0% found this document useful (0 votes)
47 views41 pages

Chap 9 Shape Functions Consist Load

The document discusses shape functions and consistent loading in one-dimensional finite element analysis. It introduces Lagrange interpolation as a method to derive shape functions for 1D elements. Specifically, it shows how to systematically construct Lagrange polynomials of degree n to represent the interpolation functions over an element. Examples are provided for first-order and second-order Lagrange elements, demonstrating how to derive the shape functions at each node using the Lagrange interpolation approach. Consistent loading is also mentioned as being important for evaluating stiffness matrices and loads.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
47 views41 pages

Chap 9 Shape Functions Consist Load

The document discusses shape functions and consistent loading in one-dimensional finite element analysis. It introduces Lagrange interpolation as a method to derive shape functions for 1D elements. Specifically, it shows how to systematically construct Lagrange polynomials of degree n to represent the interpolation functions over an element. Examples are provided for first-order and second-order Lagrange elements, demonstrating how to derive the shape functions at each node using the Lagrange interpolation approach. Consistent loading is also mentioned as being important for evaluating stiffness matrices and loads.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 41

F.E.M.

: Chapter 9
Shape functions and consistent loading in 1D
J.-P. Ponthot
University of Liège - Belgium

December 6, 2021
Introduction

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 2 / 41


Introduction

Shape function and consistent loads in 1D


● allows interpolation from nodal values to values anywhere into the element
● necessary for stiffness matrix and consistent loading evaluations
intra
● enforce inter element continuity (slope can be discontinuous)
● completeness is also capital for convergence

PROPERTIES (requirement - some exceptions do exist - see e.g. Hermitian


interpolation)
● N
PI (xJ ) = δIJ ∀xJ = nodal point
● NI (x) = 1 ∀x ∈ element (rigid body motion)
⇒ interpolation conditions
● compact support (→ sparse matrix)

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 3 / 41


Lagrange interpolation
0
C continuity

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 4 / 41


Lagrange interpolation in 1D

● So far the shape functions were computed using identification of internal


parameters with connectors

● They can also be computed using Lagrange interpolation

● Elements based on Lagrange interpolation are called ”Lagrange Elements”

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 5 / 41


Lagrange interpolation: the bar element

What we already know:

ORDER 1

u(x) = N1 (x)u1 +N2 (x)u2


N1 (x) = 1 − Lx
N1 + N2 = 1 ∀x ∈ [0, L]
N2 (x) = Lx
First order bar element
     
EA 1 −1 u1 γ̄L 1
K= L q= g= 2
−1 1 u2 1

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 6 / 41


Lagrange interpolation

ORDER 2

1 1 1
x1 x3 x 2 x1 x3 x2 x1 x3 x2

u(x) = N1 (x)u1 + N2 (x)u2 + N3 (x)u3



2x2
N1 (x) = 1 − 3x
L + L2

 3
X
4x 4x2
N2 (x) = L − L2 2
⇒ NI (x) = 1 ∀x ∈ [0, L]

N3 (x) = − Lx + 2x
L2
 i=I

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 7 / 41


Lagrange interpolation

Alternative numbering - Bar element

x1 x12 or x3 x2 x1 x2 x3
     
7 1 −8  u1   1 
E  γ̄L
K = 3L 1 7 −8  q = u2 g= 6  1 
 
−8 −8 16 u12 4

● All numbering are equivalent → They deliver same results


● You have to be consistent

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 8 / 41


Lagrange interpolation

Systematic construction of the Lagrange polynomial of degree n


(x−xI )
omitted
(x − x1 )(x − x2 ) · · · (x − xI−1 ) z}|{ (x − xI+1 ) · · · (x − xn+1 )
LnI (x) =
(xI − x1 )(xI − x2 ) · · · (xI − xI−1 ) |{z} (xI − xI+1 ) · · · (xI − xn+1 )
(xI −xI )
omitted

n : degree of the polynomial


I : node or sampling point
Basic property : LnI (xJ ) = δIJ

● note : subintervals do not have to be equal but they are generally taken so

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 9 / 41


Lagrange interpolation
Application : interpolating polynomial of degree n of a function F (x) (we just
assume we can evaluate F (x) ∀ x).

n+1
X
ψ(x) = FI LnI (x) with FI = F (xI )
I=1
Discrete values at sampling points

Basic interpolation property : ψ(xJ ) = F (xJ ) ∀J; J = 1 · · · n + 1


i.e. both ψ and F have the same values at the sampling points (if not, ψ would
just be an approximating function).
F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 10 / 41
Lagrange interpolation

● If F (x) is a complex function

Z Z Z n+1
X
→ F (x) dx ≃ ψ(x) dx = FI LnI (x) dx
I=1
n+1
X Z
= FI LnI (x) dx
| {z }
I=1
POLYNOMIAL
→ Approximation technique to
numerically evaluate integrals
→ See Gauss Legendre integration

● LnI (x) ⇒ Systematic generation of shape functions of degree n in 1D!

