Fem 02

Introduction to C 0 Continuity The Finite Element Analysis Process and Convergence Finite Element for C 1 Continuity
Finite Element Method and Computational Structural Dynamics

Finite elements of one dimension
Manish Shrikhande
[email protected]
Department of Earthquake Engineering

Indian Institute of Technology Roorkee
Based on “Shrikhande, M. Finite Element Method and Computational

Structural Dynamics, PHI Learning Pvt. Ltd., New Delhi, India.
2014. ISBN-13: 978-8120349957.”
Manish Shrikhande [email protected] Department of Earthquake Engineering Indian Institute of Technology Roorkee
Two point boundary value problems

I Many problems can be adequately modeled with only one indenpendent variable
— one coordinate direction.

I The simplest class of problems in this category are defined by a second order
differential equation — transverse deflection of a taut cable, axial deformations of
a bar, torsion of shafts, heat transfer, flow through porous media, electrostatics,
etc.

etc.
I The domain is straight line along one coordinate direction. Domain decomposition
for creating sub-domains (or, finite elements) is straight forward.

etc.
I The interpolation model for approximation of primary variable of the problem in
terms of nodal values is easily developed.

etc.

etc.
Linear Quadratic Cubic
Example problem–1
d
Let us consider the solution of − dx a dφ
dx + cφ − q = 0, in Ω : 0 < x < L with
φ(x = 0) = φ0 and a dφ
dx = Q0 at x = L.
Example problem–1
d
dx + cφ − q = 0, in Ω : 0 < x < L with
dx = Q0 at x = L.
Example problem–1
d
dx + cφ − q = 0, in Ω : 0 < x < L with
dx = Q0 at x = L.
I Weighted residual statement (strong form):

0R = h i
d d φ̂ R
Ω W − dx a dx + c φ̂ − q dΩ + Γ W̄ RΓ dΓ
Example problem–1
d
dx + cφ − q = 0, in Ω : 0 < x < L with
dx = Q0 at x = L.

0R = h i
d d φ̂ R
I Weak form:
" #L Z L
d φ̂ dW d φ̂
0 = −W · a + ·a dx
dx 0 dx dx
x=0
Z L Z L Z
+ W · c φ̂ dx − W · q dx + W̄ RΓ dΓ
0 0 Γ
Example problem–1
d
dx + cφ − q = 0, in Ω : 0 < x < L with
dx = Q0 at x = L.

0R = h i
d d φ̂ R
I Weak form:
" #L Z L
d φ̂ dW d φ̂
0 = −W · a + ·a dx
dx 0 dx dx
x=0
Z L Z L Z
0 0 Γ
Example problem–1
d
dx + cφ − q = 0, in Ω : 0 < x < L with
dx = Q0 at x = L.

I Primary variable is φ,
0R = h
d

d φ̂
i R approximation φ̂(x) shall be
an interpolation between
I Weak form: values of φ at nodes which
" #L Z L define finite elements.
d φ̂ dW d φ̂
0 = −W · a + ·a dx
dx 0 dx dx
x=0
Z L Z L Z
0 0 Γ
Example problem–1
d
dx + cφ − q = 0, in Ω : 0 < x < L with
dx = Q0 at x = L.

I Primary variable is φ,
0R = h
d

d φ̂
i R approximation φ̂(x) shall be
an interpolation between
I Weak form: values of φ at nodes which
" #L Z L define finite elements.
d φ̂ dW d φ̂ I Interpolation in terms of φ
0 = −W · a + ·a dx
dx 0 dx dx ensure continuity of φ̂(x)
x=0
Z L Z L Z across element boundaries
+ W · c φ̂ dx − W · q dx + W̄ RΓ dΓ sharing common nodes for
0 0 Γ
interpolation.
Example problem–2
Example problem–2
Example problem–2
I The highest order of derivative in the weak
form is 1 — approximation should be a
complete polynomial at least upto first degree.
Example problem–2
I A first degree polynomial interpolation within a
sub-domain (finite element) requires two nodes
with one value of primary variable (φ) at each
node: φ̂ = φ1 N1 (x) + φ2 N2 (x) = a0 + a1 x.
Example problem–2
node: φ̂ = φ1 N1 (x) + φ2 N2 (x) = a0 + a1 x.
I Contiuity of zeroth order derivative is preserved
across element boundaries — C 0 continuity.
Example problem–2
node: φ̂ = φ1 N1 (x) + φ2 N2 (x) = a0 + a1 x.
I Contiuity of zeroth order derivative is preserved
I Approximation φ̂ is piece-wise continuous over
the total domain.
Example problem–2
complete polynomial at least upto first degree. (a) Physical system
x
I A first degree polynomial interpolation within a q
00000000000000000000
11111111111111111111 dφ
11111111111111111111
00000000000000000000
00000000000000000000
11111111111111111111 111Q0 = a 0
000 1
00000000000000000000
11111111111111111111
sub-domain (finite element) requires two nodes dx
x=L
L
node: φ̂ = φ1 N1 (x) + φ2 N2 (x) = a0 + a1 x. (b) Finite elment model
1 2 3 4 5
I Contiuity of zeroth order derivative is preserved 1 2 3 4
I Approximation φ̂ is piece-wise continuous over
the total domain.
Example problem–3
Table: Nodal connectivity of finite element mesh
Element # Node i Node j xi xj xj − xi

