Thin-Walled Structures 49 (2011) 1341–1347

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Shear flow in thin-walled cellular sections

Chai H. Yoo a,n, Junsuk Kang a, Kyungsik Kim b, Kyoung C. Lee a
a
Department of Civil Engineering, Auburn University, Auburn, AL 36849-5337, USA
b
Department of Civil and Environmental Engineering, Cheongju University, Cheongju, Chungbuk 360-764, Korea

a r t i c l e i n f o abstract

Article history: Frequently flexural members are made of single or multiple cellular sections due to a variety of reasons.
Received 12 January 2011 Clearly, the superb torsional capacity of such sections is a major attraction for girders in horizontally
Accepted 23 February 2011 curved alignments. However, the ease of maintenance and esthetic considerations favor the choice of
Available online 16 March 2011
such girders.
Keywords: The determination of the shear stress at any point along the boundary of an open section can be
Analytical solution made by the evaluation of the shear flow integration from a free edge or zero surface to any surface. In
Bending the instance of a closed cell, there is no free surface and thus the problem becomes indeterminate and
Cellular section the integration cannot be evaluated to determine the shear flow.
Closed section
Although a general methodology to overcome this difficulty is available in the literature, a simple
Finite element analysis
digital computer program has been created to evaluate the bending shear flow of any multiply-
Indeterminate problem
Shear flow connected cellular sections. Several examples are solved and showed the unreasonableness of the
Torsion frequently used approximate solutions.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction Although the basic shear stress formula remains unchanged

for solid sections such as a rectangle or a circle and thin-walled
Flexural shear stresses are caused by the change in bending sections, the way to compute the first moment of area or the
stresses along the beam length. The distribution of shear stresses statical area moment must be in such a way to reflect the
may be computed by simple beam theory based on strength of condition of zero shear stresses at free boundaries such as those
materials, theory of elasticity, or by finite element analysis. inner regions of a wide flange. The execution of the equation for
Although Saint-Venant [1] studied bending of a prismatic canti- the statical area moment is straightforward for an open cross
lever by a force applied at the free end and parallel to one of the section. If the cross section contains one or more closed cells, the
principal axes of the cross section as early as in 1856 based on his point of zero shear flow, or the point of zero statical area moment
semi-inverse method, and Timoshenko and Goodier [2] intro- as the starting point of the integration is not known. In fact, the
duced the procedure for a selected few simple cross sections, the determination of the statical area moment becomes a statically
determination of the shear stresses remains complicated for a indeterminate problem for a cellular section as stated by Galam-
general cross section simply because of the difficulty in selecting bos [4] and one must invoke the condition of the compatibility of
an exact stress function that satisfies both the equations of the shear deformations. This is where the difference in the
equilibrium and compatibility of the theory of elasticity. determination of the statical area moment for open and closed
Timoshenko [3] shows that the shear stress can be assumed to cross sections is not properly treated in many elementary text-
be uniform across the thickness of a thin plate. A great many books on strength of materials including those by Hibbeler [5]
practical rolled metal sections, such as the wide flange, the angle, and Singer and Pytel [6] and is the impetus for this paper.
or the channel can be considered as thin-walled sections [4]. Many A systematic procedure to overcome the nature of indeterminacy
glued or nailed composite lumber sections can also be treated as presented for a multi-cellular cross section is presented. Several
thin-walled sections. A thin-walled section is defined by a cross example problems are solved by the presented procedure and
section having thin-walled elements, that is, the wall thickness is compared with less rigorous approximate solutions to show the
small compared with the height or width of the section. unconservative nature of the approximate solutions. The analytical
values from the present method are verified by a finite element
analysis program, ABAQUS [7]. Very good agreements are evident
n
Corresponding author. Tel.: þ1 334 844 6279; fax: þ 1 334 844 6290. between the analytical results from the present method and
E-mail address: chyoo@auburn.edu (C.H. Yoo). corresponding values obtained from the finite element analysis.

