Journal of Constructional Steel Research 128 (2017) 672–686

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Performance of mono-symmetric upright pallet racks under

slab deflections
Claudio Bernuzzi, Marco Simoncelli ⁎, Marcello Venezia
Department of Architecture, Built Environment and Construction Engineering, Politecnico di Milano, Milano, Italy

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

Article history: Goods and products are often stored in framed systems, such as pallet racks, which are used for industrial and
Received 23 May 2016 commercial activities. Recently, pallet rack provisions have been significantly improved but there are still rele-
Received in revised form 30 September 2016 vant aspects that need urgent attention for guaranteeing safe structural design. Among them, the interaction be-
Accepted 3 October 2016
tween the rack and the supporting surface (i.e. concrete foundation or floor slab) plays a non-negligible influence
Available online xxxx
in routine design, owing to the high degree of redundancy of the skeleton frames. In addition, rack design is usu-
Keywords:
ally carried out using analysis packages unable to capture the behaviour of mono-symmetric cross-section mem-
Steel storage pallet racks bers and hence all the warping effects are generally ignored.
Flexible/rigid floor slab This paper summarizes a study on the influence of floor displacements on the performance of medium-rise steel
Structural analysis storage pallet racks. In order to obtain a reasonably wide range of data of practical interest for routine design, nu-
Relative base settlements merical analyses have been carried out by varying several key parameters such as the position of the rack on the
Design approaches floor slab, the rack geometry, the floor span length and the degree of rotational stiffness of both beam-to-column
Mono-symmetric cross-section uprights and base-plate connections. Moreover, attention has been focused on the influence of the accuracy of the struc-
tural model used in the design and two commercial analysis packages, differing for the degree of refinement of
the implemented finite element beam formulation, have been considered. Finally, a suitable equation is present-
ed to estimate the reduction of performance with respect to the ideal case of racks supported by rigid floor slab.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction
Goods and products are generally stored in industrial framed sys-
tems (racks) comprised of thin-walled cold-formed members, which
represent the best solution from the logistic and economic points of
view, owing to the high strength-to-weight ratio [1,2]. Depending on
the storage density needs, in-service accessibility and picking modali-
ties, several types of commercial racks are available on the market.
The most common of them are the so-called selective pallet racks
(Fig. 1), the focus of the present paper, which are characterized by hav-
ing pallets that are always accessible, independent of the storage
sequence.
These structures are generally comprised of a set of two vertical col-
umns (uprights), often having mono-symmetric sections, that are con-
nected to each other by means of lacings to form trussed (built-up
laced) members. These are identified as upright frames, which are
placed in the cross-aisle (transversal) direction. Stored units are sup-
ported by pairs of pallet beams that are attached to two adjacent upright
frames and the pallet access is in the down-aisle (longitudinal)
⁎ Corresponding author.
direction, where in general the vertical bracings cannot be placed be-
cause of the need to optimise the number of pallet unit locations.
From the structural point of view, racks are always braced in the
cross-aisle direction. In the down-aisle direction they behave like
semi-continuous unbraced (moment-resisting) frames [3], where sta-
bility to lateral loads is provided solely by the degree of flexural continu-
ity associated with beam-to-column joints and base-plate connections.
Routine design is carried out with reference to recently updated pro-
visions in Europe [4–6], the United States [7,8] and Australia [9]. Despite
the recent updates, several aspects playing a key role in design need ad-
ditional investigations and two of them, strictly related to each other,
are considered in the present paper. The first is the assessment of the ef-
fective load carrying capacity if vertical displacements occur at the up-
right bases. In the case of flexible foundations or deformable floor
slabs, due to the high redundancy of framed systems, the relative settle-
ments are expected to have a non-negligible influence on the rack per-
formance, but code requirements dealing with the rack-base interaction
are either lacking or extremely poor. The second aspect needing urgent
attention is related to the approaches used for routine design, which are
inadequate for capturing actual rack response due to fact that they
derive strictly from approaches proposed for carpentry steel frames, i.e.
structural systems comprised of bi-symmetric cross-section members.

