Proceedings of the Institution of

Civil Engineers
Structures & Buildings 160
April 2007 Issue SB2
Pages 65–72

Paper 14648
Received 10/02/2006
Accepted 24/11/2006
Keywords: Michael Willford Peter Young Caroline Field
bridges/design methods & aims/ Director, Arup, London, UK Associate, Arup, London, UK Associate, Arup, San
dynamics Francisco, USA

Predicting footfall-induced vibration: Part 1

M. Willford MA, CEng, MIMechE, P. Young MEng, CEng, MIMechE and C. Field MSc, CEng, MICE

The current paper reviews different methods that are 2. HISTORICAL BACKGROUND AND CURRENT
presently available for the prediction of footfall-induced INDUSTRY PRACTICE
vibration in structures. Five such methods are The issue of vibration serviceability for structures has been
considered in detail and it will be shown that these vary acknowledged for many years now. Early work by Smith 8 and
considerably in formulation, applicability and predicted others related to footbridges and this led to the performance
response. Where possible, the theoretical basis, criterion that still forms the basis of BS 5400. 4 Little
complexity, robustness, rigour, reliability and limitations consideration was given to the dynamic performance of floors
of each are considered. This paper is presented in two in buildings until steel composite design became more
parts. Part 1 deals with bridges and floors having vertical prevalent in the 1960s and 1970s. Since then, architectural and
natural frequencies below 10 Hz, part 2 deals with floors economic considerations have led to floors having longer spans
whose first vertical natural frequency is over 10 Hz. and lower weight, and the issue of vibration serviceability has
become much more significant.

1. INTRODUCTION 2.1. Bridges

An important serviceability consideration in the design of For bridges, the BS 5400 methodology is based on the
many structures is the vibration induced by individuals as they calculation of resonant response of the bridge to excitation by
move around on them. Vibration may be noticed by other a single pedestrian walking at the most critical footfall rate.
people and, if excessive, can detract from the perceived quality Equations are given for simply spanning structures, but for
of a building or a bridge. It might also impair the use of more general bridges only the loading is specified and the
sensitive equipment. engineer needs to decide how the dynamic response to this
load is to be calculated. This makes the methodology robust,
For these reasons an engineer is often required to design a new and it has generally ensured the adequate vertical dynamic
structure such that the extent of footfall-induced vibration will performance of footbridges over the last 30 years. There is,
not exceed a particular criterion. A number of calculation however, some scope for refining the loading definitions in the
methodologies have been published, 1–6 which claim to light of the large amount of experimental research that has
facilitate the prediction of this vibration in service, but as parts been performed since it was developed. Also the performance
1 and 27 of the current paper will show, each has significant criterion was developed for long-span external bridges and this
limitations. In addition, the published methodologies contradict could be developed to incorporate the differing expectations of
each other, can be difficult to integrate with modern design pedestrians and bystanders on relatively short-span bridges in
methods and cannot be consistently applied across a broad different (e.g. internal) environments.
range of structures. Comparative studies (for example between
a steel and a concrete structure) are, therefore, fraught with The method most commonly used in the USA for assessing the
difficulties and serve further to mystify the issue of vibration dynamic performance of footbridges is that provided in the
serviceability. 1997 American Institute of Steel Construction (AISC) Guide. 3 As
discussed in section 3.2.2 below, the variation of harmonic force
The present paper uses worked examples to illustrate the with frequency it proposes is grossly oversimplified and is very
inconsistencies and limitations of existing methodologies and un-conservative in some frequency ranges in relation to
to emphasise the need for a predictive tool that is robust and measured data. Higher loads are given for bridges than for
readily applicable to simple and complex structures of varying floors, although no explanation is given as to why the same
structural form. individual might exert a greater vertical force when walking on
a bridge than when walking across a floor. Otherwise the
For reasons discussed below, a distinction is made between limitations of the methodology are as discussed below for floors.
low- and high-frequency structures (that is those with their
first vertical mode below and above 10 Hz). Low-frequency 2.2. Floors
floors and bridges are discussed here and floors with a first Early acceptability criteria for floor vibration related only to
mode above 10 Hz are discussed in part 2 of this paper. 7 the frequency of the floor (‘fundamental frequency should be

