Sunteți pe pagina 1din 7

Journal of Constructional Steel Research 62 (2006) 11711177

www.elsevier.com/locate/jcsr

Design of unbraced multi-storey steel frames


D. Anderson
School of Engineering, University of Warwick, Coventry CV4 7AL, UK

Abstract
Simple approximate methods for analysis of unbraced frames have played a part in encouraging the use of steel in construction. These methods
have been supported by research, to verify and to extend their scope. This paper questions their continued use and further development. It is
appropriate to do this at the present time when documentation and software are needed to introduce the Eurocodes. Approximate methods still
have a role, in initial design. For unbraced frames early consideration needs to be given to limiting sway in service conditions. A very simple
formulation is presented, to enable section sizes to be determined to satisfy this criterion.
c 2006 Elsevier Ltd. All rights reserved.

Keywords: Frames; Steel; Composite; Steelconcrete; Sway; Analysis; Second-order; Serviceability; Optimisation

1. Introduction
It is expected that Eurocode Parts required for the design of
common building structures will be published in 2005. Their
preparation has taken much longer than originally expected
but the final provisions for steel and composite building
structures [1,2] are still recognisably similar to the first drafts
published in the 1980s [3,4]. The codes as published therefore
benefit from the initial leadership given by Dowling for
steelwork and Johnson for composite construction.
During the development of the Eurocodes, the drafting
groups sought to meet two principal objectives: that the codes
be user-friendly; that the codes reflect structural behaviour
accurately. When these are in conflict, preference has been
given to accuracy. Publications in support of the Eurocodes can
provide simplified recommendations if necessary and the very
widespread acceptance of software enables the drudgery of long
hand calculations to be avoided.
Software for analysis of structural frames goes back a
long way [5]. Despite this, designers have continued to use
approximate methods and the research community has provided
technical justification. Indeed some such methods have been
developed further, to address modern forms of construction and
to be compatible with limit states design codes. The analysis

Tel.: +44 24 7652 3163; fax: +44 24 7641 8922.

E-mail address: da@eng.warwick.ac.uk.


c 2006 Elsevier Ltd. All rights reserved.
0143-974X/$ - see front matter
doi:10.1016/j.jcsr.2006.06.009

of unbraced multi-storey steel frames is a particular example


of such endeavour, which has resulted in the wind-moment
method [6] and the MerchantRankine formula [7] continuing
to be used in design. Both methods are of limited scope but
avoid taking explicit account of second-order effects due to
sway.
These approaches may have encouraged the use of steel
in construction. This paper argues though that the Eurocodes
provide opportunity for designers to move to methods of
general application not dependent on empirical justification.
For steel frames, and particularly for frames including
composite elements, an attempt to avoid direct treatment of
second-order effects resulting from sway is likely to hinder
rather than help verification of designs for ultimate limit states.
Measured by the number of design rules, serviceability limit
states receive far less attention and yet limitation of sway
deflections may well govern section sizes in an unbraced frame,
even when the joints are taken as rigid. This paper concludes by
describing a new method to generate section sizes to meet limits
on sway deflections due to horizontal loading.
2. Wind-moment method
2.1. The basis of the method and its advantages
Fig. 1 illustrates the method. In its basic form, the method
assumes:

1172

D. Anderson / Journal of Constructional Steel Research 62 (2006) 11711177

Fig. 1. Bending moments according to the wind-moment method.

under gravity loads the beam-to-column joints act as pinned


connections
under horizontal wind loads these joints are rigid. Points
of contraflexure are assumed to occur at the mid-height of
columns and the mid-length of beams.
Members and connections are proportioned initially to
withstand gravity load, factored for ultimate limit states (ULS).
The frame is then analysed under horizontal wind loads. The
internal forces and moments due to gravity load and wind
are then combined, using appropriate load factors for each
combination. Design for ULS is completed by amending the
initial section sizes and connection details where necessary, to
withstand the combined effects. Second-order effects due to
sway are allowed for by using effective lengths for the columns
that are greater than the lengths between floor levels.
The main advantages of the method are its simplicity, its
suitability for hand calculation and economy in fabrication
of connections. The assumptions render the frame statically
determinate and section sizes and connection details can be
determined without iteration. Joints are designed to withstand
only the forces and moments given by the analyses, rather than
being designed to match the resistance of connected members.
2.2. Development of the method
Despite the assumptions inherent in the method, its
advantages in practice have proved enduring. Judged by the
interest shown by designers, the method has encouraged the
use of steel in construction. This interest provided the incentive
to develop the method for use with BS 5950, the British limit
states steelwork code first published in 1985, and to address
caution in use of the method recommended by Nethercot [8].
Rules developed for steel frames [6,9] were justified by
second-order elasticplastic analysis of frames designed by the
method [911]. In these studies, the analysis accounted for
the development of plastic hinges and the semi-rigid nature of
typical beam-to-column connections (Fig. 2) in wind-moment
frames.
Some justification for the method is also given by the lowerbound theorem of plastic theory. Provided the internal moments
and forces are in equilibrium with the loads, and resistances
around the frame are not exceeded, the loads will not exceed
those for collapse. These conditions are met by the windmoment method provided that:

