16-Civ-B9 The Finite Element Method · December 2018
Question 2 of 3: Two-span beam with a seating gap at the middle support (Problem 2)
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
Notes on this paper
Paper format. National Exams, December 2018 — 16-Civ-B9 The Finite
Element Method. Three hours; four pages; three problems of equal weight, all to be
attempted. Closed book, one aid sheet written on both sides, approved Casio or Sharp
calculator. Candidates are asked to state any interpretive assumptions with the answer
paper, which matters here because Problem 2 is a contact problem whose answer depends on
one such check.
Reference texts for this subject.
Logan, D. L., A First Course in the Finite Element Method, 6th ed., Cengage
— beam elements and the Hermite shape functions (Ch. 4), the bilinear rectangle and
plane stress (Ch. 6 and Ch. 10), work-equivalent nodal loads (§4.5).
Cook, R. D., Malkus, D. S., Plesha, M. E. and Witt, R. J., Concepts and
Applications of Finite Element Analysis, 4th ed., Wiley — completeness,
rigid-body modes and the row-sum property of a stiffness matrix (Ch. 3), consistent load
vectors (Ch. 4).
Bathe, K.-J., Finite Element Procedures, 2nd ed. — the principle of
virtual work as the origin of the finite element equations (Ch. 4), and the treatment of
prescribed displacements and contact conditions.
Hibbeler, R. C., Structural Analysis, 10th ed., Pearson — fixed-end
moments, shear and bending moment diagrams and the sign conventions used to plot
them.
Check — misprint on page 3 of the paper. The
examination prints the first Hermite shape function as
$N_1(s) = \dfrac{2s^3}{L^3} - \dfrac{2s^2}{L^2} + 1$. That expression gives
$N_1(L) = 1$ instead of $0$ and makes $N_1 + N_3 = 2$ at the far node, so it cannot be a
beam shape function. The middle coefficient must be $3$:
$N_1(s) = \dfrac{2s^3}{L^3} - \dfrac{3s^2}{L^2} + 1$, which is consistent with the
$N_3(s)$ printed immediately below it and with the $[K]$ matrix given on the same page.
The corrected form is used throughout. Nothing in the numerical answer depends on it,
because the element stiffness matrix is supplied directly.
Question 2: Two-span beam with a seating gap at the middle support (Problem 2)
Given. A two-span beam, built in at node 1 and at node 3, with a rigid
support seat placed a small distance below the beam at the intermediate node 2, so that the
support can only push upward and only after the beam has deflected onto it. The uniform
load acts on span 2-3 alone; span 1-2 carries no applied load at all.
Given data
Quantity
Symbol
Value
Span length, each of 1-2 and 2-3
$L$
6 m
Uniform load, on span 2-3 only
$q$
10 kN/m (then 12 kN/m)
Young's modulus
$E$
200 GPa
Gap under the middle support
$\Delta$
5 mm
Section height
$h$
300 mm
Section width
$b$
120 mm
Boundary conditions
—
fixed at 1 and 3; unilateral (gap) support at 2
Find. The shear force and bending moment diagrams of span 1-2 under
$q = 10\ \text{kN/m}$, and then the procedure — not the full solution — that
would be followed if the load were raised to $q = 12\ \text{kN/m}$.
[Figure not reproduced: Figure 2 redrawn: fixed at nodes 1 and 3, with a seating gap of 5 mm below the beam at node 2. The uniform load acts on span 2-3 only. See the official exam paper.]
Approach. The gap makes this a contact problem rather than an ordinary
linear analysis, so it is solved in two stages: first assume the support is not touched and
solve the beam as if node 2 were free, then compare the computed deflection with the gap
and, only if the gap is exceeded, re-solve with the vertical displacement at node 2
prescribed at $-\Delta$ and recover the contact force.
Part 1 — section properties. For the rectangular section, with
$b$ the width and $h$ the height,
$$I = \frac{b\,h^{3}}{12}
= \frac{0.120 \times 0.300^{3}}{12}
= 2.70 \times 10^{-4}\ \text{m}^{4} ,$$
so the flexural rigidity is
$$EI = 200 \times 10^{6}\ \text{kPa} \times 2.70 \times 10^{-4}\ \text{m}^{4}
= 5.40 \times 10^{4}\ \text{kN}\cdot\text{m}^{2} .$$
Discretise and identify the free degrees of freedom. Use two beam
elements, 1-2 and 2-3, each of length $L$, with the element matrix printed on the paper.
