Proceedings of the Institution of

Civil Engineers
Structures & Buildings 156
May 2003 Issue SB2
Pages 141–150
Tianjian Ji Brian R. Ellis Adrian J. Bell
Paper 12770
Manchester Centre for Civil Centre for Structural Manchester Centre for Civil
Received 01/11/2001 and Construction Engineering, Building and Construction
Accepted 25/04/2002 Engineering, UMIST Research Establishment Engineering, UMIST

Keywords:
buildings, structure & design/
dynamics/structural frameworks

Horizontal movements of frame structures induced by vertical

loads
T. Ji, B. R. Ellis and A. J. Bell

This paper considers the significance of vertical loads u, u€ displacement and acceleration in the horizontal
that can produce horizontal movements of frame direction
structures. It is shown that, with a small number of Æ ratio of the column height to the span of a frame
exceptions, vertical loads can induce horizontal  rigidity ratio of beam to column
movements of symmetric, antisymmetric and ª length ratio of the left column to the right column
asymmetric frames, which represent a wide range of of a frame
engineering structures. The magnitudes of the horizontal Ł A , Ł€A rotation and rotational acceleration at node A
movements depend on both the structural form and the Ł B , Ł€B rotation and rotational acceleration at node B
location of the vertical loading. When vertical loads are 21 , 31 the second and third components of the first
applied dynamically, the movements of a structure can normal mode vector
be significantly enlarged if one of the natural frequencies
of the structure in the horizontal direction is close to 1. INTRODUCTION
one of the vertical load frequencies. These findings are When a structure moves horizontally, it is usually considered
illustrated by site measurements showing the horizontal that this is in response to horizontal loads. However, vertical
movements of a framed building induced by vertical loads can also induce horizontal movements. This is because
loading. Some implications of the findings are discussed structures are three-dimensional and movements in the
briefly for several types of structure, including cantilever orthogonal directions are often coupled. For some structures
grandstands, temporary grandstands, cable-suspended such horizontal movements can be a significant design
bridges and rail bridges, in which horizontal response consideration, especially when dynamic response is an
induced by vertical loading may need to be considered in important factor.
design.
Horizontal movements may result from the following:

NOTATION • Horizontal loading (e.g. wind loading which will generate

A1 (t) magnitude of vibration contributed by the first translational movement of tall buildings).
mode • Loading that, although primarily vertical, has a horizontal
CL load factor that relates to the type and distribution component, for example walking. The Millennium Bridge
of vertical loads in London is a structure where significant horizontal
1
CS structural factor that is a function of structural form, movements were induced by people walking.
Æ, and the distribution of member rigidities,  • Vertical loading acting on asymmetric structures. Due to
C LS equivalent horizontal load factor the structural geometry, vertical loads can induce both
EI b flexural rigidity for beam vertical and horizontal movements (i.e. vertical motion is
EI c flexural rigidity for column coupled with a horizontal response). A simple example is a
F horizontal concentrated force uniformly distributed vertical load on an inclined
M A, M B bending moments at the two ends of a fixed beam cantilever.
induced by vertical loads • Vertical loading acting asymmetrically on structures. Due
P vertical concentrated static load to their location, vertical loads can induce both vertical
P(t) vertical concentrated dynamic load and horizontal movements. An example is that of a train
PTV total vertical load crossing a bridge and producing horizontal movements
fp frequency of a harmonic load orthogonal to the rails; this will be discussed later.
h height of frame
h1 , h2 heights of the left and right columns of an This paper considers the last two situations where vertical
asymmetric frame loading can generate horizontal movements in frame
m mass density along element length structures. Initially a theoretical evaluation of the horizontal

Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al. 141
 movement of a symmetric structure subject to asymmetric 2 38 9 8 9
loading is considered and a number of load cases and EI c 4
24 6h 6h < u = < 0 =
1 6h 4h (Æ þ 1) 2h Æ
2 2 5 ŁA ¼ M A
structural combinations are examined. This is then extended to h3 : ; : ;
6h 2h2 Æ 4h2 (Æ þ 1) ŁB MB
antisymmetric and asymmetric structures and the analytical
results are verified by finite element (FE) calculations. An
equivalent horizontal load factor is determined to represent the
effect of the horizontal movements of frame structures due to
where
vertical loads. For symmetric and antisymmetric frames the
equivalent horizontal load factor can be expressed as a product
of a load factor and a structural factor. Thus the effects of the 2 Æ ¼ h=L  ¼ EI b =EI c
distribution of vertical loads and structural geometry can be
examined independently.
M A and M B are the fixed end moments of the beam arising
The examples presented consider static loading but the results from the vertical loading. The positive sign occurs when the
are equally applicable to dynamic situations although here the end moment induces clockwise rotation. Equation (1) indicates
possibility of resonance needs to be considered. An example is that the vertical loads, which cause the moments M A and M B ,
provided to illustrate that vertical dynamic loading can lead to induce rotations and horizontal displacements of the frame
a resonance that is primarily in the horizontal direction if the connections. As the stiffness matrix in equation (1) is fully
frequency of the dynamic load coincides with a horizontal populated, the horizontal displacement is coupled with the
natural frequency of the structure. To support the theoretical rotations.
findings, measurements of the response of a floor subject to a
crowd of people jumping rhythmically are presented which Expanding the first row of equation (1) gives
show that the vertical and horizontal responses of the floor
occur at the same frequency as that of the vertical load.
h(Ł A þ Ł B )
3 u¼
The implications of these findings are discussed for 4
grandstands and bridges, where horizontal responses induced
by predominantly vertical loading may be key design
considerations. from which it can be seen that u is zero only when Ł A ¼ Ł B .
This occurs when symmetric loads are applied to the beam.
2. STATIC RESPONSE Solving equation (1) gives the horizontal movement of the
frame due to the vertical loads as
2.1. A symmetric system
Consider a simple symmetric frame with no horizontal forces
but subjected to any form of vertical load, such as a (M A þ M B ) h3
4 u¼
concentrated vertical load on a beam as shown in Fig. 1. The 4(6Æ þ 1)ÆL EI c
beam has a length of L and rigidity of EI b , and the two
columns have the same length of h and rigidity of EI c .
The negative sign indicates that the movement of the frame is
If the axial deformations of the columns and the beam of the towards its left. If a horizontal force to the left, F, is applied
frame are negligible, the structure has three degrees of at one of the beam–column connections instead of the vertical
freedom, the horizontal displacement, u, and the rotations, Ł A load, solution of equation (1) gives the horizontal displacement
and Ł B , at the connections of the beam and columns. Thus the as
equations of static equilibrium of the frame are given by:
(3Æ þ 2)F h3
5 u¼
12(6Æ þ 1) EI c

For the same horizontal displacement at the

beam–column connections of the frame,
equating equations (4) and (5) gives
h

(M A þ M B )12(6Æ þ 1)
F¼
4(6Æ þ 1)(3Æ þ 2)ÆL
(M A þ M B ) 3
L ¼ PTV
6 LPTV (3Æ þ 2)Æ
(a) (b)
¼ C L C S PTV ¼ C LS PTV

Fig. 1. A symmetric frame subject to an asymmetric vertical load:

(a) frame and loading; (b) deformations
in which

142 Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al.
 MA þ MB  ¼ 0:5  ¼ 1:0  ¼ 2:0
7 CL ¼
LPTV
Æ ¼ 0:5 2·1818 1·7143 1·2000
Æ ¼ 1:0 0·8570 0·6000 0·3750
Æ ¼ 2:0 0·3000 0·1875 0·1071
3
8 CS ¼
(3Æ þ 2)Æ
Table 2. The structure factor, CS , for different ratios
of length and rigidity for a symmetric system