Fig. 2. Typical wind-moment connection.

the effects of deflections on equilibrium are accounted for


adequate rotation capacity exists for redistribution of
moment.
Based on the frame studies, the compression resistance of
a column with sway is determined using an effective length
factor of 1.5 and the resistance to lateral-torsional buckling is
calculated by the BS 5950 procedures for simple design. The
requirement for rotation capacity is met by limiting the sections
to Class 1 Plastic and Class 2 Compact and by using only
standard end-plate connections [6] whose ductility has been
proved using testing [12].
The tendency of the method to under-design columns [8]
is offset to some extent by a requirement that moments in
these members due to gravity loads be taken from an end
restraint moment to each beam equal to 10% of the free bending
moment, combined with moments due to eccentricity of beam
reactions. In accordance with British practice [7], the reactions
are assumed to act at 100 mm from the face of the column.
The frame studies also showed that under service loading
the flexibility of the connections resulted in sway deflections
significantly larger than those predicted assuming rigid joints.
This led to a recommendation that sway deflections calculated
on this assumption should be increased, for most frames by
50%.
The studies were inevitably limited to a range of geometry
and loading, steel grades and section types, which define the
scope of the method.
2.3. Composite beams
In practice multi-storey frames often include composite
beams and designers have asked that the method be extended to
cover such members. Two further studies have therefore been
undertaken.
One [10] examined the use of composite beams in conjunction with the standard ductile steel end-plate connections [6]. It
was found that simply replacing a steel beam by a composite
member of equivalent moment resistance did not always result
in a satisfactory design. With a composite beam, the shallower
depth of the steel section could result in standard connections
with inadequate resistance to the wind moments.

D. Anderson / Journal of Constructional Steel Research 62 (2006) 11711177

The solution has been to use composite connections, in


which reinforcement in the concrete slab is taken into account
in the resistance of the joint [13]. This is though a departure
from the simplicity of the method, because of the need to ensure
that the reinforcement in the composite joint is sufficiently
ductile to provide adequate rotation capacity [14].
2.4. The way forward
In view of the assumptions and the nature of the design
rules, it is not surprising that frames designed by the method
have a wide range in resistance, compared with results from
analyses used to justify the method [11]. For frames up to
four storeys, free to sway only about the major axes of the
columns, the load level for collapse under the critical load
combination varied from 1.1 to 1.70 times the design load level.
For some eight storey frames the resistance was just adequate
and the most recent rules for the method [6] limit its use to
four storeys. A wide scatter of results was also found when outof-plane stability was checked using forces and moments from
the second-order elasticplastic analysis and the BS 5950 linear
interaction formula for combined compression and bending.
With the now-ready availability of software for rigorous
second-order elasticplastic analysis, it is doubtful if use of the
wind-moment method should be encouraged any longer. For
use with Eurocode 3, further frame studies would be necessary,
particularly to develop an appropriate interpretation of the
interaction formula for buckling resistance in bending and axial
compression and to account for base flexibility. More generally,
it is doubtful if the method conforms to the Eurocodes
expectation that the calculation models and basic assumptions
reflect anticipated structural behaviour. The dependence of joint
behaviour on the type of loading (Fig. 1) is difficult to justify,
even though further frame studies would no doubt show that,
within a defined scope, a modified rules method would result in
safe structures.
There is a long history to the use of rigorous secondorder elasticplastic analysis as the basis for design of multistorey steel frames, going back to the 1960s [15]. Such
proposals were then limited to frames with rigid joints,
which are expensive to fabricate. The wind-moment method
avoids this problem, by designing joints to withstand only
the forces and moments given by its analyses. Eurocode
3 has provided the key to successfully using second-order
analysis in practice, by providing methods to predict joint
resistance and stiffness [16]. An evaluation of predicted values
for ductile end-plate connections has shown that, although the
stiffness prediction can be improved, the Eurocodes methods
are reasonable [17].
Eurocode 3 describes alternative ways to use secondorder analysis. In addition to sway effects in the frame as a
whole, second-order effects in individual members and member
imperfections may be accounted for in the global analysis.
Individual stability checks for the members are then unnecessary. At first sight this approach is attractive because of
its conceptual simplicity. However, in practice it will involve
matching the direction of the bow imperfection in each member

