

Pascal’s Triangle can be used to construct a complete polynomials for triangular and square
elements. Pi is used to indicate those required terms in triangular, while Qi are for rectangular
elements. For example P1 need three terms for linear strain triangular (constant strain), P2
requires 6 terms for parabolic distribution of displacement (linear strain. Also, Q1 require four
terms in the polynomial for linear displacement (4-node element), and Q2 require 9 terms in the
polynomial for complete parabolic presentation of displacement filed (9-node element), Q3
requires 16 terms to capture the cubic displacement distribution (16-node element).

Pascal’s Triangle for 2-D problems

Applying Rayleigh-Ritz approach in small problems such as single beam or link member requires
choosing a direct admissible function without satisfying the completeness condition, while in a
large problems, choosing such admissible function will be difficult. Step-wise Rayleigh-Ritz
approach is used in which structure is divided into finite elements in which the completeness
condition is satisfied in describing the piecewise polynomial, while the admissibility (satisfying
essential boundary conditions) is imposed in the solution of the equilibrium conditions (equations)
through assigning such conditions in the solution scheme.
Example 1

For the shown axially loaded member, with area

A, and Length L, calculate the free-end axial
displacement under the actions of the following
simultaneous loads:

 Linearly varying axial load q=cx, where x is measured from fixed end.
 Concentrated axial tensile load P @ mid span

Solution

Assumed admissible function

The essential boundary condition for such a Co element is that axial displacement @ x=0 =0.

Simple displacement function u=ax can be used, as it satisfies the above boundary condition.
Whoever, to be able to capture higher order of displacement field due to the linearly varied load a
higher order displacement field is adopted as follows:

u=ax+bx2,

It can be noticed that @ x=0 u=0.

Strain energy

𝐴𝐸 𝑑𝑢
𝑈= ( ) 𝑑𝑥
2 𝑑𝑥

du/dx= a1+2x;

𝐴𝐸
𝑈= (𝑎 + 2𝑏𝑥) 𝑑𝑥
2

Performing the integration, U=0.5AE*(a2L + 2ab L2 + 1.333b2L3)

Work done be External loads

𝑊= (𝑐𝑥)(𝑎𝑥 + 𝑏𝑥 )𝑑𝑥 + 𝑃(0.5𝑎𝐿 + 0.25𝑏𝐿 )

Performing the integration results in the following:

 W=c(aL3+0.25bL4) + 𝑃(0.5𝑎𝐿 + 0.25𝑏𝐿 )

Accordingly, the potential energy π can be expressed as follows:

𝜋 = 0.5AE ∗ (a2L + 2ab L2 + 1.333 ∗ 𝑏 L3) + c (aL3+0.25bL4) + 𝑃(0.5𝑎𝐿 + 0.25𝑏𝐿 )

According to stationary principle, the following two independent equations are to be satisfied:

𝜕𝜋
=0
𝜕𝑎

𝜕𝜋
=0
𝜕𝑏

Solving these two equations for the two unknowns, namely a, and b, in terms of the A, E and L,
result in getting the required displacement field (u=ax+bx2), Substituting for x=L, gives the end
displacement.

(Student to complete the answer)

Example 2

For the shown simply supported beam,

calculate deflection (y) and slope (θ) at
mid span under the actions of the
following simultaneous loads:

 Uniformly distributed load (w t/m’)

 Concentrated downward load P @ mid-span
 Concentrated clockwise moment M @ left end.

Solution

Assumed admissible function

Several admissible functions for the beam deflection (y) can be used to satisfy the boundary
conditions, namely deflection =0 at x=0, and at L=0. The following two function can be used:

y=ax(L-x), y’=aL-2ax, y”=-2a

y= a sin∏x/L, y’=-a(∏/L)cos(∏x/L), y” =-a(∏/L)2sin(∏x/L)

as can be noticed both equations satisfy y=0 @ x=0, and x=L.

Strain energy

𝐸𝐼
𝑈= (𝑦 " ) 𝑑𝑥
2

For the 1st function y”=-2a, and for the second function, y”=-a(∏/L)2sin (∏x/L).

Accordingly, strain energy for the two optional functions is expressed as follows:

𝐸𝐼
𝑈= 4𝑎 𝑑𝑥 = 2𝐸𝐼𝐿𝑎
2

𝐸𝐼 𝐸𝐼 𝐿 𝐸𝐼
𝑈= 𝑎 (∏/L) (sin (∏x/L) 𝑑𝑥 = ∗ 𝑎 (∏/L) ∗ = 𝑎 (∏)
2 2 2 4𝐿

∏
Since (sin (∏x/L) = 0.5(1 − cos( )), it can be found that ∫ (sin (∏x/L) 𝑑𝑥 =

Work done by external load

Work done (W) by 1st equation is expressed as follows:

∫ 𝑤 ∗ 𝑎𝑥 (𝐿 − 𝑥 )𝑑𝑥 + P*a*0.5L*(L-0.5L)+M*aL = awL3/6 + 0.25aPL-2aM

Work done (W) by 2nd equation is expressed as follows:

∏
∫ 𝑤a sin( )𝑑𝑥 + aP - M* a(∏/L) = -2waL/∏ + aP + M* a(∏/L)

Total potential energy (π)

For the 1st function π is expressed as follows:

π= 2𝐸𝐼𝐿𝑎 -[ awL3/6 + 0.25aPL-2aM]

For the 2nd function π is expressed as follows:

π= ∗ 𝑎 (∏/L) ∗ = 𝑎 (∏) – [-2waL/∏ + aP - M* a(∏/L)]

in either optional function dπ/da=0, results in solving for a in terms of EI, L, w, P and M (Student
may complete the solution). Accordingly the deflection (w) is determined, and so slope at any
point.

Problem

For the shown cantilever beam, calculate

deflection (y) and slope (θ) at free end under
the actions of the following simultaneous
loads:

 Uniformly distributed load (w t/m’)

 Concentrated downward load P @ mid-span
 Concentrated clockwise moment M @ free end.

Assumed admissible function

Several admissible functions for the cantilever beam deflection (y) can be used to satisfy the
boundary conditions, namely deflection & slope =0 at x=0. The following two function can be
used:

y=ax2, ax2+bx3, and so on of higher order polynomials.

Student is to complete the solution.

3.3 Discrete Rayleigh-Ritz Method

In the above applications of Rayleigh-Ritz approach the success of the solution depends on
adopting displacement functions (fields) which satisfy the essential (kinematic) boundary
conditions. Complicated loading conditions, having several types and locations of the applied load
results in achieving to rational adopted displacement fields to be a complicated subject.
Moreover, more complex structure, rather than one dimensional element, might result in an
impossible achievement to even close solution.
Therefore, the Rayleigh-Ritz approach was extended to be able to work with the above difficulties.
This was achieved by introducing the concept of dividing the structural problem into several
discrete elements, connected to each other at nodes. The displacement field of each element
does not need to satisfy the system essential boundary conditions, rather the essential boundary
conditions can be then imposed at the required nodes to match the physical problems. In other
words, the complicated and unknown high order displacement field, which results from the
complicated loading and geometrical configurations of the problems, is divided into piecewise low
order functions between specific points (nodes). The minimum potential energy, which results in
equilibrium equations, can be applied to the whole assembly.

According to the above general approach, different types of elements can be derived to deal with
one, two, and three dimensional structural problems. Not only this energy approach is restricted
to beams or truss elements but is extended to deal with two and three dimensional continuum
problems, such as floor slabs, walls, dams, and the complicated structures.

Two conditions are implicitly satisfied, namely the compatibility and equilibrium conditions. The
compatibility is satisfied through connecting the elements at nodes and assigning the same
deformations at the nodes to all connected elements to this nodes (i.e. no overlaps between
nodes exists). The second conditions (equilibrium) is satisfied through solving the equilibrium
equations at all nodes. These equilibrium conditions is the product of applying the stationary
energy principle to the system.

One prime steps is essential to achieved useful and meaningful finite element analytical approach.
This is to replace the generalized coordinates (parameters of the discrete element shape
functions) by real nodal deformations which is called degrees of freedom (DOF). This approach as
will be discussed later for each element results in what is called shape functions which relate the
displacement fields at any point within the element in terms of nodal degree of freedom, rather
than the polynomial parameters.

For a problem having one dependent fields variable, if 2m be the highest order of derivatives of
the variable appears in the governing equilibrium differential equation, only m derivative appears
in the corresponding energy functional. The essential (kinematic) boundary conditions are from
zeroth derivative (field variable itself) up to m-1 derivatives. The non-essential (natural or static)

boundary conditions contain m to 2m-1 derivatives of the field variable.

Problem Truss Beam

Differential equation AEu,xx+q=0 EIw,xxxx –q=0

2m,m-1,2m-1 2,0,1 4,1,3

Essential boundary condition u w & w,x

Non-essential boundary conditions σ or ε M=EI w,xx, & Q=EI w,xxx

In case of two dimensional problems where two field variables exist, the previous provisions are
still applicable. For example for two dependent field variables u, and v (depending on x & y
coordinates) their second derivatives, u,xx, u, xy, u,yy, v,xx, v,yy, and v, xy appear in the equilibrium
differential equations. In other words, 2m=2.therfore, since m-1=0, only field variables u, and v
are used in the nodal connectivity and essential boundary conditions. The non-essential boundary
conditions up to 2m-1 (1st derivatives of u, and v) are used in the non-essential boundary
conditions.

It is worth noting that giving the energy functional in terms of admissible dependent variable
along with applying the stationary principal and calculus of variation results in both the
differential equation and the non-essential boundary conditions. This can be explained in the
following example:

The Energy functional for axially loaded bar fixed at one end and free at the other end and axially
loaded with load q can be expressed as following:

𝐴𝐸
𝜋= 𝑢, 𝑑𝑥 − 𝑞𝑢𝑑𝑥
2

Invoking the stationary principal

𝛿𝜋 = 𝐴𝐸 𝑢, 𝛿𝑢, 𝑑𝑥 − 𝑞𝛿𝑢𝑑𝑥 = 0

We note that variations and the differentiations follow the same rule.

Integrating by parts the first term of the above variation, results in the following:

[𝐴𝐸𝑢, 𝛿𝑢] − 𝐴𝐸 𝑢, 𝛿𝑢𝑑𝑥 − 𝑞𝛿𝑢𝑑𝑥 = 0

Since u is admissible its variation is also admissible, i.e δu=0 at x=0; ac

cordingly the above results in the following:

AEu,xx+q=0 which is the equilibrium differential equation

U,x = 0 at x=L which is the non-essential boundary condition

From the above the energy functional has to be integrated by parts m time to get rid of the
derivative of the displacement field. This procedures along with applying the essential boundary
condition results in both the differential equation and the non-essential boundary conditions.

The rate of convergence of the finite element mesh with displacement field of order n (=1 for for
two node truss element, and four node plane stress/strain, =2 for three node truss element, and 9
node plane stress/strain) and element size h is equal to hn+1.

Problem

Follow the integration by parts twice to get the equilibrium differential equation of simply
supported beam loaded by uniform disturbed load .

3.4 Two-Node Truss-Link Element

Displacement function

The 2-D displacement function (axial displacement) (u)

of the two-node truss element is described as follows:

𝑢 = 𝑎1 + 𝑎2𝑥 (10)

Or in a matrix form

𝑎1
𝑢 = (1 𝑥) ; 𝑢 = (𝑃) ∗ (𝑔) ∗ (11)
𝑎2
Where (P) is the polynomial vector, and (g) is the generalized coordinates (constants) vector.

As can be noticed only two generalized coordinates (polynomial constants a1 to a2) are used.
These constants have no direct/physical engineering meanings and need to be replaced by nodal
displacement.

Substituting the for nodal coordinates x=-L & x=L for node 1, and 2, respectively, in the above two
equations results in the following boundary conditions:

𝑢1 = 𝑎1-a2L (12)

𝑢2 = 𝑎1 + 𝑎2𝐿 (13)

Or in a matrix form,

𝑢1 1 −𝐿 𝑎1
= (14)
𝑢2 1 𝐿 𝑎2

In a matrix form

(∆) ∗ = [𝐴] ∗ (𝑔) ∗ (15)

Therefore,

(𝑔) = [𝐴] (∆) (16)

Substituting (g) in (w) function results in the following:

𝑢 = (1 𝑥 ) [𝐴] (∆), 𝑜𝑟 𝑢 = (𝑁) (∆), 𝑤ℎ𝑒𝑟𝑒, (𝑁) = (𝑁1 𝑁2 ) (17)

Since truss/link element is categorized as Co-element, having the zero derivative of the
displacement field as nodal degrees of freedom, Lagrangian interpolation function, rather than
the above relatively complicated procedures can be used as follows:

( )
𝑁𝑖 = , 𝑖≠𝑗 (18)
( )

( ) ( ) ( ) ( )
𝑁1 = = , 𝑁2 = = (19)
( )

As can be noticed from the above equations and Fig. 1 that substituting for i th nodal coordinate
results in a unit value for the ith shape function while substituting with jth coordinate in the ith
shape function, results in a zero value.
Strain Energy

The strain energy of the link element can be described as follows:

𝑈= ∫ 𝜀 𝑑𝑥 = ∫ ( ) 𝑑𝑥 (20)

𝑢1
𝜀= = (𝑁1, 𝑁2, ) (21)
𝑢2

Or 𝜀 = [𝐵](∆) (22)

Performing the above derivatives, the strain-nodal displacement matrix [B] 1*2 is described as
follows:

[𝐵] = [ ] (23)

In a matrix form ,U can rewritten as follows:

𝑈= ∫ (∆) [𝐵] [𝐵](∆)𝑑𝑥 (24)

Stiffness matrix

Since 1st derivative of U results in [K](Δ), the stiffness matrix of the truss element is defined as
follows:

[𝐾] = 𝐴𝐸 ∫ [𝐵] [𝐵]𝑑𝑥 (25)

3.5 Three-Node Truss-Link Element

Displacement function

The 2-D displacement function (axial

displacement) (u) of the two-node truss element
is described as follows:

𝑢 = 𝑎1 + 𝑎2𝑥 + 𝑎3𝑥 (26)

Or in a matrix form

𝑎1
𝑢 = (1 𝑥 𝑥 ) 𝑎2 ; 𝑢 = (𝑃) ∗ (𝑔) ∗ (27)
𝑎3

Where (P) is the polynomial vector, and (g) is the generalized coordinates (constants) vector.

