Вы находитесь на странице: 1из 13

.

10

Axisymmetric Iso-P Elements

10–1

Chapter 10: AXISYMMETRIC ISO-P ELEMENTS

10–2

Page

 §10.1. INTRODUCTION 10–3 §10.2. ISOPARAMETRIC DEFINITION 10–3 §10.3. THE ELEMENT STIFFNESS MATRIX 10–5 §10.3.1. The Strain-Displacement Matrix 10–5 §10.3.2. The Stiffness Matrix Integrand 10–5 §10.3.3. Numerical Integration 10–6 §10.4. CONSISTENT NODE FORCES FOR BODY LOADS 10–6 §10.4.1. Body Forces 10–7 §10.4.2. Thermal Effects 10–7 EXERCISES 10–9

10–2

10–3

§10.2 ISOPARAMETRIC DEFINITION

§10.1.

INTRODUCTION

In this Chapter we consider the ﬁnite element discretization of structures of revolution (SOR). The focus in this Chapter will be on isoparametric elements, so as to tie with the introductory course.

The dimensionality reduction process described in Chapter 9 “folds” the problem into integrals taken over the generating cross section and its boundaries. The ﬁnite element discretization can be therefore conﬁnedto the r z plane andthe circumferential (θ) dimension conceptually disappears.

The resulting ﬁnite elements are called axisymmetric solid elements, SOR elements, or “ring” elements in the literature. They are deﬁned completely by the geometry of their cross section in the (r, z) plane, as illustratedin Figure 10.1. Because this cross section is plane, the element geometry deﬁnition is two-dimensional. It follows that the two-dimensional element conﬁgurations studied in IFEM: 3 and 6-node triangles, 4-, 8-, and 9-node quadrilaterals, etc, can be used.

The key difference with respect to the plane stress case is the appearance of the “hoop” strain and stress, which together contribute a term

1 2 σ θθ θθ

to the strain energy density. This in turn introduces some additional terms in the structure of the strain-displacement matrix B, which translate into additional contributions in the formation of the element stiffness matrix andconsistent node force vector.

(10.1)

§10.2.

ISOPARAMETRIC DEFINITION

Axisymmetric ﬁnite elements are developed in this Chapter using the displacement-based isopara- metric formulation. That is, the element cross section geometry anddisplacement ﬁeldare inter- polatedby the same shape functions. For an isoparametric element with n nodes:

1

r

z

u

u

r

z

=

1

r

z 1

1

u

u

r1

z1

1

r

z 2

2

u r2

u z2

···

···

···

···

···

1

r

z n

n

u rn

u zn

(e)

N

N

1

(e)

2

.

.

.

(e)

N

n

.

(10.2)

Here u r (r, z) and u z (r, z) are ﬁnite element approximations to the actual displacement ﬁeld com-

ponents, while N

(e) are the usual element shape functions in element natural coordinates deﬁned in

i

the Introduction to Finite Element Methods (IFEM) course. The same continuity and completeness

requirements apply with minor differences (see remarks below).

REMARK 10.1

Displacement basedelements are basedon the TPE functional. The variational index for the only master ﬁeld:

displacements, is one. Consequently the interelement continuity required is C 0 . That is, the displacement components u r and u z must be continuous between adjacent elements. Element boundaries lying on the axis of revolution are special: at such points the radial displacement u r must vanish if the structure is continuous there (that is, a tiny hole is precluded) although there is no need to make that condition explicit in the element formulation.

10–3

Chapter 10: AXISYMMETRIC ISO-P ELEMENTS

10–4

generating cross-section
z
r
generic "ring" finite element

Figure 10.1.

FEM discretization of a SOR with axisymmetric “ring” elements.

10–4

10–5

§10.3 THE ELEMENT STIFFNESS MATRIX

REMARK 10.2

The completeness criterion demands that all rigid body modes and strain states be exactly represented. This is met if the element shape functions can represent any displacement ﬁeld that is linear in r and z. (In fact, a more detailed analysis shows that this is a slight overkill, but will do for now.)