0263-8231/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2011.02.014
1342 C.H. Yoo et al. / Thin-Walled Structures 49 (2011) 1341–1347

2. Backgrounds For point B0 , tB0 ¼10 in (254 mm) and A0 is the shaded area
shown in Fig. 1. Thus
It is well known that the shear formula given by Eq. (1) is
QB0 ¼ y0 A0 ¼ ð12:1=20:575=2Þð10Þð0:575Þ ¼ 33:13 in3 ð542:9  103 mm3 Þ
completely general, i.e., it can be derived from the equilibrium
condition (the sum of the horizontal forces on an infinitesimal so that
free body isolated from a solid member or a thin-walled member
VQB0 60ð33:13Þ
must be equal to zero). However, the manner in which the tB0 ¼ ¼ ¼ 0:47 ksið3:24 MN=m2 Þ
ItB0 420ð10Þ
infinitesimal is being isolated should not violate the fundamental
condition in the theory of elasticity, i.e., the shear stress must be For point B, tB ¼0.345 in (8.76 mm) and QB ¼QB0 so that
equal to zero at the free boundaries. The shear stress formula is VQB 60ð33:13Þ
tB ¼ ¼ ¼ 13:72 ksið94:6 MN=m2 Þ
V Q ItB 420ð0:345Þ
t¼ y x ð1aÞ
Ix t This is where the result of using the approximate method is so
Or the expression for the shear flow appalling. As points B0 and B are so close, it is not expected that
Z there exist such drastic differences in the shear stresses.
Vy Qx Vy s
qo ¼ tt ¼ ¼ ytds ð1bÞ For point C, tC ¼0.345 in (8.76 mm) and QC ¼QB þ DQC. Thus
Ix Ix 0
3
In the case of a solid section, say, a rectangle, the integral path QC ¼ 33:13 þ 0:345ð12:1=20:575Þ2 =2 ¼ 38:30 in ð627:6  103 mm3 Þ
of the static area moment, Qx in Eq. (1b) commences from the top so that
or the bottom face of the cross section and proceeds to the
VQC 60ð38:30Þ
direction increasing the Qx value. However, in the case of a wide tC ¼ ¼ ¼ 15:86 ksið109:3 MN=m2 Þ
ItC 420ð0:345Þ
flange, the integral path may begin from any one of the four
flange tips where the shear stress is equal to zero as it is defined It would be interesting to compute the shear force carried by
to be the stress free surface. The shear flow given by Eq. (1b) is the web. The shear stress at the arbitrary location y from the
analogous to the incompressible liquid flow, in that the sum of neutral axis needs to be determined
the shear flow entered into a node of a section must be equal to
Ay ¼ 10ð0:575Þ þ0:345ð12:1=20:575yÞ ¼ 5:75 þ0:345ð5:475yÞ
the sum of the shear flow discharged from the same node.
Although Hibbeler [5] correctly points out that the shear formula, Qy ¼ QB þ DQy ¼ 33:13 þ0:345ð5:475yÞ½y þ ð5:475yÞ=2
Eq. (1a), should not be used to determine the shear stress at points
¼ 38:300:1725y2
of sudden cross-sectional changes, it is used in many elementary
textbooks on strength of materials including those by Hibbeler [5] so that
and Singer and Pytel [6] perhaps under the feeble justification of VQy 60ð38:300:1725y2 Þ
being simple or approximate. However, such a practice may lead to ty ¼ ¼ ¼ ð15:860:0714286y2 Þ ksi
Ity 420ð0:345Þ
many unconservative results. Consider W12  53 illustrating the
approximate shear stress computation using Eq. (1a) and the shear Z Z 5:475
stress based on the shear flow theory in thin-walled elements, Vw ¼ tdA ¼ 2 ð15:860:0714286y2 Þð0:345Þdy
Aw 0
Eq. (1b). The section is subjected to a shear of V¼ 60 kips (266.9 kN). ¼ 57:22 kips ð254:5kNÞ o60 kipsð266:9 kNÞ
The dimensions in fractions are given in Fig. 1.
This implies that the top and bottom flanges carried the
2.1. Approximate method remaining shearing force of 2.78 kips (12.37 kN). According to
the theory of shear flow in thin-walled elements, the net shearing
The shear-stress distribution will be parabolic throughout the forces carried by the flanges are equal to zero.
depth of the section and varies in the manner shown in Fig. 1.
Taking advantage of the section being symmetric, the shear 2.2. Shear flow in thin-walled elements
stresses at points B0 , B, and C need to be computed. Neglecting
the fillets at various edges and corners, the moment of inertia of Fig. 2 shows the shear flow, tt or the statical area moment, Q
the section about the neutral axis is computed to be diagram as per Eq. (1b). The maximum statical area moment at
1 the flange and web juncture is
I¼ ð0:345Þð12:12  0:575Þ3
12  Qf max ¼ tf ðbf =2Þðdtf Þ=2 ¼ 0:575ð5Þð12:10:575Þ=2
1
þ2 ð10Þð0:575Þ3 þ 10  0:575  ð12:1=20:575=2Þ2 ¼ 420 in4 ¼ 16:567 in3 ð271:5  103 mm3 Þ
12