As can be noticed only two generalized coordinates (polynomial constants a1 to a3) are used.
These constants have no direct/physical engineering meanings and need to be replaced by nodal
displacement.

Substituting the for nodal coordinates x=-L , x=L , and x=0 for node 1, and 2, and 3 respectively, in
the above two equations results in the following boundary conditions:

𝑢1 = 𝑎1 − 𝑎2𝐿 + 𝑎3𝐿 (28)

𝑢2 = 𝑎1 + 𝑎2𝐿 + 𝑎3𝐿 (29)

𝑢3 = 𝑎1 (30)

Or in a matrix form,

𝑢1 1 −𝐿 𝐿 𝑎1
𝑢2 = 1 𝐿 𝐿 𝑎2 (31)
𝑢3 1 0 0 𝑎3

In a matrix form

(∆) ∗ = [𝐴] ∗ (𝑔) ∗ (32)

Therefore,

(𝑔) = [𝐴] (∆) (33)

Substituting (g) in (w) function results in the following:

𝑢 = (1 𝑥 𝑥 ) [𝐴] (∆), 𝑜𝑟 𝑢 = (𝑁) (∆), 𝑤ℎ𝑒𝑟𝑒, (𝑁) = (𝑁1 𝑁2 𝑁3) (34)

Since truss/link element is categorized as Co-element, having the zero derivative of the
displacement field as nodal degrees of freedom, Lagrangian interpolation function, rather than
the above relatively complicated procedures can be used as follows:
 ( )
𝑁𝑖 = ∏ , , 𝑖≠𝑗 (35)
( )

( )( ) ( )
𝑁1 = = (36)
( )( )

( ( )) ( ) ( )
𝑁2 = = (37)
( ( ) ( )

( ( )) ( ) ( )( )
𝑁3 = ( )) (
= (38)
( )

Strain Energy

The strain energy of the link element can be described as follows:

𝑈= ∫ 𝜀 𝑑𝑥 = ∫ ( ) 𝑑𝑥 (39)

𝑢1
𝜀= = (𝑁1, 𝑁2, 𝑁3, ) 𝑢2 (40)
𝑢3

Or 𝜀 = [𝐵](∆) (41)

Performing the above derivatives, the strain-nodal displacement matrix [B] 1*3 is described as
follows:

[𝐵] = [ ] (42)

In a matrix form ,U can rewritten as follows:

𝑈= ∫ (∆) [𝐵] [𝐵](∆)𝑑𝑥 (43)

Stiffness matrix

Since 1st derivative of U results in [K](Δ), the stiffness matrix of the truss element is defined as
follows:

[𝐾] = 𝐴𝐸 ∫ [𝐵] [𝐵]𝑑𝑥 (44)

3.6 Two-Node Beam Element
Displacement function

The 2-D displacement function (in-plane

deflection) of the two-node beam element is
described as follows:

𝑤 = 𝑎1 + 𝑎2𝑥 + 𝑎3𝑥 + 𝑎4𝑥 (45)

Or in a matrix form

𝑎1
𝑎2
𝑤 = (1 𝑥 𝑥 𝑥 ) ; 𝑤 = (𝑃) ∗ (𝑔) ∗ (46)
𝑎3
𝑎4

Where (P) is the polynomial vector, and (g) is the generalized coordinates (constants) vector.

As can be noticed only four generalized coordinates (polynomial constants a1 to a4) are used.
These constants have no direct/physical engineering meanings and need to be replaced by nodal
displacement. Since Beam element is categorized as C1-element, having the first derivative of the
deflection (rotation) as nodal degrees of freedom along with nodal deflection, the rotation is
described as follows:

= 𝜃 = 𝑎2 + 2𝑎3𝑥 + 3𝑎4𝑥 (47)

Substituting the for nodal coordinates x=0 & x=L for node 1, and 2, respectively, in the above two
equations results in the following boundary conditions:

𝑤1 = 𝑎1 (48)

𝜃1 = 𝑎2 (49)

𝑤2 = 𝑎1 + 𝑎2𝐿 + 𝑎3𝐿 + 𝑎4𝐿 (50)

𝜃2 = 𝑎2 + 𝑎3𝐿 + 𝑎4𝐿 (51)

Or in a matrix form,

𝑤1 1 0 0 0 𝑎1
𝜃1 0 1 0 0 𝑎2
= (52)
𝑤2 1 𝐿 𝐿 𝐿 𝑎3
𝜃2 0 1 2𝐿 3𝐿 𝑎4
In a matrix form

(∆) ∗ = [𝐴] ∗ (𝑔) ∗ (53)

Therefore,

(𝑔) = [𝐴] (∆) (54)

Substituting (g) in (w) function results in the following:

𝑤 = (1 𝑥 𝑥 𝑥 ) [𝐴] (∆), 𝑜𝑟 𝑤 = (𝑁) (∆), (55)

𝑤ℎ𝑒𝑟𝑒, (𝑁) = (𝑁1 𝑁2 𝑁3 𝑁4)

(N) is the shape function vector (in terms of x-coordinate of any specified point) describing
deflection w of the specific point. Multiplying ( 1 x x2 x3)T by [A]-1, results in the following
definitions :

( ) ( )
𝑁1 = (56)

( )
𝑁2 = (57)

( )
𝑁3 = (58)

( )
𝑁4 = (59)

Stiffness matrix

The strain energy U of the beam element is described as follows:

𝑈= ∫ (𝑤 " ) 𝑑𝑥 (60)

𝑤" = (61)

In a matrix form, w” is described as follows:

𝑤 " = (𝑁1" 𝑁2" 𝑁3" 𝑁4" ) (∆) 𝑤ℎ𝑒𝑟𝑒, 𝑁𝑖 " = (62)

𝑤 " = (𝐵)(∆) (63)

(B)1*4 is the strain-nodal displacement vector. Accordingly, U can be determined as follows:

𝑈= ∫ (∆) (𝐵) (𝐵 )(∆) 𝑑𝑥 (64)

Also,

= [𝐾](∆) (65)
∆

Where [K]4*4 is the stiffness matrix, which is described as follows:

[𝐾] = 𝐸𝐼 ∫ (𝐵) (𝐵) 𝑑𝑥 (66)

3.7 Coordinate Transformation

Generally, the 1-D elements such as truss or beam elements are connected together in a way that
different local elements axes would require unifying the directions of the coordinate system to be
allow compatible process of summing up the forces, displacements or stiffness coefficients.
Therefore, a global coordinate system (X,Y, and Z) has to be introduced to which local coordinates
systems of different element orientations (x,y, and z) are to be transformed.

In the following the transformation processes of vectors, representing displacements or forces,

and matrix, representing stiffness matrix are presented.

3.7.1 Transformation of vectors

Considering that li,mi, and ni are the directions cosines of a x’ ,y’, and z’ axes with respect to x,y
and z, axes, respectively, where i=1, 2, and 3, is referred to local axes x’, y’, and z’, respectively.
For a vector V whose
components on x, y, and z
axes are u,v, and w,
respectively. This vector may
represent a displacement,
rotation, force, or moment
vectors. The component u’ of
the vector along x’ axis may
be expressed as sum of the
projections of u, v, and w components on x’ axis that is u’=l1u+m1v+n1w. Similar process can be
used to get v’, and w’. Accordingly, the components u’, v’, and w’ are expressed as following:

𝑢′ 𝑙 𝑚 𝑛 𝑢 𝑢′ 𝑢
𝑣′ = 𝑙 𝑚 𝑛 𝑣 𝑜𝑟, 𝑣′ = [𝜆] 𝑣 (67)
𝑤′ 𝑙 𝑚 𝑛 𝑤 𝑤′ 𝑤

Where [λ] is called the rotation matrix. The matrix [λ] is orthogonal where its inverse is equal to
its transpose. Therefore, a reverse process can be incorporated as follows:

𝑢 𝑢′
𝑣 = [𝜆 ] 𝑣′ (68)
𝑤 𝑤′

Accordingly, mutual transformations between two sets of axes can be conducted, giving direction
cosines of one set to the other.

3.7.2 Transformation of stiffness matrix

Consider that stiffness matrix [k] is developed in local coordinate system (x’,y’,z’), and it is
required to get its corresponding matrix in a global coordinate system (x,y,z). Assuming that
direction cosines of x’,y’, and z’ with respect to x,y, and z are known as [λ].

Equilibrium equation in the local axes is given as follows:

[𝑘′](𝑑′) = (𝑅′) (69)

Since (d)=[λ]T(d’) (or (d’)=[λ](d)), (R)=[λ]T(R’), or (R’) =[λ](R), the above equation can be rewritten
as following:
[𝑘 ][λ](d) = [λ](R) = [λ] [𝑘 ][λ](d) = (𝑅), 𝑜𝑟 [𝑘](𝑑) = (𝑅), (70)
Therefore, the stiffness matrix in the global coordinate system ([k]) is expressed as follows:
[𝑘] = [λ] [𝑘 ][λ] (71)
𝑤ℎ𝑒𝑟𝑒 [𝑘] = [λ] [𝑘 ][λ]
3.7.3 Transformation of strain
It can be found the following transformation:

(ε’)=[Tε](ε) (72)

Where

𝑇 𝑇
[𝑇 ] = (73)
𝑇 𝑇
Where,

𝑙 𝑚 𝑛 𝑙 𝑚 𝑚 𝑛 𝑛 𝑙 2𝑙 𝑙 2𝑚 𝑚 2𝑛 𝑛
[𝑇 ] = 𝑙 𝑚 𝑛 ; [𝑇 ] = 𝑚
𝑙 𝑚 𝑛 𝑛 𝑙 ; [𝑇 ] = 2𝑙 𝑙 2𝑚 𝑚 2𝑛 𝑛
𝑙 𝑚 𝑛 𝑙 𝑚 𝑚 𝑛 𝑛 𝑙 2𝑙 𝑙 2𝑚 𝑚 2𝑛 𝑛

𝑙 𝑚 +𝑙 𝑚 𝑚 𝑛 +𝑚 𝑛 𝑛 𝑙 +𝑛 𝑙
[𝑇 ] = 𝑙 𝑚 + 𝑙 𝑚 𝑚 𝑛 +𝑚 𝑛 𝑛 𝑙 +𝑛 𝑙 (74)
𝑙 𝑚 +𝑙 𝑚 𝑚 𝑛 +𝑚 𝑛 𝑛 𝑙 +𝑛 𝑙

Transformation of Stress

Considering stress (σ), the transformation is conducted as following:

(𝜎) = [𝑇 ](𝜎 ), 𝑎𝑛𝑑 (𝜎 ) = [𝑇 ] (𝜎) (75)
Where
𝑇 2𝑇
[𝑇 ] = (76)
0.5𝑇 𝑇

3.7.4 Transformation of Elasticity Matrix [E]

The Elasticity matrix [E’] can be transformed to [E} as following:

[𝐸] = [𝑇 ] [𝐸 ][𝑇 ] (77)

3.8 Four Node Plane stress /strain element

Displacement Function

Since the element contains four

nodes, having four nodal
displacement in x, and y directions,
namely u, and v, respectively.
Therefore four generalized
coordinates (constants) are used in
describing the displacement in each
direction. In other words, the four nodal displacement in each direction shall substitute the
corresponding four constants. From the above the u, and v displacement at any specified point
having x, and y coordinates can be described as follows:
 u=a1+a2x+a3y+a4xy (78)

v=b1+b2x+b3y+b4xy (79)

Displacement in terms of nodal displacement

Since the generalized coordinates ai, and bi have no engineering physical meaning, the task will be
to substitute them with the nodal displacement ui, and vi (where i is the total number of the
nodes). For this purpose the following will be incorporated:

𝑢 = {𝑃}{𝑔} (80)

Where {P}= {1 x y xy}, and {g}= {a1 a2 a3 a4} (81)

Considering element dimension 2L*2m, and substituting for coordinates of nodes results in the
following:

u1=a1+a2(-L)+a3*(-m)+a4 (Lm) (82)

u2=a1+a2(L)+a3*(-m)+a4 (-Lm) (83)

u3=a1+a2(L)+a3*(m)+a4 (Lm) (84)

u4=a1+a2(-L)+a3*(m)+a4 (-Lm) (85)

or in a matrix for as following:

𝑢1 1 −𝐿 −𝑚 𝐿𝑚 𝑎1
𝑢2 = 1 𝐿 −𝑚 −𝐿𝑚 𝑎2
(86)
𝑢3 1 𝐿 𝑚 𝐿𝑚 𝑎3
𝑢4 1 −𝐿 𝑚 −𝐿𝑚 𝑎4

Or

{𝑢𝑖} = [𝐴]{𝑔} (87)

Or

{𝑔} = [𝐴] {𝑢𝑖} (88)

Substituting {g} in u function, results in the following:

𝑢 = [𝐴] {𝑃}{𝑢𝑖} (89)

Or

𝑢1
𝑢2
𝑢 = {𝑁1 𝑁2 𝑁3 𝑁4} (90)
𝑢3
𝑢4

Same procedures are used for displacement v, resulting in the following:

𝑣1
𝑣2
𝑣 = {𝑁1 𝑁2 𝑁3 𝑁4} (91)
𝑣3
𝑣4

Displacements {Δ}2x1 (which are u,v) at any point within the element can be determined in terms
of nodal displacements (d)8x1 using the shape functions Matrix [N]2x8 (which are N1 to N8) as
following:

𝑢1
𝑣1
⎛𝑢2⎞
𝑢 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 0 ⎜ 𝑣2⎟
= ⎜ ⎟ (92)
𝑣 0 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 ⎜𝑢3⎟
⎜𝑣3⎟
𝑢4
⎝𝑣4⎠

Or

{𝛥} = [𝑁] {𝑑} (93)

The above procedures in getting shape function, through inversing the [A] is complicated,
especially for high order element. In Lieu of this, Lagrangian interpolation functions can be used as
following:

!( ) !
𝑁 = ∗ 𝑖≠𝑗 (94)
( )

Where ! indicates multiplier.