1 1 2 0 L/4 L/4
2 2 3 L/4 L/2 L/4
3 3 4 L/2 3L/4 L/4
4 4 5 3L/4 L L/4
I The approximate solution over the whole domain is constructed in piecewise

manner by constructing indepedent approximations over each element.
Example problem–3
Table: Nodal connectivity of finite element mesh
Element # Node i Node j xi xj xj − xi

1 1 2 0 L/4 L/4
2 2 3 L/4 L/2 L/4
3 3 4 L/2 3L/4 L/4
4 4 5 3L/4 L L/4
I The approximate solution over the whole domain is constructed in piecewise

manner by constructing indepedent approximations over each element.
I The approximation within an element is an interpolation between primary
variables at the nodes.
Example problem–4
A Generic 2-node Element
Ni Nj
1
i
j
xi e xj
φi φj Qj = (a dφ )
dφ dx x
Qi = −(a
dx )x i
j
i j
Example problem–4
A Generic 2-node Element
Ni Nj
1
i
j
xi e xj
φi φj Qj = (a dφ )
dφ dx x
Qi = −(a
dx )x i
j
i j
I The variation of primary variable φ within

an element,
e, defined by two end
φ i a0
nodes is: φ̂(e) (x) = Ni (x) Nj (x)

= 1 x
φj a1
Example problem–5
I The polynomial approximation
should
evaluate to function values at the nodes of
φi 1 xi a0
interpolation: = which leads to the unknown coefficients of
φj 1 xj a1
−1
a0 1 xi φi
polynomial approximation: =
a1 1 xj φj
Example problem–5
should
φi 1 xi a0
φj 1 xj a1
−1
a0 1 xi φi
a1 1 xj φj
I Therefore,
1 xi −1 φi

(e)
φi
φ̂ (x) = 1 x = Ni (x) Nj (x)
1 xj φj φj

1 xj −xi φi
= 1 x ·
xj − xi −1 1 φj
Example problem–5
should
φi 1 xi a0
φj 1 xj a1
−1
a0 1 xi φi
a1 1 xj φj
I Therefore,
1 xi −1 φi

(e)
φi
φ̂ (x) = 1 x = Ni (x) Nj (x)
1 xj φj φj

1 xj −xi φi
= 1 x ·
xj − xi −1 1 φj
I These shape functions can also be derived by using Lagrange interpolation:
x−x
Ni (x) = − xj −xji and Nj (x) = xx−x i
j −xi
.
Example problem–6
I The weighted residual statement for domain residual in element e is:
( ! )
d φ̂(e)
Z
d
Wk − a + c φ̂(e) − q dΩe = 0
Ωe dx dx
! " #xj
d φ̂(e) d φ̂(e)
Z
dWk (e)
·a + cWk φ̂ − Wk q dΩe − Wk · a = 0
Ωe dx dx dx
x=xi
!
Z
dWk d φ̂ (e)
·a + cWk φ̂(e) − Wk q dΩe − Wk (xj )Qj − Wk (xi )Qi = 0
Ωe dx dx

