Вы находитесь на странице: 1из 17

Lecture 20: Isoparametric Formulations.

„ Chapter #10 Isoparametric Formulation.


„ Isoparametric formulations help us solve two problems.
„ Help simplify the definition of the approximate displacement field for
more complex planar elements (4-sided elements, elements with
curved edges, …).
„ Significantly reduce the integration process by ensuring that we
always have integrals in terms of natural coordinates that are taken
over fixed bounds.
„ We start by applying the isoparametric formulation to a bar
element.
„ Move on to derivation of the stiffness matrix and ee’sfor a “plane
element” that can be used in plane stress and plane strain problems.
„ The derivations will show the set up of the integral expression.
„ We will apply Guassian quadrature to evaluate these integrals.

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ §10.1 Isoparametric Formulation of the Bar Element Euations.
„ Step 1: Set the element type.
„ Step 1 now includes the definition of a natural (or curvilinear
coordinate), s.
„ We must create a mapping between the curvilinear coordinate s and
the cartesian coordinate x.
„ This mapping will match the element boundaries/keypoints/nodes
between the natural coordinate plane and the Cartesian coordinate
system.

Element
Element nodes centroid

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ The mapping procedure is very similar to the creation of shape
functions as in Logan Ch.’s 2, 3, 4, 5, and 6.
„ Set the order of the approximation:
x = a1 + a2 s

„ Apply some kinematic constraints on the approximation.


⎧ x1 + x2 ⎫
−1
x1 = a1 + a2 (−1) ⎧ a1 ⎫ ⎡1 −1⎤ ⎧ x1 ⎫ ⎧ a1 ⎫ ⎪⎪ 2 ⎪⎪
⎨ ⎬=⎢ ⎥ ⎨x ⎬ ⎨ ⎬=⎨ ⎬
x2 = a1 + a2 (+1) ⎩ 2⎭ ⎣
a 1 1 ⎦ ⎩ 2⎭ ⎩ 2⎭ ⎪ 2
a x − x1⎪

⎩⎪ 2 ⎭⎪

1
x= ⎡⎣(1 − s ) x1 + (1 + s ) x2 ⎤⎦
2

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ Defining the shape functions that are the core of the mapping
between s and x…
(1 − s )
N1 =
⎧ x1 ⎫
x = [ N1 N2 ] ⎨ ⎬
2
⎩ x2 ⎭ (1 + s )
N2 =
2

„ The shape functions define the variation of a quantity (now a


coordinate x) over some domain of interest (now a natural
coordinate s).
„ We can apply the same shape functions to define the variation of
other values over the s ∈ ( −1.. + 1) domain…

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ Step 2: Select a displacement function:
⎧ u1 ⎫ ⎧⎪ dˆ1x ⎫⎪
u = [ N1 N 2 ] ⎨ ⎬ = [ N1 N2 ] ⎨ ⎬
⎩u2 ⎭ ˆ
⎩⎪d 2 x ⎭⎪

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ Step 3: Define the stress strain relationships.
„ Regardless of what natural coordinates we introduce, we do not
change the governing physics of the structural problem:

du
εx = Governing physics/Constitutive
dx equations.
⎛ du ⎞
du ⎜⎝ ds ⎟⎠
= How we realize the
dx ⎛ dx ⎞ differential term(s) in the
⎜ ⎟
du u2 − u1 ⎝ ds ⎠ constitutive equations.
=
ds 2
dx x2 − x1 L
= = Again, we use the
ds 2 2 B notation to
du ( u2 − u1 ) ⎧⎪ dˆ1x ⎫⎪ ⎡ 1 1⎤ refer to a matrix
∴ε x = = = B⎨ ⎬ ; B = ⎢−
( ) L ⎥⎦
that relates
dx L ⎪⎩dˆ2 x ⎪⎭ ⎣ L
displacements to
MECH 420: Finite Element Applications strains.
Lecture 20: Isoparametric Formulations.
„ Step 4: Derive the element equations:

⎡ ⎤ G G
0 = ⎢ ∫∫∫ B DB ⋅ dV ⎥ d + ∫∫∫ N b ⋅ dV + ∫∫ N TS ⋅ dS + fˆ T dˆ
T ˆ T T

⎢⎣V ( e ) ⎥⎦ V (e) S (e)

For line elements the integrals reduces to


integration in 1-D: x. (uniform cross sectional
area assumed in Step 1.)

⎡L T ⎤ˆ L G L
G
0 = A ⎢ ∫ B DB ⋅ dx ⎥ d + A ∫ N b ⋅ dx + P ∫ N TS ⋅ dx + fˆ T dˆ
T T

⎣ x =0 ⎦ x =0 x =0
1 2 3

P ≡ Perimeter around the bar cross section.

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ Consider the first term in the functional :
„ (the strain energy term that produces the stiffness matrix)

⎡ L
⎤ ⎡ s =+1

k = A ⎢ ∫ B DB ⋅ dx ⎥ = A ⎢ ∫ B DB ⋅ dx ⎥
ˆ T T

⎣ x =0 ⎦ ⎣ s =−1 ⎦
dx x2 − x1 L
= =
ds 2 2
L Jacobian
∴ dx = ds matrix.
2
⎡ s =+1
⎤ L
∴ k = A ⎢ ∫ B DB ⋅ Jds ⎥
ˆ T
J =
⎣ s =−1 ⎦ 2

