Professional Documents
Culture Documents
ch16 PDF
ch16 PDF
16–1
Chapter 16: THE TEN NODE TETRAHEDRON 16–2
TABLE OF CONTENTS
Page
§16.1. INTRODUCTION 16–3
§16.2. THE QUADRATIC TETRAHEDRON 16–3
§16.3. PARTIAL DERIVATIVE CALCULATIONS 16–4
§16.3.1. Implementation Considerations . . . . . . . . . . . 16–6
§16.4. GAUSS RULES OVER TETRAHEDRA 16–8
§16.5. THE ELEMENT STIFFNESS MATRIX 16–8
§16.6. THE CONSISTENT NODE FORCE VECTOR 16–9
EXERCISES . . . . . . . . . . . . . . . . . . . . . . 16–10
16–2
16–3 §16.2 THE QUADRATIC TETRAHEDRON
§16.1. INTRODUCTION
The technique used in Chapter 15 for the derivation of the linear triangle is now extended to the
10-node tetrahedron, also called the quadratic tetrahedron. This element can have curved faces and
edges. The extension is based on the isoparametric technique and the partial-derivative construction
procedure followed the IFEM course for the 6-node triangle.
The ten node tetrahedron shown in Figure 16.1 is the next complete-polynomial member of the
isoparametric tetrahedron family.
The element has four corners with local numbers 1 through 4, which must be traversed following the
same convention as the four node tetrahedron. It has six side nodes, with local numbers 5 through
10; nodes 5,6,7 are located on sides 1-2, 2-3 and 3-1, whereas nodes 8,9,10 are located on sides 1-4,
2-4, and 3-4. The side nodes may be located arbitrarily subjected to positive-Jacobian-determinant
constraints. Each element face is defined by six nodes, which are not necessarily on a plane.
The isoparametric element definition is
(e)
1 1 1 1 ... 1 N1
1 x1 x2 x3 x4 . . . x10 N2(e)
x y1 y2 y3 y4 . . . y10 N (e)
3
y = z1 z2 z3 z4 . . . z 10 N (e) . (16.1)
ux u x1 u x2 u x3 u x4 . . . u x10 .4
..
uy u y1 u y2 u y3 u y4 . . . u y10
(e)
u z1 u z2 u z3 u z4 . . . u z10 N10
These shape functions are similar in form to those of the six node quadratic triangle discussed in
IFEM. If the element is curved (that is, the six nodes that define each face are not on a plane), the
tetrahedron coordinates no longer fall on planes, but form a curvilinear system.
16–3
Chapter 16: THE TEN NODE TETRAHEDRON 16–4
8 10
z 7 3
1 9
5 6
x y 2
= wi = wi + + + ,
∂x ∂x ∂ζ1 ∂x ∂ζ2 ∂ x ∂ζ3 ∂ x ∂ζ4 ∂x
16–4
16–5 §16.3 PARTIAL DERIVATIVE CALCULATIONS
where all sums are understood to run from i = 1 through 10, and element superscripts on the shape
functions have been suppressed for clarity. In matrix form:
∂ Ni
wi ∂ζ
∂w ∂ζ1 ∂ζ2 ∂ζ3
∂ζ4 1
∂x ∂x ∂x ∂x ∂x ∂ Ni
∂w
wi
∂ζ4 ∂ζ2
= ∂ζ1 ∂ζ2 ∂ζ3
(16.5)
∂y ∂y ∂y ∂y ∂y ∂N
∂w wi ∂ζ3i
∂ζ1 ∂ζ2 ∂ζ3 ∂ζ4
∂z ∂z ∂z ∂z ∂z ∂ Ni
wi ∂ζ
4
Transposing both sides of (16.5) while switching the left and right hand sides, yields a form exploited
below:
∂ζ1 ∂ζ1 ∂ζ1
∂x ∂y ∂z
∂ζ2 ∂ζ2 ∂ζ2
∂ Ni ∂ Ni ∂ Ni
∂x ∂y ∂z ∂w ∂w ∂w
wi ∂∂ζNi wi ∂ζ wi ∂ζ ∂ζ4 ∂ζ3 ∂ζ3 ∂ζ3
= ∂x ∂y ∂z .