1173

to the pattern of bending moments resulting from the response


of the frame to the applied loads.
Designers are likely to prefer to continue to use separate
member stability checks to account for imperfections and
second-order effects in individual members. The member
is loaded by the end moments and forces from the frame
analysis and the check is based on an effective length
factor of unity. This is straightforward but this, and indeed
second-order analysis itself, will require further education
for some designers. Although second-order analysis is the
basis for design of struts, often its application to frames is
included only in optional modules at university. Fortunately
the implementation of Eurocode 3 follows a revision of BS
5950 [7] that gave greater prominence to sway stability. As
a consequence, practising designers, and some students also,
will already have become more familiar with second-order
analysis [18,19], irrespective of the content of their university
lecture courses.
3. The MerchantRankine method
3.1. Formulae
Although the wind-moment method is to some extent
justified by the lower-bound theorem of plastic theory, it
was originally intended as an elastic method. In contrast, the
MerchantRankine formula [20] allows as full a distribution of
plastic hinges as is possible when account is taken of secondorder effects. The load level at failure, f , is assumed to be
given by:
1/ f = 1/pl + 1/cr

(1)

where pl is the load level for rigid-plastic collapse and cr is


that for elastic critical buckling.
For rigid-jointed frames, studies to verify the formula were
summarised by Horne [20]. Its generally conservative nature
encouraged Wood to propose a modified relationship, making
some allowance for strain-hardening and stray composite
action:
1/ f = 0.9/pl + 1/cr .

(2)

For cr /pl 10, second-order effects are ignored and the


failure load is taken as that for rigid-plastic collapse. Limiting
values of cr /pl have been used in design codes [1,2,7] as a
basis for neglect of second-order effects and therefore for use
of first-order analysis.
3.2. Practical application
Clearly, verification of a trial design depends on ability to
calculate pl and cr . The most convenient procedure is to
determine cr by an approximate method e.g. [1,7]. Taking f
as the design load level, Eq. (1) or (2) can be solved for the
required value of pl . In this way exact calculation of the rigidplastic load for the trial design can sometimes be avoided, use
being made instead of the lower-bound theorem.
The MerchantRankine approach could continue to be used
for design of rigid-jointed multi-storey steel frames, provided

1174

D. Anderson / Journal of Constructional Steel Research 62 (2006) 11711177

that application is limited to frames of usual type [20].


Eurocode 3 [1] is not prescriptive in methods to account
for second-order effects in global analysis. However, the
calculation of cr becomes less straight-forward if joint
flexibility needs to be considered or the frame includes
composite members. Eurocode 4 [2] requires allowance to be
made for cracking and creep of concrete as well as for joint
flexibility. In general, these effects are dependent on internal
forces and moments and iteration will therefore be required to
determine their extent.
In practice, cr will need to be determined by the use of
software for global analysis. This not only puts into question
the usefulness of the MerchantRankine approach, but also
the value of checks to see if second-order analysis can be
avoided. As commercial software for second-order analysis is
readily available, the most straight-forward procedure is always
to include such effects in the global analysis.
4. Future role for simplified methods
Advocating general use of second-order analysis does not
mean that simplified methods, particularly the wind-moment
approach, have no continued value. An initial design is needed
and a simplified method can be used to obtain this. As more
exact analysis will now follow, there is no need for concern
if the structure is outside the scope of studies to justify
the method. Nor is there a need to establish an approved
interpretation of provisions from the Eurocodes to suit the
method.
Useful as such an approach will be, there is however no point
in refining section sizes and joint details if the resulting design
exceeds limits on deflection under service loading. To use the
MerchantRankine method, it was recommended [20] that the
frame already be designed so that sway due to unfactored
wind load did not exceed height/300. This paper concludes by
presenting a development of previous methods [21] that further
simplifies initial design.
5. Design to limit sway
5.1. General assumptions
It is assumed that:
beam loads have negligible effect on horizontal displacements
sway due to shortening and elongation of the columns on
opposite sides of the frame can be neglected.
5.2. Direct design
Anderson and Islam derived equations that give second
moments of area I for members of a rigid-jointed frame [21,
22]. They based their analysis on sub-assemblages, such as that
shown in Fig. 3 for an intermediate region of a regular frame.
In addition to the general assumptions, they also assumed:
a point of contraflexure exists at the mid-height of each
column (except in the bottom storey) and at the mid-length
of each beam