§10.3.

THE ELEMENT STIFFNESS MATRIX

§10.3.1. The Strain-Displacement Matrix

The strain-displacement matrix that appears in the computation of the element stiffness matrix is given by

B = DN

(10.3)

where D is the strain-displacement gradient matrix introduced in the previous Chapter,

r

0

1

r

z and N is the matrix of element shape functions:

D =

(e)

N = N 0

1

(e)

N

2

0

···

···

(e)

N

n

0

0

z

0

r

0

N

(e)

1

,

0

(e)

N

2

···

···

0

(e)

N

n

,

(10.4)

(10.5)

Here we have assumedthat the same shape functions are usedin both r and z directions.

Comparing these expressions with those in Chapter 15 of IFEM, we see that N is the same, and D has an extra row. Multiplying (10.3) through we obtain

B = DN =

q r

0

q

q

z

θ

0

q z

0

q r

,

where the n-component row vectors q r , q z , q θ are deﬁned by

q r = N r q z = N z

q θ = N

(e)

1

(e)

1

(e)

1

r

(e)

N

2

r

(e)

N

2

z

(e)

N

2

r

10–5

···

···

···

(e)

N

n

r

(e)

N

n

(e)

N

n

r

z

.

 (10.6) , , (10.7)

Chapter 10: AXISYMMETRIC ISO-P ELEMENTS

10–6

§10.3.2. The Stiffness Matrix Integrand

Let the constitutive (stress-strain) matrix be

E =  

E

E

E

E

11

12

13

14

E 12

E 22

E 23

E 24

E 13

E 23

E 33

0

E 14

E 24

0

E 44

,

(10.8)

in which E 34 = E 43 = 0 to maintain axisymmetry conditions. First we form the product

EB =

E

E

11 q r + 12 q r +

E 13

E 23

q θ + q θ +

E 14

E 24

E

E

13 q r +

14 q r +

E 33

E 44

q

q

θ

z

q

q

z

z

Premultiplying by B T gives

E 12

E 22

E 24

q z + q z +

E 14 q r E 24 q r

E 23 q z

q z + E 44 q r

=  

in which

B T EB = S S rr

rz

S rz S

zz

S rr = q

S rz = S

S zz = q

c 1r + q

zr = q

T

r

T

r

T

T

θ

c

3r + q

T

z

c

4r

c 1z + q

T

θ

c 3z + q

T

r

c 4z + q

T

z

c

2z

T

z

c

4z

c

c

c

c

1r

2r

3r

4r

c 1z

c 2z

c 3z

c 4z

.

(10.9)

(10.10)

(10.11)

§10.3.3. Numerical Integration

We shall assume here that the entries of the element stiffness matrix are computedby the standard numerical integration techniques coveredin IFEM. For example, the stiffness matrix of a quadrilat- eral isoparametric element integratedby a p-point Gauss rule along each isoparametric coordinate is evaluatedas

K (e) =

p p

k=1

=1

w k w l B T k ) E Bk ) r k η ) J k ).

(10.12)

in which as usual ξ k and η denote abcissae of Gauss points whereas w k and w denote integration weights.

Notice the appearance of the radius r at the Gauss point; this quantity also appears inside B in the form 1/r . This radius is obtained by the isoparametric interpolation, namely, the ﬁrst row of (10.6):

r

k ) =

n

i=1

(e)

r i N

i

k ).

(10.13)

Other quantities such as the Jacobian determinant J , are calculatedas explainedin IFEM. In particular the shape function subroutines can be reused, as explained in more detail in the following Chapter.

The stiffness of triangular elements can also be evaluatedby numerical quadrature but using inte- gration rules appropriate to the triangular geometry.

10–6

10–7

§10.4CONSISTENT NODE FORCES FOR BODY LOADS

§10.4.