ð174:8  106 mm4 Þ:

 f max = 6.86 ksi Q = 16.57 in3
f max

0.575" B′ B′ = 0.47 ksi

2Q = 33.13 in3
f max
B = 13.72 ksi
B
C
12.1" C = 15.86 ksi Q = 38.86 in3
wmax

0.345"  wmax = 15.86 ksi

10"
Fig. 1. Dimensions of W12  53 and shear stresses (notes: 1 in ¼25.4 mm; Fig. 2. Statical area moment and shear stress in thin-walled elements (notes:
1 ksi ¼6.895 MN/m2). 1in3 ¼ 16:387  103 mm3 ; 1 ksi ¼ 6:895 MN=m2 ).
 C.H. Yoo et al. / Thin-Walled Structures 49 (2011) 1341–1347 1343

Qwmax ¼ 2Qf max þ tw ðdtf Þ2 =8 ¼ 33:134 þ0:345ð12:10:575Þ2 =8 The application of Eq. (6) will be demonstrated below follow-
¼ 38:86 in3 ð636:8  103 mm3 Þ ing the general lines presented by Heins [8] for a single cell and
multi-cell section. Kollbrunner and Basler [9] give credit to
VQB 60ð33:134Þ Goodey who applied the procedure outlined in Eqs. (2)–(6) for
tB ¼ ¼ ¼ 13:53 ksi ð93:3 MN=m2 Þ
Itw 426  0:345 the first time in 1936. The procedure is equally applicable to
determining the torsional shear flow of multi-cellular sections.
VQC 60ð38:86Þ
tC ¼ ¼ ¼ 15:86 ksi ð109:4 MN=m2 Þ
Itw 426ð0:345Þ
3.1. Single cell
Vw ¼ ½13:53 þ 2ð15:8613:53Þ=3ð0:345Þð12:10:575Þ
¼ 59:97 kipsð266:7kNÞ 660 kipsð266:7kNÞ
A single cell shown in Fig. 3(a) is cut at the lower right corner,
It should be noted that the moment of inertia of the cross thus creating an open section. A separate node number is
section used in the above computations is determined from the assigned to distinguish two separate free edges of an open
idealization of the cross section as the assembly of thin-walled section. Eq. (1b) can now be applied to the open section thus
elements in order to maintain consistency. As expected, the web created. The resisting shear flow, qo is computed as shown in
carries 100% of the vertical shear in the theory of shear flow in Table 1 and plotted in Fig. 3(b). Note that the resulting shear flow
thin-walled elements. at node 7 is equal to zero as node 7 is a free edge. In order to close
the artificial cut, Eq. (2) or (6) must be applied. In applying Eq. (6),
it should be noted that qk ¼0 as there is only one cell. Therefore,
3. Multi-cellular sections Eq. (6) becomes
I I
The computation of shear flows and shearing stresses on thin- qo ds
dsþ qr ¼0 ð7Þ
walled open sections has been illustrated in the previous section i Gt i Gt
as per Eq. (1b). Eq. (1b) is applicable for thin-walled closed cross
sections. This equation permits computation of the shear stress at
any point along the perimeter coordinate, s, since the integration
commences from a free edge or zero surface to any surface of the
e dbt
cross section. Any point on the perimeter coordinate can be
4 3 2