http://dx.doi.org/10.1016/j.jcsr.2016.10.004
0143-974X/© 2017 Elsevier Ltd. All rights reserved.
 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686
Fig. 1. Typical adjustable steel storage pallet-racks.
In a previous paper [10], the effect of the base displacement on the
reduction of the pallet rack performance was discussed with reference
to racks comprised of bi-symmetric cross-section uprights. Now, atten-
tion is focussed on racks with mono-symmetric uprights whose behav-
iour is significantly affected by warping torsion, Wagner coefficients
and the non-coincidence between the shear centre and the cross-
section centroid, as shown by Teh et al. [11]. Despite the relevant studies
carried out on storage rack systems in the last years, in both numerical
and experimental fields [12–21], these aspects are still neglected in
the European routine design, leading in many cases to an overestima-
tion of the rack performance [22]. The proposed research outcomes
are based on a numerical parametric investigation, which has been de-
veloped based on cases differing in rack location on the floor slab, down-
aisle rack layout, upright geometry and the degree of flexural continuity
associated with beam-to-column joints and base-plate connections. The
direct evaluation of the reduction of the load carrying capacity due to
vertical floor deflections has been based on the cases of rigid and de-
formed slabs. The associated design calculations have been developed
with reference to two different European design procedures, with one
leading to the most conservative evaluation of rack performance and
the other leading to the least conservative [23,24]. Furthermore, two
different FE beam formulations have been considered: the first is tradi-
tionally, but incorrectly, used by manufacturing engineers for rack
design and the second is specifically developed for non bi-symmetric
beams and hence adequate to investigate the behaviour of these struc-
tures. Finally, an approximate equation, suitably accounting for the
presence of mono-symmetric cross-section uprights, is proposed for a
direct assessment of the reduction of rack performance, which has
been validated with reference to pallet racks that are significantly differ-
ent from the ones adopted in the numerical analysis.
673
2. Standard provisions and rack base movements
The cases of interest considered herein, are representative of selec-
tive pallet racks having heights up to 6–8 m and with a limited number
of bays, erected inside a multi-storey building and directly connected to
the floor slab. Generally, these racks are located in areas where the pub-
lic has access, such as supermarkets, discount stores and home handy-
man stores. A collapse could therefore result in the loss of stored
goods, injuries and potentially the loss of human life. The most common
contractual situation is that the clients and/or owners locate the rack on
an intermediate floor, without any structural check regarding the ade-
quacy of the supporting slab. Buildings are usually designed and erected
without any interaction between the building designer and the rack
manufacturing engineer, despite the fact that the EN15512 provisions
[4] state that the limiting deflection values have to be agreed with the
client/owner, taking into account the specific requirements related to
the installation. Otherwise, in-service deformability values are specified
only for pallet beams in terms of allowable vertical deflection and for
twist angles when the load is not applied to the shear centre of the
cross-section. Up to now, adequate attention does not seem to have
been paid to the floor tolerances and deformations. These are covered
by clause 5.1.5, which states that the flatness deviations and deforma-
tions of the building floor upon where the rack should be installed
may be ignored when the building floor is designed according to the rel-
evant limit values specified in EN 15620 [25]. As to these guidelines,
point 7.3.1 is the only part dealing with slab deformations due to settling
and slab deflections due to vertical loads, but no useful indications are
provided to the designers. It is stated that deflection of the floor slab results
in additional stresses and inclination of the rack structure and can be consid-
erable. Furthermore, the code also recommends that 1) the deflection of
the floor slab shall be included at the planning stage and information shall
be provided by the specifier or client to the racking supplier for evaluation
of the additional stresses in the racking and 2) the deflection of the floor
slab shall be included at the planning stage by the specifier or client and
added to the clearances and deformations as required for the specific project.
As an alternative, the floor slab could be treated as quasi-rigid, which
implies that its deformations never affect the set of displacements, in-
ternal forces and moments. As reported by the code, this can be as-
sumed if the following conditions are satisfied:
• the angular rotation at any location of the floor slab within the rack area
shall not exceed 0.5 mrad;
• the overall vertical deformation shall not exceed 1/2000 of the total stor-
age length;
• the overall vertical deformation shall not exceed 1/2000 of the total stor-
age width.
These statements rarely appear useful to manufacturing engineers
from a practical point of view, which know only general data of the
supporting surface, such as the floor span (LFS). The limitation on the an-
gular slab rotation implies an accurate knowledge of the response of the
flooring system. Very complex activities should be made to evaluate
correctly the floor displacement effects, on the rack internal forces and
moments, especially in the case of existing buildings. As a practical re-
sult, these further structural checks could appear as complex and
quite expensive if compared with the cost of the rack system itself and
usually are never included in the design budget. Furthermore, it is
clear that the condition of ‘quasi-rigid’ applied to the floor slabs of
multi-story buildings reflects a very severe design criterion associated
with serviceability limit states. To satisfy this condition (limiting deflec-
tions to LFS/2000) instead of the one [26] usually adopted for the floor
slab in building design (LFS/300) a significantly larger flexural stiffness
(approximately at least 7 times) is required. This implies very thick,
heavy and expensive slabs, needing for appropriate calculations, strictly
depending on the selected rack layout. As an alternative, it seems more
674 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686
reasonable to reduce the rack load-carrying capacity with respect to the
limit condition of a rigid floor slab, this is the approach that is more ap-
propriate for achieving a safe design but very rarely adopted by rack
manufacturer engineers. In a previous study on this topic [10], the
case of racks comprised of bi-symmetric cross-section uprights has
been deeply investigated. Now attention is focussed on uprights having
cross-sections with only one axis of symmetry, in order to consider also
the influence of warping torsion and the associated effects.
3. The numerical study
A parametric investigation has been carried out by considering cases
of practical interest for routine design. To this end, attention has been
focussed on medium-rise adjustable pallet racks, which during the in-
service life could vary their geometric layout in terms of number of
load levels and location on the supporting floor slab. In order to propose
remarks of practical interest, a double-entry rack configuration with six
bays of equal spans (Lb = 2.78 m) has been considered (Fig. 2), which is
the layout typically considered by rack designers.
Three lipped channels, identified as MM_, DD_ and TT_ types, have
been selected as uprights (Table 1). It is worth noting that despite the
fact that these cross-sections could appear slightly different to those
used in industrial storage frames, the presence of one axis of symmetry
guarantees that research outcomes are adequately representative and
useful for rack design. Table 1 reports, for each of these cross-sections,
the geometry and the values of the area (A), second moments of area
(Iy and Iz) and section moduli (Wy and Wz). The eccentricity (ys) be-
tween the shear centre and the cross-section centroid is also reported,
together with the term β, which is defined as:
Iy þ Iz
β¼ : ð1Þ
Iy þ Iz þ A  y2s
For beams and lacings, 100 × 50 × 3 mm rectangular hollow section
(RHS) and 30 × 30 × 3 mm square HS members have been used, respec-
tively. It has been assumed that all rack components belong to class 3 ac-
cording to the classification criteria reported in Eurocode 3 [27], hence
being characterized by the absence of any kind of post-elastic cross-
section strength resources, as well as of local and/or distortional buck-
ling phenomena. The numerical analysis has been focussed on the fol-
lowing parameters:
• the frame configuration: each of the three uprights has been associated
with the upright frame layout presented in Fig. 2. To account for the
Fig. 2. Cross-aisle (a) and down-aisle (b) views of the considered racks.
possibility that the owner's needs should change in terms of weight
and sizes of the stored pallets, two configurations in the down-aisle
direction have been considered for each upright type, differing in
terms of the number of load levels (nLL) and for the inter-story height
(hLL): four load levels (_4LL) with hLL = 1.8 m and five load levels
(_5LL) with hLL = 1.5 m;
• the degree of flexural stiffness of the beam-to-column joints: several
values of the joint stiffness Sj,btc have been expressed as multiples
(by means of term ρj,btc) of a reference stiffness value SEC3−LB
j ,btc via the
equation:
S j;btc ¼ ρ j;btc  SEC3−LB
j;btc ð2aÞ
− LB
where SEC3j , btc is the rotational stiffness associated with the lower
bound of the semi-rigid domain (corresponding to the transition be-
tween the flexible and semi-rigid joint regions) based on the classifica-
tion criteria of part 1–8 of EC3 [28] and defined as:
EIb
SEC3
j;btc
LB
¼ 0:5 ð2bÞ
Lb
where terms Lb and Ib are the length and the second moment of area of
the beam, respectively. Parameter ρj,btc has been assumed to range from
1 to 5, which is the field of interest for beam-to-column joints typically
adopted in medium-rise pallet racks [29]. In addition to these values, al-
so the values of 1.5, 2, 3 and 4 have been considered. It can be noted that
lowest values are related to rack standard applications while the higher
ones can be reached via the use of bolts, increasing joint performance in
seismic zones. Joint bending resistance has been considered equal to
0.40 times the flexural resistance of the beams;
• the degree of flexural stiffness associated with base-plate connections: at-
tention has been focused on two different types of base-plate connec-
tions, i.e. fixed and semi-rigid bases with a moderate degree of
flexural continuity. In the latter case, the stiffness has been assumed
equal to 0.15 times the stiffness at the boundary between the semi-
rigid and rigid joint regions for base-plate connections, i.e. SEC3−UB
j;base ¼3
EI
0 hLLy (where hLL is the height of the first level and Iy is the second mo-
ment of area of the cross-section). For these cases, joint bending resis-
tance has been considered 0.30 times the flexural resistance of
uprights;
 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686 675

Table 1
Main geometric data related to the considered cross-section uprights.

A [mm2] 5.63 · 102 7.07 · 102 8.38 · 102 4.80 · 102

Wy [mm3] 13.70 · 103 24.03 · 103 35.09 · 103 10.88 · 103
Wz [mm3] 12.65 · 103 17.33 · 103 21.63 · 103 10.29 · 103
Iy [mm4] 50.99 · 104 136.87 · 104 261.42 · 104 38.09 · 104
Iz [mm4] 48.93 · 104 66.64 · 104 83.54 · 104 36.81 · 104
It [mm4] 0.089 · 104 0.11 · 104 0.13 · 104 0.06 · 104
Iw [mm6] 1509.81 · 106 5173.80 · 106 11,400.00 · 106 1082.52 · 106
ωmax [mm2] 42.78 · 102 72.08 · 102 90.02 · 102 43.19 · 102
ys [mm] 80.3 82.2 84.2 75.8
β 0.216 0.299 0.367 0.214