Body forces (also called volume forces) arise frequently in SOR analysis. The most important loads of this type are

CONSISTENT NODE FORCES FOR BODY LOADS

1. Gravity (own weight). This effect is important in massive SOR, as encounteredin civil, geophysical andnuclear applications.

2. Centrifugal forces in rotating structures. These are important in aerospace andmechanical structures (for examble, high-speedrotating machinery such as turbines).

3. Thermal, shrinkage and prestress effects. These may be important depending on fabrication techniques, material, andthe environment to which the structure will be exposed. In the present course thermal effects are coveredbrieﬂy.

§10.4.1. Body Forces

An appliedbody force ﬁeld b is deﬁned by two components in the r and z directions, each of which is a function of those two directions only:

b(r, z) = b b r z (r, (r, z) z)

The consistent node forces f are given by

f (e) = A

N T b rdA,

(10.14)

(10.15)

where N is the matrix of element shape functions given in (10.5). This can be conveniently processed by numerical integration. Consideraing again the quadrilateral element with a p-point Gauss rule in each direction,

f (e) =

p p

k=1

=1

w k w l N T k ) bk ) r k ) J k ).

(10.16)

§10.4.2. Thermal Effects

Suppose the temperature of the structure changes axisymmetrically by T (r, z) from a reference temperature. The constitutive equations become

σ = E(e α T )

where α is an array of dilatation coefﬁcients that provide thermal strains in a mechanically un- constrained body. (A linear relation between strains and temperature changes is assumed.) For isotropic materials

(10.17)

α T = [ ααα

0 ]

(10.18)

10–7

Chapter 10: AXISYMMETRIC ISO-P ELEMENTS

10–8

where α is the usual coefﬁcient of dilatation of the material. On inserting (10.18) into the potential energy formulation for a generic element yields element stiffness equations that account for thermal effects:

(10.19)

where f M are the mechanical forces condidered so far, and f T , calledthe thermal forces or equivalent thermal loads, account for contribution from the temperature changes:

(e)

K (e) u (e) = f (e) = f

M

+ f

(e)

T

f

(e)

T

= A

B T α TrdA

(10.20)

This term can be evaluatedby numerical integration.

The sum of mechanical andthermal node forces is sometimes calledthe effective node force vector or simply effective forces in the FEM literature.

10–8

10–9

Exercises

Homework Exercises for Chapter 10

Axisymmetric Iso-P Elements

The following material is largely coveredin Chapter 24 of IFEM, postedat

but is recapitulatedhere for completeness.

The four simplest Gauss integration rules over a triangle of area A use 1, 3, 3 and7 points. These rules are tabulatedbelow. In the following expressions, F1 2 3 ) denotes the function to be integrated over the traingle, expressedin terms of the triangular coordinates ζ 1 , ζ 2 and ζ 3 , while A is the area of the element.

One point rule (exact for constant andlinear polynomials over straight sidedtriangles):

1

A

A (e)

F1 2 3 ) d A (e) F(

1

3 ,

1

3 ,

1

3 ).

(E10.1)

Three-point rules (exact for constant through quadratic polynomials over straight sided triangles):

1

A

1

A

A (e)

A (e)

F1 2 3 ) d A (e)

F1 2 3 ) d A (e)

1

3

1

3

F( 2

3

F(

1

2

,

,

1

6 ,

1

6 )

1

2 , 0)

+

+

1

3 F(

1

6 ,

1

3 F(0,

2

3

1

2

,

,

1

6

1

2

) +

)

+

1

3 F(

1

6 ,

1

6 ,

1

3

1

F( 2 , 0,

2

3

).

1

2 ).

(E10.2)

(E10.3)

The latter is also calledthe 3-midpoint rule for obvious reasons.