reached from a free edge in any open sections. In the case of a
closed cell, however, there is no free edge and thus the integration t
cannot begin as indicated by Galambos [4] and Heins [8]. There- dt dbt d 2t
d x 2
fore, an imaginary cut needs to be made in order to use Eq. (1b) at 5 V 8 2 8
a location on the cell. In the case of a multi-cells, as many cuts cut
need to be made as to make any point on the cross section can be 6 1 dbt
reached by an integration path from a free edge. This scheme is 7
y 2
applicable for cells enclosed by another cells as those can be
found in a multi-cell bridge girder with the orthotropic deck b
stiffened by U or trapezoidal stiffeners. When the cross section is
dbt
cut, the shear flows become zero at the cut which is generally not
4
the case in the real structure. Therefore, an indeterminate shear
flow must be introduced at the cut to close the gap and restore
dt2 dbt d 2t dbt
compatibility that exists in the real structure with closed cross 4
8 8 4
sections. Hence, the required compatibility condition is that the
shear deformation, summed around the perimeter coordinate of qr
dbt
each cell, must be equal to zero 4
I
gds ¼ 0 i ¼ 1,2,3,    ,n cells ð2Þ Fig. 3. Bending shear flow in a single cell: (a) node numbering and the fictitious
i cut, (b) shear flow in the open cell,  V/I, (c) redundant shear flow, and (d) final
shear flow,  V/I.
The shear strain in Eq. (2) must be the one in the real structure
t q
g¼ ¼ ð3Þ
G Gt
Table 1
where G is shear modulus of elasticity and Shear flow in open cell.
q ¼ qo þ qr ð4Þ Rs
Node y tds Q¼ 0 ytds qo ¼ (VQ/I)
where
qo is the shear flow of cut sections and qr is the indeterminate 1 0 0
or redundant shear flow. 1–2 þ d/4 dt/2 þd2t/8
2 þd2t/8(V/I)
Substituting Eq. (3) into Eq. (2) gives 2–3  d/4 dt/2  d2t/8
I
q 3 0
ds ¼ 0 ð5Þ 3–4  d/2 bt  dbt/2
i Gt
4  dbt/2(V/I)
Expanding Eq. (5) for each cell j yields 4–5  d/4 dt/2  d2t/8
I I X Z 1 5  (dbt/2 þd2t/8)(V/I)
qo 1
ds þqr ds qk ds ¼ 0 ð6Þ 5–6 þ d/4 dt/2 þd2t/8
j Gt j Gt k jk Gt 6  dbt/2(V/I)
6–7 þ d/2 bt þdbt/2
where j is the jth cell; k is the cell(s) adjacent to cell j; and qk is 7 0 0
the shear flow in kth wall.
1344 C.H. Yoo et al. / Thin-Walled Structures 49 (2011) 1341–1347

The qr diagram is shown in Fig. 3(c). Table 2 gives the result of in Fig. 3(b) and qr ¼(V/I)(dbt/4) shown in Fig. 3(c). The final shear
the integration of qo and qr diagrams. The summation of these flows thus obtained are shown in Fig. 3(d).
quantities according to Eq. (7) gives It is interesting to note that the final shear flow diagram can be
used to locate the shear center. The force along the right side of
b2 dV bd2 V 2qr
  þ ðbþ dÞ ¼ 0 the cell is
2I 2I t
   
