Journal of Constructional Steel Research 97 (2014) 10–23

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Influence of pallets on the behaviour and design of steel drive-in racks

Benoit P. Gilbert a,⁎, Lip H. Teh b, Romain X. Badet a, Kim J.R. Rasmussen c
a
Griffith School of Engineering, Griffith University, Australia
b
School of Civil, Mining and Environmental Engineering, University of Wollongong, Australia
c
School of Civil Engineering, University of Sydney, Australia

a r t i c l e i n f o a b s t r a c t

Article history: This paper analyses the influence of horizontal bracing restraints provided by the friction between pallet bases
Received 22 September 2013 and rail beams on the static behaviour and design of steel drive-in storage racks. The pallet bracing restraints
Accepted 18 January 2014 are shown to significantly influence the structural behaviour of the rack, and their effect on the bending moment
Available online 20 February 2014
distribution of the uprights is studied in the paper. The 2D single upright model proposed by Godley is improved
in this study by including the restraints provided by the plan flexural stiffness of the rail beams and the friction
Keywords:
Steel drive-in racks
between the pallets and rail beams. The improved 2D model was found to accurately reproduce the bending
Steel storage racks moment distributions obtained using 3D advanced finite element analysis. The 2D single upright model is used
Steel structures to analyse 36 drive-in racks under various load case combinations. The paper evaluates the influence of the pallet
Pallet bracing restraints bracing restraints on the ultimate capacity of drive-in racks, clarifies the loading pattern(s) governing the
structural design and determines the friction coefficient, or strength of a restraining device, required to prevent
the pallets from sliding. It is shown that while restraints from pallets could potentially be considered in design,
they would not lead to more economic structural solutions.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction bracing restraints in the analysis, showed that pallet bracing restraints
had significant influence on the non-sway buckling mode, although
Worldwide, steel storage racks are extensively used in the they had less influence on the sway buckling mode.
manufacturing, wholesale and retail industries to store goods. They However, due to the uncertainty concerning the friction between
are mostly freestanding structures and are often assembled from cold- the pallet bases and the rail beams, drive-in racks are currently designed
formed steel profiles. Two main types of racks prevail, referred to as without considering the bracing effects. If a device can prevent the
“selective racks” and “drive-in racks”. In drive-in racks, pallets are stored pallets from sliding on the rail beams or if the coefficient of friction
on rail beams one after the other, and the forklift truck drives into between the pallet bases and the rail beams can be reliably determined,
the rack to store the pallets on the “first-in last-out” principle. The rail the horizontal bracing effect provided by the pallets could be fully
beams are offset from the centreline of the uprights so that the pallets exploited in the design of a drive-in rack.
apply both bending moments and axial compressive forces to the Hua and Rasmussen [3] measured the friction coefficient between
uprights. To allow the forklift truck passage, the rack is only braced wood pallets and rail beams and found that the average static friction
horizontally at the top (plan bracing) and vertically at the back (spine coefficient between the rail beams and the pallet bases to be as
bracing) in the down-aisle direction. Due to their floor space efficiency, high as 0.576, with a recommended design static friction coefficient of
drive-in racks are usually preferred to selective racks when storing the 0.439. This friction coefficient suggests that significant horizontal forces
same goods with quick turnover, or in expensive storage spaces such can develop between the pallets and the rail beams before sliding
as industrial freezers. Fig. 1 shows an example of a drive-in rack. occurs, allowing the pallets to play a structural role in the behaviour of
Experimental tests performed by Gilbert and Rasmussen [1] have drive-in racks. It is noted, however, that this design static friction coeffi-
shown that pallets act as horizontal braces between adjacent uprights, cient does not take into account grease or ice (in the case of industrial
significantly influence the structural behaviour of drive-in racks and freezers) that may accumulate on rail beams.
must be considered in order to accurately capture the 3D behaviour of Another aspect related to pallet bracing restraints is the in-plane shear
drive-in racks. Similarly, earlier research by Salmon et. al. [2], who stiffness of the pallet base. Hua and Rasmussen [3] experimentally found
numerically investigated the buckling behaviour of symmetrically that the in-plane shear stiffness of pallet bases ranged from 5.1 N/mm to
loaded drive-in racks by alternately considering and ignoring the pallet 31.4 N/mm, depending on the pallet condition. Characteristic design
shear stiffness values of 3.9 N/mm for pallets deemed in poor condition
and 8.3 N/mm for pallets deemed in good condition were recommended.
⁎ Corresponding author at: Griffith School of Engineering, Griffith University, Gold Coast
Campus, Griffith University, QLD 4222, Australia. Tel.: +61 7 555 28577. The current paper analyses the influence of the horizontal bracing
E-mail address: b.gilbert@griffith.edu.au (B.P. Gilbert). effect of pallets on the static behaviour and design of steel drive-in

0143-974X/$ – see front matter © 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jcsr.2014.01.013
 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23 11