1 2 3
∂x ∂y ∂z
∂ζ4 ∂ζ4 ∂ζ4
∂x ∂y ∂z
(16.6)
Now make w ≡ x, y, z and stack the results row-wise:
∂ζ1 ∂ζ1 ∂ζ1
∂x ∂y ∂z
xi ∂∂ζNi xi ∂∂ζNi xi ∂∂ζNi xi ∂∂ζNi ∂ζ2
∂x
∂x ∂x
∂ζ2 ∂ζ2 ∂x ∂y ∂z
∂ Ni
1 2 3 4
∂N ∂x ∂y ∂z ∂y ∂y ∂y
yi ∂ζ i yi ∂∂ζNi yi ∂∂ζNi yi ∂ζ = ∂x .
1 2 3 4 ∂ζ3 ∂ζ3 ∂ζ3 ∂y ∂z
∂ Ni ∂ Ni ∂x ∂y ∂z ∂z ∂z ∂z
z i ∂ζ z i ∂∂ζNi z i ∂∂ζNi z i ∂ζ ∂x ∂y ∂z
1 2 3 4 ∂ζ4 ∂ζ4 ∂ζ4
∂x ∂y ∂z
(16.7)
This is a linear system with the required unknowns in the second matrix, but its coefficient matrix
is not square. To achieve that, differentiate both sides of the identity ζ1 + ζ2 + ζ3 + ζ4 = 1 with
respect to x, y and z, and insert as first row:
∂ζ1 ∂ζ1 ∂ζ1
∂1 ∂1 ∂1
1 1 1 1 ∂x ∂y ∂z ∂z ∂x ∂y
∂ Ni ∂ Ni ∂ Ni ∂ Ni
xi xi ∂ζ xi ∂ζ xi ∂ζ ∂ζ2 ∂ζ2 ∂ζ2 ∂x
∂x ∂x
∂ζ1 ∂x ∂y ∂z
∂ Ni
2
∂ Ni
3
∂ Ni
4
∂x
∂z ∂y
∂N = ∂y
∂y . ∂y
yi ∂ζ i yi ∂ζ yi ∂ζ yi ∂ζ ∂ζ3 ∂ζ3 ∂ζ3
1 2 3 4 ∂x ∂y ∂z ∂x
∂z ∂y
∂ Ni ∂ Ni ∂ Ni ∂ Ni
∂z ∂z ∂z
z i ∂ζ z i ∂ζ z i ∂ζ z i ∂ζ ∂ζ4 ∂ζ4 ∂ζ4 ∂z ∂x ∂y
1 2 3 4
∂x ∂y ∂z
(16.8)
But ∂ x/∂ x = ∂ y/∂ y = ∂z/∂z = 1 and ∂1/∂ x = ∂1/∂ y = ∂1∂z = ∂ x/∂ y = ∂ x/∂z = ∂ y/∂ x =
∂ y/∂z = ∂z/∂ x = ∂z/∂ y = 0 because x, y and z are independent coordinates. Consequently we
16–5
Chapter 16: THE TEN NODE TETRAHEDRON 16–6
arrive at a system of linear equations of order 4 with three right-hand sides. Using the summation
convention to get rid of sum symbols the system is
∂ζ1 ∂ζ1 ∂ζ1
1 1 ∂x
1 1 ∂y ∂z
∂ N ∂ N ∂ N ∂ N
xi i x i x i x i ∂ζ2 ∂ζ2 ∂ζ2 0 0 0
∂ζ1 i ∂ζ2 i ∂ζ3 i ∂ζ4 ∂ x ∂y ∂z
1 0 0
∂ Ni = . (16.9)
yi ∂ζ yi ∂∂ζNi yi ∂∂ζNi yi ∂∂ζNi ∂ζ3 ∂ζ3 ∂ζ3 0 1 0
4
1 2 3
∂ x ∂ y ∂z 0 0 1
∂ N ∂ N ∂ N ∂ N
z i ∂ζ i z i ∂ζ i z i ∂ζ i z i ∂ζ i
∂ζ4 ∂ζ4 ∂ζ4
1 2 3 4
∂x ∂y ∂z
In compact matrix notation
JP = Iaug . (16.10)
where
1 1 1 1
1 1 1 1 x i ∂ Ni x i ∂ Ni x i ∂ Ni x i ∂ Ni
J Jx2 Jx3 Jx4 ∂ζ1 ∂ζ2 ∂ζ3 ∂ζ4
J = x1 = ∂ N i ∂ N i ∂ N i ∂ Ni
. (16.11)
Jy1 Jy2 Jy3 Jy4 yi ∂ζ1 yi ∂ζ2 yi ∂ζ3 yi ∂ζ4
z i ∂∂ζNi z i ∂∂ζNi z i ∂∂ζNi z i ∂∂ζNi
Jz1 Jz2 Jz3 Jz4
1 2 3 4
and Iaug is the 3 × 3 identity matrix augmented with a zero first row. Taking the partials of the shape
functions (16.2) with respect to the tetrahedron coordinates and substituting into the above yields
Jx1 = x1 (4ζ1 − 1) + 4x5 ζ2 + 4x7 ζ3 + 4x8 ζ4 ,
Jx2 = x2 (4ζ2 − 1) + 4x6 ζ3 + 4x5 ζ1 + 4x9 ζ4 ,
(16.12)
Jx3 = x3 (4ζ3 − 1) + 4x7 ζ1 + 4x6 ζ2 + 4x10 ζ4 ,
Jx4 = x4 (4ζ4 − 1) + 4x8 ζ1 + 4x9 ζ2 + 4x10 ζ3 .