Nodes 1 and 3 are built in, so $v_1 = \theta_1 = v_3 = \theta_3 = 0$ and only the two
degrees of freedom at node 2 survive. Adding the lower-right block of element 1-2 to the
upper-left block of element 2-3,
$$\mathbf{K}_{ff}
= \frac{EI}{L^{3}}
\begin{bmatrix} 12 & -6L \\ -6L & 4L^{2} \end{bmatrix}
+ \frac{EI}{L^{3}}
\begin{bmatrix} 12 & 6L \\ 6L & 4L^{2} \end{bmatrix}
= \frac{EI}{L^{3}}
\begin{bmatrix} 24 & 0 \\ 0 & 8L^{2} \end{bmatrix} .$$
The coupling terms cancel because the two spans are equal and lie on opposite sides of the
node, so the vertical and rotational equations decouple:
$K_{vv} = 24EI/L^{3} = 6000\ \text{kN/m}$ and
$K_{\theta\theta} = 8EI/L = 72\,000\ \text{kN}\cdot\text{m/rad}$.
Form the work-equivalent nodal loads. Only element 2-3 is loaded. Its
consistent load vector for a downward uniform load is
$\left[-qL/2,\; -qL^{2}/12,\; -qL/2,\; +qL^{2}/12\right]$, of which the first two
entries act at node 2:
$$F_{v2} = -\frac{qL}{2} = -30\ \text{kN}, \qquad
F_{\theta 2} = -\frac{qL^{2}}{12} = -30\ \text{kN}\cdot\text{m} .$$
These are the fixed-end reactions of the loaded span, reversed and applied at the
node.
Stage one — solve with the gap assumed open. With no contact
force at node 2 the two decoupled equations give
$$v_2 = \frac{F_{v2}}{24EI/L^{3}} = -\frac{qL^{4}}{48EI}
= -\frac{10 \times 6^{4}}{48 \times 54\,000} , \qquad
\theta_2 = \frac{F_{\theta 2}}{8EI/L} = -\frac{qL^{3}}{96EI} ,$$
so that
$$\boxed{\,v_2 = -5.000\ \text{mm}, \qquad \theta_2 = -0.4167 \times 10^{-3}\ \text{rad}\,}$$
the negative signs meaning downward and clockwise.
Apply the contact check — this is the whole point of the problem.
The beam would have to descend 5.000 mm to reach the seat, and the gap is exactly
$\Delta = 5\ \text{mm}$. The load is therefore precisely the threshold value: rearranging
the deflection expression, the uniform load that just closes the gap is
$$q_{\text{cr}} = \frac{48\,EI\,\Delta}{L^{4}}
= \frac{48 \times 54\,000 \times 0.005}{6^{4}}
= 10.00\ \text{kN/m} .$$
Since $q = q_{\text{cr}}$, the beam just touches the seat without pressing on it, and
$$\boxed{\,R_2 = 0\ \text{kN}\,}$$
The structure to be analysed for the diagrams is therefore the single fixed-fixed beam of
total length $2L = 12\ \text{m}$ carrying the uniform load over its right-hand half; the
middle support may be ignored.
Recover the end forces of span 1-2. Span 1-2 carries no applied load,
so its member end forces follow from the element matrix alone, acting on the displacement
vector $\{0,\ 0,\ v_2,\ \theta_2\}$:
$$\mathbf{f}^{(1\text{-}2)} = \frac{EI}{L^{3}}
\begin{bmatrix} 12 & 6L & -12 & 6L \\ 6L & 4L^{2} & -6L & 2L^{2} \\
-12 & -6L & 12 & -6L \\ 6L & 2L^{2} & -6L & 4L^{2} \end{bmatrix}
\begin{Bmatrix} 0 \\ 0 \\ -0.005 \\ -0.4167\times 10^{-3} \end{Bmatrix}
= \begin{Bmatrix} 11.25 \\ 37.5 \\ -11.25 \\ 30.0 \end{Bmatrix} ,$$
in kN and kN·m, with $EI/L^{3} = 250\ \text{kN/m}$. Reading the entries: the
element is pushed up by 11.25 kN at node 1 and down by 11.25 kN at node 2, and carries
end moments of 37.5 and 30.0 kN·m.
Convert to diagram ordinates. Converting the nodal moments to the
sagging-positive convention used for plotting, the moment at node 1 is hogging and the
moment at node 2 is sagging:
$$\boxed{\;V_{1\text{-}2} = +11.25\ \text{kN (constant)}, \qquad
M_1 = -37.5\ \text{kN}\cdot\text{m}, \qquad
M_2 = +30.0\ \text{kN}\cdot\text{m}\;}$$
Because the span carries no distributed load, $dV/dx = 0$ and the shear is constant, while
$dM/dx = V$ makes the moment vary linearly. The point of contraflexure sits at
$$x_0 = \frac{37.5}{11.25} = 3.333\ \text{m}$$
from node 1, that is at $0.556L$.