racks in the down-aisle direction only, as due to the upright frames,

Notations
pallets are not believed to influence the behaviour of drive-in racks in
the cross-aisle direction. It should also be noted that the friction
Au Cross-sectional area of upright members
between pallet bases and the rail beams would prevent the pallets
Cb Coefficient depending on the BMD in the unbraced
from dropping through on account of the upright bowing deformations
segment
[4,5]. As such, the serviceability check against upright bowing deforma-
E Steel Young's modulus
tions is not considered in this paper.
f Out-of-plumb load applied to a rail beam
The 2D analysis model for drive-in racks proposed by Godley [6] is
fu,m Load applied in the translational stiffness Kuh,m
improved herein by introducing the horizontal restraints provided by
foy, foz Elastic buckling stresses about the y- and z- axes
both the rail beams and the pallet bracing restraints. The improved
h, hp Frame bracing pitch
model is checked against the 3D model developed by Gilbert and
hrail Rail beam elevation
Rasmussen [1,7] that is calibrated against laboratory test results. The
H Height of the rack
influence of the pallet restraints on the bending moment distribution
Iu Second moment of area of the upright
in the uprights is also evaluated. Thirty six drive-in racks representing
Ir Second moment of area of two rail beams.
the global sale of an Australian manufacturer over three years are then
ku Upright stiffness
analysed using the improved 2D model under all possible static loading
Kb Top rotational stiffness of the upright for a rack is sway
scenarios, alternately considering and ignoring the pallet bracing
mode
restraints. This paper evaluates the influence of pallet bracing restraints
Kc Base plate rotational stiffness of the base plate to floor
on the ultimate capacity of steel drive-in racks in the down-aisle
connection
direction, clarifies the loading scenario(s) governing the design and
Kr,i Rail beam translation stiffness at rail beam elevation i
determines the friction coefficient or the strength of a restraining device
Kt Top translational stiffness for the single upright model
required to prevent the pallets from sliding.
Kuh Horizontal translational stiffness of the upright at the
point of application of load P
2. Single upright model
Kuh,m Horizontal translational stiffness of the inner uprights
Kuh,fb Horizontal translational stiffness of the front and back
2.1. Single upright model proposed by Godley
uprights
lex, ley, lez Effective buckling lengths about the x-, y- and z- axes,
In order to reduce the computation time associated with large
respectively
models, Godley [6] developed a “single upright model” to analyse fully
L Distance between two uprights (upright frame width)
loaded drive-in racks in the down-aisle direction. The upright is
N* Design factored axial load
restrained at its base by a spring support having a rotational stiffness
Nc Nominal axial compression capacity of the upright
Kc, and at its top by another having a rotational stiffness Kb and a
Ncd Nominal axial compression distortional capacity of the
translational stiffness Kt, as shown in Fig. 2. Kc represents the restraint
upright
provided by the base plate to the floor connection, Kb the restraint pro-
Ncl Nominal axial compression local capacity of the upright
vided by the portal beams in double curvature (sway mode) having
Nce Nominal axial compression global capacity of the
semi-rigid connections to the upright, and Kt the combined restraint
upright
from the plan bracing (spanning the entire rack), spine bracing (span-
Ncrb Elastic buckling load of the upright determined from an
ning one bay) and upright frames. Pallet loads and out-of-plumb loads
elastic buckling analysis
are applied to the upright as shown in Fig. 2, where the rack is assumed
Ns Number of rail beam elevations
to be fully loaded such that loads W are applied on all rail beams.
Nu Number of uprights in the down-aisle direction
Detailed calculations for Kc, Kb and Kt, can be found in [6].
Mbx, Mby Nominal bending moments capacity of the upright
Despite its attractiveness, this model has limitations as it (i) ignores
about the x- and y-axes, respectively.
the restraint provided by the rail beams, (ii) does not take into account
Mbxd Nominal distortional bending moment capacity of the
the horizontal bracing restraint provided by pallets, and (iii) does
upright about the x-axis
not consider all possible upright loading scenarios, including partially
Mbxl Nominal local bending moment capacity of the upright
loaded racks where pallet loads are placed asymmetrically so as to
about the x-axis
induce bending of the upright. These limitations are addressed in
Mbxe Nominal global bending moment capacity of the up-
following sections.
right about the x-axis
Mo Global buckling moment
2.2. Improved single upright model
Mx*, My* Design factored bending moments about the x- and y-
axes, respectively
2.2.1. Rail beam restraints
P Horizontal load applied to the upright.
Typically, the out-of-plumb of drive-in racks is modelled by horizon-
Pb, Pc Load
tal forces at the rail beam supports that are linearly proportional to the
s Friction effect
gravity loads of the pallets (see Section 4.1.2). For a fully loaded rail
Sf Friction force
beam, the front and the back uprights are less loaded than the inner
rol Radius of gyration
uprights, resulting in smaller out-of-plumb forces being applied to the
W Pallet load
front and back uprights, as illustrated in Fig. 3 for a rack with two up-
α Out-of-plumb angle
right frames. Therefore and since rail beams link the uprights together,
Δ Total down-aisle displacement at the top of a drive-in
they restrain the deflection of the inner uprights when subjected
rack
to the out-of-plumb forces, as shown in Fig. 3(b), in which α is the
Φb Reduction capacity factor for member in bending
out-of-plumb angle with vertical.
Φc Reduction capacity factor for member in compression
Consequently, these restraints provided by the rail beams are intro-
ω Pallet uniform distributed load
duced into the single upright model by adding a horizontal translational
μ Friction coefficient
stiffness Kr,i at each rail beam elevation i, as shown in Fig. 4. An expres-
sion for Kr,i is derived in Section 2.2.1.2.
12 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23

Fig. 1. Example of a drive-in rack.

2.2.1.1. Upright down-aisle stiffness Kuh. Consider an upright of height H where

and second moment of area Iu that is subjected to a down-aisle force P
at a distance hrail from the floor, as shown in Fig. 5(a). The upright is 1 ð4ku þ K b ÞðH−hrail Þ þ 2ku hrail
¼ ð2Þ
restrained at its base by the rotational stiffness Kc provided by the Pc 12k2u þ 4ku ðK c þ K b Þ þ K c K b
base plate to floor connection, and at its top by the rotational stiffness
Kb provided by the upright to portal beam connection. The total
deformation of the upright can be resolved into three distinct parts 1 ð4ku þ K c Þhrail þ 2ku ðH−hrail Þ
¼ ð3Þ
with various boundary conditions and applied loads, as illustrated in Pb 12k2u þ 4ku ðK c þ K b Þ þ K c K b
Fig. 5(b–d), and the down-aisle stiffness Kuh of the upright (Fig. 5(e))
is then given as [8], and

EIu


   ku ¼ : ð4Þ
1 h3 ðH−hrail Þ h H 1 hrail K b H H
¼ rail K c 1− rail þ 2ku þ − ð1Þ
K uh 3EI u H2 3H hrail P c 3HP b 3

2.2.1.2. Upright down-aisle stiffness Kr,i. The stiffness Kr,i is derived herein
for the critical upright (second from the front) of a drive-in rack with
two upright frames and uniform spacing between uprights. For simplicity,
the restraints provided by all rail beams to an upright are assumed to be
independent of each other.

