22-Mec-B10 Finite Element Analysis · December 2013
Question 6 of 7: Eight-node hexahedron — shape functions, field variables, Jacobian and strains [20 marks]
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
Notes on this paper
Paper format. National Examinations, December 2013 — 07-Mec-B10 Finite Element Analysis, 3 hours, open book (any textbooks, references or notes; any non-communicating calculator). Seven questions of equal value [20 marks each]; candidates attempt any five, and all questions are to be solved within the context of the finite element method. Every one of the seven questions is worked below, because the full set is more useful as a study resource than a five-question subset.
Reference texts. The worked answers below are keyed to the standard finite-element texts used for this subject:
Logan, A First Course in the Finite Element Method, 6th ed. — Ch. 6 (constant-strain triangle), Ch. 10 (isoparametric formulation, Gauss quadrature).
Reddy, An Introduction to the Finite Element Method, 4th ed. — Ch. 2–3 (weighted-residual methods, collocation, Galerkin vs. Ritz), Ch. 5 (Timoshenko beam elements and shear locking).
Cook, Malkus, Plesha & Witt, Concepts and Applications of Finite Element Analysis, 4th ed. — Ch. 6 (isoparametric elements, the Jacobian and element distortion).
Zienkiewicz, Taylor & Zhu, The Finite Element Method: Its Basis and Fundamentals, 7th ed. — Ch. 6 (shape functions, completeness and geometric invariance), Ch. 15 (adaptivity, h- and p-refinement).
Hutton, Fundamentals of Finite Element Analysis — Ch. 6 (interpolation functions, geometric isotropy).
Question 6: Eight-node hexahedron — shape functions, field variables, Jacobian and strains [20 marks]
Given. A rectangular eight-node brick (trilinear hexahedron) whose corner coordinates are listed in the figure, loaded through a prescribed set of nodal displacements:
Given data — Question 6 (coordinates in cm)
Node
$x$
$y$
$z$
$(\xi,\eta,\zeta)$
$u$ (mm)
1
1
2
2
$(-1,-1,-1)$
0
2
6
2
2
$(+1,-1,-1)$
0.045
3
6
8
2
$(+1,+1,-1)$
0.045
4
1
8
2
$(-1,+1,-1)$
0
5
1
2
4
$(-1,-1,+1)$
0
6
6
2
4
$(+1,-1,+1)$
0.045
7
6
8
4
$(+1,+1,+1)$
0.045
8
1
8
4
$(-1,+1,+1)$
0
All $v_i=w_i=0$; element spans $1\le x\le 6$, $2\le y\le 8$, $2\le z\le 4$ cm
Find. (a) the eight trilinear shape functions; (b) $N_3$ at nodes 3 and 5 and at the centroid; (c) the displacement fields $u$, $v$, $w$ as functions of $(\xi,\eta,\zeta)$; (d) the Jacobian matrix and its determinant; (e) $\varepsilon_x$ and $\gamma_{xy}$ at the element centre.
Figure 6.1 — The eight-node hexahedron. Nodes 1 to 4 lie on the z = 2 cm face and nodes 5 to 8 on the z = 4 cm face; node 1 and its three edges are hidden and shown dashed. The element spans 1 ≤ x ≤ 6, 2 ≤ y ≤ 8, 2 ≤ z ≤ 4 cm.
Approach. Read the natural coordinates of each node from the figure’s numbering, write the trilinear shape functions in the standard product form, use them to interpolate the displacement field, differentiate the geometric mapping to build the Jacobian matrix, and finally convert natural-coordinate derivatives into physical strains through the inverse Jacobian.
(a) The eight shape functions [4 marks]
Assign natural coordinates to the nodes. Comparing the global coordinates with the parent cube, $x=1$ and $x=6$ are the $\xi=-1$ and $\xi=+1$ faces, $y=2$ and $y=8$ are the $\eta=\mp 1$ faces, and $z=2$ and $z=4$ are the $\zeta=\mp1$ faces. This gives the $(\xi_i,\eta_i,\zeta_i)$ triples listed in the Given table.