Applying Lagrangian interpolation for the 4-node element the shape functions can be found as
follows:

(𝐿 − 𝑥)(𝑚 − 𝑦) (𝐿 + 𝑥)(𝑚 − 𝑦) (𝐿 + 𝑥)(𝑚 + 𝑦) (𝐿 − 𝑥)(𝑚 + 𝑦)

𝑁1 = ; 𝑁2 = ; 𝑁3 = ; 𝑁4 =
4𝐿𝑚 4𝐿𝑚 4𝐿𝑚 4𝐿𝑚
For fast formation of the previous shape function, note that signs of x, and y, in i th N is according to
the location of the ith point.

It is worth noting that substituting with ith coordinates in Ni result in a unity, while substituting
with coordinates of other nodes results in zero value of the ith shape function.

Lagrangian interpolation functions can be adopted only for Co elements (elements having up to
zeroth derivatives of the displacement as nodal degrees of freedom).

It is worth noting that higher order element can be used, resulting in the ability to capture second
order displacement distribution. For example for nine-node element the displacement u and v are
described as follows:

u=a1+a2x+a3y+a4xy+a5x2+a6y2+a7x2y+a8xy2+a9x2y2 (95)

v=b1+b2x+b3y+b4xy+b5x2+b6y2+a7x2y+a8xy2+a9x2y2 (96)

Strain-Displacement Relationship

The strain at any specific point within the element {ε}3x1 (εx, εx, γxy) is defined in terms of the
displacements at the specified point {Δ} (or u & v) using the following differential operator [𝛛]

:
 ⎡ 0⎤
𝜀
⎢ ⎥ 𝑢
𝜀 = ⎢0 ⎥ 𝑣 (97)
𝛾 ⎢ ⎥
⎣ ⎦

Or

It is worth noting that in-plane shear strain γxy is defined as change in the right angle of the prism,
after deformation.

{𝜀} = [𝜕] {𝛥} (98)

Strain-nodal displacement matrix ([B]-matrix)

Substituting {Δ} of the above equation in the strain-displacement equation results in one of the
most important matrices in the finite element approach, namely [B], or strain-nodal displacement
matrix as follows:

𝑢1
𝑣1
⎛𝑢2⎞
𝜀 N1, x 0 N2, x 0 N3, x 0 N4, x 0 ⎜ ⎟
𝜀 𝑣2
= 0 N1, y 0 N2, y 0 N3, y 0 N4, y ⎜ ⎟ (99)
𝛾 ⎜𝑢3⎟
N1, y N1, x N2, y N2, x N3, y N3, x N4, y N4, x
⎜𝑣3 ⎟
𝑢4
⎝𝑣4 ⎠

Or

{𝜀} = [𝐵] {𝑑} (100)

Stiffness matrix

Recalling the stationary of potential energy 𝜋 = 𝑈 − 𝑊, where U is strain energy, and W is the
work done by the external load, described as follows:

𝑑𝜋 = − = 0 (101)

Where the derivatives of U with respect to nodal displacement is described as follows:

=∭ [𝐵] [𝐸] [𝐵] {𝑑}𝑑𝑣 = [𝐾]{𝑑} (102)

Since dv can be described as t*dA, where t is the element thickness, stiffness matrix [K] can be
written as follows:

[𝐾] = 𝑡 ∬ [𝐵] [𝐸][𝐵]𝑑𝑥𝑑𝑦 (103)

Instead of performing the above integration of the multiplied matrices, having a closed form stiffness
matrix, a numerical integration, discussed in section 4.3, can be adopted in which the final stiffness matrix
is the sum of the corresponding stiffness matrices of the integration process.

[𝐾] = 𝑡𝐿𝑚⅀ [𝐵𝑖] [𝐸][𝐵𝑖] (103-1)

[Bi] is the strain-displacement matrix calculated at the ith integration point. As can be seen from sec. 4.3,
the x, and y coordinates are ±0.577L, ±0.577m ; i.e one integration point in each quarter of the element.

The summation above is multiplied by L*m because a unit weight numbers are used in Gauss scheme
provided that integration is from -1 to 1.

The following excel sheet shows the process of forming the stiffness matrix of a plane stress element.

𝒕𝑳𝒎 is called element magnified thickness in the below Excel sheet.

The below excel sheet shows the formation of a complete mesh of a wall with width (L short), and height (L
long). More details of matrices assemblage will be presented later.
3.9 Eight Node solid element
Displacement Function

Since the element contains four nodes, having four nodal displacement in x, y, and z directions,
namely u, v, and w, respectively. Therefore four generalized coordinates (constants) are used in
describing the displacement in each direction. In other words, the four nodal displacement in each
direction shall substitute the corresponding four constants. From the above the u, and v
displacement at any specified point having x, and y coordinates can be described as follows:

u=a1+a2x+a3y+a4z+a5xy+a6yz+a7xz+a8xyz (104)

v=b1+b2x+b3y+b4z+b5xy+b6yz+b7xz+b8xyz (105)

w=c1+c2x+c3y+c4z+c5xy+c6yz+c7xz+c8xyz (106)

It is worth noting that higher order element can be used, resulting in the ability to capture second
order displacement distribution.

Displacement in terms of nodal displacement

Since the generalized coordinates ai, bi, and ci have no engineering physical meaning, the task will
be to substitute them with the nodal displacement ui, vi, and wi (where i is the total number of
the nodes). For this purpose the following will be incorporated:

𝑢 = {𝑃}{𝑔} (107)

Where {P}= {1 x y z xy yz xz xyz}, and {g}= {a1 a2 a3 a4 a5 a6 a7 a8} (108)

For element dimensions (L,m,n) in x, y and z, directions, respectively, substituting for coordinates
of nodes results in the following:
u1=a1+a2(-L)+a3*(-m)+a4 (-n)+a5(Lm)+a6(mn)+a7(Ln)+a8(-Lmn) (109)

u2=a1+a2(L)+a3*(-m)+a4 (-n)+a5(-Lm)+a6(mn)+a7(-Ln)+a8(Lmn) (110)

u3=a1+a2(L)+a3*(m)+a4 (-n)+a5(lm)+a6(-mn)+a7(Ln)+a8(-Lmn) (111)

u4=a1+a2(-L)+a3*(m)+a4 (-n)+a5(-Lm)+a6(-mn)+a7(Ln)+a8(-Lmn) (112)

u5=a1+a2(-L)+a3*(-m)+a4 (n)+a5(Lm)+a6(-mn)+a7(-Ln)+a8(Lmn) (113)

u6=a1+a2(L)+a3*(-m)+a4 (n)+a5(-Lm)+a6(-mn)+a7(-Ln)+a8(-Lmn) (114)

u7=a1+a2(L)+a3*(m)+a4 (n)+a5(Lm)+a6(mn)+a7(-Ln)+a8(Lmn) (115)

u8=a1+a2(-L)+a3*(m)+a4 (n)+a5(-Lm)+a6(mn)+a7(-Ln)+a8(Lmn) (116)

or in a matrix for as following:

𝑢1 1 −𝐿 −𝑚 −𝑛 𝐿𝑚 𝑚𝑛 𝐿𝑛 −𝐿𝑚𝑛 𝑎1
⎡1 𝐿 −𝑚 −𝑛 −𝐿𝑚 𝑚𝑛 −𝐿𝑛 𝐿𝑚𝑛 ⎤ 𝑎2
𝑢2
⎛𝑢3⎞ ⎢1 ⎛ ⎞
𝐿 𝑚 −𝑛 𝐿𝑚 −𝑚𝑛 𝐿𝑛 −𝐿𝑚𝑛⎥ 𝑎3
⎜𝑢4⎟ ⎢1 −𝐿 𝑚 −𝑛 −𝐿𝑚 −𝑚𝑛 𝐿𝑛
⎥⎜ ⎟
−𝐿𝑚𝑛⎥ ⎜𝑎4⎟
⎜ ⎟=⎢ (117)
⎜𝑢5⎟ ⎢1 −𝐿 −𝑚 𝑛 𝐿𝑚 −𝑚𝑛 −𝐿𝑛 𝐿𝑚𝑛 ⎥ ⎜𝑎5⎟
⎜𝑢6⎟ ⎢1 𝐿 −𝑚 𝑛 −𝐿𝑚 −𝑚𝑛 −𝐿𝑛 −𝐿𝑚𝑛⎥ ⎜𝑎6⎟
𝑢7 ⎢1 𝐿 𝑚 𝑛 𝐿𝑚 𝑚𝑛 −𝐿𝑛 −𝐿𝑚𝑛⎥ 𝑎7
⎝𝑢8⎠ ⎣1 −𝐿 𝑚 𝑛 −𝐿𝑚 𝑚𝑛 −𝐿𝑛 −𝐿𝑚𝑛⎦ ⎝𝑎8⎠

Or

{𝑢𝑖} = [𝐴]{𝑔} (118)

Or

{𝑔} = [𝐴] {𝑢𝑖} (119)

Substituting {g} in u function, results in the following

:𝑢 = [𝐴] {𝑃}{𝑢𝑖} (120)

Or
 𝑢1
𝑢2
⎛ ⎞
𝑢3
⎜ ⎟
𝑢4
𝑢 = {𝑁1 𝑁2 𝑁3 𝑁4 𝑁5 𝑁6 𝑁7 𝑁8} ⎜ ⎟ (121)
⎜𝑢5⎟
⎜𝑢6⎟
𝑢7
⎝𝑢8⎠

Same procedures are used for displacement v, and w resulting in the following:

𝑣1
𝑣2
⎛ ⎞
𝑣3
⎜ ⎟
𝑣4
𝑣 = {𝑁1 𝑁2 𝑁3 𝑁4 𝑁5 𝑁6 𝑁7 𝑁8} ⎜ ⎟ (122)
⎜𝑣5⎟
⎜𝑣6⎟
𝑣7
⎝𝑣8⎠

𝑤1
𝑤2
⎛ ⎞
𝑤3
⎜ ⎟
𝑤4
𝑤 = {𝑁1 𝑁2 𝑁3 𝑁4 𝑁5 𝑁6 𝑁7 𝑁8} ⎜ ⎟ (123)
⎜𝑤5⎟
⎜𝑤6⎟
𝑤7
⎝𝑤8⎠

Displacements {Δ}3x1 (which are u,v, and w) at any point within the element can be determined in
terms of nodal displacements (d)24x1 using the shape functions Matrix [N]3x24 (which are N1 to N8)
as following:
 𝑢1
𝑣1
⎛ ⎞
𝑤1
⎜ 𝑢2 ⎟
⎜ 𝑣2 ⎟
⎜ ⎟
⎜𝑤2⎟
⎜ 𝑢3 ⎟
⎜ 𝑣3 ⎟
⎜𝑤3⎟
⎜ 𝑢4 ⎟
⎜ 𝑣4 ⎟
𝑢 𝑁1 0 0 ⋯ 𝑁8 0 0 ⎜ ⎟
𝑤4
𝑣 = 0 𝑁1 0 … 0 𝑁8 0 ⎜ ⎟ (124)
𝑢5
𝑤 0 0 𝑁1 … 0 0 𝑁8 ⎜ ⎟
𝑣5
⎜𝑤5⎟
⎜ ⎟
𝑢6
⎜ ⎟
⎜ 𝑣6 ⎟
⎜𝑤6⎟
⎜ 𝑢7 ⎟
⎜ 𝑣7 ⎟
⎜𝑤7⎟
⎜ 𝑢8 ⎟
𝑣8
⎝𝑤8⎠

Or

{𝛥} = [𝑁] {𝑑} (125)

The above procedures in getting shape function, through inversing the [A] is complicated,
especially for high order element. Since the element is classified Co element, Lagrangian
interpolation functions can be as following:

!( ) ! !( )
𝑁 = ∗ ∗ 𝑖≠𝑗 (126)
( ) ( )

Where ! indicates multiplier.

Applying Lagrangian interpolation for the 4-node element the shape functions can be found as
follows:

( )( )( ) ( )( )( )
𝑁1 = ; 𝑁2 = ; (126-1)

( )( )( ) ( )( )( )
𝑁3 = ; 𝑁4 = (126-2)
 ( )( )( ) ( )( )( )
𝑁5 = ; 𝑁6 = ; (126-3)

( )( )( ) ( )( )( )
𝑁7 = ; 𝑁8 = (126-4)

For fast formation of the previous shape function, note that signs of x,y, and z in i th N is according
to the location of the ith point.