Structures & Buildings 160 Issue SB2 Predicting footfall-induced vibration: Part 1 Willford et al. 65
 greater than 4 Hz’) or the impulse response under a heel drop. root sum of the squares’ calculation. The force data used are
Neither approach represents a performance-based assessment of harmonic loads derived statistically from a large number of
the vibration likely to arise in normal service. The weight and measured footfall force–time histories.
stiffness of modern steel composite and post-tensioned
concrete floors are such that the need for a more robust and Reliance on fundamental dynamic principles makes the method
comprehensive assessment technique became clear. robust and applicable to any structure of arbitrary complexity
whose modal properties may be measured or calculated. This
In 1989 the Steel Construction Institute (SCI) published their rigour does, however, make the method more involved than the
Guide P076,1 which had been prepared independently by Dr others described here. For a structure with several modes that
Tom Wyatt of Imperial College London with the support of a might respond it is impractical to do the full calculation by
seven-person steering committee representing various sectors hand and a simple spreadsheet is required. CCIP-016 5 provides
of the industry. This presented the first widely recognised flowcharts to illustrate how this might be done.
performance-based approach for floors in buildings, including
calculation procedures aimed at predicting the actual 3. TECHNICAL BACKGROUND
vibration that would arise in practice as a result of walking. It
made the important distinction between lower-frequency 3.1. Quantifying response
floors that are susceptible to resonant response under repeated Absolute levels of footfall-induced vibration are normally
footfalls, and higher-frequency floors, which are not. quantified in terms of velocity or acceleration. It is, however,
Although applicable only to steel structures with a regular well documented (for example reference 13) that human
framing arrangement, its low-frequency floor prediction perception of vibration is influenced not just by amplitude but
methodology has proved to be reasonably robust and reliable also by frequency and duration. Amplitude alone is therefore
over 15 years. Its principal shortcomings are the very limited an inadequate descriptor of footfall-induced vibration.
experimental data upon which the pedestrian dynamic forces ‘Response factor’ (R) and ‘vibration dose value’ (VDV) are two
are based and the difficulty in applying the method to metrics that attempt to incorporate these additional factors.
anything other than the specific steel floor framing layouts
covered in the guide. BS 6472 13 contains curves which nominally define the
threshold of human perception to vibration. Specifically, these
In 2004 the SCI published P331,2 which purports to ‘enable define the perception threshold as a level of harmonic
designers to calculate the vibration response of sensitive floors vibration of constant amplitude that is a function of frequency
more accurately’. Unlike its 1989 predecessor, these guidelines and the orientation of the vibration to the individual. These
have not been independently reviewed and no data or curves are used to define a response factor of 1 (R ¼ 1), and the
theoretical basis that supports the methodology have been response factor of any harmonic response of known frequency
made generally available to facilitate such a review. Although and constant amplitude is found by dividing the amplitude by
it claims to be based in part on the work of the current authors, the corresponding base value at that frequency. The response
this paper shows that the P331 method can give very different factor is therefore a direct measure of perceptibility, R ¼ 2
(and often much less conservative) predictions than the Arup being twice what is perceptible and so on.
method. Further concerns relating to the theoretical basis of the
method have been discussed in recent correspondence between The VDV is also defined in BS 6472. It incorporates the
the current authors and the SCI in The Structural Engineer.9 duration of exposure to vibration and also weights higher
responses more heavily. To calculate a VDV it is necessary to
In the USA the most commonly used predictive technique for know how often and for how long a given level of vibration
low-frequency floors was published in 1997 by AISC. 3 Again, occurs. If vibration is attributed to a nearby railway and the
this guide is applicable to steel structures only and is similar in frequency of train pass-bys is known then this information
basis to SCI P076.1 It does provide a means of assessing the may well be available. In the context of footfall-induced
response in edge and corner bays (which is generally higher vibration, however, the required additional data (speed, number
than in central bays) and also considers the response of and route of pedestrians) are seldom available to the designer
footbridges. Otherwise its limitations are generally similar to and the VDV is less commonly used. In the present paper all
those of SCI P076. calculated responses will be expressed as response factors.