φ̂(e) φ̂(e)
where, Qi = −a d dx and Qj = a d dx .
x=xi x=xj
Example problem–7
Following Galerkin approach, let us first consider Wk = Ni (x) in the WR statement
with dW dNi
dx = dx = − xj −xi :
k 1
Z xj Z xj
−1 φj − φi xj − x xj − x x − xi
a dx + c φi + φj dx
xi xj − xi xj − xi xi xj − xi xj − xi xj − xi
Z xj
xj − x
− q dx − Qi = 0
xi xj − xi
The first equation can be obtained as:

a c 1
+ −xi xj + (xi2 + xi xj + xj2 ) φi
xj − xi xj − xi 3
" (
xi2 + xj2 (e)

−a c 1 2 x j − xi d φ̂
+ + − (xi + xi xj + xj2 ) φj = q −a .

xj − xi xj − xi 2 3 2 dx

xi
Example problem–8
Similarly, substituting Wk = Nj (x) in the WR statement provides the second algebraic
equation:
" ( )#
−a c xi2 + xj2 1 2 2
+ − (xi + xi xj + xj ) φi
xj − xi xj − xi 2 3

(e)

a c 1 2 xj − x i d φ̂
+ + −xi xj + (xi + xi xj + xj2 ) φj = q +a .

xj − xi xj − xi 3 2 dx
xj
These algebraic simultaneous equations can be arranged as:

K(e) Φ(e) = f (e)
the coefficient matrix K(e) is referred to as the element stiffness matrix, Φ(e) is the
array of (unknown) primary variables at element nodes (also known as degrees of
freedom) and f (e) is the vector of equivalent nodal forces.
Example problem–9
More specifically, the elements of these matrices and vectors for the two-node linear
element can be given as:

(e) (e) a c 1 2 2
kii = kjj = + −xi xj + (xi + xi xj + xj )
xj − xi xj − xi 3
( )
2
xi + xj 2
(e) (e) −a c 1 2 2
kij = kji = + − (xi + xi xj + xj )
xj − xi xj − xi 2 3

(e) xj − xi d φ̂(e) (e) xj − xi d φ̂(e)
fi = q −a and fj = q +a
2 dx 2 dx

xi xj
(e)
where, the stiffness coefficient kij denotes the (i.j)th element of the element stiffness
matrix K(e) .
Example problem–10
Element 1
Node i is the node numbered 1 and node j is the node numbered 2 in the finite
element mesh. Accordingly, we have φi ≡ φ1 , φj ≡ φ2 , and xj − xi = L/4. Thus
K(1) Φ(1) = f (1) :

d φ̂(1)
 qL 
4a cL −4a cL
+ + φ 8 − a dx qL
+ Q

L 12 L 24 1 x=0 8 1
−4a cL 4a cL =  qL
d φ̂(1)
 = qL
L + 24 L + 12
φ2 8 + a dx 8 + Q2
L
x= 4
Element 2
K(2) Φ(2) = f (2) :
 
qL d φ̂(2)
4a cL −4a cL − a qL
L + 12 L + 24 φ2 8 dx x= L 
8 + Q2

−4a cL 4a cL =
d φ̂(2)
4 = qL
φ3
 qL
L + 24 L + 12 8 + Q3

8 + a dx L
x= 2
Element 3
K(3) Φ(3) = f (3) :
 
qL d φ̂(3)
4a cL −4a cL − a qL
L + 12 L + 24 φ3  8 dx x= L 
8 + Q3

−4a cL 4a cL =
d φ̂(3)
2 = qL
φ4
 qL
L + 24 L + 12 8 + Q4

8 + a dx 3L
x= 4
Element 4
K(4) Φ(4) = f (4) :

d φ̂(4)
 qL 
4a cL −4a cL
+ + φ 8 − a dx 3L qL
+ Q

L 12 L 24 4 x= 8 4
−4a cL 4a cL = 4  = qL
L + 24 L + 12
φ5 qL
+ a d φ̂
(4)
8 + Q5
8 dx x=L
Element 4
K(4) Φ(4) = f (4) :

d φ̂(4)
 qL 
4a cL −4a cL
+ + φ 8 − a dx 3L qL
+ Q

L 12 L 24 4 x= 8 4
−4a cL 4a cL = 4  = qL
L + 24 L + 12
φ5 qL
+ a d φ̂
(4)
8 + Q5
8 dx x=L
I Piecewise approximation over individual finite element
Element 4
K(4) Φ(4) = f (4) :

d φ̂(4)
 qL 
4a cL −4a cL
+ + φ 8 − a dx 3L qL
+ Q

L 12 L 24 4 x= 8 4
−4a cL 4a cL = 4  = qL
L + 24 L + 12
φ5 qL
+ a d φ̂
(4)
8 + Q5
8 dx x=L
I Piecewise approximation over individual finite element

I Combine element-level approximations to assemble approximation for the whole
domain — from part to whole!
Example problem–Assembly of equations

I Assembly of element-level equations to
construct a global
approximation—valid for the whole
domain—is based on the concept of
node/joint equilibrium.