Fig. 3. Intermediate storey sub-assemblage for a regular frame.

the total horizontal shear is divided between the bays in


proportion to their relative widths.
These assumptions render a frame statically determinate,
except in the bottom storey, and enabled each storey to be
considered in isolation. Expressions were derived relating the
sway in a storey to the second moments of area of the storeys
columns and beams. The need to choose trial values was
avoided by introducing an element of optimisation.
In some work on optimum design, it has been assumed that
cost depends solely on the weight of structural material. To
avoid this, Anderson and Islam [22] allowed for the cost c of
a member to be represented by:
c = kI L

(3)

where L is the length of the member, I its second moment of


area and k is a correction factor to allow more realistically for
the actual cost of supply, delivery and erection. Minimising the
cost of a storey led to a unique set of second moments of area
that could then be used to determine section sizes.
The principal disadvantage arose from the assumption that
horizontal shear is divided between the bays in proportion to
their relative widths. This fixed the relative values of the second
moments of area, which hindered choice of appropriate rolled
sections from the limited sizes available.
5.3. Analysis of substitute frames
An alternative approach was proposed by Wood and
Roberts [21,23]. Their method was one of analysis, with each
storey substituted by a cell comprising a single column and
upper and lower beams (Fig. 4). For convenience, the spans
L t and L b may be taken as equal. The second moment of
area of each member was based on summation of the values
of the corresponding members across the storey. Although the
method did not generate required section properties, neither did
it prescribe fixed relative values. In this method, the sway
over a storey height h is given by:
/ h = Fh/(12E K c )

(4)

D. Anderson / Journal of Constructional Steel Research 62 (2006) 11711177

1175

Substituting Eq. (10) into Eq. (6) and then from Eq. (5):
= (3K c + K b )/K b .

(11)

Substituting Eq. (11) into Eq. (4) and re-arranging:


K c = f K b /(K b 3 f )

(12)

where:
f = Fh/(12E/ h).

The optimisation is based on that of Anderson and


Islam [22]. The objective function Z to be minimised is taken
as:

Fig. 4. Substitute frame.

where:

Z = Ic h + 0.5Ibt L + 0.5Ibb L

is a non-dimensional sway index


F is the horizontal shear in the storey
E is the modulus of elasticity and
K c is the stiffness of the substitute column.
For an unbraced frame in which the cladding stiffness is
zero, the sway index is given by:
= 1 + [3(kb + kt kb kt )/(4 3kb 3kt + 2kb kt )]

(5)

kt and kb are distribution factors for the storey concerned.


Assuming that each beam restrains the column above and below
to the same extent, these factors are defined as:
kt = (K c )/(K c + 0.5K bt )

(6a)

kb = (K c )/(K c + 0.5K bb )

(6b)

where K bt and K bb are the stiffnesses of the substitute beams


at the top and bottom of the storey.
The stiffness K c is given by:
X
Kc =
(Ic / h)
(7)
while it can be shown [21] that the stiffness K b is given by:
X
Kb =
(3Ib /L).
(8)
In Eqs. (7) and (8), the summation is over all the relevant
members in the real frame at the level being considered; the
second moments of area I and the beam spans L are those of
the members of that frame.
Values of have been plotted against kt and kb [21,23].
5.4. Direct design using the substitute frame
This analysis has been developed into a method for
direct design of rectangular frames by assuming a point
of contraflexure at the mid-height of the column and by
introducing an approximate optimisation [24].
By imposing a central point of contraflexure:
kt = kb = k

(9)

which implies that for a single storey:


K bt = K bb = K b .