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 11 / 41


Lagrange interpolation - Order 1

(x − x1 )(x − x2 ) I = 1 or 2 (the singular term


L1I (x) =
(xI − x1 )(xI − x2 ) has to be omitted)
Node x1
x − x2
L11 (x) = valid ∀ x1 6= x2
x1 − x2
Particular
 cases:
x1 = −1
if ⇒ L11 (x) = 1−x
2
 x 2 = +1
x1 = 0
if ⇒ L11 (x) = 1 − x
x2 = +1

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 12 / 41


Lagrange interpolation - Order 1

Node x2
x − x1
L12 (x) = valid ∀ x1 6= x2
x2 − x1
Particular
 cases:
x1 = −1
if ⇒ L12 (x) = x+1
2
 x 2 = +1
x1 = 0
if ⇒ L12 (x) = x
x2 = +1

Any node xI : NI (x) = L1I (x) = 12 (1 + x xI )


where xI = coordinates of node I (−1 or + 1)
⇒Compact form for interval [−1, 1]

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 13 / 41


Lagrange interpolation - Order 2

(x − x1 )(x − x2 )(x − x3 ) I = 1, 2 or 3 (the singular


L2I (x) =
(xI − x1 )(xI − x2 )(xI − x3 ) term has to be omitted)
Node x1
(x − x2 )(x − x3 )
L21 (x) =
(x1 − x2 )(x1 − x3 )

Particular
 case:
 x1 = −1
x(x−1)
if x2 = 0 ⇒ L21 (x) = 2

x3 = +1

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 14 / 41


Lagrange interpolation - Order 2

Node x2
(x − x1 )(x − x3 )
L22 (x) =
(x2 − x1 )(x2 − x3 )

Particular
 case:
 x1 = −1
if x2 = 0 ⇒ L22 (x) = (1 − x)(1 + x)

x3 = +1

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 15 / 41


Lagrange interpolation - Order 2

Node x3

(x − x1 )(x − x2 )
L23 (x) =
(x3 − x1 )(x3 − x2 )

Particular
 case:
 x1 = −1
x(x+1)
if x2 = 0 ⇒ L23 (x) = 2

x3 = +1

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 16 / 41


Element Interpolation and Local
Coordinates

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 17 / 41


Lagrange interpolation - Order 2

● So far the shape functions depend on the element geometry (at least its length,
which is defined on [0, L])
● Interest for defining a “local” or “natural” coordinate system (e.g. ξ ∈ [−1, 1]
or r ∈ [0, 1] so that the calculations for a typical element may be standardized
● By an appropriate change of variables any interval x ∈ [x1 , x2 ] can be mapped
onto ξ ∈ [−1, +1] using x = x2 −x 2
1
ξ + x2 +x1
2
L
Example: [0, L] → [−1, +1] : x = 2 (1 + ξ)
● Shape functions can thus be defined independently of the element geometry ⇒
systematic computation
● Finite Elements that have shape functions built on Lagrange interpolation
function are called LAGRANGE ELEMENTS

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 18 / 41


Element Interpolation and Local Coordinates

−1 ≤ ξ ≤ 1 0≤r≤1
A) Linear
x1 x2 (1−ξ) 1
N1 = 2 or NI = (1 + ξξI ) N1 = (1 − r)
2
(1+ξ)
N2 = 2 with I = 1 or 2 N2 = r
B) Quadratic
ξ(1−ξ)
N1 = 2 N1 = (r − 1)(2r − 1)
x1 x3 x2 ξ(1+ξ)
N2 = 2 N2 = r(2r − 1)
N3 = (1 + ξ)(1 − ξ) N3 = 4r(1 − r)
C) Cubic
N1 = (1 − ξ)(3ξ + 1) (3ξ−1)
16 N1 = (1 − r)(r − 3r) (1−3r)
2
(3ξ−1) (1−3r)
x1 x3 x4 x2
N2 = (1 + ξ)(3ξ + 1) 16 N2 = r(2 − 3r) 2
N3 = 9(1 + ξ)(ξ − 1) (3ξ−1)
16 N3 = 9r(1 − r) (2−3r)
2
(3ξ+1) (3r−1)
N4 = 9(1 + ξ)(1 − ξ) 16 N4 = 9r(1 − r) 2