I Assembly of element-level equations to
construct a global
approximation—valid for the whole
I Let us consider the equations for finite
elements 1 and 2, described by a pair
of linear simultaneous equations
K(1) Φ(1) = f (1) and K(2) Φ(2) = f (2)
where, Φ(1) ≡ [φ1 , φ2 ]T , and
Φ(2) ≡ [φ2 , φ3 ]T , are the vectors of
degrees of freedom of elements 1 and
2, respectively.

(1) (2) (1) (2)
I Assembly of element-level equations to K + K f + f
construct a global φ1 φ2 φ3
approximation—valid for the whole φ1
+ φ2 +
I Let us consider the equations for finite φ3
elements 1 and 2, described by a pair (1) (1) (1)
of linear simultaneous equations k11 φ1 + k12 φ2 = f1
(1) (1) (2) (2) (1) (2)
K(1) Φ(1) = f (1) and K(2) Φ(2) = f (2) k21 φ1 + (k22 + k11 )φ2 + k12 φ3 = f2 + f1
where, Φ(1) ≡ [φ1 , φ2 ]T , and (2) (2) (3)
k21 φ2 + (k22 + k11 )φ3 + k12 φ4
(3)
= f2
(2)
+ f1
(3)
Φ(2) ≡ [φ2 , φ3 ]T , are the vectors of (3) (3) (4) (4) (3) (4)
k21 φ3 + (k22 + k11 )φ4 + k12 φ5 = f2 + f1
degrees of freedom of elements 1 and (4) (4) (4)
2, respectively. k21 φ4 + k22 φ5 = f2

Leading to the global system of equations: KΦ = f. Let us look at assembly of terms
in the load vector f:

(1) (2) qL d φ̂(1) qL d φ̂(2)
f2 = f2 + f1 = +a + −a
8 dx L 8 dx L

x= 4 x= 4
 
qL  d φ̂(1) d φ̂(2)  ≈ qL
= + a −a
4 dx L dx L 4

x= 4 x= 4

qL d φ̂(1) qL qL d φ̂(4) qL
f1 = −a = + R and f5 = +a = + Q0
8 dx 8 8 dx 8

x=0 x=L
Example problem–Imposition of essential boundary conditions

I Essential boundary condition of φ1 = φ∗1 is to be incorporated in the system of
algebraic equations.
Example problem–Imposition of essential boundary conditions

I Essential boundary condition of φ1 = φ∗1 is to be incorporated in the system of
algebraic equations.
I Augment the system of equations, where the coefficient of the DOF to be
constrained is the diagonal entry, as:
(k11 + α)φ1 + k12 φ2 + . . . = f1 + αφ∗1
1 2
α 1 If the stiffness of the artificial spring α is assumed to be very
large, say, of the order of 104 times the largest stiffness coeffcient in the equation,
this equation can be approximated as
αφ1 + negligible terms = αφ∗1 + negligible term
which, after dropping the negligible terms, yields φ1 ≈ φ∗1 .
Element Orientation — Global and Local Coordinate Systems

A planar truss and orientation of a generic truss/bar element
vj
Y uj
vj θ
vi j uj
ui e
vi
i ui
X

vj
Y uj
vj θ
vi j uj
ui e
vi
i ui
X
I It is convenient to formulate element equilibrium equations with coordinate

direction oriented along a member — local coordinate system.

vj
Y uj
vj θ
vi j uj
ui e
vi
i ui
X

I Different members might have different orientations in a structure.

vj
Y uj
vj θ
vi j uj
ui e
vi
i ui
X

I Different members might have different orientations in a structure.
I It is necessary to transform the element equilibrium equations to a common global
frame of reference before the assembly of element equations.
The Finite Element Analysis Process — A Recapitulation

I Identification of physical domain and problem statement.