Strain-Displacement Relationship

The strain at any specific point within the element {ε}3x1 (εx, εx,εz, γxy,γyz,γxz) is defined in terms of
the displacements at the specified point {Δ} (or u , v, and w) using the following differential
operator [𝛛]:

⎡ 0 0⎤
⎢0 0⎥
𝜀
⎢ ⎥
𝜀
⎛𝜀 ⎞ ⎢0 0 ⎥ 𝑢
⎜ ⎟ =⎢ ⎥ 𝑣 (127)
⎜𝛾 ⎟ ⎢ 0⎥ 𝑤
𝛾 ⎢ ⎥
⎝𝛾 ⎠ ⎢0 ⎥
⎢ ⎥
0
⎣ ⎦

Or

{𝜀} = [𝜕] {𝛥} (128)

Strain-nodal displacement matrix [B]-matrix

Substituting {Δ} of the above equation in the strain-displacement equation results in one of the
most important matrices in the finite element approach, namely [B], or strain-nodal displacement
matrix as follows:

𝜀 𝑁1, 𝑥 0 0 ⋯ 𝑁8, 𝑥 0 0 𝑢1
𝜀 ⎡ 0 𝑁1, 𝑦 0 … 0 𝑁8, 𝑦 0 ⎛ 𝑣1 ⎞
⎤
⎛𝜀 ⎞ ⎢ 0 0 𝑁1, 𝑧 ⋯ 0 0
⎥ 𝑤1
𝑁8, 𝑧 ⎥ ⎜ ⎟
⎜ ⎟ =⎢ ⋮ (129)
⎜𝛾 ⎟ ⎢𝑁1, 𝑦 𝑁1, 𝑥 0 … 𝑁8, 𝑦 𝑁8, 𝑥 0 ⎥⎜ ⎟
𝛾 ⎜ 𝑢8 ⎟
⎢ 0 𝑁1, 𝑧 𝑁1, 𝑦 … 0 𝑁8, 𝑧 𝑁8, 𝑦⎥
⎝𝛾 ⎠ ⎣ 𝑁1, 𝑧 𝑣8
0 𝑁1, 𝑥 … 𝑁8, 𝑧 0 𝑁8, 𝑥 ⎦ ⎝ ⎠
𝑤8
Or

{𝜀} = [𝐵] {𝑑} (130)

Stiffness matrix

Recalling the stationary of potential energy 𝜋 = 𝑈 − 𝑊, where U is strain energy, and W is the
work done by the external load, described as follows:

𝑑𝜋 = − = 0 (131)

Where the derivatives of U with respect to nodal displacement is described as follows:

=∭ [𝐵] [𝐸] [𝐵] {𝑑}𝑑𝑣 = [𝐾]{𝑑} (132)

Where, dv can be described as dx dy dz. Accordingly, the element thickness, stiffness matrix [K]
can be written as follows:

[𝐾] = ∭ [𝐵] [𝐸] [𝐵] 𝑑𝑥𝑑𝑦 𝑑𝑧 (133)

Instead of performing the above integration of the multiplied matrices, having a closed form stiffness
matrix, a numerical integration, discussed in section 4.3, can be adopted in which the final stiffness matrix
is the sum of the corresponding stiffness matrices of the integration process.

[𝐾] = 𝐿𝑚𝑛 ∗ ⅀ [𝐵𝑖] [𝐸][𝐵𝑖] (103-1)

[Bi] is the strain-displacement matrix calculated at the ith integration point. As can be seen from sec. 4.3,
the x, and y coordinates are ±0.577L, ±0.577m ; ±0.577n i.e one integration point in each quarter of the
element. The multiplier Lmn is used to correct the integration limits from -1 to 1 to be –L to L, -m to m, and
–n to n.

3.10 Consistent load

3.10.1 General
The consistent nodal loads of the element subjected to external loads/pressure are the equivalent
nodal loads that produce the same work done due to nodal deformation by the external
load/pressure subjected to the element deformation. In other words, the work done by
consistent nodal load due to the nodal deformation is equal to the work done by element external
loads/pressure due to the element deformation. The consistent nodal loads is calculated as by
defining the work done (W) by external loads as follows:

𝑊=∫ 𝜌𝑑𝑣 ∗ 𝑤 + ∫ 𝑝𝑑𝐴 ∗ 𝑤 + ∫ 𝑞𝑑𝑥 ∗ 𝑤 + 𝑃 ∗ 𝑤(𝑎𝑡 𝑥1) + 𝑀 ∗ (𝜃 𝑎𝑡 𝑥2) (134)

where ρ is the load per unit volume, p is load per unit area, and q is load per unit length,
respectively. P and M are external load and moment acting at specific coordinates x1, and x2,
respectively. 𝑤(𝑎𝑡 𝑥1), 𝑎𝑛𝑑 𝑀 ∗ (𝜃 𝑎𝑡 𝑥2) are deformation , and rotation at the same
coordinates. Substituting the deformation (w) in terms of the nodal values (Δ) through shape
functions (Ni), and performing derivative with respect to (Δ), results in the nodal consistent load
vector (R).

3.10.2 Truss-link element consistent load

The shown link member with
length 2L is subjected to axial
distributed load q=xc, and
concentrated axial load P @ L/2 from
point 1.

Work done by these load can be expressed as following:

𝑊= (𝑐𝑥)𝑑𝑥 ∗ 𝑢 + 𝑃 ∗ 𝑢 /

In terms nodal values (Δ), W can be expressed as follows:

𝑊=𝑐 𝑥 (𝑁1 𝑁2)(∆) + 𝑃(𝑁1@ 𝑁2@ )(𝛥)

Performing the derivative of W with respect to (Δ), results in consistent nodal load vectore (R) 1*2
as follows:

𝒅𝑾
(𝑹) = (𝑹𝟏 𝑹𝟐) = =𝑐 𝑥 (𝑁1 𝑁2) + 𝑃(𝑁1 𝑁2 )
𝒅∆ @ @
Student may complete the above.

3.10.3 Beam element consistent load

The work done by the external loads
acting on the shown beam element can
be expressed in terms of the beam
deflection (w) as follows:

𝑊 = ∫ 6𝑑𝑥 ∗ 𝑤 + 30 ∗ 𝑤 − 20 ∗ 𝜃

In terms nodal values (Δ), W can be expressed as follows:

𝑊=6 (𝑁1 𝑁2 𝑁3 𝑁4)(∆)𝑑𝑥 + 30 ∗ (𝑁1 𝑁2 𝑁3 𝑁4 ) (∆) − 20

∗ (𝑁1, 𝑥 𝑁2, 𝑥 𝑁3, 𝑥 𝑁4, 𝑥 )(∆)

Performing the derivative of W with respect to (Δ), results in consistent nodal load vector (R) 1*4 as
follows:

(𝑅) = (𝑅1 𝑅2 𝑅3 𝑅4)

=6 (𝑁1 𝑁2 𝑁3 𝑁4)𝑑𝑥 + 30 ∗ (𝑁1 𝑁2 𝑁3 𝑁4 ) − 20

∗ (𝑁1, 𝑥 𝑁2, 𝑥 𝑁3, 𝑥 𝑁4, 𝑥 )

3.10.4 9-node plane stress element
The work done by external
triangular load as well as the
shown vertical and
horizontal load

s can be expressed as
follows:

𝑤= 3𝑥 ∗ 𝑑𝑥 ∗ 𝑣 . − 30 ∗ 𝑣 . , . + 15 ∗ 𝑢 , .

It is worth noting that substituting for y=0.75 in all shape functions (Ni) those shape function
related to points with y≠0.75 will be equal to zeros. In other words, only (N3, N4, and N7) will not
be equal to zero in case substituting for y=0.75. Similarly will be for (y=0.75, x=0.5). For specific
(x=-1, y=0.75) N8 will be equal to unity, the rest will be equal to zeros. Accordingly W can be
expressed as following:

𝑊=3 𝑥(𝑁3 . 𝑁4 . 𝑁7 . )(𝑣3 𝑣4 𝑣7) 𝑑𝑥 − 30

∗ 𝑁3 . , . 𝑁4 . , . 𝑁7 . , . ∗ (𝑣3 𝑣4 𝑣7) + 15 ∗ 𝑢8

Performing derivative of W with respect to nodal displacement results in the following load (R):

(𝑅) = 3 𝑥(𝑁3 . 𝑁4 . 𝑁7 . )𝑑𝑥 − 30

∗ 𝑁3 . , . 𝑁4 . , . 𝑁7 . , . + 15

Since external load vector (R) is 18*1, as every node may have horizontal and vertical loads, the
external load vector (R)18*1, can be expressed as following:
 0
0
⎛ ⎞ 0 0
0
⎜ ⎟ 0
⎜
0
⎟ ⎛ ⎞ ⎛0⎞
0
⎜
0
⎟ ⎜ ⎟ ⎜0⎟
0
⎜3 ⎜ ⎟ ⎜0⎟
𝑥𝑁3 𝑑𝑥 ⎟ 0
⎜
.
⎟ ⎜−30 ∗ 𝑁3 ⎟ ⎜0⎟
⎜ 0 ⎟ ⎜ . , . ⎟ ⎜0⎟
⎜ ⎟ ⎜ 0 ⎟ ⎜0⎟
⎜3 𝑥𝑁4 . 𝑑𝑥 ⎟ ⎜−30 ∗ 𝑁4 . , . ⎟ ⎜0⎟
⎜ ⎟
(𝑅 ) = ⎜ ⎟+ 0 ⎜0⎟
0 ⎜ 0 ⎟+ 0
⎜ ⎟ ⎜ ⎟
0 ⎜ ⎟ 0
⎜ ⎟ ⎜ 0
0 ⎟ ⎜ ⎟
⎜ ⎟ ⎜ 0 0
0 ⎟ ⎜ ⎟
⎜
0
⎟ ⎜ 0 ⎟ ⎜0⎟
⎜ ⎟ ⎜−30 ∗ 𝑁7 . , . ⎟ ⎜0⎟
⎜3 𝑥𝑁7 𝑑𝑥 ⎟ ⎜ 0 ⎟ ⎜15⎟
.
⎜ ⎟ ⎜ 0 ⎟ ⎜0⎟
⎜ 0 ⎟ 0 0
⎜ 0 ⎟ ⎝ 0 ⎠ ⎝0⎠
0
⎝ 0 ⎠

3.11 Stress-Strain Relationship

The third pillar of the finite element formulation, after describing the energy functional in terms
of the nodal displacement (through shape function), and defining the strain-displacement
relationship is to define stress-strain relationship. In the following two and three dimensional
stress strain relationships are presented. The privilege of adopting 2-D modelling technique is to
reduce the overall degrees of freedom of the global structure.
Three-dimensional state of stress
𝒗 𝒗
⎡𝟏 𝟏 𝒗 𝟏 𝒗
𝟎 𝟎 𝟎 ⎤
𝝈𝒙 ⎢ 𝒗 𝟏
𝒗
𝟎 𝟎 𝟎 ⎥ 𝝐𝒙
𝝈𝒚 ⎢𝟏 𝒗 𝒗 𝒗
𝟏 𝒗
⎥ 𝝐
⎛𝝈 ⎞ ⎢𝟏 𝒗 𝟏 𝟎 𝟎 𝟎 ⎥⎛ 𝒚 ⎞
𝒛 𝑬(𝟏 𝒗) 𝟏 𝒗 𝜺𝒛
⎜ = ⎢ ⎥ (135)
⎜𝝉𝒙𝒚 ⎟ 𝟎 ⎥⎜ ⎟
𝟏 𝟐𝒗
⎟ (𝟏 𝒗)(𝟏 𝟐𝒗) ⎢ 𝟎 𝟎 𝟎 𝟎 𝜸
⎜ 𝒙𝒚 ⎟
𝟐(𝟏 𝒗)
𝝉𝒚𝒛 ⎢ 𝟏 𝟐𝒗 ⎥ 𝜸𝒚𝒛
⎝ 𝝉𝒙𝒛 ⎠ ⎢𝟎 𝟎 𝟎 𝟎
𝟐(𝟏 𝒗)
𝟎 ⎥ ⎝ 𝜸𝒙𝒛 ⎠
⎢ 𝟏 𝟐𝒗 ⎥
⎣𝟎 𝟎 𝟎 𝟎 𝟎
𝟐(𝟏 𝒗)⎦

Plane stress state

Substituting σz, τyz τxz equal to zero in the above three dimensional stress state, and solving for ε z,
γxy, γxy in terms of the rest of strain components, the plane stress stress-stain relationship can be
written as follows:

𝝈𝒙 𝟏 𝒗 𝟎 𝝐𝒙
𝑬 𝒗 𝟏 𝟎
𝝈𝒚 = 𝝐𝒚 (136)
(𝟏 𝒗𝟐 ) 𝟏 𝒗
𝝉𝒙𝒚 𝟎 𝟎 𝜸𝒙𝒚
𝟐

Plane strain
Substituting for εz, γxy, γxy equal to zero in the above three dimensional stress-strain relationship,
the plane-strain stress-strain relationship can be expressed as following:
𝒗
𝝈𝒙 ⎡ 𝟏 (𝟏 𝒗)
𝟎 ⎤
𝝐𝒙
𝝈𝒚 = 𝑬(𝟏 𝒗) ⎢ ⎥
𝒗
𝟏 𝟎 𝝐𝒚 (137)
(𝟏 𝒗)(𝟏 𝟐𝒗) ⎢ (𝟏 𝒗) ⎥
𝝉𝒙𝒚 ⎢ 𝟏 𝟐𝒗 ⎥ 𝜸 𝒙𝒚
⎣ 𝟎 𝟎
𝟐(𝟏 𝒗)⎦

The plane stress state is expressed when loading is on a 2-D , and the structure dimension in the z-
direction is negligible compared to the structure dimension (e.g. shear walls), where plane strain
would be applicable if the structure dimension in the z-direction is significantly high compared to
its in-plane dimension (e.g. extended retaining walls)
3.12 Assembly of the stiffness matrix

The assembly process is incorporated

through defining what is called
“identification vector” for each element.
This vector contains the numbered
degrees of freedom (DOF) for the
corresponding nodes of the element. It is
worth noting that numbering the DOF is
recommended to be done in such a way
that maximum difference of numbers of
DOF in each element to be as minimum as
possible. This require also, numbering
both element and the corresponding
nodes in the same way that maximum
difference between node number of each
element is as minimum as possible. This
will results in narrowing what is called
Matrix Band Width (MBW) which is the
width (number) of non-zero stiffness
coefficients measured from the main
diagonal of the overall stiffness matrix.
MBW can be estimated by ( maximum
difference of nodal numbering in all
element +1) * max. number of DOF of the adopted elements.
The shown simple example shows the aforementioned basic rules in establishing the mesh of
rectangular wall.
As can be noticed the numbering of both elements and nodes are done in the short direction of
the wall. Assuming 4-node element with 2 DOF per node the MBW=(5+1)*2=12.