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 19 / 41


Consistent nodal forces in 1D

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 20 / 41


Consistent nodal forces
Work done by applied loads:

applied load
Z L nodal value
P= u(x)t̄(x)dx and since u(x) = NI (x)uI , one gets P = uI gI
0
Continuum Mechanics Discrete Approximation

Z L Z +1
L
with gI = NI (x)t̄(x) dx = NI (ξ)t̄(ξ) dξ
0 2 −1

ξ ∈ [−1, 1] L
note : ⇒ dx = 2 dξ
x ∈ [0, L]
● systematic computation of nodal forces
● Since u(x) = NI (x)uI ; u(x) is polynomial (this is a restriction) but t̄(x) is
quite arbitrary (as long as integrals remain well defined!).

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 21 / 41


Consistent nodal forces
Z +1 Z +1
L L
P= u(ξ)t̄(ξ) dξ = NI (ξ)t̄(ξ) dξ uI = g I uI
2 −1 2 −1 |{z}
| {z } | {z } Finite Element
Continuum Mechanics gI approximation
g2
t(x)

g1

x u1 u2
Continuum F.E. approx
t(x) = applied load Discrete nodal loads

● discrete nodal loads gI do represent the actually applied load in an energy


consistent way. They are called consistent nodal forces.
● gI and uI are energy conjugated.
● In 2D or 3D, both u and t̄ are vectors and loading t̄ can have any direction.
Any component of u and t̄ will result in the same component for uI and gI .
Body forces can also be applied in a similar way.

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 22 / 41


Consistent nodal forces

Actual P
load P

L L 2/3
1/2 1/2
Consistent 1/3
load linear
4/6 2/3
1/6 1/6 1/3
0
quadratic
36/60
1/8 3/8 3/8 1/8 2/60 9/60 13/60
cubic

Fraction of total load (respectively P L & P L/2) applied to each node

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 23 / 41


Consistent nodal forces

32/90 32/90
12/90
7/90 7/90

degree 4
75/288 75/288
50/288 50/288
19/288 19/288

degree 5

Fraction of total load (P L) applied to each node

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 24 / 41


Hermitian interpolation :
Bar element (1)

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 25 / 41


Hermitian interpolation : Bar element (1)

● Lagrange interpolation only enforces C 0 interelement continuity


● Aim : Enhance higher order continuity
● Example : C 1 bar element : 2 DOF’s per node
T

q = u1 u′1 u2 u′2
X4
u(ξ) = NI uI ξ ∈ [−1, 1]
I=1
Physical interpretation :
● u → displacement
● u′ → derivative, slope, ...
● NI =?? (shape functions)

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 26 / 41


Hermitian interpolation : Bar element (2)

assumed displacement and derivative :

u(ξ) = α0 + α1 ξ + α2 ξ 2 + α3 ξ 3 ξ ∈ [−1, 1]
⇒ u′ (ξ) = α1 + 2α2 ξ + 3α3 ξ 2

identification :
u1 = u(−1) = α0 − α1 + α2 − α3
u′1 = u′ (−1) = α1 − 2α2 + 3α3
u2 = u(1) = α0 + α1 + α2 + α3
u′2 = u′ (1) = α1 + 2α2 + 3α3
inversion :
⇒ αi = f ct(u1 , u′1 , u2 , u′2 )
leads to :
4
X
⇒ u(ξ) = f ct(u1 , u′1 , u2 , u′2 , ξ) = NI (ξ)uI
I=1
where the NI are the shape functions

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 27 / 41


Hermitian interpolation

Hermitian shape function plots in terms of ξ ∈ [−1, 1]

1 N1 N1 (ξ) = 14 (1 − ξ)2 (2 + ξ)
N2 (ξ) = 41 (1 − ξ)2 (1 + ξ)
N3 (ξ) = 41 (1 + ξ)2 (2 − ξ)
N2 N4 (ξ) = − 14 (1 + ξ)2 (1 − ξ)
tg q=1
P
1 N.B. NI 6= 1
N3
This is an exception as mentionned at
the beginning of the chapter.
N4 tg q=1 This property is valid for Lagrangian
interpolation, not for Hermitian inter-
x=-1 x=1
polation.