I Establishing the governing differential equations, domain of applicability, and
boundary conditions.

I Developing the weak form of the statement of weighted residuals for domain
residual for an approximate solution.

I Identifying primary and secondary variables of the problem from the boundary
terms of the weak form.

I Determining the interpolation model required for approximation of primary
variables and hence determining the type of finite element required for discretising
the domain.

I Determining the interpolation model required for approximation of primary
variables and hence determining the type of finite element required for discretising
the domain.
I Developing the finite element mesh for discretisation of the domain.
The Finite Element Analysis Process — A Recapitulation (Contd.)

I Constructing polynomial interpolation for primary variables over one element at a
time and establish element level equations by evaluating the weak form of
weighted residual statement over the element by using Galerkin approach and
using interpolation functions as weighting functions.

I Assembling element level equations and impose specified natural boundary
conditions on the appropriate secondary variables.

I Imposing essential boundary condition on the assembled global system of
equations.

equations.
I Solving the system of simltaneous equations for unknown primary variables at the
nodes of finite element mesh.

equations.
I Solving the system of simltaneous equations for unknown primary variables at the
nodes of finite element mesh.
I Determining secondary variables and other derived quantities of interest and any
point by using element level equations.
The Finite Element Analysis Process — Convergence

A monotonic convergence of the computed solution is assured if the finite element
model satisfies following conditions:
I Continuity of primary variables within the element and across element boundaries
is ensured.

is ensured.
I Compatibility between adjacent elements in a finite element mesh is maintained.

is ensured.
I Completeness of polynomial approximation for the primary variable — inclusion of
all lower order terms upto the highest degree of complete polynomial. Constant
term is necessary for representation of rigid body motion while constant derivative
terms represent the state of constant strain in the limiting case of element size
reducing to a point.

is ensured.
I Completeness of polynomial approximation for the primary variable — inclusion of
all lower order terms upto the highest degree of complete polynomial. Constant
term is necessary for representation of rigid body motion while constant derivative
terms represent the state of constant strain in the limiting case of element size
reducing to a point.
I Consistency of the finite element approximation with respect to the constraints of
the problem is ensured.
The Euler-Bernoulli Hypothesis

I Plane sections normal to the neutral axis remain plane and normal during bending.
A beam is a structural member whose length is large compared to its cross-sectional

dimenstions and supports loads and/or moments that produce deflection transverse to
its longitudinal axis. The governing differential equation may be derived as:

I Hooke’s law is applicable and Young’s modulus for tension and compression is
same.


same.
I The beam is initially straight and all longitudinal fibres bend into concentric
circles with radii of curvatures that are much larger compared to the
cross-sectional dimensions of the beam.

same.

same.
..
ρ A dx v
f f
V V+dV
M M+dM
y
x dx
L, EI

same.
its longitudinal axis. The governing differential equationmay be derived as:
.. d2 d 2v
ρ A dx v
EI =f ; 0<x <L
f
M
V
f
V+dV
M+dM
dx 2 dx 2
y
x
L, EI
dx
where, EI (> 0) is the flexural rigidity.
The Euler-Bernoulli Beam Element–1

Developing the weak form of weighted residual statement for an approximate solution
v̂ (x) ≈ v (x):
L L Z L 2 Z L
d 2 v̂ dW d 2 v̂ d W d 2 v̂

d
W EI 2 − EI 2 + EI 2 dx − Wf dx = 0
dx dx x=0 dx dx x=0 0 dx 2 dx 0
where, W (x) is a weighting function used to construct the statement of weighted

residuals. This reveals two primary variables and two corresponding secondary variables:
Primary Variable Secondary Variable

d 2
Displacement, v (x) Shear Force, dx EI ddxv2
dv 2
Slope, dx Bending Moment, EI ddxv2
The polynomial approximation v̂ (x) should be chosen such that continuity upto first
derivative is ensured across the nodes — Hermite interpolation.

vi vj
θi θj A 2-node element with 2 primary variables at each
i j x
Vi e Vj node to define the polynomial approximation v̂ (x)
Mi Mj within the element.
L, EI
There are a total of 4 constraints available as 4 primary variables at the 2 nodes — a
cubic polynomial can be defined from 4 constraints.
v̂ (x) = a0 + a1 x + a2 x 2 + a3 x 3 = N1 (x)vi + N2 (x)θi + N3 (x)vj + N4 (x)θj ; x ∈ Ωe

d v̂ (x) dN1 dN2 dN3 dN4
= θ̂ = a1 + 2a2 x + 3a3 x 2 = vi + θi + vj + θj
dx dx dx dx dx

