Thin-Walled Structures 37 (2000) 277–304

www.elsevier.com/locate/tws

Analysis and behaviour of steel storage pallet

racks
a,* b
Nadia Baldassino , Claudio Bernuzzi
a
Department of Mechanical and Structural Engineering, University of Trento, via Mesiano 77, 38050
Trento, Italy
b
Department of Structural Engineering, Politecnico di Milano, Piazza Leonardo da Vinci, 32 - 20133
Milan, Italy

Received 5 January 2000; accepted 24 March 2000

Abstract

For steel storage pallet racks, lateral stiffness in down-aisle direction is usually provided
by beam-to-column joints and base-plate connections, owing to the impracticability of using
bracing systems in selected areas of pallet racks. Because of the high slenderness of members,
second order effects influence remarkably the frame performance. The model of unbraced
semi-continuous frame is hence adopted for design.
This paper presents the main results of a numerical study on the response of pallet racks
commonly used in Europe. The influence of beam-to-column joint modelling on the overall
frame response is singled out with reference to both service condition and ultimate limit states.
The effect of an accurate definition of frame imperfections is considered. Finally, the influence
of base-plate joints on the overall stability of the rack systems is discussed focussing attention
on both stiffness and strength of the column bases.  2000 Elsevier Science Ltd. All rights
reserved.

Keywords: Pallet racks; Thin-walled perforated members; Beam-to-column joints; Column bases; Semi-
rigid joints; Unbraced frames; Second order effects; Structural analysis

* Corresponding author. Fax: +39-0461-882-505.

E-mail addresses: [email protected] (N. Baldassino), [email protected] (C.
Bernuzzi).

0263-8231/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 3 - 8 2 3 1 ( 0 0 ) 0 0 0 2 1 - 5
278 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

1. Introduction

One of the most significant uses of cold formed members is for steel storage
racking structures [1], such as pallet, drive-in, and drive-through racking systems.
With reference to the European practice for pallets racks (Fig. 1), considered in the
following, beams (stringers) are generally realised by means of boxed cross-section
members. Moreover, columns (uprights) present perforated open sections apt to
accept the hooks of beam-end-connectors, which join beams and columns together
without the need for bolts or welds. The behaviour of the perforated columns, that
are generally thin-walled members, is affected by different buckling modes (local,
distortional and global) as well as by their mutual interactions [2,3]. The response
of both beam-to-column and base-plate joints is typically non linear. In addition, the
performance of base-plate connections depends significantly on the level of the axial
load [1,4]. Moreover, bracing systems are generally placed only in the cross-aisle
direction. The need for organising pallet racks in such a way that the product is
efficiently stored and sufficiently accessible hampers the presence of bracings in the
down-aisle direction. Lateral stability is, hence, provided by the sole degree of conti-
nuity associated with beam-to-column joints as well as by base-plate connections.
Due to difficulties associated with a theoretical assessment of the performance of
thin walled perforated members, the design standards recently developed for steel
storage pallet racks [5–8] are based on an approach requiring specific tests on isolated
members and joints. As far as the structural analysis of the racks is concerned, the
design model of semi-continuous sway frame (i.e. frame with semi-rigid joints) has
hence to be adopted. Furthermore, due to the high slenderness of the members,
second order effects are generally non negligible and, as a consequence, advanced
analysis methods are required for practical design.
A research project on steel storage pallet racking systems is currently in progress
[9,10] with the goal of developing the corresponding Italian recommendations. Typi-

Fig. 1. Typical configuration of steel storage pallet racks.

 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 279

cal rack components from 17 commercial Italian manufactures have been experimen-
tally investigated. In particular, until October ’99 approximately 500 tests [11,12]
have been executed in order to identify the performance of 108 types of stocky
columns (stub column tests) and 52 different beam-to-column joints (beam-end-con-
nector tests). Tests on both base-plate joints and on perforated columns in bending
are currently in progress, in order to investigate the behaviour of other isolated key
components of pallet racks. All these types of tests are expected to provide essential
data for an accurate and reliable pallet rack design.
Complementary to the experimental phase, a numerical study on stability issues
related to pallet racks has been carried out in order to investigate the main aspects
influencing the overall frame response.
This paper deals with the numerical phase of the research, carried out by means
of a commercial non linear finite element program for the analysis of planar frames.
The influence of the joint model is analysed with reference to serviceability as well
as ultimate limit states. In particular, 16 different configurations of pallet racks com-
monly used in Italy and Europe are considered; simplified bi-linear moment-rotation
joint relationships are proposed for use in design. The effects of a suitable definition
of the frame imperfections, based on the frame geometry (i.e. on the numbers of
bays and of storeys) as well as on the initial looseness of beam-end-connectors, are
considered. Finally, attention is focused on the influence of base-plate joints on the
frame performance: the non negligible benefits associated with an accurate modelling
of these components are presented and discussed.

2. The numerical analysis

Steel storage pallet racks are three-dimensional framed systems (Fig. 1). With
reference to the most common regular configurations (i.e. frames with beams at each
floor level and at each span), braced and unbraced planar frame models have to be
considered for the cross-aisle and down-aisle direction (Fig. 2), respectively. In the
following, attention is focused on the behaviour in the sole down-aisle direction.