• the span of the floor slab: four lengths (LFS) of the floor slab have been
considered: 12 m, 18 m, 24 m and 30 m. The structural scheme of a
simply supported floor slab has been considered, or equivalently, of
a slab with other restraint conditions but characterized by the same
distance between two subsequent inflection points. It is worth noting
that the proposed research outcomes can be immediately addressed
also to multiple span beams by considering for them the distance be-
tween two contiguous contra-flexure points;
• the location of the racks on the floor slab (Fig. 3): racks were considered
in the centre of the floor or eccentrically located with respect to the
floor centre (identified in the following as ‘lateral’ location), except
for LFS = 12 m, for which only the central case was considered. It is
worth clarifying that the slab serviceability check is always carried
out with reference to the accurate position of the rack. As a conse-
quence, for the lateral position, the slab has a significant lower bend-
ing stiffness.
Racks have been assumed to be fully loaded, simulating the pallet
weight (live load) via a uniformly distributed load of 4 N/mm on each
pallet beam. Frame (sway) imperfections have been accounted for by
means of an out-of-plumb upright angle equal to 3 mrad, modeled via
suitable notional horizontal loads, which are applied in the locations
corresponding of each load level in both the cross-aisle and down-
aisle directions.
To account for the influence of the effects associated with the pres-
ence of mono-symmetric cross-section members, two different com-
mercial FE analysis packages have been used. One of them, offering a
traditional 6DOFs FE beam formulation [30], is very efficient in simulat-
ing the response of the traditional carpentry steel frame. In the other
[31], a refined FE beam formulation including the cross-section warping
as an additional (7th) degree of freedom (DOF) for each node allows for
modeling structures made of lipped channels, like the ones use in the
considered racks. Owing to the high degree of redundancy of the rack
frames, both the conditions of rigid and deflected floor slab have been
considered: in the first case, the upright bases have been assumed to
be perfectly horizontal and lying on the same plane, in accordance
with the routine manufacturing design practice.
As to the deformed condition, the floor system has been assumed to
be simply-supported on the building skeleton frame; for each value of
the floor span (LFS) and for each rack location, reference has been
made always to the maximum permitted deflection for variable loads,
i.e. LFS/300. The associated set of relative vertical displacements δbase
(differential settlements) has been imposed under upright bases (joint
load displacement option), which is depicted in Fig. 4 for the cases of
central (solid line) and lateral (dashed line) rack locations. Grade S355
steel [32] was considered for each rack component.
The layout summarizing key features of this study is presented in
Fig. 5. Attention has been focused on 72 racks, modeled by means of
the two aforementioned FE software packages [30,31] and two different
design approaches (RAM and GEM), which are described in the next
section, have been used for the verification checks. As already intro-
duced, the conditions of rigid and flexible floor slab have been consid-
ered with the racks in different positions, for a total of 2304 design
cases.
4. The considered design approaches
A very important open question concerning the design of racks, as
well as the design of steel framed structures, according to European pro-
visions is related to the choice of the method of analysis. Owing to the
great amount of data arising from the present parametric study and to
the need to identify clearly the research outcomes, attention has been
focussed on solely the uprights, neglecting all verification checks re-
quired for joints and for pallets beams. As clearly shown with reference
to traditional carpentry steel buildings [33] as well as to pallet racks
comprised of bi-symmetric cross-section uprights [23], it is possible to
obtain remarkably different estimates of structural performance, de-
pending on the selected design approach. It is worth mentioning that
the general scope of the present paper is to investigate the influence
of the vertical floor deflection and the interaction with the warping

Fig. 3. Locations of the racks on the floor slab: a) centrally and b) laterally located racks.
676 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686

Fig. 4. The set of vertical displacements (differential settlements) imposed at the upright bases.

torsion effects, independently of the analysis method. Consequently, the safety factor, subscript eff is related to the effective cross-section, and
rigorous analysis method (RAM) and the general method (GEM) have BRd is the resisting bimoment, which is defined as:
been considered, which lead to the lowest and highest values of the
load carrying capacity, respectively. Iw  f y
BRd ¼ ð3bÞ
In this study, racks have been modeled as spatial structures and ωmax  γM
hence uprights are subjected to axial load (NEd) bimoment (BEd) and
to bending moments along the principal cross-section axes, due to the where, with reference to the gross cross-section, Iw is the warping con-
flexural continuity of beam-to-column joints (My,Ed) in the down-aisle stant and ωmax is maximum value of the sectorial area.
direction and to the eccentricity of connections of the bracing members For a deeper understanding of the current paper, especially for what
forming upright frames (Mz,Ed) in the cross-aisle direction. Owing to the concerns warping influence, Eq. (3a) has been proposed in a form ap-
use of mono-symmetric cross-sections, warping torsion along the up- propriate to distinguish clearly the contribution due to the traditional
rights is reflected in the additional state of normal and tangential stress- design approach neglecting warping (subscript 6) from the one due to
es due to the bimoment, which has to be adequately accounted for in the bimoment.
the design phases and can be captured only by using a 7 DOFs FE
beam formulation. 4.1. The rigorous analysis method (RAM)
As to the resistance checks, which always have to be conducted, ref-
erence can be made to the safety index (SI7Rex) expressed by the equa- The Rigorous Analysis Method (RAM) is proposed in EN15512 provi-
tion: sions [4] and derives strictly from one of the design alternatives admitted
0 1 in EC3-1-1 [27]. It requires a second-order analysis taking into account
  the lack-of-verticality imperfections and neglecting the out-of-
B N M y;Ed M z;Ed C
B Ed C BEd straightness member imperfections. With reference to upright stability
SI 7Rex ¼B þ þ Cþ
@Aeff  f y W eff ;y  f eff ;y W z  f y A BRd checks, the following condition, which defines SIRAM, has to be fulfilled:
γM   γM γM ! !
6 BEd NEd ky My;Ed
¼ SIRex þ ≤1 ð3aÞ j
SI RAM ¼ þ
BRd χ min Aeff f y =γ M W eff ;y f y =γ M
!
kz Mz;Ed
where A and W indicate the area and the section modulus of the effec- þ ≤1 ð4Þ
W eff ;z f y =γM
tive cross-section, fy is the material yield strength, γM is the material

Fig. 5. Layout of the considered design cases.

 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686
where, in addition to the symbols already presented, coefficients ky and kz
are the appropriate factors accounting for the bending moment distribu-
tion and the term j identifies whether the internal forces have been
assessed using a 6 or 7 DOFs analysis.
As to stability, EN15512 declares that the structure shall be considered
a no-sway frame and buckling lengths shall be put equal to system (geo-
metric) lengths. It appears clearly that this is an extremely poor stability
check, with the effective length independent of the geometric charac-
teristics of the cross-section of the upright and on the degree of rota-
tional stiffness of joints. In the case of mono-symmetric cross-section
members, a flexural-torsional buckling mode generally governs design
and the code provides directly the equations to evaluate the buckling
loads for isolated members.
4.2. The general method (GEM)
Eurocode 3 in its part 1-1 [27] permits an innovative [34–36] design
approach, the so-called general method (GEM), for structural compo-
nents having geometric, loading and/or supporting irregularities.
Recently, its extension to thin-walled cold formed members for indus-
trial storage systems [37–38], has been proposed and validated via a
wide range of numerical simulations of cases of interest for practical
design. Overall buckling resistance is verified when:
j non-negligible reduction of the rack performance is associated with the in-
χ op α ult;k 1
¼ ≥1
γM j
SI GEM
where αjult,k is the minimum load multiplier based on the cross-section
resistance, χop is the buckling reduction factor related to the overall
structural system and γM is the material safety factor.
For routine rack design, the use of thin-walled uprights guarantees
that plastic hinges do not form in the members. Overall failure is gener-
ally due to the interactions between upright instability and plasticity in
joints, as confirmed also by recent experimental research, comprised of
several full-scale pushover rack tests [39–40].
The reciprocal of the ultimate load multiplier for resistance, αjult, k is
derived from the condition:
     