As a reaction to the inconsistencies and limitations of the 3.2. Resonant response of floors and bridges
published methods, Arup has developed its own methodology, It is commonly accepted by the methods reviewed here that
which is now used worldwide within the firm. It was first made ‘low-frequency’ structures are susceptible to resonant response
available outside of Arup in 2001. 10 The method has been under repeated footfall loading. Each methodology therefore
independently peer reviewed 5 and has recently been adopted seeks to predict the extent of this resonant response when
by other organisations and industrial bodies. 5,11 It is most assessing the maximum acceleration that is likely to occur in
completely described in the recent CCIP-016. 5 service. They differ, however, in the following respects

For low-frequency floors, standard modal response theory is (a) the threshold between ‘low’- and ‘high’-frequency floors
used to calculate the response to each harmonic load in exactly (i.e. the highest frequency at which a floor is susceptible to
the way that a finite element (FE) package such as Nastran 12 resonant build-up of vibration from successive footfalls)
calculates modal forced response. Total response is then (b) the magnitude of the harmonic components of footfall
calculated from the four harmonic responses using a ‘square forces

66 Structures & Buildings 160 Issue SB2 Predicting footfall-induced vibration: Part 1 Willford et al.
(c) how the resonant response is calculated for real structural they are well above or below a reasonable representative value.
systems with several modes and a limited number of Over most of the range there is a factor of at least two between
footfalls per span. the lowest and highest values selected, and at some frequencies
this difference is more than a factor of four. Even if the
The significance of these factors is discussed in turn below. dynamic properties of a floor and its response to a particular
force are known exactly, therefore, the choice of prediction
3.2.1. Threshold frequency for low-frequency floors. Most method will very significantly affect the predicted performance
modern floor vibration prediction methods consider there to be of the structure.
a cut-off frequency above which resonant build-up of vibration
response to repeated footfalls is not significant compared with As well as being very different in absolute terms, the trends in
the transient vibration generated by a single footfall. In harmonic force with frequency also differ greatly from one
justifying this, at least three approaches are possible. method to the next. Depending on the method adopted, an
engineer could be led to believe that by increasing the
First, measured response data can be analysed using a series of frequency of his structure from 5.5 Hz to 7.5 Hz there is either
band pass filters to isolate response in individual modes. The a 50% decrease (AISC) or a 20% increase (Arup) in force to
filtered time histories can then be inspected for evidence of consider. The implication in the AISC guide that someone
resonant build-up in each mode. Modes with frequencies walking very slowly at 1.5 steps per second applies a 40%
higher than the fundamental are always present in a floor. greater force (in the first harmonic) than someone who is
Therefore, if resonant excitation of the higher modes by higher walking very fast at 2.5 steps per second is counterintuitive
harmonics were possible, then this response should be and at variance with the measured data.
identifiable. In practice response in these modes is
characterised more by impulses followed by decays, and Ellis6 is alone in recommending that resonance be considered
resonant build-up over a number of cycles is not seen. for modes over 10 Hz. The conservatism associated with this
assumed loading becomes clear when comparing the predicted
Second, measured or simulated footfall force–time histories resonant response owing to a dynamic load factor (DLF) of 0.1
can be applied to a series of single mode systems of varying (as proposed by Ellis) with the measured data and impulsive
natural frequency so that the sensitivity of the maximum responses calculated for the worked examples in part 2 of the
