Rack January 2001 Opus2
Rack January 2001 Opus2
Rack January 2001 Opus2
by
Teoman Pekz and Kotha Rao
Cornell University, Ithaca, New York, USA
SUMMARY
The current Design Provisions for Steel Industrial Storage
Racks are summarized and a progress report on a current
Cornell University Research project is given.
The design of industrial steel storage racks presents several challenges to the structural
engineer. Presently the design in the United States is carried out according to the 1997
edition of the Specification (1) published by the Rack Manufacturers Institute (RMI).
The RMI first published its first Minimum Engineering Standards for Industrial Storage
Racks in 1964.
The work that resulted in the first edition of the Specification was initiated by the RMI in
1972 at Cornell University. Several editions of the Specification have been prepared
based on the work by the RMI Specification Advisory Committee and the researchers at
Cornell under the supervision of Professors George Winter and Teoman Pekz until
1979 and under the supervision of senior author since 1979. The RMI Specification is
tied closely to the AISI Specification [2] for the provisions on Cold-Formed Steel Design.
The Australian Specification for racks is based primarily on the RMI Specification with
some regional enhancements. The applicable standard in Europe [3] is described in
some detail in Reference [4].
The discussion in this article will be on the RMI Specification and the current research at
Cornell on evolving and improving the RMI Specification. The discussion on the loads
specified in the RMI Specification will be limited to Load and Resistance Factor Design.
The extensive earthquake provisions of the RMI Specification will not be discussed in
this article.
Most of the current research on racks in the United States is being conducted at Cornell
University with the senior author as the principal Investigator for the project. The focus
of the discussion in this paper is on the progress to date at Cornell University, and the
conclusions may change based upon future research.
The research uses finite element solutions verified by tests on both the component and
the global scale. On the component level, the topics focused upon are the behavior of
joints, and the interaction of the frames with their column bases.
1
1. SOME FEATURES OF THE 1997 RMI SPECIFICATION (1)
1.1 Loads
Dead and live loads on racks are only a small portion of the total load on racks. Product
load is a major portion of the loads applied on a rack structure. The Specification
defines the product load as the products or pallet loads stored on a rack. Since the
product load is usually well defined and listed on the plaques placed on the racks, lower
load factors are specified for the Load and Resistance Factor Design of racks that for
live loads. Below are two examples of load factors and load combinations out of eight
given in the Specification:
1.4DL + LL + 1.2PL
1.2DL + 1.6LL + 0.5(SL or RL) + 1.4PL
DL and LL are dead and live loads, respectively. SL and RL snow and rain loads,
respectively. PL is the product load.
Beam support connections, frame bracing, and frame bracing to column connections of
racks are to be designed for horizontal loads equal to 1.5% of the factored dead load
and factored product load. The horizontal forces include the effect of out-of-plumbness.
The tolerance for out-of plumbness is given as 0.5 inches in 10 feet of height. The
horizontal forces are to be applied separately, not simultaneously, in each of the two
principal directions of the rack. The beam support connection moments are to be
checked against the permissible moments (both positive and negative) determined from
Cantilever Tests and/or Portal Tests (Fig. 1 and 2).
1.2 Design of Steel Elements and Members
The design of steel elements and members are carried out according to the AISI
Specification [2] for cold-formed members and AISC Specification [5] for hot-rolled
members. Some exceptions and modifications to these Specifications are noted in the
Specification. Some of these exceptions and modifications will be discussed briefly.
The load carrying capacity beyond local buckling (post-buckling strength) is quite
significant for cold-formed steel racks. For this reason some extensions to the AISI
Specification had to be made.
Nominal bending moment capacity
n
M is obtained by multiplying the effective section
modulus
e
S
by the yield stress
y
F
. The effective section modulus
e
S
for perforated
members is to be determined by multiplying the elastic section modulus of the net
section by
2
0.5
2
Q _
+
,
.
