Abstract PaIIet racks, often fabricated using cold-formed steel, are used for the storage of

goods. Uprights of these racks are braced in the cross-aisle direction forming a frame, which
behaves like a built-up column. Evaluation of the shear stiffness of this frame is needed to
determine the buckling load. Currently two approaches prevail in the rack industry to determine
the shear stiffness. The RMI code uses a theoretical formula and the FEM code requires testing.
There is a considerable difference in the stiffness values determined by two approaches. The
present paper describes experimental and numerical studies conducted at Oxford Brookes
University to evaluate shear stiffness in an ongoing research project.
Introduction Pallet racks, often fabricated using cold-formed steel, improve the storage of
goods by the efficient use of the cubic space available for storage. The uprights (columns) of
these racks are braced in the cross-aisle direction as shown in Fig. 1, forming a frame, which
behaves like a built-up column. In recent years, large pallet racking systems have been used
containing tall and narrow upright frames. Such frames need to be checked for stability in the
cross-aisle direction to prevent any potential collapse, which could lead to loss of human life.
The stability of such frames in the cross-aisle direction i.e. about the z-z axis [see Fig. 2)
depends on
• Overall behaviour of the frame. This is affected by elastic flexural buckling and shear
deformations,
• Local thin walled behaviour of each upright,
• Forces in diagonal bracing members due to eccentric joints and
• The bolt slip due to the bolt-hole clearance, if the connection is bolted

Flexural deformations can be calculated using established theories. However, there is only
limited research available to find the shear deformations, particularly in cold-formed built-up
columns. Accurate evaluation of these shear deformations is needed to determine the elastic
buckling load and sway deflections of upright frames. Hence, research is being undertaken to
predict the actual shear stiffness values of upright frames. The present paper reviews the
published literature and existing design approaches for evaluating shear stiffness values of built-
up columns or pallet rack upright frames. It also describes the experimental and numerical
studies to evaluate shear stiffness of upright frames. A simple frame analysis has been carried
out on upright frames using the finite element analysis (PEA) program, LUSAS. The effects of
various factors such as the flexibility of uprights, eccentric loading on bracing members, the
aspect ratio (panel length/panel depth) of frame panels and bolt bending, on the shear stiffness
of upright frames are studied. The study is limited to upright frames with bolted lacings or
battens, which are common in Europe. In future, more refined FEA models that will be
validated against test specimens can be used to generate additional data to obtain a better
method of evaluating shear stiffness of upright frames.

State of the art Literature review The first theoretical investigation into the affect of shear on
compressive strength of columns was by Engesser (1891), who modified the Euler analysis for
axially loaded columns to account for shear [Galambos 1988]. Timoshenko (1949, 1961) was
first to include the affect of shear in built-up columns. He proposed a study of shear effects in
built-up laced and battened members. In 1952, Bleich extended Engesser's work to the critical
load analysis of built-up columns. Shear effects in battened and laced members were more
recently studied by Lin, Glauser and Johnston (1970). They recommended shear formulae
incorporating the effects of stiffened zones at the ends of built-up members, eccentricities in the
joints, and net span of chord between batten plates. Gjelsvik (1990, 1991) reviewed the different
methods for evaluating shear stiffness of built-up columns.
A review of literature indicates that Timoshenko's theory is widely used for evaluating the shear
stiffness. He assumed that a built-up column has a large number of panels and hence an
equation, derived to account for shear in solid columns, is used for the prediction of critical
loads. Shear deformations are attributed to the deformability of lacing bars or battens. Shear
stiffness equations are derived for different types of laced or battened patterns. A representative
equation for evaluation of shear stiffness, S, (for a single lacing pattern) is shown below:
where, E - Young's modulus of the lacing member
Ad Cross-sectional area of diagonal lacing members
- Angle of inclination of lacing member with the line drawn perpendicular to upright