Short side 6 Total number of element 18

Long side 18 Total number of DOF 28
number of element in short direction 3 Total number of nodes 51
number of element in Long direction 6
number of restraint points 5
Boundary conditions restraining condition
points x y
1 0 1
2 0 1
3 0 1
4 0 1
25 1 0
The below shows some data required to build the geometry of the wall upon which the node
number, element number and the element connectivity can be generated. As can be noticed 5
restrained point were assumed. A Flag is given for each restrained point to differentiate between
the restricted or the free condition. In this generation 1 means restricted, while 0 for free DOF.

Nodal pre-identification number Nodal identification number

node number x y node number x y
The next step is to identify the number of 1 0 1 1 1 0
2 0 1 2 2 0
DOF for each node. AS can be noticed the 3 0 1 3 3 0
4 0 1 4 4 0
counter of DOF is start to increase 5 0 0 5 5 6
6 0 0 6 7 8
incrementally at each node. It stop 7 0 0 7 9 10
8 0 0 8 11 12
counting if restrained DOF is countered. 9 0 0 9 13 14
10 0 0 10 15 16
11 0 0 11 17 18
According to a reasonable subroutine, this 12 0 0 12 19 20
13 0 0 13 21 22
results in the shown numbering of DOF for 14 0 0 14 23 24
15 0 0 15 25 26
each node, and the total DOF in this case 16 0 0 16 27 28
17 0 0 17 29 30
was found 51. It can be noticed that the 18 0 0 18 31 32
19 0 0 19 33 34
counter counts the zero number in the pre 20 0 0 20 35 36
21 0 0 21 37 38
allocated cells (memory) which is originally 22 0 0 22 39 40
23 0 0 23 41 42
is zero unless a unit value (1) is put at the 24 0 0 24 43 44
25 1 0 25 0 45
26 0 0 26 46 47
27 0 0 27 48 49
28 0 0 28 50 51
specified restrained node and a specified direction.

Finally the element element identification vectors

element number DOF1 DOF2 DOF3 DOF4 DOF5 DOF6 DOF7 DOF8
identification vector is 1 1 0 2 0 7 8 5 6
2 2 0 3 0 9 10 7 8
established through 3 3 0 4 0 11 12 9 10
4 5 6 7 8 15 16 13 14
extracting number of 5 7 8 9 10 17 18 15 16
6 9 10 11 12 19 20 17 18
DOF of its nodes in the 7 13 14 15 16 23 24 21 22
8 15 16 17 18 25 26 23 24
same order the node is 9 17 18 19 20 27 28 25 26
10 21 22 23 24 31 32 29 30
numbered for the 11 23 24 25 26 33 34 31 32
12 25 26 27 28 35 36 33 34
element (left bottom, 13 29 30 31 32 39 40 37 38
14 31 32 33 34 41 42 39 40
15 33 34 35 36 43 44 41 42
right bottom, right top, 16 37 38 39 40 46 47 0 45
17 39 40 41 42 48 49 46 47
and left top). 18 41 42 43 44 50 51 48 49

Now, each element has its own identification vector (put horizontally in the above design). The ID
vector can be assumed to be a ruler in both vertical and horizontal direction of the 8*8 stiffness
matrix of the element. The vertical and horizontal rulers allocates the row number and column
number, respectively, in the global stiffness matrix in which the element will be stored
(assembled). Each stiffness coefficient in the ith row, and jth column of the local stiffness matrix is
allocated to the global stiffness matrix according to the number in the horizontal ruler, and
vertical ruler, respectively.

For example for element number 14 The ID vector is (31 32 33 34 41 42 39 40). Stiffness
coefficient in the 4th row and 5th column of the local stiffness matric is allocated @ 34th row and
41st column in the global matrix.

Zero ID either in the vertical or horizontal rulers means that corresponding stiffness coefficient is
not assembled. For example element number 16 contains zero in the 7 th DOF. This means that the
7th row and the 7th columns of the local stiffness matrix of the element will not assembled (stored)

More advanced techniques in mesh generation and assemblage process have been developed to
minimize the memory sizes. This is outside the scope of this note.
The following example (quoted from Bathe 2014) may further explain the assemblage process.
4. Isoparametric Element
4.1 Concept
For irregular four-side elements (other than rectangular or square shapes) to be able to perform
the integration process required to obtain the stiffness matrix, a mapping process is required to
convert them to virtual square elements. In such elements, the new coordinates namely r and s
which are called natural coordinates, varies from -1 to 1, for convenience. Other virtual limits can
be used. The physical coordinates (x, y, or/and z coordinates) can be treated as a field variable in
such way the displacement (or any other dependent variable, such as the velocity). In other
words, the physical x, y, and/or z coordinates can be described in terms of the nodal coordinates
using the same shape functions used in describing the displacement within the elements in terms
of its nodal displacement. Using the same shape function in doing so, is the reason behind calling
such elements to be isoparametric.

4.2 Three-Node 2-D truss/link element

Displacement in terms of natural coordinates

Using Lagrangian interpolation function,

Cartesian coordinate (x) as well axial
displacement of any point (u) along the link element with length L are defined in terms of the
corresponding nodal values in terms of its natural coordinate as follows:

𝑥 = 𝑁𝑖 ∗ 𝑥𝑖 (138)

𝑢 = 𝑁𝑖 ∗ 𝑢𝑖 (139)

Where xi, and ui are Cartesian x-coordinate and nodal axial displacement (u) of the node i. and Ni
is defined as follows:

( )
𝑁𝑖 = ∏ , 𝑖≠𝑗 (140)
( )

Or

( )( ) ( )( )
𝑁1 = = = (141)
( )( ) ∗

( )( ) ( )( )
𝑁2 = = = (142)
( )( ) ( )( )
 ( )( ) ( )( )
𝑁3 = = = (1 − 𝑟 ) (143)
( )( ) ( )( )

In a matrix form the above x, and u definition can be rewritten as follows:

𝑥1
𝑥 = (𝑁1 𝑁2 𝑁3) 𝑥2 (144)
𝑥3

𝑢1
𝑢 = (𝑁1 𝑁2 𝑁3) 𝑢2 (145)
𝑢3

Or

x=(N)(xi)T u= (N)(ui)T (146)

Strain-displacement relationship

Axial strain at any point along the element εx is defined as follows:

( )
𝜀 = = (𝑢𝑖) (147)

Since d(Ni)/dx can not be calculated as N is a function in r, the following complete derivate is
incorporated:

( ) ( )
= (148)

Also, dr/dx can not be determined, rather dx/dr can be calculated. The following complete
derivative is adopted:

( ) ( )
= (149)

Now, dx/dr , called the Jacobian (J) , can be calculated as follows:

( )
𝐽= = (𝑥𝑖) (150)

Substituting for nodal Cartesian coordinates in the above equation, results in the following:

𝐽 = (2𝑟 − 1) ∗ + (2𝑟 + 1) − 2𝑟 ∗ 0 = (151)

Or
𝑑𝑥 = 𝑑𝑟 (152)

Accordingly, strain εx can be calculated as follows:

( ) ( )
𝜀 = (𝑢𝑖) = (𝑢𝑖) = [𝐵](𝑢𝑖) (153)

[B]1*3 is the strain-displacement matrix defined as follows:

[𝐵] = (𝑁1, 𝑁2, 𝑁3, ) = ( (2𝑟 − 1) (2𝑟 + 1) − 2𝑟) (154)

Accordingly, the stiffness matrix [K] is defined as follows:

/
[𝐾] = ∫ /
[𝐵] 𝐴𝐸[𝐵]𝑑𝑥 = 𝐴𝐸 ∫ [𝐵] [𝐵] ( ) 𝑑𝑟 (155)

Adopting numerical integration, two points shall be used @ locations ±0.577, with unit weight
number for each (refer to next section)

Therefore [K] can be calculated as follows:

[𝐾] = 𝐴𝐸⅀ [𝐵𝑖] [𝐵𝑖] det[𝐽𝑖] (155-A)

4.3 Four node plane stress/strain element

Displacement in terms of natural coordinates

The x and y coordinates within

the virtual square elements
can be described in terms of
nodal coordinates (which are
known from the mesh
configuration) as following:
 𝑥1
𝑦1
⎛ ⎞
𝑥2
⎜ ⎟
𝑥 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 0 ⎜𝑦2⎟
𝑦 = 0 (156)
𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 ⎜𝑥3⎟
⎜𝑦3⎟
𝑥4
⎝𝑦4⎠

Moreover, the displacement within the virtual elements can be described as following:

𝑢1
𝑣1
⎛𝑢2⎞
𝑢 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 0 ⎜ 𝑣2⎟
= ⎜ ⎟ (157)
𝑣 0 𝑁1 0 𝑁2 0 𝑁3 0 𝑁4 ⎜𝑢3⎟
⎜𝑣3⎟
𝑢4
⎝𝑣4⎠

Where Ni as per lagrangian interpolation function is described in terms of the natural coordinates
(r,s) as follows:

!( ) !
𝑁𝑖 = ∗ 𝑖≠𝑗 (158)
( )

As can be notice ri or rj are either 1 or -1, similarly are si, and sj.

Therefore,

( )( )
𝑁1 = (159)

( )( )
𝑁2 = (160)

( )( )
𝑁3 = (161)

( )( )
𝑁4 = (162)

Note that signs of r, and s are according to the location of the nodes.

Strain-displacement relationship

Regardless of being isoparametric formulation, the strain definition is the same as follows:
 ⎡ 0⎤
𝜀
⎢ ⎥ 𝑢
𝜀 = ⎢0 ⎥ 𝑣 (163)
𝛾 ⎢ ⎥
⎣ ⎦

Or in a more expanded form as follows:

𝑢,
𝜀 1 0 0 0 𝑢,
𝜀 = 0 0 0 1 (164)
𝑣,
𝛾 0 1 1 0 𝑣,

Or

(𝜀) = [𝐼](𝜕𝑥𝑦) (165)

Since u, and v are described in terms of r, and s coordinates, rather than x and y, the following full
derivatives are adopted:

= + (166)

= + (167)

= + (168)

= + (169)

Or in a matrix form can be rewritten as follows:

⎡ 0 0⎤
⎛ ⎞ ⎢ ⎥
0 0⎥⎛ ⎞
⎜ ⎟=⎢ (170)
⎜ ⎟ ⎢ ⎥⎜ ⎟
⎜ ⎟ ⎢0 0 ⎥
⎢ ⎥⎝ ⎠
⎝ ⎠ ⎣0 0 ⎦

Or

[𝛽] [0]
(𝜕𝑥𝑦) = (𝜕𝑟𝑠) (171)
[0] [𝛽]
It can be noticed that [β] cannot be determined as r, and s are not described in terms of x, and y,
rather x, and y are described in terms of r, and s. Therefore the following inverted full derivatives
have to be adopted as follows:

= + (172)

= + (173)

= + (174)

= + (175)

Or in a matrix form as follows:

⎡ 0 0⎤
⎛ ⎞ ⎢ ⎥
0 0⎥⎛ ⎞
⎜ ⎟=⎢ (176)
⎜ ⎟ ⎢ ⎥⎜⎜ ⎟
⎟
⎜ ⎟ ⎢0 0 ⎥
⎢ ⎥
⎝ ⎠ ⎣0 0 ⎦⎝ ⎠

Or

[𝑗] [0]
(𝜕𝑟𝑠) = (𝜕𝑥𝑦) (177)
[0] [𝑗]

It is worth noting that [J]2x2 which is called jacobian matrix can be calculated as x and y
coordinates are described in terms of r, and s. Moreover, the inverse of [J] is equal to [β].

Therefore

(𝜕𝑥𝑦) = [𝑗] (𝜕𝑟𝑠) (178)

The terms of the jacobian matrix [j] can be determined as follows:

 𝑥1
𝑥2
⎛ ⎞
⎛ ⎞ 𝑁1, 𝑁2, 𝑁3, 𝑁4, 0 0 0 0 𝑥3
⎜ ⎟
⎜ ⎟= 0 0 0 0 𝑁1, 𝑁2, 𝑁3, 𝑁4, ⎜ 𝑥4⎟ (179)
⎜ ⎟
⎜ ⎟ 𝑁1, 𝑁2, 𝑁3, 𝑁4, 0 0 0 0 ⎜𝑦1⎟
0 0 0 0 𝑁1, 𝑁2, 𝑁3, 𝑁4, ⎜𝑦2⎟
⎝ ⎠ 𝑦3
⎝𝑦4⎠

Moreover, (𝛛rs) can be determined in terms of the nodal displacement (d) as following:

𝑢1
𝑣1
𝑁1, 0 𝑁2, 0 𝑁3, 0 𝑁4, 0 ⎛ ⎞
⎛ ⎞ 𝑢2
⎜ ⎟
⎜ ⎟ = 𝑁1, 0 𝑁2, 0 𝑁3, 0 𝑁4, 0
⎜𝑣2 ⎟ (180)
⎜ ⎟ 0 𝑁1, 0 𝑁2, 0 𝑁3, 0 𝑁4, ⎜𝑢3⎟
⎜ ⎟
0 𝑁1, 0 𝑁2, 0 𝑁3, 0 𝑁4, ⎜𝑣3 ⎟
⎝ ⎠ 𝑢4
⎝𝑣4 ⎠

Or

(𝛛rs)=[λ](d) (181)

According to the above, the strain vector can be calculated as follows:

[𝑗] [0]
(𝜀) = [𝐼] [𝜆](𝑑) (182)
[0] [𝑗]

Or in a more compacted form

(𝜀) = [𝐵](𝑑) (183)

Where [B] is the strain-nodal displacement matrix which is equal is calculated as follws:

𝑁1, 0 𝑁2, 0 𝑁3, 0 𝑁4, 0

1 0 0 0 [𝑗] [0] 𝑁1, 0 𝑁2, 0 𝑁3, 0 𝑁4, 0
[𝐵] = 0 0 0 1
[0] [𝑗] 0 𝑁1, 0 𝑁2, 0 𝑁3, 0 𝑁4,
0 1 1 0
0 𝑁1, 0 𝑁2, 0 𝑁3, 0 𝑁4,
(184)
The stiffness matrix [K] is calculated as follows:

[𝐾] = 𝑡 ∬ [𝐵] [𝐸][𝐵]𝑑𝑥𝑑𝑦 (185)

Where dxdy=det[j] dr ds

[𝐾] = 𝑡 ∫ ∫ [𝐵] [𝐸][𝐵] det[𝐽]𝑑𝑟 𝑑𝑠 (186)

Instead of performing the above integration, the numerical integration can be used as follows:

[𝐾] = 𝑡⅀ [𝐵𝑖] [𝐸][𝐵𝑖] det[𝐽𝑖] (187)

Where the sum @ ith integration points would results in [K]. It worth noting that weight numbers
of all integration point of order 2x2 is equal to 1, and was deleted from the above summation for
convenience.