Fig. 3. (a) side view of a fully loaded rail beam and resulting axial loads in the uprights and
Fig. 2. Drive-in rack single upright model from Godley [6]. (b) top view deformed shape of the rail beam and frames under out-of-plumb forces.
 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23 13

Replacing the rail beam in Fig. 6 with its equivalent stiffness Kr,i at
the critical upright (either point B or C), using static equilibrium and
the expression for fu,m in Eq. (5), the translational restraint Kr,i provided
by the rail beam to the upright at the ith beam level is then expressed
as [8],

11K uh;fb −4K uh;m

K r;i ¼ ð6Þ
55L3 K uh;fb
þ 15:
6EIr

2.2.2. Pallet bracing restraints

The bracing effect provided by the pallets is now considered for any
loading scenario of a studied single upright. Bays not directly in the
vicinity of this upright are assumed to be fully loaded, as it would max-
imise the down-aisle displacement Δ of the rack and therefore the P-Δ
effects in the upright. Specifically, two loading scenarios are considered
for these bays, believed to represent the two design envelopes:
Fig. 4. Single upright model with rail beam restraints for a 4 storey (3 rail beam elevations)
drive-in rack. • Bay loading scenario A: all bays not directly connected to the studied
upright are fully loaded, as shown in Fig. 7(a).
• Bay loading scenario B: the two bays on each side of the two bays
directly connected to the studied upright are empty, while the
Each rail beam is restrained at its supports by the upright down-aisle remaining bays are fully loaded, as shown in Fig. 7(b). This loading
stiffness Kuh, derived in Section 2.2.1.1. Since the base plate to floor scenario aims to limit the influence of the pallets on the bending
connection stiffness Kc depends on the axial load in the upright [9] moment distribution in the studied upright, since contrary to the
and the front and back uprights are less loaded than the inner uprights, previous Bay loading scenario A, the pallets only link the studied upright
the upright down-aisle stiffness Kuh varies accordingly. This variation is and its two neighbours.
represented in Fig. 6 with the rail beam restrained at the front and back
uprights by spring supports each having the stiffness Kuh,fb and at the 2.2.2.1. Improved model for Bay loading scenario A (Model A). In a fully
inner uprights by springs having the stiffness Kuh,m. loaded rack, the influence of the pallets on the deformed shape of the
The force fu,m in the inner springs in Fig. 6 can be expressed as [8], uprights would be minimal, as all internal uprights in a row of uprights
in the down-aisle direction would identically deform. Therefore, the
11L
3
3 overall deformation of the rack at the critical row of uprights can be
þ found using the fully loaded improved single upright model introduced
12EI r 2K uh;fb
f u;m ¼ f ð5Þ in Section 2.2.1, i.e. not considering pallets, as illustrated in Fig. 8.
5L3 1 1
þ þ Moreover, if the number of bays of the rack is large, as frequently
6EI r K uh;fb K uh;m encountered in drive-in racks (see Fig. 1), removing pallets from each
side of the studied upright would have negligible influence on the over-
where 3f is the total out-of-plumb force applied to the rail beam all deformation of the rack, and the deformation of this upright would
(see Figs. 3(b) and 6), L is the distance between two uprights and Ir is be a function of both its immediate loading configuration and the over-
twice the second moment of area of the rail beam, as two rail beams all deformation of the rack imposed to the upright by the portal beams
are typically connected to the uprights. and the pallet bracing restraints.

Fig. 5. Upright deformation, with (a) total upright deformation, (b) bottom end fixed and top end released, (c) bottom end released and top end fixed, (d) bottom and top ends fixed with
applied force P and (e) equivalent stiffness Kuh.
14 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23

Fig. 6. Model of a rail beam subjected to out-of-plumb forces.

Therefore, the bracing restraint provided by the pallets for a given Step 2: The single upright model is loaded with its studied loading
loading scenario of the single upright is introduced into the model scenario, with the corresponding base plate to floor rotational
in the following manner, as illustrated in Fig. 9: stiffness Kc and rail beam stiffness Kr,I, determined for the axial
Step 1: The overall down-aisle displacements of the rack at each load in the studied upright.
rail beam elevation and at the top of the rack are determined using Step 3: The overall down-aisle displacement at the top of the rack
the fully loaded single upright model with out-of-plumb forces, (portal beam elevation) found in Step 1 is imposed at the top of
as shown in Fig. 8. The base plate to floor rotational stiffness Kc, the single upright model created in Step 2.
and rail beam stiffness Kr,i are calculated for the fully loaded Step 4: For each rail beam elevation of the model in Step 2, if there is
configuration. at least one pallet at the elevation, then the overall down-aisle

Fig. 7. Studied upright for (a) Bay loading scenario A and (b) Bay loading scenario B.
 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23 15

Fig. 8. Deformed shape of the single upright model for a fully loaded rack.

displacement at that elevation found in Step 1 is imposed on the 3. Influence of the pallet restraint on the bending moment
upright. distribution and validation of the single upright model