2.1. The considered rack frames

Five Italian Companies (in the following conventionally identified as A, B,..., E)

indicated some of the most significant typologies of planar frames (named a, b,....,f).
For each, the Companies specified the geometric data related to the member proper-
ties and the service load, herein identified by a load multiplier a equal to unity.
Moreover, in order to assess the influence of the behaviour of base-plate connections
on the overall frame response, the frame which has been selected for the practical
application proposed by the Federation Europeenne de la Manutention [13] has also
been considered (FEM frame).
In total, 16 planar frames have been modeled. Their layout is presented in Table
1, where the number of spans (ns) and of floors (nL) are reported together with the
span length (Lb), the interstorey heigth (h), the height of the first load level (h1) and
280 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Fig. 2. Elevation and plan view of steel storage pallet racks.

the total height of the rack frame (H). A uniformly distributed load has been con-
sidered on each beam; frame imperfections have been accounted for by considering
an initial out-of-plumb angle for the columns, in accordance with the approaches
recommended in modern codes for the design of steel buildings. This angle is ident-
ified as the out-of-plumb divided by the rack height.
As to the component properties, experimental results of tests carried out in accord-
ance with the procedure specified in ref. 14 were considered both for perforated
columns and for beam-to-column joints. In particular, stub column tests were carried
out to define the axial carrying capacity, which takes into account the interaction
between local and distorsional buckling and perforations. As to beam-to-column
joints, reference was made to the moment-rotation, M-⌽, curves obtained from beam-
end-connector tests. For each type of beam-end-connectors, four tests were executed
(Fig. 3) and, in particular:

앫 three tests under hogging moments (Tests I–III), to appraise the connection behav-
iour in the usual service conditions. It should be observed that reference is herein
made to the curve associated with the lower joint resistance;
앫 one test under actions generating sagging moments (Test R), to evaluate the joint
performance in the presence of accidental upward action or of frame sway.
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 281

Table 1
Layout of the planar rack configurations considered for structural analysis

Frame ns Lb (mm) nL h (mm) h1 (mm) H (mm)

Aaa 6 2780 4 1575 1575 6300

Aba 6 2780 4 1575 1575 6300
Aca 6 2780 4 1575 1575 6300

Ba 3 2830 3 2100 2020 6220

Bb 3 2790 3 1800 1735 5335
Bc 3 1860 3 1200 1160 3560

Ca 5 2700 7 1000 200 6200

Cb 5 2700 7 1000 1000 7000
Cc 5 2700 5 1500 1500 7500
Cd 5 2700 4 1900 2000 7700
Ce 5 2700 3 2500 2500 7500
Cf 5 2700 3 2400 3000 7800

Da 2 2790 3 1500 1430 4430

Db 2 2760 4 1000 1000 4000

Ea 3 2780 2 2100 2070 4170

FEM 6 2800 4 1400 1400 5600

a
Frames Aa, Ab and Ac differ for the type of beam, column and beam-to-column joint.

In accordance with the approach proposed for steel building design in Eurocode
3 [15], the frame model to be used in the design depends significantly on the degree
of flexural continuity provided by beam-to-column joints. In the case of rigid joints,
the design can be based on the rigid frame model: when joints are hinges, a simple
frame model is adopted and lateral stiffness is provided by bracing systems. Other-
wise, if joints can be considered as semi-rigid, a semi-continuous frame model has
to be adopted for the analysis.
In order to select the frame model to be used, the criterion of Eurocode 3 [15]
for joint classification in unbraced frames was applied (Fig. 4). The joint M-⌽ curves,
282 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Fig. 3. Typical experimental moment-rotation curves for beam-end connector.

adopted in the numerical analysis are directly compared in Fig. 5, where they are
expressed as m̄⫺f̄ non dimensional curves, obtained by means of the following equa-
tions:
M
m̄⫽ (1a)
Mu,b
EIb
f̄⫽⌽ (1b)
LbMu,b
where E is the Young modulus, Ib the second moment of area of the beam, Lb
the beam length (from the data of Table 1) and where Mu,b represents the flexural
beam resistance.
With reference to the hogging m̄⫺f̄ joint curves, it can be noted that:

앫 the joint curve for frame Bb is located in the domain of flexible connections (this
frame should hence be modelled as simple frame);
앫 the joints of frames Ac, Bc, Db and Ea are semi-rigid, referring to the value of
the rotational stiffness, while they must be considered as hinges if reference is
made to the bending strength;
앫 in all other cases (i.e. for Aa, Ab, Ba, Ca-Cf, Da and FEM frames), joints are
semi-rigid with reference to both stiffness and strength and their responses are
located in the vicinity of the bound between semi-rigid and flexible domain.
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 283

Fig. 4. EC3 criterion for joint classification in unbraced steel frames [15].

With reference to the behaviour under sagging moments, the previous remarks related
to joint classification are confirmed except for the joints of frame Ac, that are in the
flexible domain, and for the joints of frame Bc, which are semi-rigid for both stiffness
and strength.
As a general conclusion, it can be said that beam-to-column joints are characterised
by a very modest degree of flexural continuity and a simple frame model should be
used in most cases. However, a numerical study on the analysis models for steel
framed buildings [16] shows that the joint influence on frame behaviour is non negli-
gible, also in the case of flexible joints. Consequently, semi-continuous frame design
models are always recommended to assess more accurately the rack response, also
when the joint response falls in the flexible joint domain.

2.2. The finite element program

The numerical study on rack systems has been carried out by means of a commer-
cial finite element (F.E.) program for frame analysis, PEP micro [17], which allows
284 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Fig. 5. Classification for the joints of the considered frames (see Fig. 4).
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 285

non-linear static analysis of planar framed structures. In particular, the F.E. formu-
lation implemented in the program makes it possible to account for:

앫 elasto-plastic member behaviour and section plastification in accordance with the

“plastic hinge” concept;
앫 second order effects that are considered when creating the tangential stiffness
matrix of members in an initial Lagrange reference system. Only overall flexural
instability in the plane of the frame is considered by the implemented formulation;
앫 non linear relationships, by which both beam-to-column and base-plate joints
(multi-linear M-⌽ curve) may be simulated.