Solving for the redundant shear flow qr yields dbt V 2 d2 t V
dþ d
4 I 3 8 I
bdtV
qr ¼
4I which also equals to the force along the left side. The resultant
forces on the top and bottom are equal to zero. Summing
With the magnitude of the redundant shear flow known, the
moments about node 6 yields
final shear flow diagram can be determined by adding qo shown
 2 
d bt V d3 t V
Veþ þ b¼0
Table 2 4 I 8 I
Sum of shear flow in single cell.
 
H H V d2 bt d
Node ðqo =tÞds qr ðds=tÞ Ve ¼ bþ
I 4 3
 
1–3 2d d2 V qrd/t
3 8I  Recall that the moment of inertia of the cell is
3–4 b dbV qrb/t  2  

2 2I   
d t d
4–6 dbV 2d d2 V qrd/t
bþ
d  2 3
 2I  3 8I
6–7 b dbV qrb/t

2 2I
Therefore, e¼b/2, that is the shear center is located at the
centroid.

2a 3a 4a t
5 4 11 15
t t/2 t/2 t

2a 6 (1) 3 (2) 10 (3) 14

x

7 1 2 8 y 9 12 13
t
4
3
2

0.5 2.25 3.25 4.5

2
3 4

q1 q2 q3

1.95
0.05 1.11

1.95
1.89 1.40

2.45 2.19 2.76 3.10

Fig. 4. Bending shear flow in a multi-cell section: (a) node numbering, fictitious cuts, and integration paths, (b) shear flow in the open cell,  V/I, (c) redundant shear flow,
and (d) final shear flow,  V/I.
 C.H. Yoo et al. / Thin-Walled Structures 49 (2011) 1341–1347 1345

3.2. Multi-cell Table 6

Sum of shear flow in cell 1.

The induced bending shear flow in a multi-cell, shown in Node

H qo H ds R 4 ds
t ds q1 t q2
Fig. 4(a) will be illustrated. The dimensions of three cells, node 8 t

numbering, the location of the fictitious cuts, and the integration 1–2 3
2a (V/I) þ2q1a/t 0
paths following the right-hand coordinate system are shown 2–4 8(2a3/3)(V/I) þ4q1a/t  4q2a/t
in Fig. 4(a). The shear flows of the open cells are evaluated using 4–5  2a3/3(V/I) þ2q1a/t 0
5–7  23 a3 VI þ 2qt1 a 0
Eq. (1b) and tabulated in Tables 3–5 and plotted in Fig. 4(b). In
order to restore the compatibility at each cut, indeterminate
(redundant) shear flows are introduced as given in Fig. 4(c). The
integrations of Eq. (6) are tabulated in Tables 6–8. The sum of Table 7
integration results for each cell as per Eq. (6) is Sum of shear flow in cell 2.
H qo H R 11 ds R 2 ds
Node ds q2 ds
q3 q1
V
3 a a t t 12 t 4 t
12a þ 10 q1 4 q2 ¼ 0 ðaÞ
I t t 8–9 4.5a3(V/I) þ3q2a/t 0 0
9–11 12(2a3/3)(V/I) þ6q2a/t  4q3a/t 0
V a a a 11–4 4.5a3(V/I) þ3q2a/t 0 0
13a3 4 q1 þ 14 q2 4 q3 ¼ 0 ðbÞ 4–2  8(2/3)a3(V/I) þ3q2a/t 0  4q1a/t
I t t t