Seven point rule (exact for constant through cubic polynomials over straight sided triangles):

1

A

A (e)

F1 2 3 ) d A (e) w 0 F(

1

3 ,

1

3 ,

1

3

 ) + w 1 [F(α 1 ,β 1 ,β 1 ) + F(β 1 ,α 1 ,β 1 ) + F(β 1 ,β 1 ,α 1 )] + w 2 [F(α 2 ,β 2 ,β 2 ) + F(β 2 ,α 2 ,β 2 ) + F(β 2 ,β 2 ,α 2 )] ,

(E10.4)

where w 0 = 9/40 = 0.225, w 1 = (155 15)/1200 = 0.1259391805, w 2 = (155 + 15)/1200 =

0.1323941528, α 1 = (9 + 2 15)/21 = 0.7974269853, β 1 = (6 15)/21 = 0.1012865073, α 2 =

(9 2 15)/21 = 0.0597158718, β 2 = (6 + 15)/21 = 0.4701420641.

A Mathematica module that implement these rules is TrigGaussRuleInfo.nb postedin the index of this Chapter. The use of this module is explained in Chapter 24 of the IFEM Notes.

EXERCISE 10.1

[A/C:15] Using the minimum quadrature rule necessary for exactness, verify the following polynomial integrals over straight-sided triangles, where indices i, j, k run over 1,2,3, and r is interpolatedas r = r 1 ζ 1 +r 2 ζ 2 +r 3 ζ 3 .

A (e)

ζ i d A (e) =

1

3 A,

A (e) ζ i ζ j d A (e) =

1

12 A(1 + δ i j ),

A (e) ζ i ζ j ζ k d A (e) =

1 60 γ ijk A,

10–9

(E10.5)

(E10.6)

(E10.7)

Chapter 10: AXISYMMETRIC ISO-P ELEMENTS

10–10

A (e)

A (e)

A (e)

rdA (e) =

3 1 A(r 1 + r 2 + r 3 ) = A r 0 ,

ζ i rdA (e) =

12 1 A(r 1 + r 2 + r 3 + r i ),

r 2 d A (e) =

12 1 A[(r 1 + r 2 + r 3 ) 2 + r

1 2 + r

+ r

2

2

A (e)

ζ i r 2 d A (e) =

1 60 γ ijk A r j r k .

 (E10.8) (E10.9) 3 2 ], (E10.10) (E10.11)

In the above, δ i j = 1 if i =

Note: you can (and should) make use of previous results for expediency; to prove for example (E8.8) substitute (E8.5) as appropriate.

= k, else 2.

j else 0 (the Kronecker delta); γ ijk = 6 if i = j = k, γ ijk = 1 if i

= j

EXERCISE 10.2

[A/C:15] Apply the triangle integration formulas to the non-polynomial integral

for the two cases

A (e)

d A (e)

r

(E10.12)

 (a) r 1 = 0, r 2 = r 3 = a, andcompare to the exact integral 2A/a. (b) r 1 = 1, r 2 = 2, r 3 = 3 andcompare to the exact integral log(27/16) A/a = 0.523238A/a. (The 7 point rule shouldbe accurate to 4 digits).

EXERCISE 10.3

[A/C:15] Repeat (a) of the previous exercise for the integrals

A (e)

ζ

1

r

d

A (e) ,

A (e)

ζ

2

r

d

A (e) ,

A (e)

ζ

3

r

d

A (e) .

(E10.13)

The exact integrals are A/a for the ﬁrst one, and formula exact for the latter case?

2 1 A/a for the other two. Why are all numerical quadrature

EXERCISE 10.4

[A/C:20] The isoparametric deﬁnition of the 3-node linear SOR triangle is

1

r

z

u

u

r

z

=

111

r 2

z 2

u r2

r1

z1

u z2

r 1

z

u

u

1

r

3

z 3

u z3

u z3

ζ 1

ζ

ζ

2

3

(E10.14)

Using results statedin Exercise 10.1 as appropriate, compute the entries

K r1r1 ,

K r1z1

10–10

K z1z1

(E10.15)

10–11

Exercises

of the element stiffness matrix K (e) for the geometry deﬁned by

r 1 = 0,

r 2 = r 3 = a,

z 1 = z 2 = 0,

z 3 = b.