9 C LS ¼ C L C S

h ¼ 6:0 m, L ¼ 6:0 m, E ¼ 30 3 109 N=m2 , I b ¼ I c ¼

where PTV is the total vertical load; C L is a load factor that 0:254 =12 ¼ 3:255 3 104 m4 and P ¼ 100 kN (acting on a
relates to the type and distribution of vertical loads; C S is a quarter of the length of the beam from the right end).
structural factor that is a function of structural form, Æ, and
the distribution of member rigidities, ; and C LS is the Figure 1(b) shows the deformed shape of the frame subject to
equivalent horizontal load factor. the concentrated vertical load. The horizontal displacements
2
calculated using equation (5) and the finite element method
It can be seen that the smaller the values of Æ and  the larger are 7:406 mm and 7:405 mm, respectively.
the structural factor. It should also be noted that the load
factor and the structural factor are independent for this case. The combined effect of the load and structural factors is shown
Equation (6) indicates that the equivalent horizontal load can in Table 3, which provides the equivalent horizontal load factor
be expressed as a product of the load factor, the structural for three different load distributions on frames with Æ ¼ 0:5,
factor and the total vertical load. Table 1 provides values of the 1·0 and 2·0, and  ¼ 0:5, 1·0 and 2·0.
load factor for several vertical load distributions on the beam
with two fixed ends. Table 2 shows the structural factor for a From Tables 1, 2 and 3 it can be concluded that:
range of geometry and rigidity ratios.
• The magnitude of the horizontal displacement induced by
Consider a particular case where Æ ¼ 1,  ¼ 1 and a vertical loads (equation (4)) or the equivalent horizontal
concentrated load, P, acts at a quarter of the length of the load (equation (6)) depends on the load distribution and the
beam as shown in Fig. 1. The moments in equation (1) are structural form.
• The structural factors are significantly larger than the load
3Ph 9Ph factors.
10 MA ¼  MB ¼ • The smaller the values of Æ and , the larger the equivalent
64 64
horizontal load (i.e. if the frame is relatively low and has a
relatively large span, and/or the rigidity of the beam is
Substituting these into equation (6) gives smaller than that of the column, the frame will be
subjected to a relatively large equivalent horizontal). Hence
63P the load equivalent horizontal load for a taller frame is
11 F¼ ¼ 0:05625 P smaller than that for a similar lower frame if both are
1120
subjected to the same vertical loading.
• The height/length ratio, Æ, is more significant than the
In this case the effect of the vertical load, P, on the horizontal rigidity ratio, , in determining the magnitude of the
movement of the frame is equal to that of a horizontal load of horizontal movement.
5:625% P. • The horizontal movement of a frame due to vertical loads
is zero only when M A ¼ M B (i.e. when concentrated
2.1.1. Example 1: Consider the frame shown in Fig. 1(a) with loads act on the beam–column joints or when a symmetric
load is applied to the
beam.

Load distribution MA MB CL
2.2. An antisymmetric
system
Uniformly distributed load over full length qL2 =12 qL2 =12 0 If the left column of the
Concentrated load acting at a quarter of the 3PL=64 9PL=64 3=32 frame shown in Fig. 1(a) is
span from the right rotated through 1808 around
Uniformly distributed load over a half of the 5qL2 =192 11qL2 =192 1=16 its connection to the beam, it
span from right becomes antisymmetric as
Uniformly distributed load over three quarters 63qL2 =1024 81qL2 =1024 3=128 shown in Fig. 2(a). The
of the span from right equivalent horizontal load
can be found, as in section
Table 1. The load factor, C L , for different load distributions for a symmetric system 2.1, by solving its
equilibrium equations as

Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al. 143
 In addition to the conclusions
drawn from section 2.1,
which are also valid for the
antisymmetric system, it can
be deduced that:

• The load factor and

 ¼ 0:5  ¼ 1:0  ¼ 2:0  ¼ 0:5  ¼ 1:0  ¼ 2:0  ¼ 0:5  ¼ 1:0  ¼ 2:0
structural factor for the
antisymmetric systems
Æ ¼ 0:5 0·2045 0·1607 0·1125 0·1364 0·1071 0·0750 0·0511 0·0402 0·0281 are significantly larger
Æ ¼ 1:0 0·0804 0·0563 0·0352 0·0536 0·0375 0·0234 0·0201 0.0141 0.0088
Æ ¼ 2:0 0·0281 0·0176 0·0100 0·0188 0·0117 0·0067 0·0070 0·0044 0·0025 than those for the
symmetric system. Hence,
the magnitude of the
Table 3. The equivalent horizontal load factor, C LS , for a symmetric system
horizontal movement due
to vertical loads depends
on the structural form.
• Equation (12) indicates
that the antisymmetric frame has no horizontal movement
when M A ¼ M B , which requires a particular distribution
of antisymmetric vertical loading. For any other vertical
loading situation, there will be a resulting horizontal
h movement.

Example 2: Consider the frame shown in Fig. 2(a) with similar

data to that used for Example 1: h ¼ 6:0 m, L ¼ 6:0 m,
E ¼ 30 3 109 N=m2 , I b ¼ I c ¼ 0:254 =12 ¼ 3:255 3 104 m4
and P ¼ 100 kN (acting on a quarter of the length of the beam
h from the right end).