This expression was obtained through reasoning that the effect local buckling would be
less significant for flexural members than that for axially loaded members. The factor Q
is determined by tests on axially loaded stub columns. Using this value of Q for flexure
would be too conservative. The above modification factor accepts one half of Q for the
compression part of the section but takes the full section for the tension part.
The interaction of lateral buckling with local buckling needs to be accounted for; thus,
the expression for calculating member strength as for lateral buckling involves the
effective section modulus at lateral buckling stress
c
S
. Modulus
c
S
is calculated by
multiplying the net section modulus by the factor
/ 1
1
2
c f
y
M S Q
F
_
,
.
In this expression
c
M
is the lateral buckling moment,
f
S is the section modulus for the
full section and Q is determined by stub column test. Though the specification does not
point it out, Q could be determined by the expression
0.5
2
stubcolumntest
Q
Q
_
+
,
as reasoned out above. The above factor is derived in accordance with the approach
described in Reference [6].
As required by the AISI Specification, nominal strength of a column is determined by
multiplying the column limit-state stress
n
F by the effective area
e
A
at stress
n
F . This
accounts for the interaction of local and overall column buckling. Studies on the
interaction of the distortional buckling and overall buckling require special consideration.
A research project on this subject has just been completed. The results have not yet
been published. However, it is expected that the future editions of the RMI Specification
will have provisions on theis subject. The effective area is determined by the expression
( )
min
1 1
Q
n
e net
y
F
A Q A
F
1
_
1
1
,
]
where
min net
A
is the minimum cross-sectional area obtained by passing a plane through
the column normal to the axis of the column. This expression is derived in Reference
[6].
1.3 Frame Design
Frame design involves the use of effective length factors which are specified for various
situations and the interaction equations given in the AISI Specification.
3
Racks in general consist of upright frames and beams connecting the upright frames.
Upright frames consist of two columns braced together. There may or may not be
bracing in the down-aisle direction, namely in the direction perpendicular to the upright
frames.
The current practice is to do a linear analysis and account for the second order effects
by a magnification factor
is defined as
e
P
P
1
where
e
P
is the elastic buckling load of the column about the bending axis. Thus the
effective length is determined for buckling about the bending axis.
The terms
n
P and
n
M are the limit-state axial load and bending moment, respectively,
when each acts separately. For bending about the symmetry axis, as is the case for
buckling in the down aisle direction, the terms
n
P and
n
M are determined using effective
length factors
K
and
x
K with the nominal strength equations given in the AISI
Specification.
Effective length factor for torsional buckling
K
1
,
+
1
]
where
b
I
= the actual moment of inertia of the pallet beam
b
L
= the span of the pallet beam measured between the centroids of the columns
supporting the beam
F
= the joint rigidity determined by the Portal Test
E = modulus of elasticity
The analysis for the effective length factor for the portion of the column from the floor to
the first beam level would involve the following G values as defined in the commentary
of Ref. [5].
1 2 1
1 1
2
c
e
c c c
a b
f
b
f
b
red
I
I
L L L
G andG
I
I
L
L
_
+
,
_
,
where
c
I
= the column moment of inertia
1 c
L
= the distance from the floor to the first beam level
2 c
L
= the distance from the first beam level to the second beam level
The fixity implied at the column base by the equation above can be calculated as
2
1440
f
f
I bd
L
(all dimensions in inches)
2
3
1440(25.4)
f
f
I bd
L
(all dimensions in mm)
where
b = the width of the column (parallel to the flexure axis)
d = the depth of the column (perpendicular to the flexure axis)
For the above equation, the floor is assumed to be concrete, and the column connection
to the floor must be adequate to develop base moments consistent with this stiffness.
For other floor material the equation should be modified.
5
Similar rules are given for design of frames perpendicular to the aisle and for braced
frames.
1.4 Tests
Several tests are prescribed in the RMI Specification for the determination of
parameters that are difficult to determine computationally. Here a few of these tests will
be mentioned briefly.