2.2.2.2. Improved model for Bay loading scenario B (Model B). In order The 3D advanced Finite Element model for drive-in racks developed
to determine the bending moment distribution in the studied upright by Gilbert and Rasmussen [1,7] is used herein to (i) analyse the influ-
for a given loading scenario of the upright, three single upright models ence of the pallet restraint on the bending moment distribution in the
are used and linked together by pinned rigid elements (ties) representing upright and (ii) validate the improved single upright model introduced
the pallet bracing restraints. The following steps are carried out as in Section 2.2. The 3D model has been calibrated against experimental
illustrated in Fig. 10: test results and considers joint eccentricities, nonlinear portal beam-
to-upright connections, nonlinear base-plate connections, and pallet
Step 1: As with the previous Bay loading scenario A, the overall bracing restraints. Seven degrees of freedom (i.e. Warping considered)
displacement imposed by the rack at the top of the critical upright beam elements are used in the 3D Finite Element model. Refer to [1,7]
and its two adjacent uprights (Fig. 7(b)) is determined using the for more details. In the present 3D second-order analysis, the FE
fully loaded single upright model with out-of-plumb forces, as software Abaqus [10] is used, while the FE software Strand7 [11] is
shown in Fig. 8. The base plate to floor rotational stiffness Kc, and used to run the 2D second-order analysis of the improved single upright
rail beam stiffness Kr,i are calculated for the fully loaded configuration. model. It may be noted that while the 2D model ignores torsion and
Step 2: A three single upright model is created and loaded with the warping of the uprights (phenomena that are considered in the 3D
model), it closely predicts the overall down-aisle behaviour of the
studied loading scenario. The base plate to floor rotational stiffness
rack, as developed using the 3D model in Sections 3.1 and 3.2. Therefore,
Kc and rail beam stiffness Kr,i for each of the three uprights is deter-
torsion and warping of the uprights are likely to have a limited influence
mined separately for the axial load in the upright. on the overall drive-in rack behaviour.
Step 3: The overall down-aisle displacement at the top of the rack A rack with similar characteristics to the one tested by Gilbert and
(portal beam elevation) found in Step 1 is imposed at the top of Rasmussen [1] is used as a case study in this validation. Specifically,
the three uprights created in Step 2. the rack is 12 bays wide, 4 pallets and 2 upright frames deep, and 4
Step 4: Pallet bracing restraints are modelled using horizontal ties storeys high (i.e. featuring 3 rail beam levels). It has 3 spine bracing
between rail beams, as shown in Fig. 10. modules, each spanning one bay, and 4 plan bracing modules, each

Fig. 9. Improved single upright model for Bay loading scenario A (Model A).
16 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23

Fig. 10. Improved single upright model for Bay loading scenario B (Model B).

spanning three bays. The overall lay-out of the rack is shown in Fig. 11. beams as the static friction coefficient is assumed to be sufficiently
Each pallet load is 2 tonnes. The rack is loaded as in Bay loading scenario high to prevent sliding. Two loading scenarios are studied, with
A, described in Section 2.2.2. The shear stiffness of the pallets is taken as the out-of-plumb and other design parameters given in Section 4.1.
7.2 N/mm, which is within the range experimentally found by Hua and Further verification of the improved single upright model can be
Rasmussen [3]. The pallets are considered to be fastened to the rail found in [8].

(a) – Side view (Down-aisle) (b) – Side view (Cross-aisle)

(c) – Top view

Fig. 11. Case study rack. (a) — Side view (down-aisle) (b) — side view (cross-aisle) (c) — top view.
 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23 17

Fig. 12. Loading scenario believed to generally govern the design.

3.1. First loading scenario — maximum combined axial compression comparison between Figs. 13 and 14 that the single model upright
and bending is able to accurately reproduce the bending moment distribution
of the critical upright, with and without pallet bracing restraints. The
The load case involving the loading scenario shown in elevation difference in the design bending moment between the 3D and the single
in Fig. 12 generally represents the governing load case for combined upright models is less than 6%.
axial compression and bending of the critical upright adjacent to the
unloaded compartment and to the aisle upright [12]. 3.2. Second loading scenario — maximum bending
The down-aisle bending moment distribution of the critical upright
from the 3D model accounting for pallet bracing restraints is plotted The load case involving the loading scenario shown in Fig. 15 typical-
in Fig. 13(a), and that obtained from the 3D model ignoring pallet ly induces the largest design bending moment in the critical upright.
bracing restraints in Fig. 13(b). Fig. 13 shows that the pallet bracing The down-aisle bending moment distribution of the critical upright
restraints significantly affect the bending moment distribution of the under the second load case obtained from the 3D model accounting
critical upright, but have only a relatively minor impact on the maxi- for pallet bracing restraints is plotted in Fig. 16(a), and that obtained
mum design bending moment. This observation appears to be general from the 3D model ignoring pallet bracing restraints in Fig. 16(b).
for this type of loading scenario. Note, however, that depending on the Fig. 16 shows that the pallet bracing restraints not only significantly
rack configuration, the design moment in the critical section of the up- affect the bending moment distribution of the critical upright, but also
right may be larger when pallet restraints are considered, and hence reduce the maximum design bending moment by almost one third
lower capacities may, in fact, result from considering pallet restraints under the second load case.
compared to ignoring pallet restraints. This unexpected result is investi- The down-aisle bending moment distribution of the critical upright
gated further in Section 5.1.1. obtained from the single upright model accounting for pallet bracing
The down-aisle bending moment distribution of the critical obtained retraints is plotted in Fig. 17(a), and that obtained from the single up-
from the single upright model accounting for pallet bracing restraints right model ignoring pallet bracing restraints in Fig. 17(b). Consistent
described in Section 2.2.2.1 is plotted in Fig. 14(a), and that obtained with the results for the previous loading scenario, the comparison
from the single upright model ignoring pallet bracing restraints (i.e. ig- between Figs. 16 and 17 shows that the single model upright is able to
noring Step 4 in Section 2.2.2.1) in Fig. 14(b). It can be seen from the accurately reproduce the bending moment distribution of the critical