Write the trilinear product form. Because the parent element is a cube and the interpolation must be linear along every edge, the shape functions are the tensor product of three one-dimensional linear functions and can be written compactly as
$$\boxed{\,N_i(\xi,\eta,\zeta)=\tfrac{1}{8}\big(1+\xi\,\xi_i\big)\big(1+\eta\,\eta_i\big)\big(1+\zeta\,\zeta_i\big),\qquad i=1,2,\dots ,8\,}$$
Written out for the numbering of this element, the eight functions are
Trilinear shape functions of the eight-node hexahedron
$i$
$N_i(\xi,\eta,\zeta)$
$i$
$N_i(\xi,\eta,\zeta)$
1
$\tfrac{1}{8}(1-\xi)(1-\eta)(1-\zeta)$
5
$\tfrac{1}{8}(1-\xi)(1-\eta)(1+\zeta)$
2
$\tfrac{1}{8}(1+\xi)(1-\eta)(1-\zeta)$
6
$\tfrac{1}{8}(1+\xi)(1-\eta)(1+\zeta)$
3
$\tfrac{1}{8}(1+\xi)(1+\eta)(1-\zeta)$
7
$\tfrac{1}{8}(1+\xi)(1+\eta)(1+\zeta)$
4
$\tfrac{1}{8}(1-\xi)(1+\eta)(1-\zeta)$
8
$\tfrac{1}{8}(1-\xi)(1+\eta)(1+\zeta)$
These satisfy the two defining requirements: $N_i=1$ at node $i$ and zero at the other seven nodes, and $\sum_{i=1}^{8}N_i=1$ everywhere in the element.
(b) $N_3$ at three locations [3 marks]
Evaluating $N_3=\tfrac18(1+\xi)(1+\eta)(1-\zeta)$ at the three requested points, and noting that node 3 is at $(+1,+1,-1)$, node 5 at $(-1,-1,+1)$ and the centroid at $(0,0,0)$:
The first two results are the Kronecker-delta property in action, and the third reflects the fact that at the centroid all eight nodes contribute equally, $\tfrac18$ each, summing to unity.
(c) The displacement field in natural coordinates [3 marks]
Interpolate $u$ from the non-zero nodal values. Only $u_2=u_3=u_6=u_7=0.045$ mm are non-zero, and inspecting the natural coordinates shows these are precisely the four nodes on the $\xi=+1$ face. Hence
$$u(\xi,\eta,\zeta)=0.045\big(N_2+N_3+N_6+N_7\big)$$
Collapse the sum. Every one of these four functions carries the common factor $\tfrac18(1+\xi)$, and the remaining factors run over all four sign combinations of $\eta$ and $\zeta$:
$$N_2+N_3+N_6+N_7=\tfrac18(1+\xi)\Big[(1-\eta)(1-\zeta)+(1+\eta)(1-\zeta)+(1-\eta)(1+\zeta)+(1+\eta)(1+\zeta)\Big]=\tfrac18(1+\xi)(4)=\frac{1+\xi}{2}$$
State the three field variables. Therefore
$$\boxed{\,u(\xi,\eta,\zeta)=0.0225\,(1+\xi)\ \text{mm},\qquad v=0,\qquad w=0\,}$$
The field is a function of $\xi$ alone: it is zero on the $\xi=-1$ face, $0.0225$ mm at the centroid, and $0.045$ mm on the $\xi=+1$ face. Physically the element is being stretched uniformly in the $x$ direction with the $x=1$ cm face held fixed.
(d) Jacobian matrix and Jacobian [7 marks]
Write the geometric mapping. The element is isoparametric, so the geometry uses the same shape functions: $x=\sum N_ix_i$, $y=\sum N_iy_i$, $z=\sum N_iz_i$. Because the element is a rectangular box, these collapse to
$$x=\frac{1+6}{2}+\frac{6-1}{2}\xi=3.5+2.5\,\xi ,\qquad y=\frac{2+8}{2}+\frac{8-2}{2}\eta=5+3\,\eta ,\qquad z=\frac{2+4}{2}+\frac{4-2}{2}\zeta=3+\zeta$$
in centimetres.