The overall response of the planar frames is assessed by increasing proportionally

all the external loads, via the load multiplier a, till collapse (due to yielding of
members and/or joints, flexural instability or interaction between plasticity and
instability) is achieved.
Because of the remarkable influence of the high slenderness of rack components
on the frame behaviour, results related to second order analysis (assuming linear
behaviour for the members and either linear or non linear constitutive relationships
for joints) are presented and discussed.

3. Influence of the joint modelling

As previously mentioned, the joint response is typically non linear [1,4,11]. There-
fore, practical design requires the definition, from experimental data, of suitable sim-
plified M-⌽ relationships, generally by means of bilinear models. As a consequence,
attention has been initially directed to assess the influence of joint modelling on the
overall frame response, with the aim at defining a bilinear M-⌽ curve, capable of
simulating the joint response accurately and efficiently, for design purposes.
Reference was herein made to the first 15 frames in Table 1 (i.e., frames Aa–Ea).
For each of them, an initial out-of-plumb angle of 1/200 rad was assumed for the
columns, in order to account for the presence of geometric imperfections.
After some preliminary analysis carried out to assess the efficiency of different
types of bilinear M-⌽ models, attention was focused on the elastic-perfectly plastic
M-⌽ relationships. Two values of the elastic stiffness were considered: the initial
tangent stiffness (TS) and the secant stiffness (SS). The latter is defined as the value
of the secant stiffness corresponding to a moment level equal to 0.66 times the joint
moment resistance, Mu. For each type of relationship (i.e. TS or SS) both the sym-
metrical joint response model (s), which has been referred to the behaviour under
hogging moment, and the unsymmetrical one (u) have been used to simulate the M-
⌽ curve. Additionally, for a comparison with a more refined modelling of the joint
behaviour, also the model of multi-linear unsymmetrical relationship (ML), capable
to simulate accurately the elastic, inelastic and plastic branches of the joint response
has been considered. A summary of the different M-⌽ relationships is presented in
Fig. 6.
286 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Fig. 6. Considered moment-rotation (M-⌽) curves.

The absence of experimental data related to the behaviour of the base-plate con-
nections suggested to consider two different “ideal” and extreme restraint conditions
at the column bases: hinged bases and fixed bases.
Due to the importance of the frame flexibility to lateral loads, a preliminary evalu-
ation of the relationships between the load multiplier and the horizontal displacement
at the top of the frames (a⫺dT curves) was carried out for the different joint models,
before assessing the influence of joint modelling on the frame response.
As a general remark, the influence of the joint symmetry appears very limited.
The responses associated with the TSu and SSu joint frames are practically coincident
with the a-dT curves associated with frames modeled by means of the TSs and SSs
joint models.
Different trends can be observed, for the different joint laws. In particular:

1. in many cases (40% of the considered frames), the tangent stiffness model (TS)
provides a lateral stiffness greater than that related to the ML model. The greater
flexibility is associated with the SS law, as it appears from Fig. 7, related to the
Cf frame with hinged bases;
2. in other cases (33%), the trend previously observed is associated only with low
values of the load multiplier. Increasing a, the SS joint model provides a stiffer
response than the one associated with the ML joint model (Fig. 8);
3. in a more limited number of cases, the trend of the a-dT curves is quite different
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 287

Fig. 7. Load multiplier-displacement relationships for Cf frame with hinged bases.

from those described at points (1) and (2). An example is in Fig. 9: after the
initial part of the response, which confirms trend (1), the more rigid joint model
TS provides a severe state of stress at some joints. These achieve the plastic
moment prematurely with respect to the joint modelled with other relationships.
As a consequence, the frame response associated with the TS joint law becomes
more flexible. This behaviour is shown in Fig. 10, which presents the state of
stress on frame AC with fixed bases (i.e. the frame considered in Fig. 9) for three
different values of the load multiplier. In particular, it can be noted that for a load
multiplier equal to the half of the ultimate load multiplier (i.e. to 0.5 au) and with
the TS joint law, approximately 33% of the joints are plastic while for the frame
modeled with an SS joint model each joint is elastic. By increasing the load multi-
plier, the number of joints with a bending moment M equal to the ultimate joint
moment Mu increases remarkably for the TS model and in a more limited amount
for the SS model. In correspondence with the ultimate load multiplier, the number
of joints in the plastic range for the two bilinear relationships is, however, equal.

Limited differences can be observed in the initial branch on the frame responses.
Collapse is generally due to the interaction between column flexural instability and
joint plasticity. Columns never achieved their ultimate strength, while in a very lim-
ited number of cases, a plastic hinge occurred in the vicinity of the midspans of some
288 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Fig. 8. Load multiplier-displacement relationships for frame Ea with hinged bases.

beams. However, frame response is significantly affected by the yielding sequence of

joints.

3.1. Serviceability limit state

A lateral displacement at the top of the frame of H/300 and an interstorey drift
of h/200 were assumed as the serviceability limit values, where H is the total frame
height and h represents the frame interstorey. These values were considered in
accordance with current design recommendations for pallet racks [6–8]. A load multi-
plier as,d was, hence, evaluated (Fig. 11) by considering the minimum multipliers
associated with:

앫 the load multiplier corresponding to a maximum horizontal top displacement dT

of H/300 (as(H/300));
앫 the load multiplier associated with a maximum interstorey drift di.d.j of h/200
(as(h/200)).