Check the result independently. Treating the beam as fixed-fixed over
$2L = 12\ \text{m}$ and integrating the standard point-load fixed-end formulae
$M_A = \int P a b^{2} / L_t^{2}$ and $R_A = \int P b^{2}(3a+b)/L_t^{3}$ over the loaded
half returns $M_1 = 37.5\ \text{kN}\cdot\text{m}$,
$M_3 = 82.5\ \text{kN}\cdot\text{m}$ and $R_1 = 11.25\ \text{kN}$, matching the
element result exactly. Span 1-2 also closes on its own:
$M_1 + M_2 - V L = 37.5 + 30.0 - 11.25 \times 6 = 0$. The agreement is exact rather than
approximate because the cubic Hermite functions are the exact solution of
$EI\,v'''' = 0$, which is the governing equation of the unloaded span.
Shear force and bending moment diagrams for span 1-2 at q = 10 kN/m. The shear is constant because the span is unloaded; the moment runs linearly from 37.5 kN·m hogging at the fixed end to 30.0 kN·m sagging at node 2.
Check — the load sits exactly on the contact
threshold. At $q = 10\ \text{kN/m}$ the computed free deflection at node 2 is
5.000 mm against a gap of 5.000 mm, so the beam touches the seat with zero contact pressure.
The diagrams above are drawn for that reading, which is the only self-consistent one: any
non-zero reaction would have to lift the beam clear of the support that is supposed to be
producing it. An examiner would expect this to be stated explicitly as an assumption, and
in a real structure the tolerance on a 5 mm shim would decide the answer, so the honest
engineering statement is that node 2 is at the point of first contact and carries between
0 and a very small force.
Procedure if the load increases to q = 12 kN/m
Raising the load past the threshold changes the character of the problem: the support
becomes active and the beam becomes continuous over three supports. The gap makes the
response piecewise linear, so the procedure below is a two-stage linear analysis rather
than an iteration.
Repeat the gap-open analysis at the new load. Keep the same
$\mathbf{K}_{ff}$ and rescale the consistent load vector. The free deflection at node 2
becomes $v_2 = -qL^{4}/(48EI) = -6.000\ \text{mm}$.
Compare with the gap and decide the contact state. Because
$6.000\ \text{mm} \gt \Delta = 5\ \text{mm}$, the beam lands on the seat. The
gap-open solution is inadmissible and must be discarded — it would have the beam pass
through a rigid support.
Re-solve with a prescribed displacement. Impose
$v_2 = -\Delta = -5\ \text{mm}$ as a support settlement and leave $\theta_2$ free. The
remaining scalar equation is
$$K_{\theta\theta}\,\theta_2 = F_{\theta 2} - K_{\theta v}\,v_2 ,$$
which for equal spans has $K_{\theta v} = 0$ and therefore
$\theta_2 = -qL^{3}/(96EI) = -0.500 \times 10^{-3}\ \text{rad}$.
Recover the contact force and confirm the state is admissible. From
the vertical equation,
$$R_2 = K_{vv}\,v_2 + K_{v\theta}\,\theta_2 - F_{v2}
= 6000 \times (-0.005) + 0 + 36 = +6.00\ \text{kN} .$$
The sign test is the whole check: a unilateral support may only push, so a positive
(upward) $R_2$ confirms contact. Had $R_2$ come out negative, the correct answer would have
been the gap-open solution of step 1.
Redraw the diagrams from the new end forces. With $v_2$ and
$\theta_2$ known, the span 1-2 end forces follow from the element matrix exactly as in
step 6 above, giving
$$\boxed{\;V_{1\text{-}2} = +10.5\ \text{kN}, \qquad
M_1 = -36.0\ \text{kN}\cdot\text{m}, \qquad
M_2 = +27.0\ \text{kN}\cdot\text{m}\;}$$
with the point of contraflexure moving out to $36.0/10.5 = 3.429\ \text{m}$. The shape of
the diagrams is unchanged — constant shear and a straight moment line — because
span 1-2 is still unloaded; only the ordinates change.
The step that carries the marks is the second one. Increasing the load by twenty per
cent does not increase the span 1-2 moments by twenty per cent; it reduces them
slightly, from 37.5 to 36.0 kN·m at the fixed end, because the newly active support
takes 6.00 kN of the load and relieves span 1-2. A gap support therefore makes the
structure non-linear in the loading even though every material remains elastic, and
superposition may not be used across the contact threshold. In a general problem, where the
several gap supports do not all close at the same load, the same two stages are repeated
support by support in the order in which the gaps close.