For a polynomial
described from -1 to 1, the
below are the locations,
order, and weight
numbers of Gauss
integration point. For
other element dimensions
in a specific direction, the below locations (for example (1/√3) has to be multiplied by half the
element dimension in that specific direction.

It is worth noting that a polynomial of order 2n-1 can be exactly integrated by n Gauss point.

Example

The below Excel sheet shows the procedures to construct

stiffness matrix for four-node element.
 u,x-u,r matrix BT
0.805177265 0.129882 0 0 -0.1363534 0 -0.24243106
0.129881823 0.580079 0 0 0 -0.24243 -0.1363534
0 0 0.805177 0.129882 0.07141249 0 -0.04760833
0 0 0.129882 0.580079 0 -0.04761 0.071412493
unity matrix 0.33117614 0 0.11254924
1 0 0 0 0 0.112549 0.33117614
0 0 0 1 -0.26623523 0 0.177490152
0 1 1 0 0 0.17749 -0.26623523

B-Matrix
-0.1363534 0 0.07141 0 0.331176 0 -0.266235228 0
0 -0.24243106 0 -0.047608 0 0.11254924 0 0.177490152
-0.24243106 -0.1363534 -0.0476 0.0714125 0.112549 0.33117614 0.177490152 -0.266235228

[B]T[E] [B]T[E][B]
-314.661703 -94.3985 -195.81 90.37555 49.5844514 -13.1486 -9.48910387 -126.247 -75.472 49.01974 35.37664
-167.8368902 -559.456 -110.132 49.58445 150.646404 -6.74247 18.77000759 -67.9788 -99.4393 25.13682 -69.9771
164.7980607 49.43942 -38.4529 -13.1486 -6.74246952 13.59932 -5.09974415 50.24934 -7.17031 -50.7001 19.01252
-32.95961214 -109.865 57.67932 -9.4891 18.7700076 -5.09974 9.349530945 -4.42367 6.736751 19.01252 -34.8563

764.2526298 229.2758 90.90516 -126.247 -67.9788008 50.24934 -4.42367332 263.3335 55.91043 -187.336 16.49204
77.91870482 259.729 267.4884 -75.472 -99.4393384 -7.17031 6.736750586 55.91043 117.8181 26.73186 -25.1155
-614.3889875 -184.317 143.3574 49.01974 25.136819 -50.7001 19.01252134 -187.336 26.73186 189.0165 -70.8812
122.8777975 409.5927 -215.036 35.37664 -69.9770732 19.01252 -34.8562891 16.49204 -25.1155 -70.8812 129.9489
The below is [K] calculated at point 4 (-0.577, 0.577)

[k]=t*Det*[B]T[E][B]wrws
20.07467 11.01395 -2.92 -2.1078 -28.04 -16.7642 10.88850883 7.85803515
11.01395 33.46233 -1.5 4.1693 -15.1 -22.088 5.583515919 -15.5436574
-2.92063 -1.497671 3.021 -1.1328 11.162 -1.5927 -11.26175014 4.223156303
-2.10777 4.169288 -1.13 2.0768 -0.983 1.496401 4.223156303 -7.74245322
-28.0425 -15.09979 11.16 -0.9826 58.493 12.41911 -41.61205413 3.663294338
-16.7642 -22.08796 -1.59 1.4964 12.419 26.17034 5.937813569 -5.57878
10.88851 5.583516 -11.3 4.2232 -41.61 5.937814 41.98529544 -15.7444858
7.858035 -15.54366 4.223 -7.7425 3.6633 -5.57878 -15.74448579 28.86489062
[K] for point -0.577, -0.577 is as below

[k]=t*Det*[B]T[E][B]wrws
68.78837 25.79564 -75 -18.441 -18.45 -6.91919 24.61580927 -0.43577826
25.79564 47.29201 -16.2 -28.501 -6.919 -12.6852 -2.710297486 -6.10613834
-74.953 -16.16616 92.2 3.1335 20.105 4.336262 -37.35094203 8.696401545
-18.4407 -28.50069 3.133 32.491 4.9464 7.644765 10.36082462 -11.6352029
-18.4512 -6.919186 20.1 4.9464 4.9492 1.855939 -6.602718656 0.116889158
-6.91919 -12.68517 4.336 7.6448 1.8559 3.402555 0.726985312 1.637854482
24.61581 -2.710297 -37.4 10.361 -6.603 0.726985 19.33785142 -8.37751245
-0.43578 -6.106138 8.696 -11.635 0.1169 1.637854 -8.377512447 16.1034868

Summing up the four point results in the following [K]

Total Stiffness Matrix

152 35 -155.3 1.01 -62.614 -39.3 65.6 3.24
35 105 6.7753 -48.4 -39.268 -43 -2.53 -13.3
-155 6.78 229.25 -43.6 14.327 -0.11 -88.3 36.9
1.01 -48.4 -43.61 158 5.6637 -48.6 36.9 -60.7
-62.6 -39.3 14.327 5.66 126.7 36.2 -78.4 -2.63
-39.3 -43 -0.106 -48.6 36.237 87.1 3.14 4.5
65.6 -2.53 -88.25 36.9 -78.418 3.14 101 -37.6
3.24 -13.3 36.944 -60.7 -2.632 4.5 -37.6 69.5
4.5 Beam element
In the present section, the isoparamteric formulation of beam element considering shear deformation is
presented. In this element plane section normal to mid-surface remains plane after bending but not

necessary normal to the mid-surface. In other words, the rotation (β) of the vertical plane which is
originally normal to the unreformed neutral axis is no longer equal to the rotation of the neutral axis
(dw/dx) , rather it is equal to only the difference between dw/dx and the shear deformation (ϒ).
For simplicity both shear stress and shear strain will be assumed to be constant across the section with a
modification factor k which is determined to preserve the equality of the total shear strain energy. For
rectangular section k can be found equal to 5/6.

Since the rotation of the element is not dependent only to dw/dx (rather it depends on the shear
deformation and dw/dx) it can be considered as Co element. Hence Lagrangian interpolation shape
function can be adopted.

x-coordinate of any point can be fefined as following:

𝑥1
𝑥 = [𝑁1 𝑁2 𝑁3 ] 𝑥2
𝑥3

Where N1=0.5r(r-1); N2=0.5r(r+1), and N3=(1+r)(1-r)

Substituting the x coordinates of x1, x2, and x3 x results in

X=0.5L(1+r), therefore, dx/dr=0.5L, or dx=0.5dr

Similarly both vertical displacement (w) and rotation (β) can be defined in terms of the corresponding
nodal values. i.e.
𝑤1
𝑤 = [𝑁1 𝑁2 𝑁3 ] 𝑤2
𝑤3

𝜃1
𝛽 = [𝑁1 𝑁2 𝑁3 ] 𝜃2
𝜃3

Accordingly, the deformation of arbitrary point {u}2x1 = (w β)T is defined as following:

 𝑤1
𝜃1
𝑤 ⎛ ⎞
{𝑢} = 𝛽 = 𝑁1 0 𝑁1 0 𝑁3 0 𝑤2
⎜ ⎟=[N]2x6(d)6x1
0 𝑁1 0 𝑁2 0 𝑁3 ⎜ 𝜃2 ⎟
𝑤3
⎝ 𝜃3 ⎠

Strain energy

Two types of strain energy exits due to normal strain, ε, across the section (Un) and due to shear strain (ϒ)
across the section (Us). The first is produced due to ε=du/dx=-y*dβ/dx, and the second due to shear
strain(ϒ=dw/dx-β). The normal strain (flexural) strain energy can be found equal to

𝑈 = ∫ ( ) 𝑑𝑥; 𝑈 = 0.5𝑘𝐴𝐺 ∫ ( − 𝛽) 𝑑𝑥

Where, y is the location of the points measured from the centroid of the section, G is the shear modulus, A
is the cross section area. (refer to Eqs. 4 to 6 for Un).

Therefore, the strain-displacement relationships is as following:

𝜖 0 𝑤
{έ} = 𝛾 = 𝛽 = [𝛤]{𝑢}
−1

Accordingly the strain energy U=Un+Us can be expressed as following:

1 𝐸𝐼 0 {𝜀 }𝑑𝑥
𝑈= {𝜀 }
2 0 𝑘𝐺𝐴

Substituting {u}=[N](d) in {έ} results in

{𝜀 } = 𝑁1 0 𝑁2 0 𝑁3 0 (𝑑) = [𝐵](𝑑)
𝑁1 𝑁1 𝑁2 0 𝑁3 𝑁3

Where 𝑁𝑖 = = =𝑗 , 𝑤ℎ𝑒𝑟𝑒 𝑗 = =

Therefore, 𝑁𝑖 =

Substituting dx=0.5L*dr, U can be expressed as following:

1 𝐸𝐼 0
𝑈= (𝑑) [𝐵} [𝐸 ][𝐵](𝑑)(0.5𝐿𝑑𝑟), 𝑤ℎ𝑒𝑟𝑒 [𝐸 ] =
2 0 𝑘𝐺𝐴

Applying stationary principle, the stiffness matrix [K] can be expressed as following:

[𝐾] = [𝐵] [𝐸 ][𝐵](0.5𝐿)𝑑𝑟

Student may apply concept of consistent load to get nodal external load equivalent to applied load P.
4.6 Plate Element
The plate element is one of the most common elements used in structural engineering soft wears.
Deformation within the element is characterized by nodal transverse displacement “w” (z-
deflection), and two orthogonal rotations (θx, and θy) about two in-plane orthogonal axes (x,y) In
other words, three degrees of freedom are used in each node. It sustains transverse loads, causing
bending and shear stresses. The bending stresses are normal stresses to the element transverse
sections, varying linearly across the element thickness. Similar to beams, the resulting axial force
due to transverse load is zero. Shear stresses are composed of in-plane and transverse shear
stresses (τxy or τyx)), and (τxz, and τyz), respectively. The in-plane shear stress varies linearly across
the section, causing twisting moment about either x oy y axes, and zero in-plane shear force, while
the transverse shear stresses vary parabolically across the thickness, causing shear forces in x-z, and
y-z planes.

Two main plate elements are used, namely Kirchhoff, and Mindlin elements. The first ignore the
transverse shear deformation, while the second account for it. In the following the two finite
element formulations are presented.

4.6.1 Kirchhoff plate element

Deformation characteristics

In Kirchhoff element it is assumed that

plane normal to the mid-surface before
rotation remains plane and normal to the
mid-surface after deformation (rotation). In
other words, as shown in the figure, the
rotation of both mid-surface and its
normal surface about x, and y axes,
namely θx, and θy can be described as follows:

𝜃𝑥 = , 𝑎𝑛𝑑 𝜃𝑦 = (188)

Basically and according to the vector notations, dw/dx is the rotation about y-axis (θy), and dw/dy
is the rotation about x-axis (θx). However, as will be shown below, dw/dx results in in-plane
displacement in x-direction (u-displacement), and dw/dy causes in-plane displacement in y-
direction (v-displacement). This why θx, and θy where assigned to dw/dx, and dw/dy, respectively.

Referring to x-z section, the in-plane displacement u of a layer of distance z from the mid-surface
due to θy can be determined as follows:

𝑢 = −𝑧 = −𝑧 𝑤, = −𝑧𝜃𝑥 (189)

Similarly in y-z plane the in-plane displacement v of a layer at distance z from the mid-surface due
to θy of the mid-surface can be determined as follows:

𝑣 = −𝑧 = −𝑧 𝑤, = −𝑧𝜃𝑦 (190)

Strain-displacement relationship

The in-plane strains is defined as follows:

𝜀 ⎡ ⎤
⎢ ⎥
𝜀 =⎢ (191)
⎥
𝛾 ⎢ ⎥
⎣ + ⎦

Substituting for u, and v as function of w,x, and w,y, respectively, results in the following:

𝜀 𝑤,
𝜀 𝑤,
= −𝑧 (192)
𝛾 2𝑤,

Finite element formulation of four-node plate element

For rectangular four-node element 2a*2b, with thickness t, (Fig. ), the deflection w at any point
within the element, having x,y coordinates is described as following:

w=a1+a2x+a3y+a4xy+a5x2+a6y2+a7xy2+a8yx2+a9x2y2+a10x3+a11y3+a12x3y3 (193)

or w=(s)1*12(g)12*1T,

where (s)= (1 x y xy x2 y2 xy2 yx2 x2y2 x3 y3 x3y3)1*12 and,

(g)=( a1 a2 a3 a4 a5 a6 a7 a8 a9 a10 a11 a12), order of (g) T is 12*1

dw/dx=θx= a2+a4y+2a5x+a7y2+2a8yx+2a9xy2+3a10x2+3a12x2y3 (194)

dw/dy=θy= a3+a4x+2a6y+2a7xy+a8x2+2a9x2y+3a11y2+3a12x3y2 (195)

Substituting for the nodal values (Δ)1*12=(w1, θx1,θy1, ……….. w4, θx4,θy4) in the above equations
results in the following:

(Δ)T12*1=[A]12*12(g)T12*1 or (g)T12*1 = [A]-1 (Δ)T12*1, where [A] is the A-matrix in terms of the nodal
coordinates (or element dimensions).