The case of ultimate limit states governing design was also considered: the service
load multiplier associated with resistance as,R, was evaluated by dividing au for a
safety factor gF. This factor was assumed equal to 1.5, under the assumption that
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 289

Fig. 9. Load multilpier-displacement relationships for Ac frame with fixed bases.

the dead load can be considered negligible, when compared with the rack live load,
in accordance with [6–8].
The service load multiplier, as, was, then, assumed as the minimum value between
as,d and as,R. Table 2 presents, for the cases of hinged and fixed bases, the values
of as associated with the bilinear joint models, non dimensional with reference to
that related to the ML relationship. The limit state governing design is indicated for
each case. It should be observed that:

앫 the symmetry of the joint model affects only moderately the value of the service
load multiplier. Only for frame Cf with fixed bases non negligible differences (up
to 7%) between as(TSu) and as(TSs) can be observed;
앫 in some cases, great differences (up to 28%) are associated with the TS and SS
models, mainly for frames with hinged bases. Generally, as(SS) is lower than
as(TS). If frames with fixed bases are considered, these differences are more lim-
ited and as(SS) is slightly greater than as(TS), due to the influence of plasticity
spreading in the joints of the frame;
앫 the difference between as associated with the simplified joint models and the ML
one are more evident in the case of frames with hinged bases (up to 1.24 for
frame Bb).
앫 in the case of frames with fixed bases, the ultimate limit state governs design,
generally independently from the joint law. Otherwise, in addition to ultimate
290 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Fig. 10. Plasticity in Ac frame with fixed bases for different load multipliers.
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 291

Fig. 11. Definition of the service load multiplier as.

load, also the interstorey drift governs design (51% of the considered cases). The
limitation on the lateral top displacement is seldom associated with the definition
of as.

3.2. Elastic critical load

The elastic critical load represents an index of the influence of the joint behaviour
on the overall response of the racks. Moreover, classification in sway or non sway
frames depends on the ratio between the service load multiplier and the elastic critical
one. The value of the elastic critical multiplier, acr, was determined through a second
order elastic analysis, assuming a linear joint response, characterised by the value
of the initial stiffness. The main results are summarised in Table 3, where the values
292 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Table 2
Influence of the beam-to-column joint modeling on the service load multiplier as and limit condition
governing designa

Frame Fixed bases Hinged bases

as(SSu) as(SSs) as(TSU) as(TSs) as(SSu) as(TSu) as(TSu) as(TSs)

as(ML) as(ML) as(ML) as(ML) as(ML) as(ML) as(ML) as(ML)

Aa 쐌 쐌 쐌 쐌 쐌 쐌 쐌 쐌
0.95 0.95 0.95 0.95 1.00 1.01 0.99 0.97
쐌 쐌 쐌 쐌 쐌 쐌 쐌 쐌

Ab 쐌 쐌 쐌 쐌 쐌 쐌 쐌 쐌
1.07 1.08 1.08 1.07 1.00 1.00 1.06 1.04
䊏 䊏 䊏 䊏 쐌 쐌 쐌 쐌

Ac 쐌 쐌 䊏 䊏 쐌 쐌 쐌 쐌
1.04 1.04 0.92 0.93 1.12 1.12 1.13 1.18
䊏 䊏 䊏 䊏 왖 왖 왖 왖

Ba 쐌 쐌 쐌 쐌 왖 왖 쐌 쐌
1.10 1.09 1.04 1.04 0.92 0.92 1.14 1.15
쐌 쐌 쐌 쐌 쐌 쐌 쐌 쐌

Bb 쐌 쐌 쐌 쐌 왖 왖 쐌 쐌
1.03 1.03 0.98 0.98 0.98 0.96 1.24 1.24
쐌 쐌 쐌 쐌 왖 왖 왖 왖

Bc 䊏 䊏 䊏 䊏 왖 왖 왖 왖
1.00 0.99 0.99 0.98 0.80 0.84 0.80 0.80
䊏 䊏 䊏 䊏 쐌 쐌 쐌 쐌

Ca 쐌 쐌 쐌 쐌 쐌 쐌 쐌 쐌
0.98 1.01 1.00 1.01 1.00 1.00 1.00 1.00
쐌 쐌 쐌 쐌 쐌 쐌 쐌 쐌

Cb 쐌 쐌 쐌 쐌 왖 왖 쐌 쐌
0.98 0.99 1.00 1.01 0.94 0.93 1.05 1.06
쐌 쐌 쐌 쐌 쐌 쐌 쐌 쐌

Cc 쐌 쐌 쐌 쐌 왖 왖 쐌 쐌
0.96 0.96 0.97 0.97 0.89 0.89 1.04 1.04
쐌 쐌 쐌 쐌 왖 왖 왖 왖

Cd 쐌 쐌 쐌 쐌 왖 왖 왖 왖
1.01 1.01 1.01 1.01 0.89 0.89 1.02 1.02
쐌 쐌 쐌 쐌 왖 왖 왖 왖

Ce 쐌 쐌 쐌 쐌 䊏 䊏 쐌 쐌
1.04 1.04 1.03 1.04 0.87 0.87 1.00 1.00
쐌 쐌 쐌 쐌 쐌 쐌 쐌 쐌

Cf 쐌 쐌 쐌 쐌 왖 왖 왖 왖
1.03 1.03 1.10 1.03 0.91 0.91 1.02 1.02
쐌 쐌 쐌 쐌 왖 왖 왖 왖

Da 쐌 쐌 쐌 쐌 왖 왖 왖 왖
1.02 1.02 1.02 1.02 0.93 0.93 1.01 1.02
쐌 쐌 쐌 쐌 왖 왖 왖 왖

Db 쐌 쐌 쐌 쐌 왖 왖 왖 왖
1.01 1.01 1.01 1.01 0.97 0.98 0.98 0.98
쐌 쐌 쐌 쐌 왖 왖 왖 왖

Ea 쐌 쐌 쐌 쐌 왖 䊏 쐌 왖
1.01 1.00 1.01 1.01 0.95 0.94 1.15 1.14
쐌 쐌 쐌 쐌 왖 왖 왖 왖
a
Limit condition governing design: 쐌, ultimate load; 䊏, lateral displacement; 왖, interstorey drift.
Table 3
Influence of the beam-to-column joint modeling on the elastic critical load multiplier acr