Fig. 13. Bending moment distribution in the critical upright under vertical and out-of- Fig. 14. Bending moment distribution in the critical upright under vertical and out-of-
plumb loads for the loading scenario shown in Fig. 12 using 3D advanced analysis for plumb loads for the loading scenario shown in Fig. 12 using 2D analyses for (a) pallets
(a) pallets considered and (b) pallets ignored. considered and (b) pallets ignored.
18 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23

Fig. 15. Loading scenario inducing maximum bending moment in a row of uprights.

upright, with and without pallet bracing restraints. The difference in the 4.1. Design parameters
design bending moment between the 3D and the single upright models
is less than 7%. 4.1.1. Base plate to floor connection stiffness
Base plates are generally bolted to the floor, and the strength and
4. Parametric studies initial rotational stiffness of the base plate to floor connection depend
on the axial load in the upright [9]. Numerical investigations on the
Thirty-six rack configurations, representing the global sale of an non-linear behaviour of a typical storage rack base plate assembly [14]
Australian manufacturer over three years and designed using industry showed that (i) the connection strength is proportional to the upright
practice [12], are analysed using the single upright models. The racks width, (ii) in the presence of axial load in the upright, the initial
are considered to be 4 pallets deep, with rail beams equally spaced rotational stiffness of the base plate to floor connection is proportional
along the rack height. The uprights are referred to as “SD” for standard to the cube of the upright width and (iii) when no axial load is applied
uprights and “RF” for rear flanged uprights, their widths range from to the upright, the initial rotational stiffness is independent of the
70 mm to 150 mm and their thicknesses from 1.2 mm to 2.4 mm. upright width.
Table 1 summarises the rack configurations including the rack height, The rules described above, combined with the test results in [14]
design pallet load, number of storeys and upright type. More details applicable to a 125 mm wide base plate assembly, are used in the
can be found in [8,13]. following sections to determine the initial stiffness and strength
Specifically, three different single upright models are considered of base plate to floor connections as functions of base plate width.
and their member action-to-capacity ratios are used to quantify the The detailed moment-rotation curves used in the present work
influence of pallet restraints on the design of drive-in racks: are given in [8].

• Model A considers the pallet bracing restraints and represents the Bay 4.1.2. Out-of-plumb
loading scenario A. The model is described in Section 2.2.2.1 and illus- The main international racking specifications [15–17] consider the
trated in Fig. 9. initial looseness in the member connections as well as the initial out-
• Model B considers the pallet bracing restraints and represents the Bay of-plumb as frame imperfections, which are generally accounted for in
loading scenario B. The model is described in Section 2.2.2.2 and illus- the design by means of horizontal forces Fout-of-plumb applied at each
trated in Fig. 10. rail beam elevation as,
• Model C is based on the current industry practice of neglecting pallet
bracing restraints. The model is similar to Model A with the exception
of Step 4 in Section 2.2.2.1. F out of plumb ¼ αW ð7Þ

Fig. 16. Bending moment distribution in the critical upright under vertical and out-of- Fig. 17. Bending moment distribution in the critical upright under vertical and out-of-
plumb loads for the loading scenario shown in Figs. 15 and 3D advanced analysis for (a) plumb loads for the loading scenario shown in Figs. 16 and 2D analyses for (a) pallets
pallets considered and (b) pallets ignored. considered and (b) pallets ignored.
 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23 19

Table 1
Rack configurations.

Upright

Rack no Height (mm) Nb storeys Design pallet load (kg) Type Width (mm) Thk. (mm)
1 3775 2 950 SD 70 1.2
2 3 950 SD 90 1.5
3 1210 RF 90 1.2
4 4 690 RF 90 1.2
5 5025 2 950 SD 90 1.2
6 1210 SD 90 1.2
7 3 690 RF 90 1.2
8 950 RF 90 1.2
9 1210 RF 90 1.5
10 4 950 SD 110 1.5
11 1210 RF 110 1.5
12 1470 RF 125 1.5
13 6275 2 1470 RF 90 1.2
14 3 950 RF 90 1.5
15 1210 SD 110 1.5
16 1470 RF 110 1.5
17 4 430 SD 90 1.5
18 950 SD 110 1.2
19 1210 RF 110 1.9
20 1470 RF 125 1.5
21 5 950 RF 110 1.9
22 6 690 RF 110 1.5
23 950 RF 125 1.5
24 7525 3 1210 SD 110 1.5
25 1470 RF 110 1.5
26 4 430 RF 90 1.2
27 950 RF 110 1.9
28 5 950 SD 125 1.9
29 8775 3 1210 SD 125 1.5
30 4 430 RF 90 1.7
31 950 SD 125 1.9
32 1210 RF 125 1.9
33 5 950 SD 150 1.9
34 1210 RF 150 1.9
35 1470 RF 150 2.4
36 6 950 RF 150 1.9

where α is the out-of-plumb angle and W is the vertical load applied to per rack is a function of the number of rail beams and is equal to 4Ns,
the upright by the pallets at the rail beam elevation. The out-of-plumb where Ns is the number of rail beam elevations.
angle α is typically a function of the number of interconnected bays
and the looseness in the portal beam to upright connections. A typical 4.3. Ultimate capacity
out-of-plumb angle of 0.0044 rad (about 1/250) is used in the present
work, see [8] for more details. For each of the three rack models (A, B, C) and each upright load
case, the Australasian/New Zealand cold-formed steel structures
standard AS/NZS 4600 [18] is used to calculate the member action-to-
4.1.3. Other parameters
Other design parameters used in the present work, which corre-
spond to drive-in rack configurations currently commercialised in
Australia, are given in [8]. The height of the rack H, the number of pallet
levels Ns and the cross-sectional area of the upright Au depend on the
studied rack characteristics and are given in Table 1 and [8].