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ Some notes on the ever-present Jacobian matrix:
„ Main Entry: Ja·co·bi·an
Pronunciation: j&-'kO-bE-&n, yä-
Function: noun
Etymology: K. G. J. Jacobi died 1851 German mathematician
: a determinant defined for a finite number of functions of the same
number of variables in which each row consists of the first partial
derivatives of the same function with respect to each of the
variables
„ Main Entry: de·ter·mi·nant
Pronunciation: di-'t&r-m&-n&nt
Function: noun
1 : an element that identifies or determines the nature of something
or that fixes or conditions an outcome

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ Some more notes on the ever-present Jacobian matrix:
„ Referred to as a “Jacobian Matrix” or just “Jacobian”.
„ Any matrix that defines the conversion from one ‘rate’ to another.
„ Like a matrix of currency conversions.
„ J in this case is a scalar entity:
dx change in the cartesian coordinate 'x'
J= ≡
ds change in the natural coordinate 's '

„ Robotics: defines how a collection of joint rates maps over a


translational and rotational velocity of an end effector.
„ Basis for kinematic and dynamic robotics problems.
„ Dynamics: how do changes in the state of a body affect the
dynamic equilibrium of that body.
„ How we measure the ‘stiffness’ of a set of dynamic equations.
„ Has consequence in how we apply numerical integrators to solve
dynamics problems.

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ Completing the evaluation of k.

⎡ s =+1

k = A ⎢ ∫ B DB ⋅ Jds ⎥
ˆ T

⎣ s =−1 ⎦
⎡ −1 ⎤
s =+1 ⎢
AL L ⎥ ⎡ −1 1⎤
= ∫ ⎢ ⎥E⎢
2 s =−1 ⎢ 1 ⎥ ⎣ L ⎥
L⎦
⋅ ds

⎣⎢ L ⎦⎥
AE ⎡ 1 −1⎤
=
L ⎢⎣ −1 1 ⎥⎦

„ So what’s the big deal about introducing s?

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.

„ Look at the second term in the original functional…


L G

ˆf B = A N T b ⋅ dx
x =0

⎡1 − s ⎤
s =+1 ⎢
2 ⎥ ⎛L ⎞
=A∫ ⎢ ⎥ bx ⋅ ⎜ ds ⎟
s =−1 ⎢
1+ s ⎥ ⎝2 ⎠
⎢⎣ 2 ⎥⎦
Still pretty easy to evaluate but we saw
for the LST how the integrals get a lot
more complicated.
But now we know the limits of the
integration and the shape functions are
well defined over the -1 to +1 interval

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ §10.4. Guassian Quadrature (Numerical Integration).
„ Numerical integration methods provide a value for an integral I:
x2

I= ∫ y( x) ⋅ dx
x1

„ If we turn to numerical approaches it is because the integrand is to


complex to get a useful expression for I.
„ Numerical methods sample the integrand, y(x), and approximate
the integral, I, with a weighted summation of the sample values.
„ The choice of the weighting factors is what defines the particular
integration scheme.
„ Numerical methods effectively replace the original function, y(x),
with a polynomial approximate.

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
x2 n
I= ∫ y( x) ⋅ dx ≈ ∑W ⋅ y( x )
x1 i =1
i i

„ y(xi) = value of the function at the sample point xi.


„ Wi = weighting applied to this sample.
Guassian Quadrature
uses ‘n’ sample
values to create a
polynomial of order
‘2n-1’ which replaces
the original function.

For MECH 420 we


are integrating over
domains:
s ∈ ( −1.. + 1)
and/or
t ∈ ( −1.. + 1)
MECH 420: Finite Element Applications
Lecture 20: Isoparametric Formulations.
„ Guassian quadrature:
„ Uses a symmetric distribution of sample points over the ‘-1’ to

‘+1’ s domain.
„ Symmetric points carry the same weight.

„ Any number of points can be chosen.

„ One-point Guassian quadrature,


„ Two-point Guassian quadrature, …
„ If an odd number of points is chosen then one of the points is
the element centroid.
„ If we consider how the two-point quadrature formula is obtained
then we can repeat this procedure to obtain quadrature formulas
that are not tabulated.

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.

MECH 420: Finite Element Applications


Lecture 20: Isoparametric Formulations.
„ Using the two point formula we could quickly discretize the body
fixed load we looked at earlier:
⎡1 − s ⎤
G s =+1 ⎢ ⎥
ˆf B = A N T b ⋅ dx = A ⎢ 2 ⎥ b ⋅ ⎛ L ds ⎞
L


x =0
∫ 1 + s ⎥ x ⎜⎝ 2 ⎟⎠
s =−1 ⎢
⎢⎣ 2 ⎥⎦
⎛ ⎡1 − s ⎤ ⎞ ⎛ ⎡1 − s ⎤ ⎞
AL ⎜ ⎢ 2 ⎥ ⎟ AL ⎜ ⎢ 2 ⎥ ⎟
= (1.000 ) ⎜⎢ ⎥ bx ⎟ + (1.000 ) ⎜⎢ ⎥ bx ⎟
2 ⎜ ⎢1 + s ⎥ ⎟ 2 ⎜ ⎢1 + s ⎥ ⎟
⎜⎢ ⎟ ⎜⎢ ⎟
⎝ ⎣ 2 ⎥⎦ ⎠ s =−0.5773 ⎝ ⎣ 2 ⎥⎦ ⎠ s =+0.5773
„ Note: we would have to evaluate bx at the points:
1
x= ⎡(1 + 0.57735 ) x1 + (1 − 0.57735 ) x2 ⎤⎦
2⎣
and
1
x = ⎡⎣(1 − 0.57735 ) x1 + (1 + 0.57735 ) x2 ⎤⎦
2

MECH 420: Finite Element Applications

Вам также может понравиться