Frame Hinged bases Fixed bases

acr(SSu) acr(TSu) acr(SSu) acr(SSs) acr(SSs) acr(TSs) acr(SSu) acr(TSu) acr(SSs) acr(TSs) acr(SSs) acr(TSs)
acr(ML) acr(ML) acr(ML) acr(ML) acr(SSu) acr(TSu) acr(ML) acr(ML) acr(ML) acr(ML) acr(SSu) acr(TSu)

Aa 0.84 1.00 0.87 0.97 1.04 0.97 0.81 1.00 0.84 0.97 1.04 0.97
Ab 0.81 1.00 0.85 0.97 1.04 0.97 0.77 1.00 0.80 0.95 1.05 0.95
Ac 0.76 0.95 0.76 1.00 1.00 1.05 0.62 0.91 0.63 0.97 1.01 1.06

Ba 0.61 0.83 0.61 0.76 1.00 0.92 0.53 0.73 0.53 0.67 1.00 0.92
Bb 0.67 0.89 0.60 0.80 0.90 0.89 0.63 0.86 0.57 0.77 0.91 0.90
Bc 0.94 0.97 0.94 0.97 1.00 1.00 0.94 0.99 0.94 0.96 0.99 0.97

Ca 0.68 0.82 0.68 0.83 1.01 1.00 0.75 0.89 0.74 0.86 1.00 0.97
Cb 0.74 0.89 0.74 0.86 1.00 0.97 0.74 0.89 0.74 0.86 1.01 0.97
Cc 0.72 0.90 0.76 0.87 1.05 0.97 0.76 0.90 0.76 0.87 1.00 0.96
Cd 0.78 0.92 0.80 0.88 1.01 0.96 0.77 0.92 0.78 0.89 1.01 0.97
Ce 0.79 0.92 0.80 0.88 1.01 0.96 0.77 0.92 0.78 0.89 1.01 0.96
Cf 0.82 0.93 0.83 0.92 1.00 0.98 0.79 0.92 0.79 0.89 1.00 0.97
N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Da 0.68 0.83 0.62 0.67 0.92 0.81 0.71 0.88 0.64 0.70 0.91 0.80
Db 0.93 0.93 0.98 0.98 1.06 1.06 0.93 0.93 0.98 0.98 1.05 1.05

Ea 0.76 0.99 0.71 0.83 0.94 0.84 0.72 0.99 0.68 0.84 0.95 0.85
293
294 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

of acr associated with the considered unsymmetrical (SSu and TSu) and symmetrical
(SSs and TSs) M-⌽ relationships are presented; values are adimensionalized with
respect to acr(ML), obtained by considering the multi-linear joint relationship. The
ratios acr(SSs)/acr(SSu) and acr(TSs)/acr(TSu) are also reported, in order to allow
appreciation of the influence of asymmetry.
The main outcomes of the elastic critical analysis can be summarised in the follow-
ing points:

앫 the influence of the joint symmetry is very limited; the use of a simplified sym-
metrical relationship allows determination of acr with a degree of accuracy similar
to that associated with an unsymmetrical relationship; this is possible because of
the limited number of joints under sagging moments at collapse in the sway-mode;
앫 with reference to the sole symmetrical M-⌽ relationships, the choice of the elastic
value for the joint stiffness may remarkably influence the term acr (Fig. 12). The
values of acr associated with the SSs relationship are, in fact, significantly lower
(up to 58%) than the one related to the TSs relationship, which gives the elastic
critical multiplier with a more than satisfactory degree of accuracy. The mean
value of the ratio acr(TSs)/acr(ML) is 0.88 and its standard deviation is 0.092.
Differences are due to the choice of initial stiffness of the ML law adopted: in
order to simulate more accurately the elastic response, either 2 or 3 linear branches
may be used;

Yet, it should be remarked that, generally, a bilinear model based on the secant
stiffness allows to evaluate acr on the safe side.

3.3. Ultimate load carrying capacity

The ultimate load multiplier, au, was determined by means of second order non
linear analyses, already used to assess the service load multiplier (Section 3.1). As
previously mentioned, the collapse condition was due to the interaction between
second order effects and plasticity mainly in beam-to-column joints. The results
related to the ultimate load multiplier, au, are summarised in non dimensional form
in Table 4. They indicate that:

앫 the influence of the joint symmetry is very limited: the mean value of the ratio
between the multiplier associated with the bilinear curve and the one related to
the ML relationship is in the range 0.99–1.04;
앫 the use of the tangent stiffness leads in some cases to a non negligible overestim-
ation of the load multiplier (up to 15%);
앫 the ultimate load multiplier based on the secant stiffness is, on the average, slightly
lower than the one associated with the multi-linear joint relationship (Fig. 13).
The mean value of au(SSs)/au(ML) is 0.99 and its standard deviation is 0.05.