This interpolation will have correct limiting behaviour if it reproduces the nodal
displacements and slopes when evaluated at the nodes. Therefore, assuming that
xi = 0, and xj = L:
1 xi xi2 xi3
       
vi a0 1 0 0 0 a0
 θi  0 1 2xi 3xi2  a1  0 1 0 0  a1 
  = 
1 xj xj2 xj3  a2  = 1 L L2 L3  a2 
    
vj 
θj 0 1 2xj 3xj2 a3 0 1 2L 3L2 a3
   4  
a0 L 0 0 0 vi
a1  1  0
 L4 0 0   θi 
 
or, 
a2  = L4 −3L2 −2L3 3L2 −L3  vj 
 
a3 2L L2 −2L L2 θj

Substituting for the coefficients of the cubic polynomial approximation for v̂ (x):
 4  
L 0 0 0 vi
1 0 L 4 0 0 θ
v̂ (x) = b1 x x 2 x 3 c 4    i
   
2 3 2 3
L −3L −2L
 3L −L   vj 
2L L 2 −2L L 2 θj
 
2 3
2 3
2 3
2 3
vi
3x 2x 2x x 3x 2x x x  θi 
= 1− 2 + 3 x− + 2 − − +  
L L L L L2 L3 L L2 vj 
θj
 
vi
 θi  (e)
= bN1 (x) N2 (x) N3 (x) N4 (x)c  vj  = Nv

Manish Shrikhande [email protected]

θj
Department of Earthquake Engineering Indian Institute of Technology Roorkee

Properties of Cubic Hermitian Interpolation N1(x) N3(x)
Functions 1.0
at x = xi at x = xj
dN
N(x) dx N(x) dN dx
i
N2(x)
j
π
N1 (x) 1 0 0 0 i 4 j
π
N2 (x) 0 1 0 0 4
N3 (x) 0 0 1 0 N4(x)
N4 (x) 0 0 0 1 Cubic Hermitian polynomials as shape

functions for 2-node beam element
Finite elements based on use of cubic Hermite shape functions ensure inter-element
continuity of the function being approximated as well as its first derivative, and hence
are known as C 1 continuous elements.
Development of Element Equilibrium Equations–1

Substituting the finite element approximation for v̂ (x) into the weak form and
choosing the element shape functions (Ni (x), i = 1, 2, 3, 4) as the weighting functions:
Z L 2 Z L L
d2 d2

d Ni (e) d (e)
2
EI 2 Nv dx = Ni (x)f dx − Ni (x) EI 2 Nv
0 dx dx 0 dx dx x=0
2
L
dNi d
+ EI 2 Nv(e) ; ∀i = 1, 2, 3, 4 .
dx dx x=0
which, on simplification lead to the familiar equilibrium equations of the beam problem:
      
6 3L −6 3L vi 6 Vi
2EI  3L 2L 2 3L L2   θi  fL  L   Mi 
   =  + 
L3 −6 3L 6 −3L vj  12  6   Vj 
3L L2 −3L 2L2 θj −L Mj
K(e) v(e) = f (e) + Q(e)

where, K(e) is the element stiffness matrix for Euler-Bernoulli beam element, whose
i, jth element is given by
L
d 2 N i d 2 Nj
Z
kij = EI dx ; ∀ i, j = 1, 2, 3, 4
0 dx 2 dx 2
v(e) is the vector of element nodal DOF; f (e) is the element load vector, whose ith
element is given by
Z L
fi = Ni (x)f dx ; ∀ i = 1, 2, 3, 4
0
Vi , Mi and Vj , Mj , are the shear force and bending moment at nodes i and j and are
related to the transverse displacement of the beam by
d 2v d 2v

d d
Vi = EI 2 Vj = − EI 2
dx dx x=xi dx dx x=xj
d 2v d 2v

Mi = − EI 2 Mj = EI 2 .
dx x=xi dx x=xj

I The elements of the load vector f (e) are nodal work equivalent forces doing same
amount of work in moving through the nodal DOF as the work done by external
imposed loads in moving through the transverse displacement v̂ (x).