dynamic response to exact synchronisation of a harmonic of current paper, 7 which deals with high-frequency structures.
footfall rate to a modal frequency can be examined. CCIP-016 5
presents the results of such an exercise. 3.2.3. Dynamic properties of floors and response calculation
method. The calculation of dynamic response requires some
Third, a sensible limit on the highest likely harmonic can be estimate of the natural frequency, effective mass and damping
estimated by considering the natural variability of the footfalls of a floor. The SCI P076, SCI P331 and AISC methods are based
of an individual. Steady state harmonic response analysis upon ‘hand calculation’ techniques that attempt to approximate
assumes a periodically repeating force input where every cycle the true situation (in which there are usually several or many
is identical. The natural variation between left and right foot modes of lowish natural frequency) by a single equivalent
and perhaps a 10% (
0.04 s) variation in exact pacing rate mode. This inevitably introduces a degree of approximation,
between steps can be used to show that effective harmonic which may result in significant error, as is illustrated in the
excitation beyond the fourth harmonic (
10 Hz, 0.1 s period) worked example presented later in the current paper.
will be very low.
A more rigorous approach is to derive modal properties of a
The assumed cut-off frequency for each of the methods is shown structure by finite element analysis (FEA), which is the basis of
in Fig. 2 (see below). The lowest recommendation is 5 Hz in BS the Arup method. This approach enables a much more reliable
5400, 4 the highest, proposed by Ellis,6 is at least 19.2 Hz. It is assessment to be made, including the effect of simultaneous
the current authors’ experience that the practical significance of response in the many closely spaced modes that will be present
resonance above the fourth harmonic is very small. in large floor plates. It also enables the response to be
calculated at points distant from the walking route and the
3.2.2. Harmonic forces. The harmonic components of footfall effects of partitions to be included explicitly. Full-height
forces from over 800 measurements are plotted in Fig. 1. It can partitions can significantly increase the stiffness and damping
be seen that there is a very substantial scatter in these values, of floors, but this benefit must be evaluated on a case by case
and it is therefore important to realise that there is no single basis. It should also only be included in a response calculation
‘correct’ value, but rather that values can be selected within the if the associated compromise in space planning flexibility is
range to be more or less conservative. Ideally, therefore, the agreed with the client.
values selected should have a probability of exceedance
assigned to them. Most of the published methods select values As mentioned in section 3.2 above, the generally accepted
based on a much smaller data set and with no statistical basis. design process for low-frequency floors should consider the
The Arup method selects values with a 25% probability of maximum level of resonant response that can be induced by a
exceedance based on this large data set. person walking on the structure. There is, however,
considerable variation between the methods in how the
The harmonic force components that are assumed by each magnitude of this response should be calculated. The following
method are plotted in Fig. 2. It can be seen that while for the factors need to be considered in any reliable form of
most part they fall within the measured data set, sometimes assessment

Structures & Buildings 160 Issue SB2 Predicting footfall-induced vibration: Part 1 Willford et al. 67
 1st harmonic 2nd harmonic
1·0 0·25
0·9
0·8 0·20
0·7
0·6 0·15
DLF

DLF
0·5
0·4 0·10
0·3
0·2 0·05
0·1
0·0 0·00
0·0 0·5 1·0 1·5 2·0 2·5 3·0 3·5 0·0 1·0 2·0 3·0 4·0 5·0 6·0 7·0
Frequency: Hz Frequency: Hz

3rd harmonic 4th harmonic

0·18 0·16

0·16 0·14
0·14
0·12
0·12
0·10
0·10
DLF

DLF 0·08
0·08
0·06
0·06
0·04
0·04

0·02 0·02

0 0
0·0 2·0 4·0 6·0 8·0 10·0 0·0 2·0 4·0 6·0 8·0 10·0 12·0 14·0
Frequency: Hz Frequency: Hz