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 28 / 41


Hermitian interpolation

du
● Subtlety: So far we have both u(ξ) and u′ (ξ) = that are polynomials in ξ.

They have thus the same physical dimensions (e.g. meters) and so have the
NI .

2x
● The strains in the bar have to be computed as (remember: ξ = − 1)
L
 
du(ξ) du(ξ) dξ 2 2 X dNI
ε(ξ) = = = u′ (ξ) = qI
dx dξ dx L L dξ
| {z } |{z} I
u′ Jacobian

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 29 / 41


Hermitian interpolation
Element shape functions in terms of r ∈ [0, 1]
d()
x = Lr ⇒ dx = Ldr ()′ =
dx
C 1 continuity : u(r) = H1 u1 + H2 u′1 + H3 u2 + H4 u′2
H1 (r) = (2r3 − 3r2 + 1)
H2 (r) = (r3 − 2r2 + r)L
H3 (r) = (3r2 − 2r3 )
H4 (r) = (r3 − r2 )L
where
du du dr 1 du
u′1 = = = L
dx x=0 dr dx dr r=0
du du dr 1 du
u′2 = = =
dx x=L dr dx L
|{z} dr r=1
Explains the
appearance of L
in H2 and H4

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 30 / 41


Hermitian interpolation

C 2 continuity :
U (r) = H1 u1 + H2 u′1 + H3 u′′1 + H4 u2 + H5 u′2 + H6 u′′2

H1 (r) = (1 − 10r3 + 15r4 − 6r5 )


H2 (r) = (r − 6r3 + 8r4 − 3r5 )L
2 3 4 5 L2
H3 (r) = (r − 3r + 3r − r ) 2
H4 (r) = (10r3 − 15r4 + 6r5 )
H5 (r) = (7r4 − 3r5 − 4r3 )L
3 4 5 L2
H6 (r) = (r − 2r + r ) 2

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 31 / 41


Hermitian interpolation

C 3 continuity :
U (r) = H1 u1 + H2 u′1 + H3 u′′1 + H4 u′′′ ′ ′′ ′′′
1 + H5 u 2 + H6 u 2 + H7 u 2 + H8 u 2

H1 (r) = (1 − 35r4 + 84r5 − 70r6 + 20r7 )


H2 (r) = (r − 20r4 + 45r5 − 36r6 + 10r7 )L
2 4 5 6 7 L2
H3 (r) = (r − 10r + 20r − 15r + 4r ) 2
3 4 5 6 7 L3
H4 (r) = (r − 4r + 6r − 4r + r ) 6
H5 (r) = (35r4 − 84r5 + 70r6 − 20r7 )
H6 (r) = (10r7 − 34r6 + 39r5 − 15r4 )L
4 5 6 7 L2
H7 (r) = (5r − 14r + 13r − 4r ) 2
7 6 5 4 L3
H8 (r) = (r − 3r + 3r − r ) 6

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 32 / 41


EXAMPLE :
1
BAR ELEMENT WITH C
CONTINUITY

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 33 / 41


Hermitian interpolation for bars (1)
● Displacement interpolation
4
X 2x dξ 2
u(ξ) = N I qI = N q ξ= −1⇒ = ξ ∈ [−1, 1]
L dx L
I=1

with
● Generalized displacements
T

q = u 1 ε1 u 2 ε2

N.B. Nodal values for strains:

du du dξ 2 ′ L
ε1 = = = u1 ⇒ u′1 = ε1
dx ξ=−1 dξ ξ=−1 |{z}
dx L 2
| {z }
2/L
u′1
du du dξ 2 ′ ′ L
ε2 = = = u 2 ⇒ u 2 = ε2
dx ξ=+1 dξ ξ=+1 |{z}
dx L 2
| {z }
2/L
u′2

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 34 / 41


Hermitian interpolation for bars (2)

To interpolate the displacement in terms of εI , instead of u′I , we have to introduce


a modification in the notation:
u(ξ) = N1 u1 + N2 u′1 + N3 u2 + N4 u′2
L L
= N 1 u 1 + N 2 ε1 + N 3 u 2 + N 4 ε2
2 2