Substituting for (g), results in

w=(s)1*12[A]-112*12(Δ)T = (N)1*12 (Δ)T (196)

where (N)=(N1 N2 N3…………….N12)1*12

Ni are in terms of element dimensions, (a, b) as well as point coordinates (x,y)

Strain-nodal displacement relationship (Matrix)

Substituting definition of w in terms of nodal deformation (Δ) results in the following:

𝜀
𝜀
𝛾
𝑤1
𝜃𝑥1
𝑁1, 𝑁2, 𝑁3, 𝑁4, 𝑁5, 𝑁6, 𝑁7, 𝑁8, 𝑁9, 𝑁10, 𝑁11, 𝑁12, ⎛ ⎞
𝜃𝑦1
⎜ ⎟
= −𝑧 𝑁1, 𝑁2, 𝑁3, 𝑁4, 𝑁5, 𝑁6, 𝑁7, 𝑁8, 𝑁9, 𝑁10, 𝑁11, 𝑁12, ⎜ ⋮ ⎟
2𝑁1 2𝑁2 2𝑁3 2𝑁4 2𝑁5 2𝑁6 2𝑁7 2𝑁8 2𝑁9 2𝑁10 2𝑁11 2𝑁12 ⎜ 𝑤4 ⎟
𝜃𝑥4
⎝𝜃𝑦4⎠

(197)

Or

(ε)3*1=[B]3*12(Δ)T12*1 (198)

Where [B] is the strain-nodal displacement matrix.

Strain Energy (U), and stiffness matrix [K]

The strain energy U is defined as following:

𝑈 = 0.5 ∫ (𝜎) (𝜀) 𝑑𝑉 = 0.5 ∫ (𝜀) [𝐸](𝜖) 𝑑𝑉 =
/
0.5 ∫ / ∫ ∫ 𝑧 (∆)[𝐵] [𝐸][𝐵](∆)𝑑𝑧 𝑑𝑥𝑑𝑦 (199)

/
=∫ /
∫ ∫ 𝑧 [𝐵] [𝐸][𝐵](∆) 𝑑𝑧 𝑑𝑥𝑑𝑦 (200)
(∆)

(∆)
= ∫ ∫ [𝐵] [𝐸][𝐵](∆) 𝑑𝑥 𝑑𝑦 = [𝐾](∆), (201)

𝑤ℎ𝑒𝑟𝑒

[𝐾] = ∫ ∫ [𝐵] [𝐸][𝐵] 𝑑𝑥 𝑑𝑦 (202)

Multiplying the three matrices and perform the integration will results in stiffness matrix [K].

Moment at any point

Similar to beam theory Mx, My, and Mxy can be described as following:

/ / /
𝑀𝑥 = ∫ /
𝜎 𝑧 𝑑𝑧 , 𝑀𝑦 = ∫ /
𝜎 𝑧 𝑑𝑧 , 𝑀𝑥𝑦 = 𝑀 = ∫ /
𝜏 𝑧 𝑑𝑧 (203)

Or in a matrix form

𝑀𝑥 𝜎
/ /
𝑀𝑦 = ∫ 𝑧 𝜎 𝑑𝑧 = ∫ /
𝑧 [𝐸](𝜖) 𝑑𝑧 = ∫ /
𝑧 [𝐸] [𝐵](∆)𝑑𝑧 = [𝐸][𝐵](∆) (204)
𝑀𝑥𝑦 𝜏

Or

(M)3x1=t3/12*[E]3*3[B]3*12*(Δ)12*1 (205)

Since [B] is in terms of x, and y, the moment (M) at any point depends on its coordinates.
4.6.2 Mindlin plate element
Deformation characteristics

In Mindlin element, it is assumed that

normal plane to the mid-surface
before bending remains plane but not
necessary normal to the mid-surface.
In other words, as shown in Fig. the
rotation θx, and θy are not equal to
dw/dx, and dw/dy, respectively. It is
worth noting as mentioned in
Kirchhoff element, the rotation about
y-axis, and x-axis are assigned θx, and
θy, respectively as they cause in plane stresses in x, and y direction, respectively.

The transverse shear deformation is accordingly defined as the change in the right angle after
deformation as follows:

𝛾 = − 𝜃𝑥, 𝑎𝑛𝑑 𝛾 = − 𝜃𝑦 (206)

Strain-displacement relationship

The in-plane strains is defined as follows:

𝜀
⎛ ⎞
𝜀 =⎜ (207)
⎟
𝛾
⎝ + ⎠

In contrast to Kirchhoff element, the in-plane displacement u and u are defined as follows:

u = -z θx, (208)

v = -z θy (209)

Substituting for u, and v as function of θx and θy, respectively, results in the following:
 𝜀 𝜃𝑥,
𝜀 = −𝑧 𝜃𝑦, (210)
𝛾 𝜃𝑥, + 𝜃𝑦,

Accordingly, the collective strain vector (ε) can be described as follows:

−𝑧𝜃𝑥,
𝜀 −𝑧𝜃𝑦,
𝜀 ⎛ ⎞
⎛ ⎞ ⎜−𝑧(𝜃𝑥, + 𝜃𝑦, )⎟
⎜𝛾 ⎟ = ⎜ ⎟ (211)
𝛾, ⎜ − 𝜃𝑦 ⎟
⎝ 𝛾, ⎠
⎝ − 𝜃𝑥 ⎠

4.5.3 4- node element formulation

In the following the finite element formulation (forming stiffness matrix) for rectangular plate
element 2a*2b is presented. In contrast to Kirchhoff element (C1-element) in which the rotations
θx, and θy about in-plane axes are equal to the derivatives of the deflection “w” with respect to x,
and y, respectively, in Mindlin element θx, and θy about in-plane axes are independent to “w”.
Therefore, it can be considered as Co-element for which Lagrangian interpolation function can be
used to derive the required shape function.

The three independent displacement field (Φ), namely w, θx, and θy at any specified point within
the element can be defined in terms of the nodal values (Δ) as follows:

𝑤1
𝜃𝑥1
⎛ ⎞
𝜃𝑦1
⎜ ⎟
𝑤2
⎜ ⎟
𝜃𝑥2
𝑤 𝑁1 0 0 𝑁2 0 0 𝑁3 0 0 𝑁4 0 0 ⎜𝜃𝑦2⎟
𝜃𝑥 = 0 𝑁1 0 0 𝑁2 0 0 𝑁3 0 0 𝑁4 0 ⎜ ⎟ (212)
𝜃𝑦 ⎜ 𝑤3 ⎟
0 0 𝑁1 0 0 𝑁2 0 0 𝑁3 0 0 𝑁4
⎜𝜃𝑥3⎟
⎜𝜃𝑦3⎟
⎜ 𝑤4 ⎟
𝜃𝑥4
⎝𝜃𝑦4⎠

Where in terms of Cartesian coordinates, and natural coordinates, Ni is defined as follows,

respectively:
 !( ) !
𝑁 = ∗ 𝑖≠𝑗 (213)
( )

!( ) !
𝑁𝑖 = ∗ 𝑖≠𝑗 (214)
( )

Where ! indicates multiplier.

Strain-displacement relationship

The previously defined strain (ε) can be rewritten as follows:

⎡0 0 ⎤
𝜀 ⎢ ⎥
𝜀 ⎢0 0 ⎥ 𝑤
⎛ ⎞ ⎢ ⎥
𝛾 (215)
⎜ ⎟ = ⎢0 ⎥ 𝜃𝑥
𝛾, ⎢ ⎥ 𝜃𝑦
−1 0 ⎥
⎝𝛾, ⎠ ⎢
⎢ ⎥
⎣ 0 −1 ⎦

Or

(ε)=[𝛛](Δ) (216)

It can be noticed that the in-plane strain components εx, εy, γxy are function of x,y, and z coordinates,
while the transverse shear strain γyz, and γxz are functions only in x and y. accordingly in two
different integration schemes will be used for the strain energy determination over the volume.
Moreover, some numerical problems will be raised in the upcoming formulation which require
separately dealing with the in-plane strain components (εb), and those of the transverse shear
strains components (εs) as follows:
 𝜀 𝜀 ⎡0 0⎤
⎡0
0 0 0
⎤
0 0 0
𝜀 𝜀 ⎢0 0 ⎥ 𝑤 ⎢0 ⎥ 𝑤
⎛ ⎞ ⎛ ⎞ ⎛ 0 ⎞ ⎢ ⎥ 0 0
𝛾 𝜃𝑥 +⎢ ⎥ 𝜃𝑥
⎜ ⎟ = (𝜖𝑏) + (𝜀𝑠) = ⎜𝛾 ⎟ + ⎜ 0 ⎟ = −𝑧 ⎢ ⎥ −1 0
𝛾, 0 𝛾, ⎢0 ⎥ 𝜃𝑦 ⎢ ⎥ 𝜃𝑦
𝛾, ⎢ ⎥
⎝𝛾, ⎠ ⎝0 ⎠ ⎝ ⎠ ⎢0 0 0⎥ 0 −1⎦
⎣0 ⎣
0 0⎦
(217)

Substituting for (Φ)=[N](Δ) in the above equations results in the following:

(ε)=(-z[Bb]+[Bs])(Δ), where [Bb], and [Bs] are strain-displacement matrix for in-plane shear (εb), and
transverse shear (γs),respectively, which can be described as follows:

0 𝑁1, 𝑥 0 0 𝑁2, 𝑥 0 0 𝑁3, 𝑥 0 0 𝑁4, 𝑥 0

⎡0 0 𝑁1, 𝑦 0 0 𝑁2, 𝑦 0 0 𝑁3, 𝑦 0 0 𝑁4, 𝑦⎤
⎢ ⎥
[𝐵𝑏] = −𝑧 ⎢0 𝑁1, 𝑦 𝑁1, 𝑥 0 𝑁2, 𝑦 𝑁2, 𝑥 0 𝑁3, 𝑦 𝑁3, 𝑥 0 𝑁4, 𝑦 𝑁4, 𝑥 ⎥
⎢0 0 0 0 0 0 0 0 0 0 0 0 ⎥
⎣0 0 0 0 0 0 0 0 0 0 0 0 ⎦
(218)

0 0 0 0 0 0 0 0 0 0 0 0
⎡ 0 0 0 0 0 0 0 0 0 0 0 0 ⎤
⎢ ⎥
[𝐵𝑠] = ⎢ 0 0 0 0 0 0 0 0 0 0 0 0 ⎥(219)
⎢𝑁1, 𝑥 −𝑁1 0 𝑁2, 𝑥 −𝑁2 0 𝑁3, 𝑥 −𝑁3 0 𝑁4, 𝑥 −𝑁4 0 ⎥
⎣𝑁1, 𝑦 0 −𝑁1 𝑁2, 𝑦 0 −𝑁2 𝑁3, 𝑦 0 −𝑁3 𝑁4, 𝑦 0 −𝑁4⎦

Strain Energy (U), and stiffness matrix [K]

The strain energy U is defined as following:

𝑈 = 0.5 ∫ (𝜎) (𝜀) 𝑑𝑉 = 0.5 ∫ (𝜀) [𝐸](𝜀) 𝑑𝑉 = 0.5 ∫ ((𝜀𝑏) +

(𝜖𝑠) )[𝐸] (𝜀𝑏) + (𝜀𝑠) 𝑑𝑉 (220)

Or

𝑈 = 0.5 ∫ 𝑧 (∆) [𝐵𝑏] [𝐸][𝐵𝑏] 𝑑𝑉 + 0.5 ∫ (∆) [𝐵𝑠] [𝐸][𝐵𝑠] 𝑑𝑉 (221)

Or

/
𝑈 = 0.5 𝑧 (∆) [𝐵𝑏] [𝐸][𝐵𝑏] 𝑑𝑧 𝑑𝑥𝑑𝑦 + 0.5𝑡 (∆) [𝐵𝑠] [𝐸][𝐵𝑠] 𝑑𝑥𝑑𝑦
/

(222)

Or

𝑈= ∗ ∫ ∫ (∆) [𝐵𝑏] [𝐸][𝐵𝑏] 𝑑𝑥𝑑𝑦 + ∫ ∫ (∆) [𝐵𝑠] [𝐸][𝐵𝑠] 𝑑𝑥𝑑𝑦 (223)

(∆)
= ∫ ∫ [𝐵𝑏] [𝐸][𝐵𝑏](∆) 𝑑𝑥 𝑑𝑦 + 𝑡 ∫ ∫ [𝐵𝑠] [𝐸][𝐵𝑠](∆) 𝑑𝑥 𝑑𝑦 = ([𝐾𝑏] +

[𝐾𝑠])(∆), (224)

𝑤ℎ𝑒𝑟𝑒 [𝐾𝑏] = ∫ ∫ [𝐵𝑏] [𝐸][𝐵𝑏] 𝑑𝑥 𝑑𝑦, 𝑎𝑛𝑑 [𝐾𝑠] = 𝑡 ∫ ∫ [𝐵𝑠] [𝐸][𝐵𝑠] 𝑑𝑥 𝑑𝑦

(225)

Multiplying the three matrices and perform the integration for both [Kb], and [Ks], and summing
them up results in the total stiffness matrix [K].