1 NEd My;Ed Mz;Ed BEd 1 BEd
¼ SI7Rex ¼ þ þ þ ¼ 6 þ
α 7ult;k NRk M y;Rk M z;Rk BRk α ult;k BRk
where NRk is the squash load and My,Rk and My,Rk are the first yielding
moment along y- and z-axis, respectively.
The reduction factor χop depends on the value of the relative slen-
derness λop of the whole structure and on the minimum elastic buckling
multiplier (αjcr,op), which is defined as:
vffiffiffiffiffiffiffiffiffiffiffiffi
u j
u α ult;k
λop ¼t j :
α cr;op
As already discussed, this design option appears to be highly efficient
when applied to members having mono-symmetric cross-section be-
cause: i) the term χop is based on the effective overall buckling mode,
including also the overall flexural-torsional mode of the rack, and ii)
the resistance multiplier α7ult, k accounts adequately for the presence of
the cross-section bimoment, otherwise ignored.
5. Base displacement influence
The influence of the slab floor displacements has been directly ap-
praised by herein making reference to the increment of the safety
index (SI) of the more highly stressed upright. It corresponds to the re-
ciprocal of the reduction of the load carrying capacity and can be
677
expressed by the ratio between the SI associated with the flexible
floor (SIFF RF
k ) over the one related to a rigid floor (SIk ):
SI FF
ΔSI kj ¼ k
ð8Þ
SI RF
k
where the subscript k identifies alternatively the RAM or the GEM ap-
proach and the superscript j is a number (6 or 7) related to the use of
a 6 or 7 DOFs FE beam formulation.
5.1. The RAM approach application
As already mentioned and demonstrated in refs. [23,24], the RAM
approach applied to medium-rise pallet racks leads to SI values that
are lower than those associated with the other design alternatives per-
mitted in accordance with the European practice. For this reason, it is in
many cases the preferred approach of designers, despite the fact that, in
the authors' opinion, its use should be urgently forbidden by rack codes
because it is dangerously unsafe.
On the basis of the processed data, a common remark is that the
ΔSIjRAM always increases with the increase of the degree of semi-
continuity of the beam-to-column joints: in the case of central racks, in-
creasing the span of the floor slab (LFS), ΔSIjRAM decreases, otherwise a
ð5Þ crease of LFS. Further remarks on the rack performance reduction are sup-
ported by Fig. 6, which proposes the domains defined by the maximum
(solid line) and minimum (dashed line) values of ΔSIjRAM versus the
beam-to-column joint stiffness (ρj,btc) for central and lateral racks. In partic-
ular, the figure has been divided in two parts, related to the use of the FE
beam formulation having 6 (a) or 7 (b) DOFs per node. Independent
of the presence/absence of the 7th DOF, it can be noted that:
• due the high degree of redundancy of the racks, the term ΔSIjRAM is al-
ways greater than unity. An exception is represented by a very limited
number of cases related to the lowest values of the joint stiffness, in
which no relevant variation with respect to the performances associ-
ated with the rigid floor slab condition can been appraised;
• the reduction in performance for racks positioned in a lateral location
is always significantly greater than for the central location, owing to
ð6Þ the greater influence of the more severe set of rigid displacements
simulating the floor deflections imposed at the upright bases (Fig. 4);
• the domains have a significantly different width, reflecting directly
the dispersion of the data. Increasing ρj,btc, the differences for the
ΔSI6RAM for a considered stiffness value ranges from 34% to 57% and
from 5% and 24% for the lateral and central racks, respectively. Includ-
ing the warping as an additional DOF, the width is slightly reduced, es-
pecially for central racks, with the ΔSI 7RAM differences from 4%
(ρj,btc = 1) to 16% (ρj,btc = 5).
ð7Þ • the stiffness of the base-plate joint has only a very moderate influence
on the load carrying capacity reduction: maximum differences are al-
ways lower than 8% and 10% for central and lateral racks, respectively.
A more accurate appraisal of the influence of the FE formulation is
allowed from Tables 2 and 3, which report the mean value (mean) and
standard deviation (dev) of ΔSIjRAM for each set of frames having the
same joint stiffness, for central and laterally located frames, respective-
ly. As a general remark, it can be stated that independently on the re-
finement of FE formulation, the performance reduction due to the
floor settlements are absolutely non-negligible, with the SI differences
of the mean values up to 26% and 52% for central and lateral racks,
respectively.
For central racks, the values of ΔSIjRAM are always very similar, quite
independently of the cross-section type and on the number of load
678 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686

Fig. 6. Domains associated with central and lateral racks (RAM approach) for 6 (a) and 7 (b) DOFs FE beam formulation.

levels, as it appears from the very low dev value, limited to 0.05. For increase of the floor slab length. Otherwise, for lateral racks, ΔSIjRAM sig-
ρj.btc = 1, ΔSIjRAM is very low, practically independent of the floor length nificantly increases with the increase of the LFS, depending also on the
while, for stiffer joints, the performance reduction increases with the cross-section type and by the number of load levels, as it can be deduced

Table 2 Table 3
ΔSIjRAM for racks centrally located. ΔSIjRAM for racks laterally located.

ΔSI 6 RAM ΔSI7 RAM ΔSI 6 RAM ΔSI7 RAM

ρj,btc 12 m 18 m 24 m 30 m 12 m 18 m 24 m 30 m ρj,btc 18 m 24 m 30 m 18 m 24 m 30 m