4.2. Upright load cases

According to the FEM specification for the design of drive-in racks

[12], the load case involving the loading scenario depicted in Fig. 12
and a fully loaded upright are usually “sufficient to consider the pattern
load effects” for the Ultimate Limit State (ULS) design in the down-aisle
direction. However, it is currently unclear if a different load case may
govern the design. Moreover, in light of the horizontal bracing effect
offered by the pallets, the load case involving the loading scenario
depicted in Fig. 12 and a fully loaded upright may not always be
sufficient for the ULS design of the upright.
Consequently, every possible load case is investigated in the present
work for the 36 drive-in racks given in Table 1. Similar to Section 2,
second-order geometric analyses are carried out using the general
purpose FE software Strand7 [11]. The number of load cases analysed Fig. 18. Frame bracing pitches h and hp, unbraced segment.
20 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23

nominal moment capacities Mbxe, Mbxl, Mbxd, respectively,

Mbx ¼ minðMbxe ; M bxl ; Mbxl Þ: ð11Þ

The global nominal capacity Mbxe is a function of the bending mo-

ment distribution in the upright through the elastic buckling moment
Mo,
qffiffiffiffiffiffiffiffiffiffiffiffiffi
Mo ¼ C b Au r ol f oy f oz ð12Þ

where Cb is a coefficient depending on moment distribution in the

unbraced segment of the upright, Au is the gross cross-sectional area,
rol is the polar radius of gyration about the shear centre and foy and foz
are the elastic buckling stresses for flexural buckling about the y-axes
(perpendicular to the symmetry axis) and torsional buckling,
respectively.
Detailed rules to determine Nce, Ncl, Ncd, Mbxe, Mbxl, and Mbxd are given
in [8,18].

4.3.1. Effective buckling lengths

For the calculation of Nce, the Australian Standard AS 4084 [15] rec-
ommends effective lengths ley and lez for buckling about the y- (down-
aisle) and z- (torsional) axes equal to and 0.7 times the upright frame
Fig. 19. Example of a rack model used in the rational frame buckling analysis for each of
bracing pitch, respectively. These effective lengths are for buckling in
Models A, B and C.
the plane of the upright frames. Consistent with [12], the rail beams
capacity ratios of the critical upright. When second order-geometric are not considered to form part of the upright frame and ignored in
analyses are used, members subjected to combined axial compression the calculation of the frame bracing pitch. The upright frame bracing
and bending must satisfy the ULS design check in Eq. (8), pitches (h and hp) are defined in [15] and shown in Fig. 18. These values
are adopted in the present work, although there are peculiarities associ-
ated with this 2D approach [20].
N Mx  My 
þ þ ≤1 ð8Þ Also, as per [15], for each member of the simple upright model, the
ϕc Nc ϕb M bx ϕb Mby
effective length lex for buckling about the x-axis (in the down-aisle
plane) is calculated as
where N* is the design axial compression load, and Mx* and My* are the sffiffiffiffiffiffiffiffiffi
design bending moments about the x- (cross-aisle) and y- (down-aisle) EIx
axes (see Fig. 11(c) for axes details), respectively, Nc is the nominal axial lex ¼π ð13Þ
N crb
compression member capacity, Mbx and Mby are the nominal member
bending moment capacities about the x- and y- axes, respectively, ϕc
where Ix is the second moment of area about the x axis, and Ncrb is the
and ϕb are reduction capacity factors for members in compression and
elastic buckling load of the upright determined from a rational frame
bending, taken as 0.85 and 0.90, respectively. As the present work is
buckling analysis.
concerned with the design of drive-in racks in the down-aisle direction,
The structural model used in the rational frame buckling analysis for
the bending moment about the down-aisle axis is considered negligible
each of Models A, B and C is exemplified in Fig. 19. It should be noted
and for the 2D single upright model, Eq. (8) becomes,
that the horizontal translational restraints at the two pallet levels do
not represent actual physical restraints, whether rigid or partial (due
N Mx  to pallet base friction, for example). That is, horizontal translational
þ ≤1: ð9Þ
ϕc Nc ϕb M bx

This approach is consistent with AS 4084 [15] which does not

require a design to account for bimoments for racks in general. It
is only when the rack is subjected to primary torsion action that the
Standard requires bimoments to be considered.
A load factor of 1.4 is used for the pallets to determine the design
loads N* and bending moments Mx*. The self-weight of the rack is
ignored.
The Direct Strength Method [19] in Section 7 of the AS/NZS 4600 [18]
is used in the present work to calculate the nominal capacities Nc and
Mbx of the upright. Specifically, the axial capacity in compression Nc is
defined as the lesser of the axial global, local and distortional nominal
capacities Nce, Ncl and Ncd, respectively,

Nc ¼ minðNce ; Ncl ; Ncd Þ ð10Þ

and the nominal bending moment capacity Mbx about the x-axis Fig. 20. Influence of the horizontal pallet restraint on the action-to-capacity ratio for h =
of bending is defined as the lesser of the global, local and distortional 1500 mm.
 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23 21

Fig. 21. Bending moment distribution for the critical load case for rack 25 for (a) Model A (pallets considered) and (b) Model C (pallets ignored).