= N1 u1 + N2∗ ε1 + N3 u2 + N4∗ ε2

with

L 1
N2∗ = N2 = L(1 − ξ)2 (1 + ξ)
2 8
L 1
N4∗ = N4 = − L(1 + ξ)2 (1 − ξ)
2 8

N.B. To alleviate the notation, we will omit the ∗

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 35 / 41


Hermitian interpolation for bars (3)
● So finally, we write the displacement interpolation as
4
X
u(ξ) = N I qI = N q
I=1

with:
● Generalized displacements (yields continuity for both u and ε)
T

q = u 1 ε1 u 2 ε2

● Shape functions
N = [N1 , N2 , N3 , N4 ] ξ ∈ [−1, 1]
with
N1 (ξ) = 14 (1 − ξ)2 (2 + ξ)
N2 (ξ) = L8 (1 − ξ)2 (1 + ξ)
N3 (ξ) = 41 (1 + ξ)2 (2 − ξ)
N4 (ξ) = − L8 (1 + ξ)2 (1 − ξ)

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 36 / 41


Hermitian interpolation for bars (4)

This leads to Kq = g with:


● Stiffness matrix
 
36 3L −36 3L
EA  3L 4L 2 −3L −L 2  X
K=   N.B. K IJ 6= 0!!
30L  −36 −3L 36 −3L 
I
3L −L2 −3L 4L2
N.B. RBM = qT = [ 1 0 1 0 ]

● Generalized loading for constant load γ̄

T γ̄L  L
g = 1 6 1 − L6 (uniform loading)
2

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 37 / 41


Hermitian interpolation for bars (5)

Continuum mechanics Discrete model

    
36 3L −36 3L u1 1
EA  3L 4L 2 −3L −L 2  ε 
1 γ̄L  L 
  =  6 
30L  −36 −3L 36 −3L   u2  2  1 
3L −L2 −3L 4L2 ε2 − L6
γ̄L2 ∂u γ̄L ∂u
⇒ u2 = ε1 = = ε2 = =0
2EA ∂x 1 EA ∂x 2

EXACT SOLUTION AT THE NODES

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 38 / 41


Hermitian interpolation for bars (6)
Displacement field:
u(ξ) = N1 u1 + N2 ε1 + N3 u2 + N4 ε2
Strain field :  
du(ξ) du dξ 2 dN1 dN2 dN3 dN4
ε(ξ) = = = u1 + ε1 + u2 + ε2
dx dξ dx L dξ dξ dξ dξ
with ξ = −1 ξ = +1
dN1 3 2 dN1
= (ξ − 1) 0 0
dξ 4 dξ
dN2 L dN2
= (ξ − 1)(3ξ + 1) L/2 0
dξ 8 dξ
dN3 3 dN3
= (1 − ξ 2 ) 0 0
dξ 4 dξ
dN4 L dN4
= (ξ + 1)(3ξ − 1) 0 L/2
dξ 8 dξ
which leads to
ε(−1) = ε1 and ε(+1) = ε2

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 39 / 41


Hermitian interpolation for bars (7)

Displacement field:

u(ξ) = N1 u1 + N2 ε1 + N3 u2 + N4 ε2
L 2 γ̄L 1 2 γ̄L2
= (1 − ξ) (1 + ξ) + (1 + ξ) (2 − ξ)
8 EA 4 2EA
γ̄L2  2 2

= (1 − ξ) (1 + ξ) + (1 + ξ) (2 − ξ)
8EA
γ̄L2 2x
= (1 + ξ)(3 − ξ) with ξ= −1
8EA L
γ̄L2 2x x
= 2(2 − )
8EA L L
γ̄x
= (2L − x) EXACT SOLUTION
2EA

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 40 / 41


Hermitian interpolation for bars (8)

Strain field:
 
2 dN1 dN2 dN3 dN4
ε(ξ) = u1 + ε1 + u2 + ε2
L dξ dξ dξ dξ
 2

2 dN2 γ̄L dN3 γ̄L
= +
L dξ EA dξ 2EA
 
2 γ̄L L 3
= (ξ − 1)(3ξ + 1) + (1 − ξ 2 )L
L EA 8 8
2γ̄L 2x
= 2(1 − ξ) with ξ= −1
8EA L
γ̄
= (L − x) EXACT SOLUTION
EA

F.E.M. - Chapter 9 - Shape functions and consistent loading in 1D 41 / 41

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy