BEAM ELEMENT
CHAPTER 816
8 BEAM ELEMENT
In the analysis of beam deformations, two different types of behaviour need to be
distinguished. For slender beams we negleet shea effects and assume thatthe plane which
isinitialy normal othe neutral ais remains normal o this axis after deformation has taken
place. This assumption islustrated in Figure 8.
Figure 8.1: Beam with no shear deformation
In this discussion v denotes the vertical movement of ofthe beam (positive upwards) and
8 denotes the rotation of the normal to the neutral axis (positive anticlockwise about
yo).
For the case of thick beams, shear deformations cause warping across the beam section,
‘To model this analytically we still assume tha plane sections remain plane during losding,
‘but allow the plane normal to the neutral axis to rotate as shown in Figure 82.
neutral axis
Figure 82: Beam with shear deformationn
‘The rotation ¢ defines a constant shearing strain through the depth ofthe beam, Since the
actual shearing stress and strain are not constant (they vary parabolically throughout a
rectangular cross-section), ¢ is more precisely defined as an equivalent shearing strain
acting on ¢ shear area 4y. ‘Thus we can write
1)
‘where ris the shear stress, G = £/2(1 + »)is the shear modulus, Q isthe shear force, A
is the cross-ectional area, A, is the effective shear ares, and k= A,/A ~ 5/6 for
‘rectangular sections. From elementary beam theory, the moment at any section is given
by
8,
m= ard 62)
‘where isthe moment of inertia and £/is the flexural stiffness. Equations (8.1) and (8.2)
define the “strese-sirain” relations for the beam and may be combined in the matrix forma
{a [> cal a
‘This can also be written inthe more general form
onDe (84)
‘where the generalised “strain” and “stress” vectors are given by
65)
66)
HO
Plo Gak en
Note that in this cas, the stress vector is actually comprised of stress resultants (the
‘moment and shear force) whilst the strain vectoriscomprised ofa curvature andan average
shear strain. For this reason these vectors are, respectively, often referred to as the
generalised stress vector and the generalised strain vector.
(Se (oe)
Hl ][o5
postive shear force negative shear force
and the “stress-strain” matrix is
Figure 83: Sign conventions for moment and shear force%
‘The sign conventions forthe moments and shear forces are shown in Figure 83.
8.1 A Linear Beam Element with Shear Deformation
‘To derive the stiffness equations for a general beam element, ve allow forthe effect of
shear deformations and use the principle of virtual work. Consider the two-noded linear
‘beam clement shown in Figure 8.4 where each node has two degrees of freedom, a vertical
isplacement v (positive upwards) anda rotation 6 (postive anticlockwise)
yy
Figure 8.4: Two-noded beam with shear deformation
‘The linear shape functions sre expressed in terms of the model coordinate £ according to
N= -E-1)
1
m- de+n
Once their nodal values are known, the displacements and rotations at any point inside a
‘beam element can be found from
Nyy + Non = [Mi afr} 68)
A we} 69)
‘These equations can be expressed more compactly in the matrix form of|
Y) fm ox, oe
lof “] ON, 0 NA] y¥2p = NO (810)
e:
= NO; + N30y
where
N, ON, 0
N=low, ow,
fr % v2 &} (au)Bp
In order to compute the sifiness matrix for the beam element itis necessary to derive the
strain-displacement matrix B. This matrix relates the strain vector at any point inside the
‘beam, as defined by equation (8.5), to the nodal displacement vector of equation (8.11),
Differentiating (8), we obtain
(0;
It} ee
Differentiating (8.8) and using (8.9), this quantity may also be expressed as
o-[ f} -[K ofr as)
Aisplacement relations as
ay, ("4
“ |I0,
Ny lg
ce le,
Bu eu)
Where the strain vector eis defined in (8.5), the element displacement vector wis defined
in (8.11) and the strain-displacement matrix Bis
an, aN
Cao =e
Bel, (15)
ae
For an isoparametric formulation, the shape functions that define the displacements are
also used to define the coordinates. Thus thox-coordinate at any point within the clement
can be written in similar form according to
TEN, +My, = [My wf} 616)
Differentiating (8.16) with respect to the model coordinate & gives the determinant of the
Jacobian as80
‘To compute the B matrix, we require the partial derivatives of the shape functions with
‘espectto the cartesian coordinates andy, These derivatives cannotbe evaluated directly
8s the shape functions are written in terms ofthe model coordinates and 7, s0 we invoke
the chain rule according
‘The partial derivatives of the shape functions with respect to the model coordinates,
together with the det J, enable the required values to be found.
8.2 Formulation of Linear Beam Element Using Virtual Work
For the beam clement, the virtual work equation may be written as
4 (elf +560) a= [ances my +P C0, +6380, (18
Where A denotes the cross-sectional are, 8 2is the vistwal curvature, Mis the moment,
8 he vital equivalent ser sai, Q's the sea oc, sth ital verte
aiplacenent te dstrutedou por ot eng (ry) ae he ial verte
dipdements athe nodes, (P,P, he vera! pon loud applied athe nde
(28; .68,) are the virtual rotationsat thenodes, and (C, ,;) are the couples applied a the
nodes.
a ee
affsteu [ars aw
where
for 40, 40 604)
Now, from (810) and (8.4) it follows that
eat
outNtaL
oe = ouTBT
‘Substituting these equations in (8.19), together withthe stress-strain relations (8.4), and
rearranging gives,
ota womae sare = af NTeaerae eo a0
Ly
Since the variation Buin the nodal displacements is arbitrary, and (8.20) must hol forall
such variations it follows that
a
‘This expression defines the clement sifiness relations
winced w= f Ntecouae +P
ka=t
where the element stiffness matrix may be identified as
ke 4f BtDp det 34f (62
and the element force vertor is
[ wrccuae oe 2
The nodal displacement vector for the beam element is given by (811).
8.3 Numerical Integration of Element Stiffness Matrix
‘To derive the stiffness equations for a numericallyintegrated element, we substitute 8.15),
(827) and (8.17) into equation (8.21) and apply numerical integration. This gives
k= AS BIDS, det Jp, ey)
‘where nis the number of integration points, w; the value ofthe weigh for integration
Point i, B, = B(,) is the strain-isplacement matrix evaluated at integration point i,
et J, = 4etJ,) is the determinant of the Jacobian evaluated at integration point i, end
is the value ofthe model coordinate for integration point
8.4 Formulation of Element Force Veetor
In beam analysis, the loads may be applied ina distributed manner or as concentrated
forces/couples atthe nodes. These types of loading are shown, respectively, in Figure 85
and Figure 86. Note thatthe positive direction ofthese actions is as shown. Since the
‘wo-noded beam element uses linear shape functions to model the displacements and
rotations, its posible to compute consistent nodal forces fr a lineal varying distributed2
lod. For problems involving more complicated distributed load, it would be necessary
to.se an element of higher order
yy a)
Figure 86: Beam clement with forces and couples applied at nodes
In the most general case the element force vector comprises both forces and moments. If
the mesh ofbeam elements islozded onlybypoint forces and couples at the nodes, as show
in Figure 8.6, then the global force vector can be assembled ditecly (inthe same way a3
for the truss element discussed in Chapter 7). For example, consider the mesh of two linear
beam elements shown in Figure 87.
Py
a Ne,
aa ae
® ® ®
Figure 8.7: Mesh of beam elements with forces and couples applied at nodes
Ignoring boundary con:
(624)
For the case of distributed loading, the first term on the let hand side of (622) may be
integrated numerically using the Gauss rule to give the element force vector as
DyNT det (625)where isthe numberof integration points, wis the value of the weight fr integration
pointi, N, = NG) isthe shape funtion matrix evaluated a integration point isthe
ttaction vector evaluated atintegraton point det J, = detJ(,)'s the determinant ofthe
Jacobian evaluated at integration point and §, the value ofthe model coordinate for
integration point: Once each element force vector i formed, itis then inserted into the
slobal force vector using the element tering Veto.
8.5 Worked Example of Numerically Integrated Beam Element
To illustrate the steps involved in forming the stiffness relations by numerical
integration, 2 worked example is presented for the twomnoded beam element. In
performing the numerical integration by hand, we first note that
aNidé _ ad (1 rs
he - 8-4 ag
dt
Ba 00+ 9)e- 3 ago t
Sain (826)
Using (8.15) ths gives the strain-displacement matrix as
4 1
a | ae ma
1 1
Hi -, by
THE all 2 j thy ow
-n,|
‘To choose an appropriate rule for evaluating (8.28, itis necessary to determine the
highest order polynomial that occurs inside the summation sign. Inthe ese ofthe linear
clement, the B matrix comprises some linear terms and some constant terms and thus
corresponds toa polynomial of degree one. Provided the beam element isnot curved, the
Jacobian matrix, and hence its determinant, is always constant. Thus we see that the highest
4egrec term in the element stiffness matrix is equal to the product of terms which are of
‘degree one. The highest order term inside the summation fork is therefore of degree two
and itis necessary to use a two-point Gauss rule.
For
, equation (8.28) becomesry
TH Oph e
x
-eme
where, from Table 41, the coordinates and weights are
Ae UB w=
ey wat
Substituting
ME) = M0) 3)
MED) = NCU)
NG) = NAGI) =
Nab) = NCA)
and the integration point weights gives
Gak
As expected, the element stiffness matrix is symmetric,
For the element force vector, consider the case shown in Figure 88 where the beam
clement i subject to linearly varying distributed load defined by
9 = Na, + Node (830)Figure 88: Beam clement with linear distributed loading
Using equation (8.2) the element force vectors given by
[MGd 0
Shey fee") ae
0 NG)
PED)
0
Inseeg @@2] * % xm,
°
“8
rs
Dd |wstep [Ga + Mae d93} x E xm,
aM
sG0(uEda + NAED9R)
o
2 WENGE Dan + NaEi)aa)
o
Inside the summation sign, the highest order terms are products of the shape functions.
Since these functions are of degree one, thisimplies thatthe highest order term is quadratic
‘and a two-point rue is again sufficient. Using the two-point rule scheme in Table 4.1 we
obtain
fegp(mche + wide)
°
t=
dy (x Avg, + nxAy
Nat (side weds)
0
xke
ie8%
(Ay weg, + nc4
IN (me Bn + NL +p)
0
xh
)
+
Which simplifies to
es + do
a
5lhe +e
5
tek
‘As an immite check on this lement force vector we nts hat
Sonoda foves = 4(qy + 43) = toa load apple to beam
‘Asaspecal cat, the lament force vector fora unifrm dtbuted load cane obtained
tysuisinaingg, = 42 = atone
ft
alo
toh