1 Mean 1.021 1.009 1.006 1.005 1.015 1.007 1.005 1.004 1 Mean 1.132 1.223 1.288 1.126 1.216 1.280
Dev 0.0087 0.0082 0.0066 0.0052 0.0082 0.0080 0.0069 0.0057 Dev 0.0451 0.0851 0.1080 0.0484 0.0884 0.1128
1.5 Mean 1.063 1.026 1.013 1.010 1.047 1.020 1.010 1.008 1.5 Mean 1.203 1.296 1.362 1.193 1.285 1.350
Dev 0.0141 0.0115 0.0109 0.0088 0.0078 0.0073 0.0083 0.0070 Dev 0.0597 0.1075 0.1367 0.0638 0.1106 0.1421
2 Mean 1.105 1.055 1.032 1.020 1.080 1.041 1.023 1.013 2 Mean 1.254 1.348 1.413 1.236 1.330 1.395
Dev 0.0167 0.0133 0.0134 0.0120 0.0065 0.0084 0.0088 0.0082 Dev 0.0727 0.1253 0.1575 0.0758 0.1267 0.1602
3 Mean 1.172 1.105 1.073 1.049 1.135 1.083 1.054 1.036 3 Mean 1.325 1.419 1.483 1.299 1.394 1.458
Dev 0.0248 0.0148 0.0142 0.0188 0.0100 0.0075 0.0077 0.0085 Dev 0.0935 0.1517 0.1880 0.0975 0.1536 0.1899
4 Mean 1.223 1.144 1.104 1.076 1.174 1.114 1.079 1.057 4 Mean 1.376 1.468 1.531 1.341 1.435 1.498
Dev 0.0345 0.0168 0.0142 0.0161 0.0155 0.0106 0.0076 0.0079 Dev 0.1116 0.1727 0.2111 0.1115 0.1705 0.2083
5 Mean 1.264 1.174 1.129 1.098 1.204 1.137 1.098 1.072 5 Mean 1.414 1.504 1.567 1.369 1.463 1.525
Dev 0.0458 0.0226 0.0177 0.0158 0.0261 0.0183 0.0131 0.0104 Dev 0.1268 0.1895 0.2301 0.1269 0.1877 0.2270
 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686
by the rather high standard deviation, up to 0.23, significantly greater
than for centrally located racks.
As to the direct comparison, to highlight the influence of the
FE beam formulation, SI6RAM results are always equal to or slightly
greater than SI7RAM and this is mainly due to the different values of
internal bending moments along the down-aisle direction, greater
for 6DOFs, and to the absence of warping effects in the 6DOFs veri-
fication rules. For better appreciating the effects of the interaction
between the settlements and the degree of refinement of the FE for-
mulation, Figs. 7 and 8 (for central and lateral located racks, respec-
tively) can be considered. In particular, each figure is divided in
three parts: part a), for the MM_racks, part b), for the DD_racks
and part c), for the TT_racks.
The bi-sector line defines two sub-domains and the generic point
identified by the ΔSI6 RAM and ΔSI 7 RAM co-ordinates lies in one of these
sub-domains, the reference axis of which identifies directly the more
conservative formulation. A dashed line is also reported to specify the
maximum value of the difference between the considered formulations.
It can be noted that the overestimation of the rack performance ignoring
slab flexibility is always large. This is particularly true for racks located
laterally, with up to 70% and 90% for DD_ and TT_racks, respectively,
while the maximum difference, quite independent of the FE formula-
tion, amounts to only 30% for central racks. Furthermore, as already ob-
served, the estimation of the 6DOFs load carrying capacity reduction is
in general greater than for the 7DOFs: differences are up to 10% and
5%, for central and lateral cases respectively. From a general overview
ΔSI 6RAM
of the data, it is worth noticing that 86% of the ratios fall in the
ΔSI 7RAM
range of 1–1.04. The 7DOFs SI increment is greater only in a very limited
number of cases, (15 over 504 cases in total) and the reduction amounts
to no more than 3%.
Fig. 7. Comparison between ΔSI6RAM and ΔSI7RAM for racks in central position.
679
5.2. The GEM approach application
Similar to what was done for the RAM approach, the values of
the safety index associated with the General Method have been
considered with reference to the cases of rigid (RF) and flexible
(FF) floor slab, hence allowing for an evaluation of the associated
reduction of the rack performance in term of ΔSIjGEM, according to
Eq. (8).
All the remarks already presented for the RAM approach are in gen-
eral confirmed also with reference to this more conservative design ap-
proach. A reduction of the rack performance is always observed and the
term ΔSIjGEM increases with the increase of the degree of semi-continuity
of beam-to-column joints. A general appraisal of the reduction of the
rack performance associated with the GEM approach is allowed for by
Fig. 9, proposing the same types of domains already presented in Fig. 6
for the RAM approach: once again, it is possible to observe a wide dis-
persion of the data, especially in case of laterally located racks. Further-
more, from the figure, it appears that:
• in the case of 6DOFs, the domains are practically equivalent to the
ones introduced for the RAM approaches. Otherwise, considering
ΔSI7GEM, the width of the domains varies from 29% to 68% and from
3% to 28% with the increase of ρj,btc for the lateral and central racks, re-
spectively.
• in the case of central racks, increasing the span of the floor slab,
ΔSIjGEMdecreases, otherwise the reduction of the rack performance
increases with the increase of LFS;
• contrary to what happened with the RAM approach, the 7DOFs anal-
ysis gives a reduction of load carrying capacity greater than for the
6DOFs, up to 10%, owing to the non-negligible presence of the
bimoment in Eq. (2a, 2b);
680 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686
Fig. 8. Comparison between ΔSI6RAM and ΔSI7RAM for racks in lateral position.
• also in the GEM approach, the stiffness of the base-plate joints has
a very moderate influence on the load carrying capacity reduc-
tion: no more than 6% and 10% for central and lateral racks,
respectively.
Tables 4 and 5 present the mean value and the standard deviation
of ΔSIjGEM for each set of frames with the same joint stiffness, for cen-
tral and lateral locations, respectively. The remarks already proposed
with reference to Tables 3 and 4, are again confirmed here: also the
values of ΔSIjGEM are similar to each other for the racks placed in the
central position. Otherwise in lateral located rack the differences be-
tween SI6 and SI7 increase remarkably with the increase of floor slab
and very great values of the standard deviation can be observed (up
to 0.25).
Furthermore, it is confirmed the relevant influence of the floor set-
tlements on the safety index, however, in case of GEM approach, more
limited if compared with the RAM approach values. In addition, Figs.
10 and 11 (for central and lateral rack, respectively) can be considered,
which are equivalent to those already discussed in Figs. 7 and 8. It can
be stated, in general, that the increment of the safety index is absolutely
non-negligible, up to 38% and 90% for central and lateral racks,
respectively.
As a concluding remark it can be noted that in general the 7DOFs re-
duction is greater than the 6DOFs one: differences are up to 12% and
10%, for central and lateral locations, respectively. Only in the cases of
racks having MM-type uprights and four load levels (MM_4LL racks)
the lowest performance is always associated with 6DOFs analysis be-
cause of the more reduced warping influence. A great concentration of
ΔSI 6GEM
the ΔSI 7GEM
values is in the range 0.95–1, where approximately 82% of
the data is found.
6. A practical proposal for routine design
The discussed results underline clearly the non-negligible influence
of floor settlements on the effective load carrying capacity of medium-
rise pallet racks. The set of data associated with the parametric study
has been used to calibrate an equation to predict the reduction of the
rack performance (or, equally, the increasing of the safety index),
which should be of great interest from a design point of view. A mathe-
matical function fitting the numerical data has been developed based on
the key parameters of influence. In particular, the proposed equation
allows one to account directly for the relative position and the length
of the rack (LR), the flexural upright proprieties, the length of the
supporting floor slab (LFS) and the degree of flexural continuity associat-
ed with beam-to-column joints, expressed by the ρj,btc parameter.
Neglecting the influence of base-plate joints, because they play only a
minor role, the reduction of the rack performance has be assumed to
be null when beam-to-column joints are hinges (i.e. ρj , btc = 0) while
it is considered proportional to the beam-to-column joint stiffness. In
detail, the following equation has been defined to predict the increment
of the safety index ΔSIpred:
!
pred LR LFS  e pffiffiffiffi
ΔSI ¼1þ  1þ 1:5  Rsma β  ρ j;btc ð9Þ
20  LFS L2R
where e is the distance from the floor midspan to the centre of the rack,
always comprised in the range ½0  ðL2FS − L2R Þ, Rsma is the ratio between
the maximum and the second moment of area of along the principal
direction of the cross section upright and the term β accounting for
the cross-section properties is defined by Eq. (1).
It is worth noticing that this equation, which is based on a previous
study on by bi-symmetric upright racks [10], has been suitably
 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686 681

Fig. 9. Domains associated with central and lateral racks (GEM approach) for both 6 (a) and 7 (b) DOFs FE beam formulation.

improved to account for the presence of one axis of symmetry by means SI 7 FF

The calibration phase has been based on the SI6k RF ratio, hence making
pffiffiffiffi
of the term β and has a general validity only when the limit state k

reference to two terms characterized by different levels of accuracy in

governing design is associated with failure of uprights.
assessing the actual rack performance. The 6DOFs FE analysis on rigid

Table 4 Table 5
ΔSIjGEM for central rack. ΔSIjGEM for lateral rack.