The equivalent horizontal load can be evaluated using equation

(12) as 18:75 kN. The finite element method is used to calculate
the horizontal displacements induced by the vertical load of
L
100 kN and the horizontal load of 18·75 kN, respectively. These
(a) (b) displacements are identical and have a value of 34:56 mm.
Fig. 2(b) shows the deformed shape of the frame subject to the
Fig. 2. An antisymmetric frame subject to an concentrated vertical load.
asymmetric vertical load: (a) frame and loading; (b)
deformations 2.3. An asymmetric system
If the lengths of the columns of the frame shown in Fig. 1(a)
are different, the frame becomes asymmetric as shown in Fig.
3(a). The ratios given in equation (9) are redefined as follows:
(M B  M A ) 12(2Æ þ 1)
F¼
4(2Æ þ 1)ÆL (Æ þ 2)
15 Æ ¼ h1 =L  ¼ EI b =EI c ª ¼ h1 =h2
12
(M B  M A ) 3
¼ PTV ¼ C L C S PTV ¼ C LS PTV
LPTV (Æ þ 2)Æ
and the equivalent horizontal load becomes

where 3[(Æ(2  ª2 ) þ 2ª]M A þ 3[Æ(2ª2  1) þ 2ª2 ]M B

F¼
ÆL[4(Æ þ 1)ª þ Æ(3Æ þ 4)]PTV
16
MB  MA
13 CL ¼ 3 PTV ¼ C LS PTV
LPTV

3 where
14 CS ¼
(Æ þ 2)Æ
3[(Æ(2  ª2 ) þ 2ª]M A þ 3[Æ(2ª2  1) þ 2ª2 ]M B
17 C LS ¼
ÆL[4(Æ þ 1)ª þ Æ(3Æ þ 4)]PTV
and C LS is defined by equation (9). Equations (12), (13) and
(14) have the same form as equations (6), (7) and (8). For
comparison similar tables for the load factor, structural factor C LS is the equivalent horizontal load factor, which is a
and equivalent horizontal load factor of the antisymmetric function of load distribution, location and structural form. In
frame are given in Tables 4, 5 and 6. contrast to the symmetric and antisymmetric frames considered

144 Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al.
 The equivalent horizontal
Load distribution MA MB CL
load for this case is evaluated
using equation (16) and is
Uniformly distributed load over full length qL2 =12 qL2 =12 1=6 15:73 kN. The horizontal
Concentrated load acting at a quarter of the 3PL=64 9PL=64 3=16 displacements calculated
span from the right using the FE method for the
Uniformly distributed load over a half of the 5qL2 =192 11qL2 =192 1=6 vertical and horizontal loads
span from right have the same value of
Uniformly distributed load over three quarters 63qL2 =1024 81qL2 =1024 3=16 10:59 mm. Fig. 3(b) shows
of the span from right the deformed shape of the
frame.
Table 4. The load factor, C L , for different load distributions for an antisymmetric
system Comparing the results in
Tables 3 and 7, it can be seen
that the equivalent horizontal
load factors for the
asymmetric frame are significantly larger than those for the
 ¼ 0:5  ¼ 1:0  ¼ 2:0
symmetric frame. This again shows that structural form affects
the magnitudes of horizontal movements of frame structures
Æ ¼ 0:5 2·6667 2·4000 2·0000 subject to vertical loads.
Æ ¼ 1:0 1·2000 1·0000 0·7500
Æ ¼ 2:0 0·5000 0·3750 0·2500
2.4. Further comparison
Table 8 summarises the ranges of the equivalent horizontal
Table 5. The structure factor, CS , for different ratios
of length and rigidity for an antisymmetric system load factors for the three types of frame subject to three types
of vertical loading limiting the variations of Æ and  between
0·5 and 2·0. From Table 8 it can be seen that:

in the previous two sections, the load factor and the structural • The equivalent horizontal load factors for the
factor are coupled for the asymmetric frame. antisymmetric frame have the largest values, but this type
of structure may not be common.
Consider ª ¼ 3=2. The equivalent horizontal load factors for • The equivalent horizontal load factors of the asymmetric
the same loading cases, length ratios and rigidity ratios, as for frame are at least double those of the symmetric frame for
the symmetric and antisymmetric frames, are given in Table 7. the same loading conditions.