The above approach was derived for hot-rolled built-up columns and is not directly related to
pallet rack structures. Very few studies have been conducted on cold-formed steel built-up
columns to check its validity. Djafour, Megnounif and Kedral (1999) analysed elastic stability of
cold-formed built-up columns using the finite strip method and found it can predict buckling
loads accurately. Chwan (2001) under the supervision of Beale and Godley carried out shear
tests on pallet rack upright frames at Oxford Brookes University. Dubina et al (2002) carried out
a numerical study on behaviour of built-up columns made of coldformed C-sections connected
with bolted C-stitches, and validated design approach proposed by Rondal and Niazi (1990,
1993) for determining buckling strength of battened cold-formed built-up columns. Their study
was not extended to pallet rack upright frames. Design practice in the rack industry Currently
there are two approaches prevailing in the rack industry to consider the effect of shear. One
approach is to use a theoretical formula based on Timoshenko, and the other is to determine the
shear by testing. The codes considered for the review in the present study are Storage
Equipment Manufacturers' Association (SEMA) code, Rack Manufacturers' Institute (RMI)
code, Federation Europeenne de la Manutention (FEM) code and the Australian code (AS4084-
1993). The SEMA code is the only code where shear in upright frames is not considered. The
RMI and Australian codes recommend a theoretical approach based on Timoshenko (Equation
I). The FEM code adopts testing to evaluate shear stiffness per unit length of the frame
structure. The code requires the test sample to be a frame assembly with a number of bracing
panels. The testing procedure recommends using a minimum of three panels in the case of laced
upright frames and a whole number of panels in the case of battened frames. There is a
considerable difference in the shear stiffness values determined by these two approaches. The
test values based on the FEM code are sometimes 20 times lower than the theoretical values
[Chwan 2001] calculated using the RMI code. A review of the available literature and design
recommendations in various rack industry codes was carried out. This indicates that substantial
research was not carried out on cold-formed built-up columns to arrive at the appropriate design
approach for the evaluation of shear. Hence, the present research was undertaken in which both
experimental and numerical studies were carried out. The experimental program conducted at
Oxford Brookes University for the evaluation of shear is described in the following section.
Experimental program Test specimens Tests were conducted on full sized upright frames.
Uprights (columns) of the frames were open perforated lipped channels with additional bends
and the bracing members were simple lipped channels. In total. 21 tests were performed (Chwan
& the authors) by changing upright size, number of panels in the frame (2.5 or 3 panels), aspect
ratio of the panel (panellengthldepth varying from 1.14 to 3.23) and lacing pattern (channels
back to back or front to front). Typical upright and bracing members are shown in Fig. 3. Note
that all the dimensions mentioned in Fig. 3 are in mm.
In this paper, the upright with a 2.35 mm thickness is called the heavy upright and the upright
with a 1.6 mm thickness is called the light upright. Though series 1 and series 3 uprights are
light uprights, their cross-sectional properties are different, as there are small changes in their
profiles. 1.3 mm thick bracing members were used in series 1 and 2 tests whereas 1.8 rom thick
bracing members were used in series 3 tests. Table 1 shows cross sectional properties of upright
and bracing members that were used for testing. Gy is the distance of the centroid of the upright
from its back face centre line.
Test Arrangement The basic arrangement of the test upright frame can be seen in Fig. 4. The
frame was placed in the horizontal plane between rollers, which coincides with the points of
intersection of the bracing members. The positions of the rollers were adjusted so that the frame
just fits snugly between them with no looseness. The roller condition at the nodes was achieved
by putting two PVC sheets in between upright sections of the frame and the packing of the test
rig. The test layout and arrangement of displacement transducers (LVDTs) are shown
schematically in Fig. 5. One leg of the frame was pinned at one end so that it was prevented
from moving horizontally, as at point A in Fig. 5. The load was applied along the centroid of the
other leg, at point B in Fig. 5.