Despite the fact that the M-⌽ curves considered differ remarkably in terms of elastic
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 295

Fig. 12. Influence of the joint model on the elastic critical load multipler acr.
 296

Table 4
Influence of the beam-to-column joint modeling on the ultimate load multiplier au

Frame Hinged bases Fixed bases

au(SSu) au(TSu) au(SSs) au(TSs) au(SSs) au(TSs) au(SSu) au(TSu) au(SSs) au(TSs) au(SSs) au(TSs)
au(ML) au(ML) au(ML) au(ML) au(SSu) au(TSu) au(ML) au(ML) au(ML) au(ML) au(SSu) au(TSu)

Aa 1.00 0.99 1.01 0.97 1.01 0.98 0.95 0.95 0.95 0.95 1.00 1.00
Ab 1.00 1.06 1.00 1.04 1.00 0.98 0.97 0.99 0.99 0.98 1.01 0.99
Ac 1.03 1.04 1.03 1.08 1.00 1.04 0.97 0.94 0.97 0.95 1.00 1.01

Ba 1.05 1.14 1.05 1.15 1.00 1.01 1.10 1.04 1.09 1.04 1.00 1.00
Bb 1.07 1.14 1.05 1.14 0.99 1.00 1.03 0.98 1.03 0.98 1.00 1.00
Bc 1.00 1.03 1.00 1.03 1.00 1.00 0.97 0.98 0.98 0.97 1.00 0.99

Ca 1.00 1.00 1.00 1.00 1.00 1.00 0.98 1.00 1.01 1.01 1.03 1.03
Cb 0.96 1.05 0.97 1.06 1.01 1.01 0.98 1.00 0.99 1.01 1.00 1.01
Cc 0.92 1.00 0.92 0.99 1.00 0.99 0.96 0.97 0.96 0.97 1.00 1.00
Cd 0.93 0.99 0.93 0.99 1.01 1.00 1.01 1.01 1.01 1.01 1.00 1.00
Ce 0.93 1.00 0.93 0.99 1.00 0.99 1.04 1.04 1.04 1.04 1.00 1.00
Cf 0.94 1.01 0.94 1.01 1.01 1.00 1.03 1.10 1.03 1.03 1.01 0.94
N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Da 1.00 1.04 0.98 1.03 0.98 0.99 1.02 1.02 1.02 1.02 1.00 1.00
Db 1.06 1.06 1.06 1.06 1.00 1.00 1.01 1.01 1.01 1.01 1.00 1.00

Ea 1.02 1.06 1.02 1.06 1.00 1.00 1.01 1.01 1.00 1.01 1.00 1.00
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 297

Fig. 13. Influence of the joint model on the ultimate load multipler au.
298 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

stiffness, the influence of the considered joint relationships on the load multiplier
au is generally small.
It should be noted again that in many cases design is governed by limit displace-
ments, as it appears from Section 3.1.

4. Influence of frame imperfections

The main scope of the above analyses was at investigating the influence of beam-
to-column joint modelling on the frame performance with reference to both service
and ultimate limit states. Geometrical imperfections have been considered by means
of an initial out-of-plumb angle of the columns equal to 1/200 rad, constant for
all frames.
Recent codes for steel building design define geometrical imperfections on the
basis of the number of floors (ns) and the number of columns (nc). In rack systems,
the possible initial looseness of beam-end-connectors ⌽l (Fig. 14), due to the absence
of bolts or welds in the connection systems, often induces a larger lateral deformation
of the frame in correspondence of very limited values of the vertical loads, and,
consequently, increases second order effects.
An accurate definition of the out-of-plumb angle ⌽imp has been recently proposed

Fig. 14. Definition of the initial looseness of beam-end connectors.

 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 299

in [8] by considering also the effect due to the initial looseness ⌽l. In particular, the
term ⌽imp is evaluated by means of the following equation:

冪2+n 冪5+n (2⌽ ⫹⌽ )

1 1 1 1
⌽imp⫽ s 1 (2)
c s

where nc is either the number of columns in the down-aisle direction or that of

connected frames in the cross-aisle direction, ns is the number of beam levels, ⌽s
is the maximum specified out-of-plumb divided by the height (assumed in these
analyses as 1/200 rad) and ⌽l represents the experimental looseness of the beam-
end-connector (Fig. 14).
Some of the frames previously considered with an imperfection column angle
equal to 1/200 rad (i.e. equal to the considered ⌽s value) have been re-analysed with
⌽imp accordingly with Eq. (2)) and by considering both the ML and SSs laws to
reproduce the joint behaviour. These frames are listed in Table 5, together with the
values of ⌽l, determined from beam-end-connector tests, and the value of ⌽imp evalu-
ated by assuming for ⌽s =1/200 rad. In the same table, the values of both service
and ultimate load multipliers associated with frame imperfection ⌽imp (as(⌽imp) and
au(⌽imp), respectively) are reported in non dimensional form with reference to the
corresponding multipliers associated with ⌽s. For the two aforementioned joint laws,
both the cases of hinged and fixed bases have been considered. The limit condition
governing design is also specified.
It should be noted that the term ⌽imp is generally significantly greater than ⌽s.
The mean value of the ratio ⌽imp/⌽s is 1.74, the upper extreme is up to 3.6 (frame
Db). The influence of the imperfections on the frame performance seems, however,
quite limited. This is more evident with reference to the service load multiplier. As
to the ultimate load multiplier, the use of term ⌽s is associated with an overestimation
of load multiplier au(⌽imp). This overestimation is not greater than 27.5% (with a
mean value of 10.9% and a standard deviation 6.5%).
The use of the SS joint laws in case of frames with fixed bases leads to a moderate
overestimation of the design multiplier evaluated using the ML joint law mainly with
reference to as. If frames with hinged bases are considered, αs(⌽imp) with ML law
is slightly lower than as(⌽imp) with the SSs law, while differences associated with
the evaluation of the ultimate load multiplier are very modest.