3. DYNAMIC RESPONSE
Example 3: Consider the frame shown in Fig. 3(a) with When a structure is subjected to cyclic dynamic loading,
h1 ¼ 6:0 m, h2 ¼ 4:0 m, L ¼ 6:0 m, E ¼ 30 3 109 N=m2 , resonance may occur with a consequent, and potentially
I b ¼ I c ¼ 0:254 =12 ¼ 3:255 3 104 m4 and P ¼ 100 kN significant, increase in response. The possibility of vertical
(acting on a quarter of the length of the beam from the right loading resulting in a resonant horizontal response therefore
end). must be considered.

Consider the frame discussed

in section 2.1 and shown in
Fig. 1(a) subjected to a simple
sinusoidal vertical load, P(t),
with maximum amplitude P0 :

18 P(t) ¼ P0 sin 2 f p t

where f p is the frequency of

the load and t is time. The
mass densities for the columns
and the beam are assumed to
be m and 10m respectively,
 ¼ 0:5  ¼ 1:0  ¼ 2:0  ¼ 0:5  ¼ 1:0  ¼ 2:0  ¼ 0:5  ¼ 1:0  ¼ 2:0
with the high density of the
beam representing added
Æ ¼ 0:5 0·5000 0·4500 0·3750 0·4444 0·4000 0·3330 0·5000 0·4500 0·3750 loads that may arise from
Æ ¼ 1:0 0·2250 0·1875 0·1406 0·2000 0·1667 0·1250 0·2250 0·1875 0·1406
floors, etc. The equation of the
Æ ¼ 2:0 0·0938 0·0703 0·0469 0·0833 0·0625 0·0417 0·0938 0·0703 0·0469
undamped forced vibrations
Table 6. The equivalent horizontal load factor, C LS , for an antisymmetric system of the frame is, when
Æ ¼  ¼ 1:0,

Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al. 145
 joints respectively. The response in the first mode of the
3
frame is

21 M A þ 31 M B 1
h2 20 A1 (t) ¼ sin 2 f p t
K1 1  ( f p = f 1 )2
h1

where A1 (t) is the amplitude of the horizontal motion of the

frame and 21 M A þ 31 M B is the modal load for the first
mode, which acts in the horizontal direction. Equation (20)
L indicates that if the modal load is not equal to zero and the
(a) (b) load frequency, f p , is close to the fundamental natural
frequency, f 1 , the vertical load will induce resonant vibration
of the frame in the horizontal direction. This conclusion can be
Fig. 3. An asymmetric frame subject to a
symmetric vertical load: (a) frame and loading; (b) verified numerically.
deformations
Example 4: Consider the frame defined in example 1 with
m ¼ 2400 kg=m3 3 (0:25 m)2 ¼ 150 kg=m, P ¼ P0 sin 2 f p t
and P0 ¼ 100 kN.

2
Dynamic analysis was carried out using LUSAS with the
2 38 9 critical damping set to zero. Fig. 5 shows the time history of
4512 22h 22h > u€ >
>
< > =
mh 6 7 the horizontal motion of the frame, up to 10 s, due to the
6 22h 30h2 7 €
420 4
44h2 5> Ł A > vertical load. A typical resonance situation is encountered.
>
: > ;
22h 30h2 44h2 Ł€B
2 38 9 Although the example is simple, it illustrates the important
24 6h 6h > u >
> > phenomenon that if the frequency of a vertical load is close to
EI c 6 7< =
6 6h 8h2 2h2 75> Ł A >
19 þ one of the horizontal natural frequencies of a structure,
h3 4 >
: > ; resonance in the horizontal direction can occur as a result of
6h 2h2 8h 2
ŁB
vertical excitation. This situation should be recognised in the
8 9
> 0 > design of some structures.
>
< >
=
¼ M A sin(2 f p t)
>
> >
> The necessary condition for no horizontal movement occurs
: ;
MB when the vertical loads are applied either symmetrically on
the beam or at the beam–column joints. For any other
distributions of vertical dynamic loads, resonance can occur in
The elements in the mass matrix are obtained in the same the horizontal direction.
manner as those in the stiffness matrix. The mode shapes and
frequencies of the structure can be found by solving the 4. SITE MEASUREMENTS
eigenvalue problem associated with equation (19). Taking the In the introduction the example of an inclined cantilever was
mass density, m, equal to 150 kg=m and other data as used in used to provide an illustration where a vertical load can lead to
example 1, the three natural frequencies of the frame are both vertical and horizontal movements. With cantilever
1:39 Hz, 5:00 Hz and 14:5 Hz, and the corresponding mode grandstands the potential problems induced by people jumping
shapes are shown in Fig. 4. The first mode shows horizontal or bouncing in time to music is topical. This type of cyclic
movements of the frame while the two other modes give loading induces a characteristic type of response, which occurs
symmetric and antisymmetric rotations of the beam–column at the load frequency (the jumping frequency) and whole
number multiples thereof. So
when the structural response
is presented as a spectrum
(i.e. response plotted against
frequency), the response at
specific frequencies is shown.
Measurements on a cantilever
grandstand are presented in
reference 4 and this
 ¼ 0:5  ¼ 1:0  ¼ 2:0  ¼ 0:5  ¼ 1:0  ¼ 2:0  ¼ 0:5  ¼ 1:0  ¼ 2:0
characteristic response could
be seen in both the vertical
Æ ¼ 0:5 0·4269 0·3800 0·3146 0·3197 0·2892 0·2442 0·2251 0·2162 0·1952
and the horizontal directions.
Æ ¼ 1:0 0·1900 0·1573 0·1184 0·1446 0·1221 0·0938 0·1081 0·0976 0·0796
Æ ¼ 2:0 0·0786 0·0592 0·0400 0·0616 0·0469 0·0322 0·0488 0·0398 0·0285
For this paper, another,
Table 7. The equivalent horizontal load factor, C LS , for an asymmetric system perhaps less obvious,
example is considered. A