ΔSI6GEM ΔSI7GEM ΔSI6GEM ΔSI7GEM

ρj,btc 12 m 18 m 24 m 30 m 12 m 18 m 24 m 30 m ρj,btc 18 m 24 m 30 m 18 m 24 m 30 m

1 Mean 1.013 1.008 1.006 1.005 1.037 1.023 1.017 1.016 1 Mean 1.073 1.122 1.157 1.090 1.135 1.170
Dev 0.0054 0.0052 0.0041 0.0034 0.0096 0.0090 0.0072 0.0069 Dev 0.0291 0.0544 0.0692 0.0423 0.0724 0.0859
1.5 Mean 1.044 1.020 1.012 1.009 1.070 1.044 1.032 1.027 1.5 Mean 1.134 1.193 1.235 1.160 1.216 1.257
Dev 0.0112 0.0100 0.0080 0.0063 0.0112 0.0128 0.0107 0.0099 Dev 0.0420 0.0754 0.0960 0.0645 0.1005 0.1228
2 Mean 1.081 1.045 1.027 1.019 1.115 1.071 1.048 1.037 2 Mean 1.185 1.250 1.297 1.219 1.281 1.323
Dev 0.0146 0.0151 0.0131 0.0107 0.0172 0.0086 0.0113 0.0107 Dev 0.0518 0.0912 0.1158 0.0833 0.1239 0.1510
3 Mean 1.149 1.094 1.066 1.045 1.202 1.134 1.098 1.073 3 Mean 1.266 1.337 1.387 1.310 1.376 1.422
Dev 0.0194 0.0188 0.0172 0.0192 0.0328 0.0206 0.0136 0.0157 Dev 0.0671 0.1144 0.1447 0.1153 0.1602 0.1933
4 Mean 1.205 1.135 1.099 1.073 1.267 1.184 1.135 1.103 4 Mean 1.326 1.398 1.449 1.382 1.443 1.493
Dev 0.0226 0.0186 0.0164 0.0185 0.0406 0.0313 0.0236 0.0227 Dev 0.0813 0.1332 0.1672 0.1364 0.1872 0.2232
5 Mean 1.252 1.169 1.126 1.097 1.331 1.230 1.173 1.131 5 Mean 1.373 1.444 1.495 1.432 1.493 1.541
Dev 0.0271 0.0193 0.0162 0.0169 0.0511 0.0388 0.0305 0.0286 Dev 0.0937 0.1486 0.1850 0.1547 0.2121 0.2483
682 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686

Fig. 10. Comparison between ΔSI6GEM and ΔSI7GEM for racks in central position.

Fig. 11. Comparison between ΔSI6GEM and ΔSI7GEM for racks in lateral position.
 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686
floor (SI6_RF
k ) is in fact adopted in the design manufacturing offices in-
stead of SI7_FF
k , which represents a more accurate evaluation because it
is influenced by the deformability of the slab and accounts adequately
for the presence of mono-symmetric uprights.
Accuracy of the proposed equation can be evaluated in the case of
central racks via Fig. 12 where ΔSIpred is plotted versus ΔSI7k for the
RAM (a) and the GEM (b) approaches. If the representative point is in
the sub-domain delimited by the ΔSIpred axis and the bi-sector line pre-
diction is on the safe side, otherwise the prediction is non-conservative.
Looking at the figure, the accuracy appears to be good: an overestima-
tion of the rack reduction performances can be noted; however, this is
quite limited, up to 4% and 8% for the RAM and GEM approaches, respec-
tively. Increasing the number of load levels, i.e. reducing the inter-storey
height, the accuracy of the prediction equation results improved. The
RAM approach is in general on the safe side. In a very limited number
of cases, the term ΔSI7RAM is slightly greater than the ΔSIpred, i.e. the rep-
resentative point is below the bisector, maximum errors are however
limited to 2%. The GEM approach should be improved because is on
the safe side only for MM_4LL racks. For the other cases ΔSI7GEM results
are generally greater than the ΔSIpred, but with maximum errors limited
to up to 10%;
For what concerns the lateral racks, the efficiency of the proposed
equation is significantly reduced, for both methods, as can be seen in
Fig. 13 and in a great number of cases the reduction of performance, is
quite significantly under-estimated, up to 22% for the RAM approach,
and up to 20% for the GEM method.
An attempt to improve the accuracy of the prediction has hence been
carried out: a suitable safety index γeRFD has been calibrated, allowing to
consider also a general rack location in between the central and lateral
positions, being defined as:
 
  2e
γ eRFD ¼ γCRFD þ γ LRFD −γ CRFD
LFS −LR
where the term e, already present in Eq. (9), accounts for the eccentric
location and correction factors have been introduced to best fit the
limit cases of central (γCRFD) and lateral (γLRFD) located racks.
To predict more accurately the increment of the safety index, a more
efficient evaluation of ΔSI to be used in practical design, is proposed by
amplifying the ΔSIpred given by Eq. (9) via the term γeRFD:
ΔSI DES ¼ γ eRFD  ΔSI pred
The actual safety index of the racks over a flexible slab, SIFF
k , can
hence be directly obtained as:
SI FF
k ¼ ΔSI
DES
 SI 6k RF
Fig. 12. Accuracy of Eq. (9)) according to RAM and GEM approaches for the central racks.
683
where the subscript k identifies alternatively the RAM or the GEM
approach.
A key role in the accuracy of the prediction is played by the
γCRFD and γLRFD correction factors: as expected, high values should lead
to a conservative assessment of the effective rack performance; howev-
er, this result obviously does not allow for competitive and optimal use
of components and material. To this purpose, reference can be made to
Table 6, which presents the mean, the maximum (Max), the minimum
ΔSI7
(min) values of ΔSIDES
k
ratio together with the associated standard devia-
tion (dev) for γCRFD
ranging from 1.0 to 1.05 and for γLRFD between 1.10
and 1.30. All the proposed data are related to the RAM and GEM ap-
proaches considered together.
It can be noted that the mean value of the considered ratio is never
greater than unity and the standard deviation is always lower than
0.085. A moderate influence of the correction factor can be noted for
the central locations, with the prediction being more accurate also with-
out the use of the γeRFD factor. A more important role is played by γLRFD,
influencing significantly the accuracy of the prediction. As an example,
with reference to the choice of γCRFD =1.04 and γLRFD =1.15 (highlighted
ΔSI 7
area in Table 6), the ΔSIDES
k
values are in the range of 0.79–1.10. For a better
appraisal of the proposed improvement via correction factors, one can
refer to Fig. 14. The figure is related to central and lateral racks, respec-
ΔSI 7
tively, where the frequency of the distribution of the ratio ΔSIDES
k
is plotted
for these correction factors values.
As for central racks, it can be noted that the improvement via
γeRFD leads to a safe prediction of the rack performance reduction in
many cases. A large number of values can be observed in the range
0.95–1.00. Otherwise, with reference to lateral racks, a slightly greater
dispersion of results can be appraised: a large amount of data is in be-
tween 0.90 and 1.05. Overestimation of the effective rack performance
ð10Þ is always limited to a small number of cases and never greater than
10%, which although it represents an unsafe value would still be accept-
able from the design point of view.
An additional set of medium-rise pallet racks, differing from the ones
presented in Section 3, has been considered for the evaluation of the ef-
fective accuracy of the predicted equation on cases similar to the ones
typically encountered in routine design. In particular, a set of four bay
racks (GG_4LL and GG_3LL) have been defined, differing in terms of
geometric layout (length of the span equal to 2600 mm instead of
ð11Þ 2780 mm) and for the upright frame geometry (the height of the Z-
panel is 1500 mm); uprights present the GG_ cross-section described
in Table 1. Also for these racks, the beam-to-column joint stiffness pa-
rameter ρj , btc ranges from 1 to 5 and ρj , base takes the values of 0.15
and ∞. As to the correction coefficient, γCRFD = 1.04 and γLRFD = 1.15
have been assumed and the accuracy of Eq. (12) can be appraised
ð12Þ through the data sketched in Fig. 15 for both RAM and GEM approaches.
684 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686