V a a
12a3 4 q2 þ 14 q3 ¼ 0 ðcÞ Table 8
I t t
Sum of shear flow in cell 3.
H qo H R9
Node t ds q3 ds
q2 ds
t 11 t
Tablee 3
Shear flow in open cell 1.
12–13 8a3(V/I) þ4q3a/t 0
Rs 13–15 8(2a3/3)(V/I) þ2q3a/t  4q2a/t
Node y tds Q¼ 0ytds qo ¼ (VQ/I)
15–11 8a3(V/I) þ4q3a/t 0
11–9  12(2/3)a3(V/I) þ2q3a/t 0
1 0 0
1–2 þa 2at þ 2a2t
2 þ 2a2t þ 2a2t(V/I)
2–3 þ a/2 at/2 þ (a2t/4) Solving Eqs. (a), (b), and (c) simultaneously gives
3 2.25a2t þ 2.25a2t(V/I)
3–4  a/2 at/2  (a2t/4) q1 ¼ 1:9543a2 tðV=IÞ
4 2a2t 2
þ 2a t(V/I)
4–5 a 2at  2a2t q2 ¼ 1:8858a2 tðV=IÞ
5 0 0
5–6  a/2 at  (a2t/2) q3 ¼ 1:3959a2 tðV=IÞ
6  (a2t/2)  a2t/2(V/I)
6–7 þ a/2 at þ (a2t/2) The values are all negative, thus the assumed directions of
7 0 0 redundant shear flows shown in Fig. 4(c) are incorrect. The
summation of shear flows given in Figs. 4(b) and (c) determines
the final shear flows shown in Fig. 4(d). The final shear flows
Table 4
in Fig. 4(d) can be used to determine the shear center of the three-
Shear flow in open cell 2.
cell section as was done for a single cell.
Rs
Node y tds Q¼ 0 ytds qo ¼ (VQ/I)

8 0 0 4. Computer programs
8–9 þa 3at þ 3a2t
9 þ 3a2t þ 3a2t(V/I) The procedure to determine the bending shear flows in multi-
9–10 þ (a/2) at/2 þ a2t/4
cell cross sections has been shown. Although the procedure is
10 3.25a2t þ 3.25a2t(V/I)
10–11  a/2 at/2  a2t/4 straightforward, the arithmetic operations can be cumbersome
11 3a2t þ 3a2t(V/I) and tedious, particularly for cross sections having many closed
11–4 a 3at  3a2t cells with varying thin walls along the perimeter. Therefore, it is
4 0 0
desirable to carry out the arithmetic by digital computations.
A software program to accomplish this task is preferred to a
spreadsheet application as it requires coding on each problem.
Table 5
Shear flow in open cell 3. Yoo and Acra [10] developed a computer program, SECP, to
compute cross-sectional properties of composite open and closed
Rs
Node y tds Q¼ 0 ytds qo ¼ (VQ/I) thin-walled cross sections including bending and torsional shear
flows. Coordinates of nodes, member incidences, and cell wall
12 0 0
thickness and orientation are needed as input for the program.
12–13 þa 4at þ4a2t
13 þ4a2t þ 4a2t(V/I)
13–14 þa/2 at þa2t/2
14 4.5a2t þ 4.5a2t(V/I) 5. Maximum shear flows
14–15  a/2 at  a2t/2
15 4a2t þ 4a2t(V/I) Consider the rectangular tube shown in Fig. 3(a). A symmetric
15–11 a 4at  4a2t
11 0 0
structure will resist the symmetrically applied load in a sym-
metric manner. Therefore, the shear flow at the mid-height of the
1346 C.H. Yoo et al. / Thin-Walled Structures 49 (2011) 1341–1347