At point A, the fi'ame was pinned in all three translational displacements. Two displacement
transducers were also placed at A in the direction of the upright to determine any movement of
the support. At point B, a load cell of 6 kN capacity was connected to a jack of 230 kN capacity
and an L VDT was placed there to control the loading. Two LVDTs were placed at point C to
measure the displacement of the loaded upright along its own axis. L VDTs were placed at
bottom and top of the upright base plate. The mean value of the two LVDTs placed at C was
considered for further calculation. Loading was applied gradually using the jack with a load cell
at the rate of 0.1 kN/sec. The readings from the L VDTs were recorded using a data acquisition
system. The maximum load applied in the test was kept low (5 kN) so that, there was no visible
damage on the specimens. After reaching the maximum load, the frames were unloaded to 0.5
kN. The frames were reloaded and unloaded between loads 0.5 kN and 5 kN for 5 to 6 cycles in
each test. This was carried out to avoid any en'or in evaluating shear due to bolt slip. Full-scale
test results After acquiring the data from the data acquisition system, the load applied on upright
was plotted against the corresponding deformation to arrive at a loaddeformation curve. A
typical load-deformation curve for a tested frame is shown in Fig. 6.The slope, k,i, was obtained
by fitting a linear trend line to the cyclic loading applied in the test program omitting the first
cycle. The first cycle results were not considered to avoid errors due to initial settlement of the
joints. Then, the transverse shear stiffness of the frame, S, was calculated by
Deformation in nnn Fig. 6: A typical load-deformation curve The graph shown in Fig. 6 is for an
upright frame with back to back bracing pattern, heavy uprights and centre-to-centre depth of
1050 mm. In this case, the slope of trend line is 3.8369 and hence kti is 3.8369. After getting kli
values from graphs, shear stiffness values can be easily determined using the equation 2. For the
case shown in the graph, the length of the frame (l) was 3000 mm and the distance between the
centroidal axes of the upright sections (D) was 1050 mm. Hence, the shear stiffness value for
the case is 1413 kN. Similar types of graphs were drawn for all other tests and finally shear
stiffness values were obtained and tabulated in Table. 2
Based on the test results reported in Table 2, it is very clear that test results do not compare with
theoretical values. At higher values of aspect ratios of the panel, the shear stiffness values differ
significantly between theory and tests. The factors affecting the experimental results that are not
considered in the theoretical formula [equation 1] are:
• The eccentricity induced due to bracing pattern, which has a major role to play in shear
stiffness of pallet rack upright frames. The lip-to-lip to bracing pattern has more eccentricity in
the connection and hence has a lower shear stiffness values compared to the back-to-back
bracing pattern. • The cross-sectional properties of the upright, which also contribute to the
shear stiffness of the frame Timoshenko's theory does not consider the above two factors
whereas tests show they have a role to play in shear stiffness of frames. Hence, Timoshenko's
theory should not be used for the evaluation of shear stiffness in upright frames. The affect of
each factor is further studied using the finite element method. Numerical modelliug A linear
analysis was earned out on upright frames using the LUSAS finite element software. For the
purpose of illustration, the 1050 mm deep heavy upright frames that have been tested are
presented here. Both back-to-back and lip-to-lip bracing pattern cases were studied. The elastic
modulus of the steel was taken to be 209000 N/mm2 in accordance with the manufacturer's
specifications. The boundary conditions were kept the same as in the testing Loading was
monotonic and bolt slip was not considered; hence these linear analysis results can be compared
with the test results obtained from the trend line. Initially the frame was modelled as a simple
two-dimensional truss system, wherein both upright and bracing members were modelled using
bar elements (BAR2). Results of this analysis were compared with hand calculation and results
obtained by another frame analysis program, SAND, to check the validity of the FE model.
Later upright frames were modelled as two-dimensional rigid and pin jointed frames. In the case
of a rigid frame, both upright and bracing members were modelled using thin beam elements
(BM3) i.e. shear in beams was neglected. In the pin jointed frame model uprights were analysed
using thin beam elements and bracings with bar elements. In the case of above three models the
joints were concentric. However, in practice, there is eccentricity in connections along with
other factors such as bending in the bolt due to forces coming at the joint, rotational degree of
freedom for the bracing members about bolt axis, etc. The modelling of these parameters is
discussed below. Connection eccentricities The connection detail at a joint in upright frame can
be seen in Fig. 7. In linear analysis of the frame, there are three eccentricities; (i) due to upright
centroidal distance from bolt, a, (ii) due to actual force transfer between bracing and upright, b,
and (iii) due to bracing centroidal distance from load transfer point, c.
Eccentricities due to upright centroidal distance from bolt (a) and due to bracing centroidal
distance from load transfer point (c) are same in both back-to-back and lip-to-lip bracing pattern
cases. However, the eccentricity due to actual force transfer between bracing and upright (b) is
different. It is negligible i.e. half the thickness of bracing member in the case of back-to-back
braced frame and large in case of lip-to-lip braced frame [see Fig. 7). Hence, a significant
reduction in shear stiffness was seen in lip-lo-lip braced frames. Thin beam elements were used
to model all the three eccentricities in the FE model
Bolt bending In a frame with a single layer of bracing members, one bracing member is
sUbjected to tension and the other will be sUbjected to compression at each joint as shown in
Fig. 8. These force components induce bending in the bolt. The effect of this bolt bending in the
FE model was achieved by modifying the stiffness of the thin beam element connecting upright
and bolt. Bending in the bolt will be more predominant in back-to-back braced frames. 308
compression Brace under tension Fig. 8: Forces at a joint Rotational degree of freedom about
bolt axis Bracing members in the uprights were connected using single bolts. Hence, the joint
can be considered as a pin and bracing members are free to rotate about bolt axis. Three-
dimensional joint elements (JSH4) available in the LUSAS were used to arrive at this condition
in the FE model. These joint elements have three rotational and three translational degrees of
freedom wherein stiffness values can be given to achieve the desired condition of connectivity.
In addition, constraint equations can be used to enforce displacement restraints. In the present
study, constraint equations were written for joint elements such that the rotational degree of
freedom about the bolt axis was released and all other rotational and translational displacements
were arrested. In Table 3, the shear stiffness values obtained using numerical analysis of all
these models are reported.
The results obtained for truss. pin-jointed frame and rigid frame in the PE analysis are 15%
lower than the Timoshenko's theory (RMI). It is due to inclusion of axial and flexural stiffness
of uprights in PEA model whereas Timoshenko's theory is independent of stiffness of uprights.
In the models illustrated. the shear stiffness values from the numerical analysis are 2.3 times
higher than the test values for back-to-back braced frame and 4.9 times the test values for lip-to-
lip braced frame. From Table 3, it can also be noted that the effect of connection eccentricities is
significant in the case of lip to lip bracing pattern case and the affect of bolt bending is more
pronounced in back to back braced frames. Rotational release about the bolt axis has more effect
on back-to-back braced frames. The current models do not consider all the effects and further
study has to be carried out to find the significance of joint flexibility.
Conclusions A review of literature indicates that there are two approaches prevailing in the rack
industry to determine the shear stiffness of upright frames. The RMI code uses a formula based
on Timoshenko's theory and the FEM code requires testing. There is a considerable difference in
the stiffness values determined by the two approaches. Hence, research has been undertaken at
Oxford Brookes University. Experimental studies were conducted and primitive FE models
were developed using linear analysis. The effects of various parameters such as connection
eccentricities, bolt bending and rotational release about bolt axis was identified. Further study
needs to be carried out to find the significance of joint flexibility and to propose a better
procedure for the evaluation of the shear stiffness of upright frames.