Fig. 13. Accuracy of Eq. (9) according to RAM and GEM approaches for the lateral racks.

It can be noted that the proposed equation is always on the safe side
and never overestimates the effective rack performances. The differ-
ences with respect to the ΔSI7k increment (appraised via a 7DOFs design)
are limited to 12% for central racks and rather large for lateral racks, up
to 30% for RAM and 28% for GEM approaches, confirming the efficiency
and utility of the prediction for routine rack design.
7. Concluding remarks
The interaction between the skeleton frame and the supporting floor
slab is an aspect of relevant importance for the safe use of medium-rise
selective pallet racks located inside multi-storey buildings. Previously,
attention has been paid to bi-symmetric cross-section uprights; the
present paper has dealt with the influence of the floor slab deflections,
at the same time accounting for the mutual influence between the rela-
tive settlements and the mono-symmetry of the cross-section in terms
of warping effects, currently neglected in routine design.
Table 6
ΔSI 7
Dependence of ΔSIDES
k
by the correction factors γCRFD and γLRFD.
It has been demonstrated, based on 2304 practical cases, that the in-
fluence of the relative deflections of the upright bases can significantly
reduce the load carrying capacity and should never be neglected. Inde-
pendent of the design approach used to assess rack performance, the
overestimation of the safety index when floor deflections are ignored
could be up to 90% with mean values of 9% and 32% for centrally and lat-
erally located racks, respectively.
Furthermore, being of paramount importance to guarantee a safe de-
sign, especially for racks open to the public, an equation to predict the re-
duction of the rack performance with respect to rigid floors, has been
proposed. Its calibration has been based on data associated with the tra-
ditional 6DOFs FE approach, being the one commonly adopted by rack
manufacturing engineers and the validation has been based on a different
typology of racks, showing that the degree of accuracy appears more than
adequate for design purposes. Prediction of the reduction of the rack per-
formance is in fact always on the safe side and no additional calculations
are required when compared to those carried out under the assumption
of rigid floor slabs. Finally, it is worth underling that the overestimation
of the safety index associated with the proposed equation is acceptable
from the design point of view and, in any case, it is greatly preferable a
slightly more conservative design instead of a greatly unsafe design car-
ried out neglecting base settlements and warping effects.

γ LRFD
1.10 1.15 1.20 1.25 1.30 Appendix A. List of symbols. (checks)
C
γ RFD

mean 1.00 0.98 0.96 0.95 0.93 Latin upper case letters
dev 0.04908 0.04983 0.05812 0.07001 0.08307
1.01
Max 1.15 1.10 1.10 1.10 1.10 A gross cross-section area.
min 0.83 0.79 0.76 0.73 0.70 E modulus of elasticity of steel.
mean 0.99 0.97 0.96 0.94 0.93 EC3 EN 1993-1-1 Eurocode 3 “Design of Steel Structures”.
dev 0.04974 0.04837 0.05517 0.06627 0.07892 G shear material modulus.
1.02
Max 1.15 1.10 1.09 1.09 1.09 GEM GEneral Method.
min 0.83 0.79 0.76 0.73 0.70 LL load levels.
mean 0.99 0.97 0.95 0.94 0.92 L member length.
dev 0.05083 0.04738 0.05255 0.06274 0.07493 It the Saint-Venant torsion constant.
1.03 Iw warping constant.
Max 1.15 1.10 1.08 1.08 1.08
min 0.83 0.79 0.76 0.73 0.70
Iy,I, second moment of area.
mean 0.98 0.96 0.95 0.93 0.92
MEd, My,Ed, Mz,Ed design bending moment.
MRk characteristic bending resistance.
dev 0.05230 0.04686 0.05028 0.05945 0.07111
1.04 Ncr critical load for the i-member.
Max 1.15 1.10 1.07 1.07 1.07
N, NEd member axial load.
min 0.83 0.79 0.76 0.73 0.70
NRk characteristic axial resistance.
mean 0.98 0.96 0.94 0.92 0.91
Nb,Rd. axial stability resistance.
dev 0.05408 0.04678 0.04839 0.05640 0.06746
1.05 RAM Rigorous Analysis Method.
Max 1.15 1.10 1.06 1.06 1.06
Sj,btc, Sj,base stiffness of connection.
min 0.83 0.79 0.76 0.73 0.70
SEC3−LB
j,btc lower bound of EC3.
 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686 685

ΔSI 7
Fig. 14. Frequency distribution of the ΔSIDES
k
ratio for lateral and central racks.

−UB
SEC3
j,btc upper bound of EC3. Latin lower case letters
EC3 −UB
Sj,base upper bound of EC3 for base-plate connections.
SI design safety index. e rack eccentricity.
W cross-section modulus. hLL interstorey height.
LR length of the racks. hu panel height.
LFS length of the floor slab. Max maximum value.
Lb beam length. min minimum value.
S355 grade of the steel. fy specified minimum yield stress strength.
ΔSIk increment of safety index due to the displacement, for the kth-
approach. Greek letters
SIRF
k safety index associated with the rigid floor assumption, for the
th
k -approach. α imperfection coefficient associated with the relevant buckling
SIFF
k safety index associated with the flexible floor assumption, for curve.
the kth-approach. αcr buckling overall frame multiplier obtained via a finite ele-
ΔSIpred increment of safety index due to the Eq. (8). ment buckling analysis.
ΔSIDES increment of safety index due to the Eq. (12). αult,k minimum load multiplier evaluated with reference to the
Ie index to account the position of the rack. cross-section resistance.
Rsma ratio between maximum and minimum second moment of δbase imposed displacements at upright bases.
area. λop relative slenderness of the whole structure.
Htot total height of the racks. ρj,btc parameter to define the elastic rotational stiffness of beam-
NLL number of load levels. to-column joints.