(E10.16)

The material is isotropic with ν = 0 for which the stress-strain matrix is

E = E

1000

0100

0010

000

1

2

,

Note: For integrals that contain 1/r , use one of the 3-point rules.

K r1r1 =

2 1 Eb.

(E10.17)

Partial answer (if midpoint rule used)

EXERCISE 10.5

[A/C:20] The triangle of Exercise 10.4 is subjected to the radial-centrifugal body force ﬁeld

b r = ρω 2 r,

b z = 0

(E10.18)

where ρ and ω are constant (ρ is the mass density while ω is the angular velocity.)

Compute the consistent node force vector f (e) . In doing so interpolate the radial component b r linearly over the element:

(E10.19)

where b ri is ρω 2 r evaluatedat corner i. Partial answers: f r1 = ωa 3 /10, f z1 = 0.

b r 1 2 3 ) = b r1 ζ 1 +

b r2 ζ 2 + b r3 ζ 3 ,

EXERCISE 10.6

[A/C:20] Repeat the previous Exercise for the own-weight body force ﬁeld

b r = 0,

b z = −ρg

where ρ and g are constant. Partial answer: f z2 = −ga/4.

(E10.20)

EXERCISE 10.7

[A/C:20] For the triangle geometry of the preceding 3 exercises, ﬁnd the consistent thermal forces pertaining to a uniform temperature increase T , assuming an isotropic material with zero Poisson’s ratio.

EXERCISE 10.8

[A/C:15] Show that u z = c (c is a constant) is the only possible rigid body mode of a SOR element. (Hint:

consider the presence of the circumferential strain). Hence deduce that the correct rank of the stiffness matrix of a SOR element with n nodes and 2 DOFs per node is 2n 1.

EXERCISE 10.9

[A/C:15] Find the displacement ﬁelds that separately generate the following constant strain states:

where c zz and c rz are constants.

e zz = c zz ,

γ rz = c rz ,

 others zero (E10.21) others zero (E10.22)

10–11

Chapter 10: AXISYMMETRIC ISO-P ELEMENTS

10–12

Figure E10.1.

z

r

Variable-section circular shaft conveying torque: the subject of Exercise 10.11.

EXERCISE 10.10

[A/C:15] Show that there are no displacement ﬁelds that separately generate the following constant strain states:

e rr = c rr ,

e θθ = c θθ ,

 others zero (E10.23) others zero (E10.24)

where c rr and c θθ are constants. Hint: integrate the appropriate strain-displacement relations.

EXERCISE 10.11

[A/C:25] The SOR sketchedin Figure E10.1 (a circular shaft with varying cross section) is subjectedto torsional loading as indicated. According to Saint-Venant’s torsion theory, the displacement components for this case are entirely circumferential, that is, u r = u z = 0 and u θ = u θ (r, z). The torsional shear strains

γ rθ = u r θ

u θ

r

,

10–12

γ zθ = u z θ

,

(E10.25)

10–13

Exercises

are nonzero andfunctions of r, z only; all other strains (e rr , e zz , e θθ and γ rz ) vanish. Assuming the shaft is fabricatedwith an isotropic material, the only nonzero stress components are the shear stresses

where G =

σ rθ = G γ rθ ,

2 1 E/(1 + ν) is the shear modulus.

σ zθ = G γ zθ ,

(E10.26)

 (a) Explain why this problem can be discretized by two-dimensional “ring” ﬁnite elements by laying out a mesh over the (r, z) plane (r ≥ 0), although the element type is different from that considered previously in this Chapter. How many degrees of freedoms would these “torqued ring” elements have per node? (b) If an isoparametric formulation is usedfor the torqued-ring elements of (a), the element stiffness matrix, on suppressing the 2π factor, is given by the usual expression

But how would B and E look like?

K (e) = A (e)

B T EB r dA (e) .

10–13

(E10.27)