1.4.1 Stub Column Tests
Because of the interaction of local buckling, perforations and cold-forming effects it is
necessary to carry out stub column tests on a short segment of a column to determine
the behavior. These tests are carried out in accordance with the rules given in the AISI
Specification. The rules are modified as applicable to the rack columns. The RMI
Specification also gives rules for evaluating the test results.
1.4.2 Cantilever Tests
The test setup shown in Fig. 1 is used to determine the moment-rotation behavior of the
mechanical joints of racks.
1.4.3 Portal Tests
The test setup shown in Fig. 2 is used to determine the moment rotation behavior of
mechanical rack joints as well. The portal test reflects the effect of the vertical forces on
the connection more accurately. However, this test is more difficult to conduct. The
hinges at the column bases are prevented from moving in the plane of loading.
2. CURRENT CORNELL UNIVERSITY RESEARCH ON COLD-FORMED STEEL
FRAMES
As seen in the discussion of the current RMI Specification above, the behavior and
ultimate strength of a typical rack frame is characterized by many parameters. These
parameters include: flexibility of beam-to-column joints, column base fixity, perforations
in the columns, local buckling of member components, geometric and material
imperfections and complex bucking behavior of column members. Important features
and the tentative results of ongoing research are discussed below. The discussion
below is a progress report on the findings and may change on the basis of the findings
during the rest of the project.
The studies make extensive use of ABAQUS a commercial finite element method
(FEM) software to validate design approaches. Whenever possible the FEM solutions
are compared with physical test results to gain confidence in the modeling technique.
6
Various types of shell elements and beam elements are used. The shell-contact
element model and beam-spring element model are used to model the various features
of the components and entire frame behavior.
While the shell and contact element model is accurate and reliable, it has been found
that modeling of full frames by shell elements is tedious, computationally expensive and
requires experienced analysts. This model best serves in validation studies (as an
alternative to experiments) to evaluate the performance of other simpler numerical and
analytical models.
2.1 ANALYSIS
2.1.1 Behavior of Frame Components
2.1.1.1 Beam-to-column joint flexibility:
The connection between the shelf beam and column members of pallet rack frames is
generally flexible and influences the frame behavior significantly.
The current RMI specification accounts for the effect of joint flexibility on column
strength by modifying the pallet beam stiffness and in turn modifying the column end
restraint offered by the beam member as discussed in Section 1.3 above.
The joint stiffness is to be determined experimentally by individual manufacturer using
the test setup that is mentioned in Sections 1.4 and 1.5. The specification suggests
using the secant stiffness corresponding to 0.85 times the ultimate moment capacity of
the joint as determined from physical tests. While the specification procedure is simple
to use, there may be cases where the above assumption/simplification does not hold
true. Hence, it is always rational and safe to use the correct joint stiffness value in the
frame analysis by adopting joint M relationship valid through the entire load history.
In the present study such expressions are developed by making use of the experimental
data available in the form of moment-rotation history of a variety of joints as provided by
different manufacturers. A total of 6 joint types typical to the ones used in the United
States are considered in the present study. The FEM idealization used and its accuracy
is illustrated in Fig. 3.
2.1.1.2 Column Base Fixity:
The column base stiffness of pallet rack frames is characterized by the base plate
dimensions, number, dimensions and layout of bolts, ratio of moment to axial load at the
column base and foundation characteristics. The degree to which each parameter
effects the base stiffness depends on the way the column is connected to the
foundation.
7
In the present study, the base stiffness characteristics due to bending of the base plate
are studied by means of Finite Element Method. FEM idealization of the column base is
illustrated in Fig. 4. Contact elements, shell elements and spring elements are used in
the finite element modeling of the problem. The following observations have been made
based on a large parametric study on an isolated lipped channel column member.