5. Influence of the base-plate joint

Because of the absence of experimental data on base-plate connections, the results

related to joint tests presented in the FEM document [13] have been used to assess
the influence of the behaviour of the base-plate connections on the overall frame
response.
The FEM frame (Table 1) has been considered and a bilinear symmetrical relation-
ship, the stiffness of which is secant to the experimental curve, has been assumed
to simulate the response of beam-to-column joints. Also for the base-plate connec-
300 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Table 5
Influence of the frame imperfections on both serviceability and ultimate load multiplier and limit condition
governing designa

Frame ⌽s ⌽imp Fixed bases Hinged bases

ML law Sss law ML law Sss law

as(⌽imp) au(⌽imp) as(⌽imp) au(⌽imp) as(⌽imp) au(⌽imp) as(⌽imp) au(⌽imp)

as(⌽s) au(⌽s) as(⌽s) au(⌽s) as(⌽s) au(⌽s) as(⌽s) au(⌽s)

Aa 1/200 1/120 0.92 䊏 0.95 0.95

쐌 0.95 0.82
왖 0.94 0.78
왖 0.90
쐌 쐌 쐌 쐌

Ab 1/200 1/120 0.86䊏 0.94 0.92

쐌 0.92 0.84
왖 0.93 0.80
왖 0.93
䊏 쐌 쐌 쐌

Ac 1/200 1/156 0.79 䊏 0.93 0.93

쐌 0.93 0.82
왖 0.97 0.76
왖 0.92
䊏 쐌 왖 쐌

Ca 1/200 1/144 0.96 쐌 0.96 0.95

쐌 0.95 0.98
쐌 0.98 0.97
쐌 0.97
쐌 쐌 쐌 쐌

Cb 1/200 1/144 0.94 쐌 0.94 0.94

쐌 0.94 0.87
왖 0.93 0.83
왖 0.93
쐌 쐌 쐌 왖

Cc 1/200 1/133 0.93 쐌 0.93 0.91

쐌 0.91 0.80
왖 0.91 0.79
왖 0.93
쐌 쐌 왖 왖

Cd 1/200 1/126 0.90 䊏 0.96 0.81

䊏 0.90 0.75
왖 0.91 0.74
왖 0.91
쐌 쐌 왖 왖

Ce 1/200 1/115 0.86 쐌 0.97 0.75

䊏 0.94 0.66
왖 0.89 0.71
왖 0.89
쐌 쐌 䊏 䊏

Cf 1/200 1/115 0.80 䊏 0.96 0.71

䊏 0.90 0.71
왖 0.90 0.71
왖 0.91
쐌 쐌 왖 왖

Db 1/200 1/58 䊏 0.96 䊏 0.96 왖 0.81 왖 0.80

0.81 0.78 0.43 0.44
쐌 쐌 왖 왖
a
Limit condition governing design: 쐌, ultimate load; 䊏, lateral displacement; 왖, interstorey drift.

tions, a bilinear secant M-⌽ relationship has been adopted. Reference was made to
a stiffness value of 0.15 kNm/mrad and to a base joint resistance of 0.639 kNm, in
accordance with the FEM document. These values are mentioned in the following
as B* and Mb*, respectively. A parametric study has been carried out by varying
either the base stiffness (B), expressed as a fraction of B* or the base resistance
(Mb), expressed as a fraction of Mb*.
Tables 6 and 7 present the most significant data for this parametric analysis in
terms of load multiplier associated with second order non linear analysis for service
as well as ultimate limit state.
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 301

Table 6
Service and ultimate load multiplies for different values of the stiffness (B) of base-plate jointa

B (kNm/mrad)
as as(B∗) au(B∗)
au as(B) au(B)

0 0.000 0.64 왖 1.60 1.55

B*/50 0.003 0.65 왖 1.52 1.49
B*/10 0.015 0.66 왖 1.32 1.30
B*/5 0.030 0.67 쐌 1.17 1.17
B*/2 0.075 0.67 쐌 1.06 1.06

B* 0.150 0.67 쐌 1.00 1.00

2B* 0.300 0.67 쐌 0.95 0.95

5B* 0.750 0.67 쐌 0.92 0.92
10B* 1.500 0.67 쐌 0.90 0.90
50B* 7.500 0.67 쐌 0.90 0.90
100B* 15.000 0.67 쐌 0.90 0.90
⬁ ⬁ 0.67 쐌 0.89 0.89

a
Limit condition governing design: 쐌, ultimate load; 䊏, lateral displacement; 왖, interstorey drift.

Table 7
Service and ultimate load multiplies for different values of the moment resistance (Mb) of base plate jointa

Mb (kNm)
as as(Mb∗) au(Mb∗)
au as(Mb) au(Mb)

0 0.000 0.64 왖 1.60 1.54

Mb*/50 0.013 0.67 쐌 1.54 1.54
Mb*/10 0.064 0.67 쐌 1.50 1.50
Mb*/5 0.128 0.67 쐌 1.46 1.46
Mb*/2 0.320 0.67 쐌 1.30 1.30

Mb* 0.639 0.67 쐌 1.00 1.00

2Mb* 1.278 0.67 쐌 0.91 0.91

5Mb* 3.195 0.67 쐌 0.91 0.91
⬁ ⬁ 0.67 쐌 0.91 0.91

a
Limit condition governing design: 쐌, ultimate load; 䊏, lateral displacement; 왖, interstorey drift.
302 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