Assemble the Jacobian matrix. By definition
$$[J]=\begin{bmatrix}\dfrac{\partial x}{\partial \xi} & \dfrac{\partial y}{\partial \xi} & \dfrac{\partial z}{\partial \xi}\\[6pt] \dfrac{\partial x}{\partial \eta} & \dfrac{\partial y}{\partial \eta} & \dfrac{\partial z}{\partial \eta}\\[6pt] \dfrac{\partial x}{\partial \zeta} & \dfrac{\partial y}{\partial \zeta} & \dfrac{\partial z}{\partial \zeta}\end{bmatrix}=\begin{bmatrix}\sum \dfrac{\partial N_i}{\partial \xi}x_i & \sum \dfrac{\partial N_i}{\partial \xi}y_i & \sum \dfrac{\partial N_i}{\partial \xi}z_i\\[6pt] \sum \dfrac{\partial N_i}{\partial \eta}x_i & \cdots & \cdots\\[6pt] \sum \dfrac{\partial N_i}{\partial \zeta}x_i & \cdots & \cdots\end{bmatrix}$$
Differentiating the mapping of step 1 gives the diagonal result
$$\boxed{\,[J]=\begin{bmatrix}2.5 & 0 & 0\\ 0 & 3 & 0\\ 0 & 0 & 1\end{bmatrix}\ \text{cm}\,}$$
The off-diagonal terms vanish because the element edges are parallel to the global axes, and every entry is constant because the element is undistorted.
Evaluate the Jacobian (its determinant).
$$\boxed{\,|J|=\det[J]=(2.5)(3)(1)=7.5\ \text{cm}^{3}\,}$$
This is the volume scale factor between the parent cube and the physical element: $dV=|J|\,d\xi\,d\eta\,d\zeta$. Integrating the constant over the parent cube of volume 8 gives $7.5\times 8 = 60\ \text{cm}^{3}$, which matches the physical volume $5\times 6\times 2=60\ \text{cm}^{3}$ — the check that confirms both the mapping and the determinant.
Note the invertibility condition. $|J|=7.5\gt 0$ everywhere and is constant, so the mapping is one-to-one throughout the element and $[J]^{-1}=\mathrm{diag}(0.4,\ 1/3,\ 1)\ \text{cm}^{-1}$ exists everywhere. This is the ideal case; a distorted brick would give a $|J|$ that varies with position, and a badly distorted one could drive it to zero or negative.
(e) Strains at the element centre [3 marks]
Convert natural derivatives to physical derivatives. The chain rule gives $\{\partial/\partial x,\ \partial/\partial y,\ \partial/\partial z\}^{T}=[J]^{-1}\{\partial/\partial \xi,\ \partial/\partial \eta,\ \partial/\partial \zeta\}^{T}$. With the diagonal $[J]$ this reduces to
$$\frac{\partial}{\partial x}=\frac{1}{2.5}\frac{\partial}{\partial \xi} ,\qquad \frac{\partial}{\partial y}=\frac{1}{3}\frac{\partial}{\partial \eta} ,\qquad \frac{\partial}{\partial z}=\frac{\partial}{\partial \zeta}$$
Compute the normal strain. From part (c), $\partial u/\partial\xi = 0.0225$ mm, so
$$\varepsilon_x=\frac{\partial u}{\partial x}=\frac{1}{2.5}\,(0.0225\ \text{mm})=0.009\ \frac{\text{mm}}{\text{cm}}=\frac{0.009}{10}=9.0\times10^{-4}$$
where the final step converts centimetres to millimetres so that the strain is dimensionless.
Compute the shear strain. Since $u$ depends only on $\xi$ and $v\equiv 0$,
$$\gamma_{xy}=\frac{\partial u}{\partial y}+\frac{\partial v}{\partial x}=\frac{1}{3}\frac{\partial u}{\partial \eta}+0=0+0=0$$
giving the two requested strains
$$\boxed{\ \varepsilon_x=9.0\times10^{-4}\ \ (900\ \mu\varepsilon),\qquad \gamma_{xy}=0\ }$$
Both are independent of position, so the values at the centre hold everywhere in the element. The result is confirmed directly from the geometry: the $\xi=+1$ face moves $0.045$ mm relative to the fixed $\xi=-1$ face over a physical span of $5\ \text{cm}=50\ \text{mm}$, so $\varepsilon_x=0.045/50=9.0\times10^{-4}$, and no face displaces transversely, so there is no shear.