Kerr Galbraith and Barton Wheeler Ohlssen Rainer Ellis Mean Value Design Value

Fig. 1. Measured harmonic loads for walking

bays a very large number of modes will be required to

DLFs for walking obtain a reliable prediction
SCI P076
0·6
SCI P331 (d ) on some floors modes of different natural frequency might
Arup design values be excited by different harmonics of the footfall force.
0·5
(Ave ⫹ 1SD)
0·4 AISC floors
Adequate consideration of these factors will improve the
AISC bridges
reliability of any predictive methods.
DLF

0·3 Ellis
BS 5400 (bridges)
0·2 4. REVIEW OF RESPONSE CALCULATION
19·2 Hz PROCEDURES: WORKED EXAMPLES
0·1
The inconsistency of the various methods is illustrated by
0 example below. Unless otherwise specified, a 750 N person
0 2 4 6 8 10 12 walking at between 90 and 150 paces per minute (1.5 to
Forcing frequency: Hz
2.5 Hz) with a third harmonic force of 75 N is assumed.
Fig. 2. DLFs for various footfall prediction methodologies Damping of 3% of critical damping has been used throughout.

4.1. Example 1: A Slimdek floor on a 7.5 m 3 7.5 m grid,

of varying number of bays
(a) the number of footsteps required to cross a span may be This example is identical with the worked example used in SCI
insufficient to generate a calculated ‘steady state’ response P331 and illustrates well the effects of various assumptions
(b) not all footsteps are applied at the point of maximum within the published methods. A floor summary is presented in
mode-shape and therefore are not as effective in exciting a Table 1.
mode
(c) on floors there are often several modes of lowish frequency The Slimdek system used here essentially comprises a 90 mm
which will all be excited to some degree and will slab on deep ribs at 600 mm centres, which span 7.5 m
contribute to the overall response; on floors with many between secondary beams. The secondary beams span 7.5 m

68 Structures & Buildings 160 Issue SB2 Predicting footfall-induced vibration: Part 1 Willford et al.
 Grid 7.5 m 3 7.5 m
Beam layout: 300ASB(FE)153 (asymmetric beam) secondaries spanning in one direction between columns
Decking Comdek SD225, 1.25 mm thick spanning 7.5 m between secondaries
Concrete Normal weight with overall slab depth of 316 mm
Floor size 16 bays in the concrete spanning direction, 2 to 16 bays in secondary spanning direction (the SCI P331
assumes 16 bays in its worked example)

Table 1. Floor summary

between columns. Various approximate hand calculations 1,3,5

Slimdeck floor 16 ⫻ 3 bays
can be performed to assess the natural frequency of the
structure and, depending on the assumptions made, these yield
first mode frequencies between 5 Hz and 6 Hz. To ensure,
however, that the interaction of all contributing structural
components is included, this calculation is most accurately
done using a FE model. As the ribs are deep compared with
the slab, rather than considering the slab to have a smeared Mode : 1 Mode : 2
Freq 5·81 Hz Freq 5·83 Hz
overall thickness, it is better to model the ribs explicitly as
beam elements, offset below the 90 mm slab modelled as
shells (see Fig. 3). Z_DISP
⫺1·00
⫺0·85
The first three modes of a 16 3 3 bay floor are shown in Fig. 4. ⫺0·69
⫺0·54
Each has a modal mass of approximately 320 t and the natural ⫺0·39
frequencies are 5.81 Hz, 5.82 Hz and 5.86 Hz. As these are ⫺0·23
⫺0·08
above 5.0 Hz the floor is not susceptible to the first and second 0·08
harmonics of walking forces for pacing rates up to 2.5 Mode : 3
0·23
0·39
footfalls/s. Third harmonic excitation is therefore appropriate. Freq 5·86 Hz 0·54
0·69
0·85
The calculated response factors for this example floor are 1·00

plotted in Fig. 5. The x axis represents the number of bays in

Fig. 4. The first three modes of a 16 3 3 bay Slimdek floor
the beam spanning direction (16 is used in the example in SCI
(example 1)
P331, three in the mode shapes above). Sixteen bays are
assumed in the decking spanning direction. The following
points should also be considered.