4.5.4 Isoparametric formulation for Mindlin Element

In case of non-rectangular elements, the integration shown in the calculations of [Kb], and [Ks] are
no longer valid, as there is no specified integration limits (a or b). Accordingly, the isoparametric
formulation has to be performed. In the isoparametric formulation, both the displacement field
variables (w, θx, θy) as well as the Cartesian (x, y )coordinates are defined in terms of nodal values,
considering element virtual dimension (2*2) in what is called natural coordinates temporary a
transformed system (r,s coordinates). Therefore, all previous derivatives of shape functions with
respect to x, and y coordinates are to be through the corresponding derivatives with respect to r, s
as following:

Ni,x=Ni,r * r,x + Ni,s * s,x; or 𝛛Ni/𝛛x = 𝛛Ni/𝛛r * 𝛛r/𝛛x + 𝛛Ni/𝛛s * 𝛛s/𝛛x (226)
Ni,y=Ni,r * r,y + Ni,s * s,y or 𝛛Ni/𝛛y = 𝛛Ni/𝛛r * 𝛛r/𝛛y + 𝛛Ni/𝛛s * 𝛛s/𝛛y (227)

Or in a matrix form as following

𝑁𝑖, 𝑥 𝑟, 𝑥 𝑠, 𝑥 𝑁𝑖, 𝑟
= 𝑟, 𝑦 𝑠, 𝑦 or (𝛛i,xy)=[Т](𝛛i,rs) (228)
𝑁𝑖, 𝑦 𝑁𝑖, 𝑠

Since [Т] cannot be calculated as r, or s are not functions of x, and y, rather it is x, and y are functions
of r, and, the above complete derivatives is inversed as following:

Ni,r=Ni,x * x,r + Ni,y * y,r; or 𝛛Ni/𝛛r = 𝛛Ni/𝛛x * 𝛛x/𝛛r + 𝛛Ni/𝛛y * 𝛛y/𝛛r (229)

Ni,s=Ni,x * x,s + Ni,y*y,s; or 𝛛Ni/𝛛s = 𝛛Ni/𝛛x * 𝛛x/𝛛s + 𝛛Ni/𝛛y * 𝛛y/𝛛s (230)

Or in a matrix form as following

𝑁𝑖, 𝑟 𝑥, 𝑟 𝑦, 𝑟 𝑁𝑖, 𝑥
= 𝑥, 𝑠 𝑦, 𝑠 or (𝛛i,rs)=[J](𝛛i,xy) (231)
𝑁𝑖, 𝑦 𝑁𝑖, 𝑦

As can be notices [J], called the Jacobian matrix can be calculated where x,r=Ni,r*xi ; y,r=Ni,r*yi;
x,s=Ni,s*xi; y,s=Ni,s*yi can be calculated in terms of derivatives of N with respect to r,and s as well
as nodal x, and y coordinates.

The above tensor notation implies the following

Ni,r*xi=N1,r*x1 + N2,r*x2 + N3,r*x3 + N4,r*x4 (232)

Ni,r*yi=N1,r*y1 + N2,r*y2 + N3,r*y3 + N4,r*y4 (233)

Ni,s*xi=N1,s*x1 + N2,s*x2 + N3,s*x3 + N4,s*x4 (234)

Ni,s*yi=N1,s*y1 + N2,s*y2 + N3,s*y3 + N4,s*y4 (235)

Where x1, x2, x3, and x4 are x coordinates of the element four nodes, and y1, y2, y3, and y4 are y-
coordinates of the element four nodes.

Substituting for [Т]=[J]-1,

(𝛛,xy)=[J]-1(𝛛,rs) (236)

𝑁𝑖, 𝑟
It is worth mentioning that = (𝛛i,rs) can be calculated in terms of r, and s coordinates of the
𝑁𝑖, 𝑠
specified point (later will be assigned to the Gauss integration point in the numerical integration
process), For example N1,r=-(1-s)/4.
According to Eq.228, 231 the following can be calculated

N1,x=j11-1*N1,r+j12-1*N1,s; N1,y=j21-1*N1,r+j22-1*N1,s

N2,x=j11-1*N2,r+j12-1*N2,s; N2,y=j21-1*N2,r+j22-1*N2,s

N3,x=j11-1*N3,r+j12-1*N3,s; N3,y=j21-1*N3,r+j22-1*N3,s

N4,x=j11-1*N4,r+j12-1*N4,s; N4,y=j21-1*N4,r+j22-1*N4,s

Therefore, both [Bb], and [[Bs] can be calculated.

From the above transformation process, every term of [Bb], and [Bs] can be calculated in the
isoparametric formulation. Accordingly the stiffness matrix [Kb], and [Ks] are defined as follows:

[𝐾𝑏] = ∫ ∫ [𝐵𝑏] [𝐸][𝐵𝑏] 𝑑𝑒𝑡. [𝐽]𝑑𝑟 𝑑𝑠, 𝑎𝑛𝑑 [𝐾𝑠] =

𝑡 ∫ ∫ [𝐵𝑠] [𝐸][𝐵𝑠] 𝐷𝑒𝑡. [𝐽]𝑑𝑟 𝑑𝑠 (237)

Numerical integration

The above integration process in terms of r,s might be complicated to be achieved. Therefore,
rather to get closed forms for [Kb], and [Ks], Gauss integration process can be used as following:

[𝐾𝑏] = ∑ [𝐵𝑏𝑖] [𝐸][𝐵𝑏𝑖] 𝑤𝑖 ∗ 𝑑𝑒𝑡 [𝐽] (238)

[𝐾𝑠] = 𝑡 ∑ [𝐵𝑠𝑖] [𝐸][𝐵𝑠𝑖] 𝑤𝑖 ∗ 𝑑𝑒𝑡 [𝐽] (239)

where i is the ith Gauss integration point, n is number of integration points. It is worth noting that
for the four node element n=4 (i.e. 2*2 integration technique is used in which r,s = ±0.577, and w=1)

4.5.5 Spurious shear mode

As can be noticed that [Kb], and [ks] includes t3, and t, respectively. In case of large span to thickness
ratios, the shear stiffness terms are much higher than those of bending stiffness terms, resulting to
working the shear terms as penalty numbers (without corresponding penalty forces). This leads to
imposing spurious (unrealistic) zero shear deformation which is defined by difference between
dw/dx-θx or dw/dy-θy. This results in locking the mesh (i.e. both w, and rotations are vanished).
To be able to solve this problem, a selective integration scheme was proposed in which 1x1
integration rule for [Ks], and 2x2 for [Kb] are used for four node elements, while 2x2 integration
rule for [Ks], and 3x3 for [Kb] are used for nine-node elements. Also, a reduced integration
technique in which 1x1 is used for both [Kb],, and [Ks].

The below tables and graphs (from Cook text book) shows the reduced, selected, and full integration
rules for different elements. The deflection of clamped square plate also shows the success of
adopting the selective integration technique in achieving close results to the theoretical results.
The below is a comparative study between 4-node Mindlin element and the theoretical study. The element
size in the cantilever direction of the coarse mesh is 1m, and 0.333m for the fine mesh.

The below are some data used to construct [Kb], and [Ks] for square element 1mx1m.

Derivatives w.r.t. x, y
N1,x -0.5 N2,x 0.5 N3,x 0.5 N4,x -0.5
N1,y -0.5 N2,y -0.5 N3,y 0.5 N4,y 0.5

[E]-Mindlin Element…..Eq. 11-1-10, Eq.11.1.12 Cook

1458.333333 291.666667 0 0 0
291.6666667 1458.33333 0 0 0
0 0 583.3333 0 0

0 0 0 145833.33 0
0 0 0 0 145833.3
 [Bb]
0 -0.211325 0 0 0.211325 0 0 0.788675 0 0 -0.788675 0
0 0 -0.788675 0 0 -0.211325 0 0 0.211325 0 0 0.788675
0 -0.788675 -0.211325 0 -0.211325 0.211325 0 0.211325 0.788675 0 0.788675 -0.788675
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0

[Kb]=[Bb]T[E][Bb]
0 0 0 0 0 0 0 0 0 0 0 0
0 427.964772 145.8334 0 32.09564 -84.19694 0 -340.2779364 -375.863474 0 -119.7824802 314.227015
0 145.833401 933.146 0 -22.5605 217.005 0 -207.4697237 -340.277936 0 84.19680701 -809.87311
0 0 0 0 0 0 0 0 0 0 0 0
0 32.0956448 -22.56048 0 91.17727 -39.07597 0 217.0050198 -84.196943 0 -340.2779364 145.833401
0 -84.196943 217.005 0 -39.076 91.17727 0 -22.56048466 32.09564477 0 145.8334013 -340.27794
0 0 0 0 0 0 0 0 0 0 0 0
0 -340.27794 -207.4697 0 217.005 -22.56048 0 933.1460219 145.8334013 0 -809.8731052 84.196807
0 -375.86347 -340.2779 0 -84.1969 32.09564 0 145.8334013 427.9647719 0 314.2270153 -119.78248
0 0 0 0 0 0 0 0 0 0 0 0
0 -119.78248 84.19681 0 -340.278 145.8334 0 -809.8731052 314.2270153 0 1269.933522 -544.25722
0 314.227015 -809.8731 0 145.8334 -340.2779 0 84.19680701 -119.78248 0 -544.2572237 1269.93352

[Bs]
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
-0.5 -0.25 0 0.5 -0.25 0 0.5 -0.25 0 -0.5 -0.25 0
-0.5 0 -0.25 -0.5 0 -0.25 0.5 0 -0.25 0.5 0 -0.25

[Ks]=[Bs]T[E][Bs]
72916.66667 18229.1667 18229.17 0 18229.17 18229.17 -72916.7 18229.16667 18229.16667 0 18229.16667 18229.1667
18229.16667 9114.58333 0 -18229.17 9114.583 0 -18229.2 9114.583333 0 18229.17 9114.583333 0
18229.16667 0 9114.583 18229.167 0 9114.583 -18229.2 0 9114.583333 -18229.2 0 9114.58333
0 -18229.167 18229.17 72916.667 -18229.2 18229.17 0 -18229.16667 18229.16667 -72916.7 -18229.16667 18229.1667
18229.16667 9114.58333 0 -18229.17 9114.583 0 -18229.2 9114.583333 0 18229.17 9114.583333 0
18229.16667 0 9114.583 18229.167 0 9114.583 -18229.2 0 9114.583333 -18229.2 0 9114.58333
-72916.66667 -18229.167 -18229.17 0 -18229.2 -18229.17 72916.67 -18229.16667 -18229.1667 0 -18229.16667 -18229.167
18229.16667 9114.58333 0 -18229.17 9114.583 0 -18229.2 9114.583333 0 18229.17 9114.583333 0
18229.16667 0 9114.583 18229.167 0 9114.583 -18229.2 0 9114.583333 -18229.2 0 9114.58333
0 18229.1667 -18229.17 -72916.67 18229.17 -18229.17 0 18229.16667 -18229.1667 72916.67 18229.16667 -18229.167
18229.16667 9114.58333 0 -18229.17 9114.583 0 -18229.2 9114.583333 0 18229.17 9114.583333 0
18229.16667 0 9114.583 18229.167 0 9114.583 -18229.2 0 9114.583333 -18229.2 0 9114.58333
 [Kb]=Σ[Kb]
0 0 0 0 0 0 0 0 0 0 0 0
0 680.555 218.75 0 -388.89 -72.917 0 -340.2779 -218.75 0 48.6112698 72.91667
0 218.75 680.56 0 72.917 48.611 0 -218.75 -340.2779 0 -72.9166667 -388.889
0 0 0 0 0 0 0 0 0 0 0 0
0 -388.889 72.917 0 680.56 -218.75 0 48.61127 -72.91667 0 -340.277936 218.75
0 -72.9167 48.611 0 -218.75 680.56 0 72.916667 -388.8887 0 218.75 -340.278
0 0 0 0 0 0 0 0 0 0 0 0
0 -340.278 -218.75 0 48.611 72.917 0 680.5554 218.75 0 -388.88873 -72.9167
0 -218.75 -340.28 0 -72.917 -388.89 0 218.75 680.5554 0 72.9166667 48.61127
0 0 0 0 0 0 0 0 0 0 0 0
0 48.6113 -72.917 0 -340.28 218.75 0 -388.8887 72.916667 0 680.555397 -218.75
0 72.9167 -388.89 0 218.75 -340.28 0 -72.91667 48.61127 0 -218.75 680.5554
[Kb]+[Ks]
36458.33333 9114.6 9115 0 9115 9114.6 -36458 9114.583 9114.58 0 9114.5833 9114.6
9114.583333 5237.8 218.8 -9115 4168 -72.92 -9114.6 4217.014 -218.75 9115 4605.9029 72.917
9114.583333 218.75 5238 9114.6 72.92 4605.9 -9114.6 -218.75 4217.01 -9115 -72.91667 4168.4
0 -9114.6 9115 36458 -9115 9114.6 0 -9114.58 9114.58 -36458 -9114.583 9114.6
9114.583333 4168.4 72.92 -9115 5238 -218.8 -9114.6 4605.903 -72.9167 9115 4217.0137 218.75
9114.583333 -72.917 4606 9114.6 -218.8 5237.8 -9114.6 72.91667 4168.4 -9115 218.75 4217
-36458.33333 -9114.6 -9115 0 -9115 -9115 36458 -9114.58 -9114.58 0 -9114.583 -9114.6
9114.583333 4217 -218.8 -9115 4606 72.917 -9114.6 5237.847 218.75 9115 4168.4029 -72.917
9114.583333 -218.75 4217 9114.6 -72.92 4168.4 -9114.6 218.75 5237.85 -9115 72.916667 4605.9
0 9114.6 -9115 -36458 9115 -9115 0 9114.583 -9114.58 36458 9114.5833 -9114.6
9114.583333 4605.9 -72.92 -9115 4217 218.75 -9114.6 4168.403 72.9167 9115 5237.8471 -218.75
9114.583333 72.917 4168 9114.6 218.8 4217 -9114.6 -72.9167 4605.9 -9115 -218.75 5237.8

The below is the constructing of the nodal identification number for DOF

Nodal pre-identification number Nodal identification number

node number w θy θx w 0 0
1 1 1 1 0 0 0
2 1 1 1 0 0 0

3 0 0 0 1 2 3

4 0 0 0 4 5 6

element identification vectors

element number DOF1 DOF2 DOF3 DOF4 DOF5 DOF6 DOF7 DOF8 DOF9 DOF10 DOF11 DOF12
1 0 0 0 0 0 0 4 5 6 1 2 3
9-Node Mindlin element

The Shape functions of the element are as following:

h1= 0.25rs(1-r)(1-s)

h2= 0.25rs(1+r)(1-s)

h3= 0.25rs(1+r)(1+s)

h4= 0.25rs(1-r)(1+s)

h5= -0.5s(1-s)(1-r2)

h6= 0.5r(1+r)(1-s2)

h7= 0.5s(1+s)(1-r2)

h8= -0.5r(1-r)(1-s2)

h9= (1-s2)(1-r2)