I The element equilibrium equations may be assembled into a global system of
equations by accumulating contributions from all elements sharing a node and
DOF in the same manner as done in the case of assembly of C 0 elements.

I Similarly, the essential boundary conditions of the problem can be incorporated as
outlined in the context of C 0 elements.

I Similarly, the essential boundary conditions of the problem can be incorporated as
outlined in the context of C 0 elements.
I The resulting global system of equations, properly constrained, can then be solved
for nodal DOF and subsequently stress resultants may be computed from element
equilibrium equations during the post-processing stage.
Elements with Arbitrary Orientation

uj
Y vj θj
X
Y j
vi θi ui
α
v
i X
θ
u
I Beam elements of C 1 continuity can be combined with truss/bar elements of C 0

continuity for finite element modeling of framed structures.

uj
Y vj θj
X
Y j
vi θi ui
α
v
i X
θ
u

I Different members may have different orientation and therefore, the element
equilibrium equations are first developed in local coordinate frame with
longitudinal axis oriented along the principal coordinate axis.

uj
Y vj θj
X
Y j
vi θi ui
α
v
i X
θ
u

I Different members may have different orientation and therefore, the element
equilibrium equations are first developed in local coordinate frame with
longitudinal axis oriented along the principal coordinate axis.
I Equilibrium equations in local coordinate system are transformed to a common
global coordinate frame before assembly to build global model for entire structural
system.

Fem 02

Uploaded by

Copyright:

Available Formats

Fem 02

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fem 02

Uploaded by

Copyright:

Available Formats

Introduction to C 0 Continuity The Finite Element Analysis Process and Convergence Finite Element for C 1 Continuity

Finite Element Method and Computational Structural Dynamics

Department of Earthquake Engineering

Based on “Shrikhande, M. Finite Element Method and Computational

Two point boundary value problems

Two point boundary value problems

Two point boundary value problems

Two point boundary value problems

Two point boundary value problems

Two point boundary value problems

I Weighted residual statement (strong form):

I Weighted residual statement (strong form):

I Weighted residual statement (strong form):

I Weighted residual statement (strong form):

I Weighted residual statement (strong form):

Table: Nodal connectivity of finite element mesh

Element # Node i Node j xi xj xj − xi

I The approximate solution over the whole domain is constructed in piecewise

Table: Nodal connectivity of finite element mesh

Element # Node i Node j xi xj xj − xi

I The approximate solution over the whole domain is constructed in piecewise

I The variation of primary variable φ within

These algebraic simultaneous equations can be arranged as:

I Piecewise approximation over individual finite element

I Piecewise approximation over individual finite element

Example problem–Assembly of equations

Example problem–Assembly of equations

Example problem–Assembly of equations

Example problem–Assembly of equations

Example problem–Assembly of equations

Example problem–Assembly of equations

Example problem–Imposition of essential boundary conditions

Example problem–Imposition of essential boundary conditions

Element Orientation — Global and Local Coordinate Systems

Element Orientation — Global and Local Coordinate Systems

I It is convenient to formulate element equilibrium equations with coordinate

Element Orientation — Global and Local Coordinate Systems

I It is convenient to formulate element equilibrium equations with coordinate

Element Orientation — Global and Local Coordinate Systems

I It is convenient to formulate element equilibrium equations with coordinate

The Finite Element Analysis Process — A Recapitulation

The Finite Element Analysis Process — A Recapitulation

The Finite Element Analysis Process — A Recapitulation

The Finite Element Analysis Process — A Recapitulation

The Finite Element Analysis Process — A Recapitulation

The Finite Element Analysis Process — A Recapitulation

The Finite Element Analysis Process — A Recapitulation (Contd.)

The Finite Element Analysis Process — A Recapitulation (Contd.)

The Finite Element Analysis Process — A Recapitulation (Contd.)

The Finite Element Analysis Process — A Recapitulation (Contd.)

The Finite Element Analysis Process — A Recapitulation (Contd.)

The Finite Element Analysis Process — Convergence

The Finite Element Analysis Process — Convergence

The Finite Element Analysis Process — Convergence

The Finite Element Analysis Process — Convergence

The Euler-Bernoulli Hypothesis

A beam is a structural member whose length is large compared to its cross-sectional

The Euler-Bernoulli Hypothesis

A beam is a structural member whose length is large compared to its cross-sectional