restraints are applied wherever pallets are present, irrespective of illustrated in Fig. 12). Fig. 21 shows the bending moment distribution
whether friction is ignored or considered in the design. The horizontal in the upright for Models A and C, and the coefficient Cb in Eq. (12),
translational restraints applied where pallets are present are “notional”, for the critical load case for Rack 25. It can be seen that ignoring the
and reflect the fact that notional horizontal loads, i.e. out-of-plumb pallet restraints leads to a design bending moment in the critical upright
loads, have been applied there in conjunction with second-order analy- of 1326 kN.mm, that is 12% less than the design bending moment of
sis to account for the bending moment amplifications in the upright, 1502 kN.mm when considering pallet restraints, but with similar Cb co-
and is consistent with the use of an effective flexural length determined efficient. As specified in Section 4.3.1, it should be noted that second-
from Eq. (13) in the member design check. order bending moments have been accounted for in the analysis, and
the effective lengths used in the member design checks are not affected
by the pallet restraints.
5. Results
Table 2 summarises the average maximum member action-to-
capacity ratios given in Fig. 20.
5.1. Effects of pallet restraint

In this study, two values for the frame bracing pitches (with h = hp
5.1.2. Frame bracing pitch h = 2000 mm
in Fig. 18), being 1500 mm and 2000 mm, are considered.
Fig. 22 plots the ratios of the maximum member action-to-capacity
ratio of Model C (current industry practice) to that of Model A, and to
5.1.1. Frame bracing pitch h = 1500 mm that of Model B, for the 36 racks given in Table 1 having a frame bracing
Fig. 20 plots the ratios of the maximum member action-to-capacity pitch h of 2000 mm. Detailed results can be found in [8].
ratio, calculated using Eq. (9), of Model C (current industry practice) Similar conclusions to those in Section 5.1.1 can be drawn. Results
to that of Model A, and to that of Model B, for the 36 racks given in show that for 11 racks out of 36, considering the horizontal restraining
Table 1 having a frame bracing pitch h = hp of 1500 mm. Detailed results effect provided by the pallets would provide more economical designs
can be found in [8]. A ratio greater than 1.0 in Fig. 20 indicates that the than the current industry practice, with a decrease in the member
current industry practice results in uneconomical designs. action-to- capacity ratio of up to 5% (Rack 1) and an average decrease
Fig. 20 shows that for 12 racks out of 36, incorporating the horizontal of 2%. For the remaining 25 racks, ignoring the pallet restraints leads
restraining effect provided by the pallets would provide more econom- to less conservative designs, with a maximum increase in the member
ical designs than current industry practice, with a decrease in the mem- action-to-capacity ratio of 5% (Rack 25) and an average increase of 3%.
ber action-to-capacity ratio of up to 6% (Rack 1). On average for the 12 Table 2 summarises the average maximum member action-to-capacity
racks, the decrease is 2%. ratios given in Fig. 22.
For the remaining 24 racks, ignoring the pallet restraints would lead
to less conservative designs, with an increase in the member action-to-
capacity ratio of up to 7% (Rack 30). On average for the 24 racks,
ignoring the pallet restraints increases the design capacity by 3%. This
counterintuitive result is due to the role of the pallet restraints in in-
creasing the design bending moment of the upright under the critical
load case. While considering the pallet restraints resulted in larger
bending moments at the upright bases, it resulted in smaller maximum
bending moments at the critical sections (which are usually between
the floor and the first rail beam elevation for the critical load case

Table 2
Ratio of the maximum member action-to-capacity ratios of Model C (current industry
practice) to Models A and B.

Model C Model C
(current practice)/model A (current practice)/model B

h (mm) Average ratio CoV Average ratio CoV

1500 mm 0.99 0.029 1.03 0.034
Fig. 22. Influence of the horizontal pallet restraint on the action-to-capacity ratio for h =
2000 mm 0.99 0.026 1.04 0.016
2000 mm.
22 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23

5.2. Required friction

The minimum friction coefficient μ needed to prevent the pallets

from sliding on the rail beams is investigated herein for Model A. The
friction forces Sf developing between the pallets and the rail beams are
extracted from the horizontal reactions at each loaded rail beam eleva-
tion of the single upright model. The friction effect s is then calculated
as,

Sf
s¼ ð14Þ
W

where W is the axial load applied by the pallets to the upright at the rail
beam elevation.
Fig. 24 shows the friction effect s for all loading cases and for the 36
drive-in racks defined in Table 1. All values in Fig. 24 are less than the
design static friction coefficient of 0.439 recommended by Hua and
Fig. 23. Specific load cases governing the design.
Rasmussen [3] (see Section 1), indicating that, under normal operating
conditions, sliding is unlikely to occur between the pallets and the rail
5.1.3. Critical load cases beams, and that pallet bracing restraints could be considered in the
When the pallet restraints are considered in the analysis (Models A design of drive-in racks. Moreover, the friction effect s is dependent on
and B), the load case involving the loading scenario illustrated the number of storeys (or rail beam elevations), as seen in Fig. 24. The
in Fig. 12, which corresponds to a fully loaded rack except for one com- more storeys for a given rack height, the greater the friction effect and
partment at the first rail beam elevation, is found to generally govern hence the more likely the pallets are to slide. Results show that, for
the design. However, for the 4-storey drive-in racks numbers 26 and a given number of storeys, the friction effect decreases approximately
30, the load case shown in Fig. 23(a) is found to provide an action-to- linearly with the height of the rack.
capacity ratio of up to 13% higher than the load case involving the load-
ing scenario shown in Fig. 12. Despite the fact that a lower axial load oc- 6. Conclusions
curred in the critical upright, the loading scenario induces a buckling
length lex about twice that for the loading scenario shown in Fig. 13, This paper analyses the influence of horizontal bracing restraints
and therefore leads to a reduced axial capacity. provided by the pallets on the behaviour and design of steel drive-in
When the pallet restraints are ignored in the analysis, the load case racks. The pallets are shown to significantly influence the bending
involving the loading scenario shown in Fig. 12 is also found to generally moment distribution in the uprights. The single upright model present-
govern the design. However, the load cases shown in Fig. 23(b) for the ed by Godley was improved by including the restraints provided by the
4-storey drive-in rack number 4, Fig. 23(c) for the 5-storey rack number rail beams and the pallets. Comparison with advanced 3D Finite
21 and Fig. 23(d) for the 6-storey drive-in racks numbers 21, 22 and 36 Element Analyses showed that the improved single upright model
govern the design with action-to-capacity ratios of 2%, 3% and 4.5% was able to accurately reproduce the bending moment distribution
higher than those for the loading scenario shown in Fig. 12, respectively. in the upright in the down-aisle direction under gravity and out-of
In view of the above results, for ULS design ignoring pallet bracing plumb loads.
effects, limiting the analysis to the load case involving the loading Using the improved single upright model, analyses were run for 36
scenario shown in Fig. 12 and a fully loaded rack, would only induce a drive-in rack configurations. All possible loading cases were analysed.
limited error in the action-to-capacity ratio and may be considered to Results showed that ignoring the pallet restraints in the design usually
be “sufficient for considering the pattern load effects” as stated in [12]. leads to design bending moments in the critical upright less than the
design bending moments obtained when considering pallet restraints.
Specifically, ignoring the pallet bracing effects in design, as in the cur-
rent industry practice, was shown to lead to a less conservative design
with an action-to-capacity ratio for the critical upright being reduced
in the order of 4%.
The standard load case involving a fully loaded rack except for one
compartment at the first rail beam elevation was found to govern the
ultimate limit state design of most racks. However, other loading
scenarios were found to govern the design of some drive-in racks having
4 to 6 storeys, with action-to-capacity ratios of up to 5% greater than
those produced by the standard load case.
Results show that under normal operating conditions, the friction
coefficient between the pallets and the rail beams is sufficient to
prevent the sliding of the pallets, and therefore the effect of pallets
could be considered in the design of drive-in racks. However, as
demonstrated in the paper, overall there is no benefit to be gained
from including the effect of pallet restraint in the design.