146 Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al.
 Although this is given as an
Symmetric frame Antisymmetric frame Asymmetric frame
example of vertical loading
on an asymmetric structure,
Concentrated load acting at a 0·2045 – 0·0100 0·5000 – 0·0469 0·4269 – 0·0400 it is worthwhile considering
quarter of the span from the
right whether people jumping
actually generate horizontal
Uniformly distributed load over a 0·1364 – 0·0067 0·4444 – 0·0417 0·3197 – 0·0322
forces. For example, it is
half of the span from right
recognised that people
Uniformly distributed load over 0·0511 – 0·0025 0·5000 – 0·00469 0·2251 – 0·0285 walking actually generate a
three quarters of the span from
right horizontal force, normal to
the direction of walking, of
approximately 10% of the
Table 8. Comparison of the ranges of the equivalent horizontal load factor
vertical dynamic force. This
is a result of a horizontal
push from each foot with
each step; a process of
continued correction to
maintain a balanced progress.
Equally an individual
jumping will produce a small
horizontal force simply to
correct for any lateral
movement and so maintain a
selected jumping location.
However, with a group
jumping on a level floor, the
(a) (b) (c) overall sum of these apparent
random horizontal forces
Fig. 4. The three normal modes of the symmetric frame: (a) horizontal movement; must tend to zero as the
(b) symmetric rotation; (c) antisymmetric rotation
number of people jumping
increases.

0·4
0·3
0·2
Displacement: m

0·1
0
–0·1
–0·2
–0·3
–0·4 Time: s

Fig. 5. Resonant response of the frame

Fig. 6. 64 students jumping on a floor in response to

composite floor of area 9 m by 6 m, was tested and the music
structural response was measured for a group of people
jumping. Sixty-four students, evenly distributed over the floor,
were asked to jump in time to a musical beat (Fig. 6). At the
centre of the floor, the vertical acceleration was recorded for 5. IMPLICATIONS
just over 16 s, as was the horizontal acceleration in the It has been shown that vertical loading can produce both
direction orthogonal to the direction in which the students vertical and horizontal motion and, if the loading is dynamic,
were facing. The peak vertical acceleration was 0:48g and resonance can occur in either the horizontal or vertical
the corresponding horizontal acceleration was 0:03g. The direction. Actual frame structures will be more complicated
autospectra for these records are shown in Fig. 7, and the than the simple frames studied, but they can be considered as
characteristic response can be seen in both directions. The test an assembly of these basic units and will possess the features
area was in fact part of the much larger flooring system shown revealed in sections 2 and 3. It is therefore useful to consider
in Fig. 8 and the loading was thus applied asymmetrically on the implications of these findings for some common types of
the whole structure, which induced the horizontal motion. structure.

Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al. 147
 and front-to-back vibration
of the cross-section of a
2·0 grandstand in one typical
mode of vibration. It can be
seen that the front-to-back
movement is larger than the
Spectral acceleration: g2/Hz × 1000

vertical movements of the

1·5
two tiers for this particular
mode. Based on this
observation and the finding
in section 3, the frame is
1·0 likely to experience
resonance in the front-to-
back direction if one of the
frequencies of vertical
0·5 loading on a tier is close to
the natural frequency
associated with the mode,
even though the vertical
movement will be small.
0
The fact that the whole
(a)
0·0010
grandstand moves in this
mode means that the modal
mass will be large and hence
the resonance may not
0·0008 always lead to excessive
movement.
Spectral acceleration: g2/Hz × 1000

Although the coupling

0·0006 between vertical movements
and horizontal (front-to-
back) movements is easily
understood, measurements
0·0004
have indicated that sway
(side-to-side) movements can
also be induced by vertical
loading on some grandstands.
0·0002
The significance of the sway
movement due to vertical
loads depends on the
0 structural form. To date there
0 2 4 6 8 10 have been few measurements
Frequency: Hz of sway movement of
(b)
permanent grandstands, but
this, somewhat less obvious
Fig. 7. The acceleration spectra for 64 people jumping on a floor: (a) vertical coupling, should not be
direction; (b) side-to-side direction
overlooked.

As a simple alternative to
evaluating structural
Although the mechanisms have been illustrated using static response, frequency limits are sometimes given suggesting
models, it is likely to be the dynamic situations that are of that structures with fundamental frequencies above the limit
principal concerns, especially when resonance occurs. Hence, will not encounter problems from the specific form of
most of the following examples relate to cyclic dynamic loading. For example, guidance was issued for permanent
loading. grandstands indicating that structures with a vertical
5
frequency above 6 Hz should be suitable for concerts. The
5.1. Cantilever grandstands rationale was that this avoids resonance from the first or
The inclined cantilever was used in the introduction to second Fourier components of the cyclic loading, as the
illustrate a simple example of vertical loading inducing higher Fourier components have not been observed to cause
horizontal movement. Cantilever grandstands are frequently problems on this type of structure. However, if this logic is
subjected to dynamic crowd loads, which at certain events, like applied to horizontal movement, a difficulty arises and this
concerts, are cyclic in nature. Fig. 9 shows the coupled vertical is discussed in the next section.

148 Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al.
 A B C D E F
45 000
9000 9000 9000 9000 9000
2000 1200800 5000 9000 1050 2650 800 1850 1850 800 9000 5000 800 3200

North

4 4
3000

3000
6000

6000
3000

3000
3 3
2500

1500 500 2500

4500

4500
1000 700 1500 500

21 000
21 000

700
2000 1000 800
800
4500

4500
2000

2 2

3000
3000

6000
6000

3000
3000

Test area
1 1

2000 1200 800 5000 9000 1050 2650 800 1850 1850 800 9000 5000 800 3200

9000 9000 9000 9000 9000

45 000

A B C D E F

Fig. 8. Floor plan of the test building showing the test area

limit to avoid resonance needs to consider both vertical and

horizontal forms of loading.

When horizontal movements of a grandstand subject to human

loads are observed, they are likely to be induced by vertical
components of human loads rather than their horizontal
components. The study in section 2 shows that the equivalent
horizontal load factor for an asymmetric frame (Table 8) is
larger than the ratio of the horizontal to vertical components
of the loads. From measurements on many temporary
6
grandstands, Littler has shown that the largest horizontal
accelerations exhibited the same frequencies as the vertical
loading similar to the measurements of the floor response to
jumping.

5.3. Cable-suspended bridges

Fig. 9. Typical mode of vibration of a frame model of
Consider a cross-section of a cable-suspended bridge where the
a cantilever grandstand showing coupled vertical and
front-to-back movements two cables are perpendicular to the bridge deck and a vertical
load is applied asymmetrically on the deck as shown in Fig.
10(a). It can be shown that there is no horizontal movement
under the vertical load. This is because the vertical and
5.2. Temporary grandstands horizontal movements are not coupled and the vertical load
Temporary grandstands may also encounter dynamic crowd produces vertical movements, which are due to elastic
loads but here horizontal motion in the sway direction is elongation of the cables.
usually critical. To avoid dynamic problems the fundamental
natural frequencies of the structures should be greater than the If the two cables are inclined as shown in Fig. 10(b), it can be
related load frequencies. For these structures it has been shown that the deck will experience both horizontal and
suggested that sway motion of the crowd will generate rotational movements when it is subjected to an
5
horizontal loads with a frequency up to 0·9 Hz. However, if asymmetrically applied vertical load. Due to the inclination of
vertical jumping or bouncing is encountered this would the cables the vertical and horizontal movements of the deck
generate horizontal motions in a higher frequency range, are coupled. Thus an asymmetrically applied vertical load will
7
typically between 1·8 and 2·3 Hz. . Thus setting a frequency induce both vertical and horizontal movements, which are

Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al. 149
 occurs. Therefore it is appropriate to check horizontal as well
as vertical natural frequencies of bridges to ensure that both
are above the likely loading frequencies associated with trains
running at higher speeds.