(13)

(10)

(14)

where L is the span of the beams in the substitute frame, taken


as the average span of the beams in the real frame. The second
moments of area I are those of the members of the substitute
frame (Fig. 4). The factor 0.5 is introduced so that the objective
function for the complete frame is the sum of those for the
individual storeys. The factor is taken as 0.5 because it has been
assumed that each beam restrains the column above and below
to the same extent.
Eq. (14) can be re-written in terms of the stiffnesses defined
in Eqs. (7) and (8). By also taking account of Eq. (10), it follows
that:
Z = K c h 2 + (K b /3)L 2 .

(15)

Substituting for K c from Eq. (12), differentiating with respect


to K b and equating to zero for minimum Z :
K b = 3 f (1 + h/L).

(16)

From Eq. (12):


K c = f (1 + L/ h).

(17)

5.5. Application
The very simple Eq. (17) can be applied to each storey in
turn. However, as usual practice in multi-storey frames is to
run the same section through two or more storeys Eq. (17) will
only need to be applied to the lowest storey in each column lift.
Sections for the columns in the real frame can then be chosen
using Eq. (7).
As it stands, Eq. (16) leads to the same beam stiffness at the
top and the bottom of the storey. The use of the same sections at
two levels will increase repetition in fabrication and therefore
could be desirable. If Eq. (16) were to be applied to adjacent
storeys, two values of stiffness would be given for each floor
beam. This difficulty does not arise if the equations are only
applied to alternate storeys.
An alternative is to take the value for Ib calculated from Eq.
(16) and to treat this as an average value for the top and bottom
beams. Anderson and Islam [21,22] developed a relationship
between the top and bottom beams in a storey (I1 and I2 in
Fig. 3), recognising that a floor beam contributes to the stiffness
of adjacent storeys and accounting for the increasing horizontal
shear as one proceeds from storey to storey down the frame

1176

D. Anderson / Journal of Constructional Steel Research 62 (2006) 11711177

(and also any variation in storey height). This can be used in


conjunction with the average value of Ib to determine required
second moments of area for upper and lower beams in each
storey.
It may be objected that this would violate the assumption
that kt and kb are equal. However, in theory one would also
be changing column sizes from storey to storey, to reflect
increasing horizontal shear and any changes in storey height. It
has been shown [24] that kt and kb then remain equal provided
that they are redefined to account for continuity in connection
between storeys [7,21,23].
Anderson and Islam [22] provided special equations for the
top storey of a frame, where the roof beams restrain only the
columns beneath, and for the bottom two storeys, with fixed
or pinned bases. The equations to control top storey sway
were found to be unnecessary, because with low horizontal
shear the beam sections were governed by resistance to vertical
load; column sizes were usually governed by the design of
the storey(s) below. With pinned bases, it is difficult to limit
the sway of the bottom storey to reasonable values and such
conditions should be avoided if possible. With fixed bases, it
was found that usually the sections are governed by the sway
limit between first and second floor. For these reasons the Wood
and Roberts model has not been developed to account specially
for these extremities to a frame.

concrete. Although rigorous second-order analysis is not


without difficulties [25], it is more suitable for the future than
approximate methods of limited scope developed mainly with
hand calculation in mind. Courses, both at undergraduate level
and as continuing professional development, need to provide a
wider understanding of rigorous methods.
Approximate methods that are easy to apply do still have
uses though, particularly to generate initial section sizes and
connection details. These may be governed by the need to
control deflection. A very simple method has been presented
for design to limit sway in unbraced multi-storey frames.