two vertical walls is the same as shown in Fig. 3(d). Many flanges with relatively shallow webs similar to that shown
elementary textbooks on strength of materials including those in Fig. 5(b). In all cases examined, the approximate solution is
by Hibbeler [5] and Singer and Pytel [6] adopt this rule in that the always in the unconservative side.
only way to have the identical shear flow at the mid-height of Three-dimensional finite element analyses by ABAQUS [7]
the walls is to have a zero shear flow at the center of the top or yield solutions somewhere in between those two values obtained
the bottom flange so that the result of the integration of Eq. (1b) from the approximate method and the shear flow method given
will give the same value at the mid-height of the two walls. This by Eq. (6). The models run are two cantilever beams of the length
logic, however, will encounter a dilemma in the case when the 20 times the wall depth having a cross section shown in Fig. 5. In
section is not symmetrical, i.e., when the thickness of one wall is order to overcome the inherent numerical limitations involved in
different from that of the other wall. The procedure presented in the finite element analysis, each wall is subdivided into 20
Eq. (6) can be applied to with any unsymmetrical multi-cell cross ABAQUS S4 flat shell elements and the aspect ratio of the element
sections. is kept equal to one. The values reported in Table 9 are computed
Essentially the same classical method as applied to computing from the shear stresses at the midspan of the cantilever beams in
the statical area moment, Q, of solid sections is applied to multi- order to avoid local disturbances. Thus, the accuracy and relia-
cell sections in many elementary textbooks on strength of bility of the finite element results are considered to be sufficiently
materials including those by Hibbeler [5] and Singer and Pytel [6]. accurate. It is noted that both the approximate method and the
Consider, for example, two two-cell sections shown in Fig. 5. The shear flow method are of the strength of materials type analyses
dimensions of these two examples were adopted from textbooks whereas the finite element analysis of ABAQUS is a three-dimen-
referenced without modifications for consistent comparisons. sional evaluation without introducing any simplifying assump-
Table 9 shows the maximum statical area moment of the two tions. The effects of the cross-sectional distortions and Poisson’s
two-cell sections comparatively. Singer and Pytel [6] give the ratio appear to alleviate somewhat the shifting of shearing forces
same maximum statical area moment at the mid-height of each to the interior wall in the examples examined. For a Poisson’s
web shown in Fig. 5(a) equal to 11.8 in3 (193.4 mm3). The ratio of 0.15, the ABAQUS outputs approach very close to the
procedure given in Eq. (6) programmed in SECP gives the max- values predicted by the shear flow method.
imum statical area moment at the mid-height of the center web It has been observed that both the shear flow method and the
equal to12.2 in3 (199.9 mm3). Thus, the approximate solution is finite element results give internal resisting shearing force of
about 3.3% unconservative. Hibbeler [5] also gives the same 99.95% and 99.9%, respectively, of the external force applied,
maximum statical area moment at the mid-height of each web while the approximate method gives the internal resisting shear-
shown in Fig. 5(b) equal to 0.444  106 mm3. SECP computes the ing force of 95.4% of the external load carried by the webs equally.
maximum area moment at the mid-height of the center web It can be readily shown that the approximate solution applied to
equal to 0.500  106 mm3. Thus, the approximate solution in this very wide cellular sections will significantly underestimate the
case is about 11.2% unconservative. It has been observed that the maximum shear stress developed in the interior webs.
difference between the approximate solutions and the solutions It is generally perceived that an approximate solution usually
obtained from Eq. (6) is small when the cellular section is narrow yields a conservative value because simplifying conservative
in width with deep webs similar to that shown in Fig. 5(a) while assumptions are introduced in the development of an approx-
the difference is greater in a cellular section that has wider imate solution. However, an unconservative assumption is intro-
duced in the application of the classical method of computing the
statical area moment of solid sections to thin-walled sections. In
the classical method, as pointed out by Hibbeler [5], the interior
faces of flanges in both the open and closed cross sections shown
2 2 1 150 150 15
in Figs. 1 and 5 will have nonzero shear stresses, which is
contradictory to an elementary theory of elasticity, i.e., no shear
stress can develop on the free surface of a body in the absence of
externally applied shearing forces.
Fig. 6 shows the deviation of the maximum statical area
moment of doubly symmetric two-cell sections similar to those
12 280
shown in Fig. 5 computed by the approximate method from that
15
1