6. CONCLUSIONS
The paper has examined the horizontal movements of frame
systems induced by vertical loads. The significance of such
horizontal movements is represented by the equivalent
Fig. 10. Cross-sections of cable-suspended bridges horizontal load factor. It is concluded that:

(a) With few exceptions, vertical loads acting on frame

structures induce horizontal movements of the structures.
mainly due to the geometry of the system. This can be Exceptions are symmetric structures subject to symmetric
demonstrated using a simple model. Fig. 11 shows a metal vertical loads and (rarely) antisymmetric structures subject
plate held by two inclined strings. When a concentrated load to antisymmetric loads.
acts at a quarter point of the plate, it produces horizontal and (b) The magnitudes of the horizontal movements of frame
rotational movements. structures due to vertical loads depend on the load
distribution and the structural geometry.
The behaviour of cable-suspended bridges is different to that of (c) Structural form is more significant than load distribution to
the frame structures studied in sections 2 and 3. However, the magnitude of the horizontal movements.
cable-suspended bridges are likely to be more sensitive to (d) The taller the frame, the smaller the equivalent horizontal
horizontal movements induced by vertical loads than frame loading.
structures. Thus it is pertinent to examine the horizontal (e) When the frequency of a vertical dynamic load is close to
movements induced by both vertical and horizontal loads. one of the natural frequencies of a structure in its
horizontal direction, resonance in the horizontal direction
5.4. Railway bridges can occur.
Horizontal movements of some railway bridges in China have (f ) It is likely to be dynamic load situations that are of
been observed due to the increasing speed of trains and a principal concern, especially when resonance may occur.
number of bridges are now being reassessed for safety. As there
are often two or more rail tracks on a bridge, the loading from These findings may be useful when examining the horizontal
any one train is effectively asymmetrical on the structure and response of structures, such as grandstands and bridges, which
hence horizontal motions as described in section 2.1 are are subjected to predominantly vertical loads. The implications
generated. There will also be some horizontal forces generated of the findings in practical structures have been discussed. An
by lateral movement of the railway vehicles, even along awareness of these findings may help to identify and avoid
straight tracks. With the increasing speed of trains, the loading some potential problems.
frequency will increase and this may be a problem if resonance
REFERENCES
1. DALLARD P., FITZPATRICK T., FLINT A., LOW A., SMITH R. R. and
WILLFORD M. Pedestrian induced vibration of footbridges.
The Structural Engineer, 2000, 78, No. 23/24, 13–15.
2. LUSAS. Manual. FEA Ltd., 1995.
3. CLOUGH R. W. and PENZIEN J. Dynamics of Structures.
McGraw-Hill, 1975.
4. ELLIS B. R., JI T. and LITTLER J. The response of grandstands
to dynamic crowd loads. Proceedings of the Institution of
Civil Engineers: Structures and Buildings, 2000, 140,
No. 4, 355–365.
5. THE INSTITUTION OF STRUCTURAL ENGINEERS. Dynamic
Performance Requirements for Permanent Grandstands
Subject to Crowd Action: Interim Guidance for Assessment
and Design. ISE, 2001.
6. LITTLER J. D. Temporary Demountable Grandstands: Dynamic
Response. Building Research Establishment, 2000, IP 3/00.
7. GINTY D., DERWENT J. M. and JI T. The frequency ranges of
Fig. 11. Demonstration of horizontal movement
induced by a vertical load dance-type loads. The Structural Engineer, 2001, 79, No. 6,
27–31.

Please email, fax or post your discussion contributions to the secretary by 1 November 2003: email: [email protected];
fax: þ44 (0)20 7799 1325; or post to Daniela Wong, Journals Department, Institution of Civil Engineers, 1–7 Great George Street,
London SW1P 3AA.

150 Structures & Buildings 156 Issue 2 Movements of frame structures Ji et al.