For the elastic base stiffness (Table 6), influence of base-plate connection on the
overall frame response is remarkable and slightly greater if service limit state are
considered. The limit state governing design is associated with resistance, except for
values of the base stiffness lower than B*/10 for which the lateral deflection influ-
ences the service load multiplier. The values considered for B* increase the ultimate
load multiplier by up to 55% with reference to the frame restrained by base hinges.
However, the increase in au is remarkable also with limited stiffness values. For
instance, with B=0.1B* the increment of au is 19%. Fig. 15 presents the relationship
between au and as and B. A stiffness value of 10B* practically allows reaching of
the ultimate load with an embedded base (i.e., with B=⬁).
As to the influence of the column base resistance (Table 7), the limit condition
governing design is associated with resistance, except for the hinged bases. Fig. 16
presents the relationship between load multipliers (as and au) and base resistance
Mb. The influence of the base joint resistance appears non negligible, especially with
reference to the service load multiplier. Even a very limited resistance increases
remarkably the global frame performance, with reference to the case of hinged bases.

6. Conclusions

A numerical study on the analysis and behaviour of steel storage pallet rack sys-
tems is summarised in the paper.

Fig. 15. Influence of the base-plate joint stiffness on the service and ultimate load multiplier.
 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304 303

Fig. 16. Influence of the base-plate joint resistance on the service and ultimateload multiplier.

As a general remark and on the basis of an extensive experimental analysis, joints

should be, in most cases, modelled as hinges, if classified in accordance with Euroc-
ode 3 criteria. However, the actual response of beam-end-connectors provides a non
negligible degree of lateral stiffness to the frame. As a consequence, a semi-continu-
ous frame model is always suggested for a more refined and “optimal” design analy-
sis.
Elastic second order linear and non linear analyses have been carried out on several
planar frames representative of pallet racks commonly used in Italian and European
practice. Attention has been mainly paid to the influence of joint modelling. The
analyses have pointed out a relatively low sensitivity of the rack “ultimate” capacity
in comparison with that at the service condition. The need for including the additional
sway associated with the joint lack of fit when expressing geometrical sway imper-
fections in the design approach has been stressed, despite the limited influence on
the frame behaviour in comparison with the one associated with a constant out-of-
plumb angle of 1/200 rad.
Finally, a parametric investigation confirms the remarkable influence of the base-
plate joint on the overall rack response and, as a consequence, the need for test data
related to the behaviour of the column base restraint.
304 N. Baldassino, C. Bernuzzi / Thin-Walled Structures 37 (2000) 277–304

Acknowledgements

The data related to the pallet rack frames have been kindly supplied by 5 Italian
Companies involved in the activities of ACAI-CISI (Italian Association of Steel
Constructors, Rack Manufacturing Companies Group).
The authors wish to thank Mr. Camillo Berardi for help in the preparation of
figures and tables.

References

[1] Godley MHR. Storage racking, chapter 11. In: Rhodes J, editor. Design of Cold Formed Steel
Members, 1991 Elsevier Applied Science, Amsterdam p. 361–99.
[2] Hancock GJ. Distorsional buckling of steel storage rack columns. Journal of Structural Engineering
ASCE 1985;111(12):2770–83.
[3] Davies JM, Jiang C. Design for distorsional buckling. Journal of Constructional Steel Research
1998;46:1–3.
[4] Markazi FD, Beale RG, Godley MHR. Experimental analysis of semi-rigid boltless connectors. Thin-
Walled Structures 1997;28(1):57–87.
[5] RAL. Storage and Associated Equipment, RAL Deutsches Institut fur Gutersicherung und
Kennzeichnung (German Institute for Quality Assurance and Marketing), 1990.
[6] AS. Steel Storage Racking AS4084, Australian Standards, 1993.
[7] RMI. Specification for the Design, Testing and Utilization of Industrial Steel Storage Racks, Rack
Manufactures Institute, 1997.
[8] FEM. Reccomandation for the Design of Steel Pallet Racking and Shelving, Section X of the Feder-
ation Europeenne de la Manutention, 1997.
[9] Baldassino N, Bernuzzi C, Zandonini R. Experimental and Numerical Studies on Pallet Racks, Pro-
ceeding of the Conference “Professor Otto Halase-Memorial Session”, Technical University of Bud-
apest, TUBudapest Publ. (to appear) 1999.
[10] Baldassino N, Bernuzzi C, Zandonini R, Hancock G. Overall, local and distortional buckling in
pallet racks, Structural Stability Research Concil Conference, Atlanta USA, September 1998.
[11] Bernuzzi C, Baldassino N. Experimental Analysis on Stub Columns with Perforations for Steel
Storage Pallet Rack Systems, in preparation, 1999.
[12] Bernuzzi C, Baldassino N. Experimental Analysis on Beam-End Connectors for Steel Storage Pallet
Rack Systems, in preparation, 1999.
[13] FEM. FEM 10.2.02 — Design Example, Section X of the Federation Europeen de la Manuten-
tion, 1997.
[14] FEM. Reccomandation for the design of Steel Pallet Racking and Shelving, Section X of the Feder-
ation Europeenne de la Manutention, 1986.
[15] UNI ENV, 1993-1-1. Eurocode 3: EC3, Design of Steel Structures — General Rules and Rules for
Buildings, European Commitee for Standardization.
[16] Bernuzzi C, Zandonini R. Serviceability and Analysis Models of Steel Buildings. IABSE — Inter-
national Colloquium on Structural Serviceability of Buildings, Goteborg, Sweden, June 1993:195–
200.
[17] Galea Y, Bureau A. PEP MICRO — user’s manual — version 2.0. Centre Technique Industriel de
la Construction Metallique (CTiCM), Saint Remy-les-Chevreuse France, 1991.