The moment-rotation relationship of a typical pallet rack frame base is generally
nonlinear
The higher the axial load on the column, the stiffer is the column base.
While the axial load increased the base stiffness by about 20% when the base plate
is thin (0.25``, 6.25 mm, thinnest of plates studied), its effect is found to be
insignificant in the case of thicker plates
When the axial load on the column is accompanied by only a small amount of lateral
load (1.6% of axial load), the column base stiffness is found to be very close to that
of RMI specification value
For lateral load to axial load ratios other than 0.016, the initial stiffness of column
base is found to vary from 0.3-0.7 times the RMI specification value, depending on
the base plate configuration and amount of axial load on the column.
The effects of base plate configuration (plan dimensions) and number of bolts on the
base stiffness are found to be negligible. However in the case of thinner plates,
smaller plate configurations seem to help slightly increase the base stiffness.
2.1.2 Frame Behavior:
First, a parametric study has been carried out on four types of commonly used pallet
rack frames tested at Cornell University in the 1970s. This was done to establish
guidelines to prepare FEM models to study the behavior accounting for the influence of
various parameters discussed above. A FEM idealization and a view of a physical test is
illustrated in Fig. 5.
First the effect of warping constraint at column bases was studied. It was found that the
influence of the warping constraint on the frame strength depends on the type of column
members. The difference between the strengths of warping free and warping fixed
cases may vary from 4% to 25%. A limited parametric study showed that the warping
fixed case simulates the actual condition better.
Frames were modeled using shell and beam elements. It was found that beam
elements with proper care are capable of estimating the frame strengths accurately
when compared to experimental and shell element based FEM results.
Based on physical test results and FEM studies it was found that the current RMI
Specification may under estimate the strengths of pallet rack frames up to about 50%.
This establishes the need to review and improve the current design procedure.
8
While the Specification procedure has been made conservative in the absence of
knowledge of the effect of various parameters on the system behavior, it is not known to
what magnitude each of these parameters affect the accuracy of the design procedure.
A large parametric study involving nonlinear finite element analysis of 5 types of pallet
rack frames has been carried out to quantify the conservatism of the current procedure.
The factors focused on are the use of approximate effective length factors, linear beam-
column interaction equation and linear amplification factor. When the column base
stiffness is taken as specified in the RMI as discussed in Section 1.3 above, it was
observed that:
The strength estimates by taking
0.8 K
and 1.7
x
K are satisfactory for the
gravity together with lateral load cases. The calculated capacities in these cases are
conservative by 10%-15%. However, for the Gravity load only case, the specification
is found to be conservative by as much as 40%.
The strength estimates by taking
0.8 K
and determining
x
K from alignment
charts are found to be 20-50% conservative when load on the frame is checked for
either gravity or gravity together with small lateral loads. The estimates become 15-
25% conservative for large lateral loads.
When the column base stiffness is taken as one half of what is specified in the RMI
Specification as discussed in Section 1.3 above, it was observed that:
Taking
0.8 K
and 1.7
x
K is conservative by 10-35%.
Taking
0.8 K
and determining
x
K from alignment charts P is conservative by 20-
40% when the load on the frame is either gravity or gravity together with small lateral
loads and about 15-25% for large lateral loads.
In order to improve the current design procedure, several design methods using the
present AISI and AISC procedures for frames and beam columns are being tried. It was
found that the frame design procedure given in the AISC-LRFD Specification [5] with
1.0 K
and 1.0
x
K in conjunction with either initial beam-to-column joint stiffness or
secant stiffness might predict the frame strengths with sufficient accuracy. The AISC-
LRFD equations can be summarized as follows:
1 0.2
2
n n n
P M P
for
P M P
+
8
1 0.2
9
n n n
P M P
for
P M P
+ >
1
1
lin
M M
P
H L
9
The terms used in these equations are the same as the ones defined in Section 1.3
above. Furthermore is the inter-story deflection for horizontal forces
determined by linear analysis,
H