References

[1] Gilbert BP, Rasmussen KJR. Drive-in steel storage racks I: stiffness test and 3D load
Fig. 24. Friction effect s. transfer mechanisms. ASCE J Struct Eng 2012;138:135–47.
 B.P. Gilbert et al. / Journal of Constructional Steel Research 97 (2014) 10–23 23

[2] Salmon MA, Welch RE, Longinow A. Analysis of drive-in and drive-thru storage [10] Abaqus, Abaqus ver. 6.10 — user manual. Providence, U.S.A.: ABAQUS, Inc.; 2010.
racks. In: Yu WW, editor. Proceedings of the 2nd Specialty Conference on Cold- [11] Strand7, using Strand7 – user manual – release 2.4.4. Sydney, Australia: G + D Com-
Formed Steel Structures; 1973. p. 617–39 [St Louis, Missouri, U.S.A.]. puting Pty Ltd.; 2010.
[3] Hua V, Rasmussen KJR. Static friction coefficient between pallets and beam rails and [12] FEM 10.2.07, the design of ‘drive-in’ and ‘drive-through’ pallet racking. Brussels,
pallet shear stiffness tests, Research Report 914. Australia: School of Civil Engineering, Belgium: Federation Europeenne de la Manutention; 2012.
The University of Sydney; 2010. [13] Gilbert BP. The behaviour of steel drive-in racks under static and forklift
[4] Gilbert BP, Rasmussen KJR. Impact tests and parametric impact studies on drive-in truck impact forces. [PhD Thesis] Sydney, Australia: School of Civil Engineering,
steel storage racks. Eng Struct 2011;33:1410–22. The University of Sydney; 2010.
[5] Gilbert BP, Rasmussen KJR. Determination of accidental forklift truck impact forces [14] Gilbert BP, Rasmussen KJR. Determination of the base plate stiffness and strength of
on drive-in steel storage rack structures. Eng Struct 2011;33:1403–9. steel storage racks. J Constr Steel Res 2011;67:1031–41.
[6] Godley MHR. The behaviour of drive-in storage structures. In: LaBoule RA, Yu WW, [15] AS 4084. Steel storage racking. Sydney, Australia: Standards Australia; 2012.
editors. Proceedings of the 16th International Specialty Conference on Cold-Formed [16] EN 15512. Steel static storage systems – adjustable pallet racking systems –
Steel Structures; 2002. p. 340–52 [Orlando, Florida, U.S.A.]. principles for structural design. Brussels, Belgium: European Committee for Stan-
[7] Gilbert BP, Rasmussen KJR. Finite element modelling of steel drive-in rack structures, dardization (CEN); 2009.
research report R901. Australia: School of Civil Engineering, The University of [17] RMI. Specification for the design, testing and utilization of industrial steel storage
Sydney; 2009. racks. Charlotte, U.S.A.: Rack Manufacturers Institute; 2008.
[8] Gilbert BP, Teh LH, Badet RX, Rasmussen KJR. Determination of the influence of the [18] AS/NZS 4600. Cold-formed steel structures. Sydney, Australia: Standards Australia;
pallets on the design of drive-in steel storage racks, research Report CIEM/2013/R04. 2005.
Australia: Centre for Infrastructure Engineering and Management, Griffith University; [19] Schafer BW. Designing cold-formed steel using the direct strength method. In:
2013. LaBoule RA, Yu WW, editors. Proceedings of the 18th International Specialty
[9] Godley MHR, Beale RG, Feng X. Rotational stiffness of semi-rigid baseplates. Conference on Cold-Formed Steel Structures; 2006. p. 475–90 [Orlando,
In: Yu WW, LaBoule RA, editors. Proceedings of the 14th International Florida].
Specialty Conference on Cold-Formed Steel Structures; 1998. p. 323–35 [St [20] Teh LH, Hancock GJ, Clarke MJ. Analysis an]d design of double-sided high-rise steel
Louis, Missouri, U.S.A.]. pallet rack frames. ASCE J Struct Eng 2004;130:1011–21.