Fig. 15. Accuracy of the prediction of the performance reduction for GG_racks.
686 C. Bernuzzi et al. / Journal of Constructional Steel Research 128 (2017) 672–686
ρj,base parameter to define the elastic rotational stiffness of base-
plate joints.
χ reduction factor for the relative buckling curve.
χop buckling reduction factor referred to the overall structural
system.
ω sectorial area.
γM material safety factor.
γeRFD = γCRFD = γLRFD correction factors.
References
[1] T. Peköz, G. Winter, Cold-formed Steel Construction, IABSE Periodica, 1/1980(Febru-
(February) 1980.
[2] M.H.R. Godley, in: E. Rhodes (Ed.), Design of Cold Formed Steel Members 1991,
pp. 361–399.
[3] ECCS, European Convention for Structural Analysis and Design of Steel Frames With
Semi-rigid Joints. Publication No. 67, 1992.
[4] CEN, EN 15512, Steel Static Storage Systems - Adjustable Pallet Racking Systems –
Principles for Structural Design, CEN European Committee for Standardization,
2009 137.
[5] FEM 10.2.08, Recommendations for the Design of Static Steel Storage Pallet Racks in
Seismic Conditions, Federation Européenne de Manutention, Version 1.00, 2010.
[6] prEN 16681, Steel Static Storage Systems –Adjustable Pallet Racking System –
Principle for Seismic Design, CEN European Committee for Standardization, 2016.
[7] RMI MH 16.1, Specification for the Design, Testing and Utilization of Industrial Steel
Storage Racks, Rack Manufacturers Institute, 2012 59.
[8] Commentary on RMI, Specification for the Design, Testing and Utilization of Indus-
trial Steel Storage Racks, Rack Manufacturers Institute, 2008.
[9] AS 4084, Steel Storage Racking, Australian Standards, Australia, 2012.
[10] C. Bernuzzi, D. Persico, M. Simoncelli, Influence of floor deflections on the perfor-
mance of steel storage pallet racks, Eng. Struct. 123 (2016) 434–450.
[11] L.H. Teh, G.J. Hancock, M.J. Clarke, Analysis and design of double-sided high-rise
steel pallet rack frames, J. Struct. Eng. ASCE 130 (7) (2004) 1011–1021.
[12] F.S. Cardoso, K.J.R. Rasmussen, Finite element modelling of storage rack frames, J.
Constr. Steel Res. 126 (2016) 1–14.
[13] M. Casafont, M.M. Pastor, F. Roure, J. Bonada, T. Peköz, Design of steel storage rack
columns via the direct strength method, J. Struct. Eng. ASCE 139 (5) (2013) 669–679.
[14] B.P. Gilbert, K.J.R. Rasmussen, Determination of the base plate stiffness and strength
of steel storage racks, J. Constr. Steel Res. 67 (6) (2011) 1031–1041.
[15] K.K. Sangle, K.M. Bajora, R.S. Talcotti, Elastic stability analysis of cold-formed pallet
rack structures with semi-rigid connections, J. Constr. Steel Res. 71 (2014) 245–262.
[16] P.B. Dinis, B. Young, D. Camotim, Local-distortional interaction in cold-formed steel
rack-section columns, Thin-Walled Struct. 81 (2014) 185–194.
[17] F. Petrone, P.S. Higgins, N.P. Bissonnette, A.M. Kanvinde, The cross-aisle seismic per-
formance of storage rack base connections, J. Constr. Steel Res. 122 (2016) 520–531.
[18] A.N. Trouncer, K.J.R. Rasmussen, Ultra-light gauge steel storage rack frames. Part 1:
Experimental investigations, J. Constr. Steel Res. 124 (2016) 57–76.
[19] A. Kanyilmaz, C.A. Castiglioni, G. Brambilla, G.P. Chiarelli, Experimental assessment
of the seismic behaviour of unbraced steel storage pallet racks, Thin-Walled Struct.
108 (2016) 391–405.
[20] B.P. Gilbert, K.J.R. Rasmussen, N. Baldassino, T. Cudini, L. Rovere, Determining the
transverse shear stiffness of steel storage rack upright frames, J. Constr. Steel Res.
78 (2012) 107–116.
[21] A. Crisan, V. Ungureanu, D. Dubina, Behaviour of cold-formed steel perforated sec-
tions in compression. Part1 – experimental investigations, Thin-Walled Struct. 61
(2012) 97–105.
[22] C. Bernuzzi, A. Gobetti, G. Gabbianelli, M. Simoncelli, Warping influence on the resis-
tance of uprights in steel storage pallet racks, J. Constr. Steel Res. 101 (2014)
224–241.
[23] C. Bernuzzi, N. Draskovic, M. Simoncelli, European and United States approaches for
steel storage pallets rack design. Part 2: practical applications, Thin-Walled Struct.
97 (2015) 321–341.
[24] C. Bernuzzi, European and United States approaches for steel storage pallets rack de-
sign. Part 1: discussion and general comparisons, Thin-Walled Struct. 97 (2015)
308–320.
[25] UNI EN 15620, Steel Static Storage Systems - Adjustable Pallet Racking – Tolerances,
Deformations and Clearances, 2009.
[26] CEN, ENV 1993-1-1, Eurocode 3 - Design of Steel Structures , Part 1-1: General Rules
and Rules for Building, CEN European Committee for Standardization, 1992.
[27] EN 1993–1-1, Eurocode 3 – Design of Steel Structures – Part1–1: General Rules and
Rules for Buildings, CEN European Committee for Standardization, 2005.
[28] EN 1993–1-8, Eurocode 3 – Design of Steel Structures –Part 1-8: Design of Joints,
CEN European Committee for Standardization, 2005.
[29] N. Baldassino, C. Bernuzzi, Analysis and behaviour of steel storage pallet racks, Thin-
Walled Struct. 37 (4) (2000) 277–304.
[30] SAP2000 v.17, Integrated Software for Structural Analysis and Design, SAP2000 CSI
computer & structures.inc, 2015 (http://www.csiamerica.com).
[31] ConSteel 9.0., Finite-element-program, ConSteel Solutions Ltd, 2014 (http://www.
consteel.hu).
[32] UNI EN 10025, Hot Rolled Products of Structural Steels, 2004.
[33] C. Bernuzzi, B. Cordova, M. Simoncelli, Unbraced steel frame design according to EC£
and AISC provisions, J. Constr. Steel Res. 114 (4274) (2015) 157–177.
[34] F. Bijliard, M. Feldmann, J. Naumes, G. Sedlacek, The “General Method” for Assessing
the Out-of-plane Stability of Structural Members and Frames in Comparison With
Alternative Rules in EN1993 – Eurocode 3 – Part 1-1, Steel Construction, vol. 3, no.
1, Ernest & Sohn ed., 2010
[35] J. Szalaj, The “General Method” of EN1993-1-1, NSC. April 2011, pp. 30–31.
[36] Papp F. Global stability analysis using “general method” http://www.consteelsoftware.
com/files/sharedUploads/Pdf/General_Stability_analysis.pdf.
[37] C. Bernuzzi, M. Simoncelli, European design approaches for isolated cold-formed
thin-walled beam-columns with mono-symmetric cross-section, Eng. Struct. 86
(2015) 225–241.
[38] C. Bernuzzi, F. Maxenti, European alternatives to design perforated thin-walled cold-
formed beam-columns for steel storage systems, J. Constr. Steel Res. 110 (2015)
121–136.
[39] C.A. Castiglioni, A. Kanyilmaz, M. Angeretti, G. Brambilla, G.P. Chiarelli, C. Bernuzzi,
Experimental Results of Full Scale Push Over Tests of Project SEISRACK2 (Seismic
Behaviour of Steel Storage Pallet Racking Systems), 2nd European Conference on
Earthquake Engineering, Istambul Aug. 25–29, 2014.
[40] C. Bernuzzi, G. Gabbianelli, A. Di Gioia, M. Simoncelli, Pushover analyses of hand-
loaded steel storage shelving racks, J. Earthq. Eng. (2016) 1–27 September.