For Jyi and Jzi replace xi by yi and z i , respectively, into the above formulas.
Summarizing, to compute the x-y-z partials for a function w interpolated as per (16.3) the recipe
is: form the linear system (16.10) from the geometric data, solve for the 12 tetrahedron coordinates
partials, and substitute these into (16.5).
REMARK 16.1
By analogy with the isoparametric brick elements, matrix J of (16.10) may be called a Jacobian matrix.
However, the factor J that appears in the element-of-volume transformation
d V (e) = J dζ1 dζ2 dζ3 ζ4 , (16.13)
is not det J = |J|, but J = 16 |J. If the element has straight sides with side nodes at the midpoints, J is constant
and equal to the volume V of the tetrahedron, which is given by the usual formula
1 1 1 1
x x2 x3 x4
J = V = 16 1 , (16.14)
y1 y2 y3 y4
z1 z2 z3 z4
as in Chapter 6. In this case J is constant over the element.
16–6
16–7 §16.3 PARTIAL DERIVATIVE CALCULATIONS
For the 4-node tetrahedron, J = 16 V was constant over the element, but now it generally will vary
with position unless the tetrahedron has planar faces.
16–7
Chapter 16: THE TEN NODE TETRAHEDRON 16–8
Four-point rule (exact for constant through quadratic polynomials over plane-face tetrahedra):
1
F(ζ1 , ζ2 , ζ3 , ζ4 ) d V (e) ≈ 14 F(α, β, β, β)+ 14 F(β, α, β, β)+ 14 F(β, β, α, β)+ 14 F(β, β, β, α),
V V (e)
(16.21)
where α = 0.58541020, β = 0.13819660. More details and exact values for the integration rules
are posted in Chapter 17: A Compendium of Gauss Integration Rules for FEM.
16–8
16–9 §16.6 THE CONSISTENT NODE FORCE VECTOR
where p is the number of Gauss points, (ζik ) denotes the coordinate quartet (ζ1 ,ζ2 ,ζ3 , ζ4 ) at the k th
integration point, and wk are the corresponding weights. The stress-strain matrix E is the same as
in Chapter 6. The four-point rule integrates this element with the correct rank.
where N is the 3 × 30 matrix of shape functions that relates element displacements to node dis-
placements:
ux
u y = Nu. (16.28)
uz
For this element and the node displacement ordering (16.22)
N1 ... N10 0 ... 0 0 ... 0
N= 0 ... 0 N1 ... N10 0 ... 0 , (16.29)
0 ... 0 0 ... 0 N1 ... N10
16–9
Chapter 16: THE TEN NODE TETRAHEDRON 16–10
EXERCISE 16.1
[A:15] The 10-node tetrahedron element is converted into an 11-point tetrahedron by adding node point 11
(e)
located at the centroid ζ1 = ζ2 = ζ3 = ζ4 = 1/4. What is the shape function N11 ? (You do not need to write
the full element definition).
EXERCISE 16.2
[A:15] The next full-polynomial, isoparametric member of the tetrahedron family is the cubic tetrahedron,
which has 21 node points. Where do you think the nodes are located?
EXERCISE 16.3
[A/C:25] Derive the shape functions for the 21-node tetrahedron.
EXERCISE 16.4
[A:15] Justify the rule (16.18).
EXERCISE 16.5
[A/C:20] Compute f(e) for a straight-face 10-node tetrahedron if the body forces bx = b y = 0 and bz is
constant, using the 4-point rule (16.21) to evaluate the integral in (16.27). (You need to give only the z force
components).
16–10