32
28
25 25 25
Percent Difference

1/4 1/4 1/4 24

Fig. 5. Two-cell sections: (a) dimensions in inch and (b) dimensions in mm (note: 20
1 in ¼25.4 mm).
16
12
Table 9 8
Maximum statical area moments in cellular sections.
4
Approximate Eq. (6) ABAQUS Dimensions
0
Fig. 5(a) 11.8 12.2 11.9 in3 0 2 4 6 8 10 12
Fig. 5(b) 0.444  106 0.500  106 0.471  106 mm3 Aspect Ratio
Note: 1 in3 ¼ 16:387 mm3 : Fig. 6. Percent difference.
 C.H. Yoo et al. / Thin-Walled Structures 49 (2011) 1341–1347 1347

obtained by Eq. (6) as a function of the aspect ratio of the overall onerous for a multi-cellular wooden box section when the load
cross-sectional dimensions. In order to minimize the data hand- carrying capacity is determined by the shearing strength of the
ling, the thickness of the thin-walled elements was fixed at 0.5 in vertical walls as the shearing capacity parallel to grain of the
(12.7 mm) and other dimensions were varied. As the effect of the sawn lumber is notoriously weak.
unconservative assumption will be greater in a wider section, the It is hoped that the material presented in this paper may
deviation of the maximum statical area moment computed from alert casual practices of applying a classical method of computing
the approximate method is expected to be larger in the cross the statical area moment and thereby avoiding unexpected
section with a high aspect ratio as shown in Fig. 6. Experience has consequences.
shown that the aspect ratio of the cross section is the dominant The application of procedures illustrated in Eq. (6) can be very
parameter but variables such as the thicknesses of flanges and time consuming and cumbersome, particularly for a cross section
walls and their combination appear to affect the percent deviation with a large number of cells. For such situations, the computer
of the approximate method a little. Therefore, Fig. 6 should be program, SECP can be readily utilized.
viewed as a trend rather than an exact difference between the
two methods.
References

6. Summary and concluding remarks [1] Saint-Venant Bd. Memoir on bending. J Math Liouville 1856;1:89–189.
[2] Timoshenko SP, Goodier JN. Theory of elasticity. 2nd ed. New York: McGraw
Hill, Inc.; 1951.
Shearing stresses in multi-cellular cross sections computed [3] Timoshenko SP. Theory of bending, torsion, and buckling of thin-walled
based on statical area moment determined by the classical members of open cross section. J Franklin Inst 1945;239:3–5.
method that was developed for solid sections can lead to [4] Galambos TV. Structural members and frames. New Jersey: Prentice Hall;
1968.
unexpected consequences. As illustrated in this paper the differ- [5] Hibbeler RC. Mechanics of materials. 6th ed. New Jersey: Prentice Hall; 2005.
ences between the shear stresses computed based on the classical [6] Singer FL, Pytel A. Strength of materials. 3rd ed. New York: Harper & Row,
method and those evaluated by Eq. (6) are perhaps negligibly Inc; 1980.
[7] ABAQUS analysis user’s manual version 6.5. ABAQUS, Inc. Pawtucket, RI;
small when the depth of the cross section is several times greater
2004.
than the width of the cross section. On the other hand, if the [8] Heins CP. Bending and torsional design in structural members. Lexington,
dimensions of the cross sections are reversed such that the width MA: Lexington Books; 1975.
[9] Kolbrunner CF, Basler K. Torsion in structures. An engineering approach.
of the cross section is several times greater than the depth, the
Berlin/Heidelberg: Springer-Verlag; 1969.
classical method may underestimate the induced maximum [10] Yoo CH, Acra SV. Cross sectional properties of thin-walled multi-cellular
shearing stress significantly. The situation can be particularly section. Comput Struct 1986;22(1):53–61.