5.6. Scope for further work

[1] BS EN 1993-1-1. Eurocode 3: Design of steel structures: General rules


and rules for buildings. London: British Standards Institution; 2005.
[2] BS EN 1994-1-1. Eurocode 4: Design of composite steel and concrete
structures: General rules and rules for buildings. London: British
Standards Institution; 2005.
[3] Eurocode No. 3: Common unified rules for steel structures. Report EUR
8849. Luxembourg: Commission of the European Communities; 1984.
[4] Eurocode No. 4: Common unified rules for composite steel and concrete
structures. Report 9886. Luxembourg: Commission of the European
Communities; 1985.
[5] Livesley RK. The application of an electronic digital computer to some
problems of structural analysis. Struct Eng 1956;34(1):112.
[6] Salter PR, Couchman GH, Anderson D. Wind-moment design of low rise
frames. Publication 263, Steel Construction Institute, Ascot; 1999.
[7] BS 5950: Structural use of steelwork in building Part 1: Code of practice
for design Rolled and welded sections. London: British Standards
Institution; 2000.
[8] Nethercot DA. Joint action and the design of steel frames. Struct Eng
1985;63A(12):3719.
[9] Anderson D, Reading SJ, Kavianpour K. Wind-moment design for
unbraced frames. Publication 082, Steel Construction Institute, Ascot;
1991.
[10] Tahir MMd. Structural and economic aspects of the use of semi-rigid
joints in steel frames. Ph.D. thesis, University of Warwick; 1997.
[11] Brown ND, Anderson D, Hughes AF. Wind-moment steel frames with
standard ductile connections. J Constr Steel Res 1999;52:25568.
[12] Bose B, Hughes AF. Verifying the performance of standard ductile
connections for semi-continuous steel frames. Proc ICE Structs & Bldgs
1995;110:44157.
[13] Hensman J, Way A. Wind-moment design of unbraced composite frames.
Publication 264, Steel Construction Institute, Ascot; 2000.
[14] SCI/BCSA Connections Group. Joints in steel construction: Composite
connections. Publication 213, Steel Construction Institute, Ascot; 1998.
[15] Majid KI, Anderson D. Elasticplastic design of sway frames by
computer. Proc ICE 1968;41:70529.

It should be noted that the objective function (Eqs. (14) and


(15)) at present neglects the correction factors k previously
proposed [22] to allow for actual cost. These were omitted for
simplicity but could be included. The derivation has also been
for frames with rigid joints and only first-order behaviour has
been considered. Depending on the intended joint details and
other aspects, this may well be accurate enough for calculations
under service loading, particularly as the primary purpose of
the proposals is to generate an initial design, recognising that
refinement can be undertaken using software that permits a
more rigorous approach. The flexibility of semi-rigid joints
could though be allowed for by using reduced effective beam
stiffness and enhanced horizontal loads used to account for
second-order effects due to sway. The latter are easy to
determine because the sway can be taken as the intended
limiting value.
6. Conclusions
Architectural preference for little or no bracing encourages
use of sway frames but these commonly experience significant
second-order effects under ultimate design loads. Approximate
analysis methods have remained popular but it is questionable
whether resources should be allocated to justify their use
with the Eurocodes. There is much software available that
is capable of carrying out rigorous second-order analysis.
Further investment in software development will be made
to assist introduction of the new codes, including possible
non-linear behaviour of joints and cracking and creep of

Acknowledgements
The author acknowledges the contributions made by former
postgraduate students who have worked with him on the topic
of multi-storey frames: Dr. N.D. Brown, Dr. M.A. Islam,
Dr. K. Kavianpour, Mr. S.J. Reading, Dr. A.A. Saim and
Dr. M.Md. Tahir; and by the late Professor M.R. Horne, Mr.
A.F. Hughes, Professor K.I. Majid, and the late Dr. R.H. Wood,
who encouraged the author to carry out research on this topic.
References

D. Anderson / Journal of Constructional Steel Research 62 (2006) 11711177


[16] BS EN 1993-1-8 Eurocode 3: Design of steel structures: Design of joints.
London: British Standards Institution; 2005.
[17] Brown ND, Hughes AF, Anderson D. Prediction of the initial stiffness of
ductile end-plate connections. Proc ICE Structs & Bldgs 2001;146:1729.
[18] Rathbone AJ. Second-order effects who needs them? Struct Eng 2002;
80(21):1921.
[19] Brown D. Unbraced frames sway stability. New Steel Construction
2003;11(3):245.
[20] Horne MR. Frame instability and the plastic design of rigid frames.
In: Narayanan R, editor. Steel framed structures: Stability and strength.
Elsevier Applied Science; 1985.

1177

[21] Anderson D. Design of multi-storey steel frames to sway deflection


limitations. In: Narayanan R, editor. Steel framed structures: Stability and
strength. Elsevier Applied Science; 1985.
[22] Anderson D, Islam MA. Design of multi-storey frames to sway deflection
limitations. Struct Eng 1979;57B(1):117.
[23] Wood RH, Roberts EH. A graphical method of predicting side-sway in the
design of multi-storey buildings. Proc ICE Pt. 2 1975;59:35372.
[24] Saim AA. Aspects of continuity in steel and composite frames. Ph.D.
thesis, University of Warwick; 1998.
[25] Rathbone AJ. Second order analysis of 3D structures. Struct Eng 2004;
82(21):156.

S-ar putea să vă placă și