(a) The ‘FE’ solution is a forced response calculation assuming Slimdek floor of varying size,
a 75 N harmonic force applied in the middle of a bay using 16 bays in beam spanning direction SCI P331
8 SCI P076
7 Arup 750N walker
B AISC
300 ASB153 6
Response factor

FE
A A 5
22·5 m
3 bays 4
3

B 2
120 m, 16 bays
1
Section A–A 91 mm
0
0 2 4 6 8 10 12 14 16 18
316 mm Number of bays in beam spanning direction

300
mm 7·5 mm Fig. 5. Predicted response factors for a Slimdek floor of
varying number of bays
91 mm
Section B–B

225 mm all modes up to 15 Hz. For a given set of modes this will
calculate the theoretically correct, full resonant response
400 200
mm mm
for this load using standard modal response theory.
Resonant response to a harmonic load of 75 N is the basis
of the SCI P331 method and so the FE and SCI P331
Fig. 3. Slimdek floor. Three bays in the spanning direction are
calculations would be expected to give very similar results.
shown
(b) The Arup method uses the mode-shapes from the FE model

Structures & Buildings 160 Issue SB2 Predicting footfall-induced vibration: Part 1 Willford et al. 69
 (modes up to 15 Hz) to calculate the response using the different construction types. This is a considerable weakness of
method described in reference 5. both the SCI and AISC techniques, which makes like for like
(c) All calculations are for mid-span in a bay in the middle of comparisons very difficult.
the floor. (The AISC and Arup methods allow calculation of
response in edge and corner bays, where the response will A comparable concrete design to the Slimdek floor of example
be higher. The FE method could similarly calculate 1 is a 220 mm post-tensioned flat slab, and the first three
response away from the middle of the floor.) modes of the 16 33 bay floor predicted by FEA are shown in
Fig. 6. Each mode has a modal mass of approximately 430 t.
The following points arise from this graph. Data on the floor summary are presented in Table 2.

The natural frequencies of this post-tensioned slab floor are a

(a) Regardless of the floor size, the four published methods
little higher than those of the Slimdeck floor and third
give very different results. For a floor that is 16 bays 3 16
harmonic excitation is again the critical case. Response factors
bays, there is a factor of 5 between the largest and smallest
calculated using the Arup and FE methods are shown in Fig. 7.
predicted responses.
These can be compared directly against those for the steel
(b) The SCI P331 hand-calculation method uses this worked
design in such a way that will be useful to designers wishing to
example to show that a 16 3 16 bay floor (a very unusual
evaluate the relative benefits of different schemes (Fig. 8). It
layout in practice) is adequate as an operating theatre
can be seen that the trend in response with increasing floor
according to NHS Estate guidelines 14 with R , 1.0. Yet all
size is consistent between the steel and concrete designs. In this
the other methods show that such a floor is suitable only as
case the selected concrete floor results in slightly lower
an office (R , 8) or a day ward (R , 4). SCI P331 is clearly
vibration levels principally because it is approximately 25%
inconsistent and very unconservative in comparison with
heavier than the Slimdeck floor.
the other methods.
(c) The SCI P331 hand-calculation method claims to predict a
resonant response consistent with subjecting the 16 3 16
bay structure to a harmonic load of 75 N, yet the predicted 5. SUMMARY AND CONCLUSIONS
response is a factor of 4 lower than that owing to the The current paper has reviewed the basis of a number of
equivalent FE method using the same load. This leads to published techniques for predicting footfall-induced vibration
significant concerns about the validity of the method. for structures with a first natural frequency lower than 10 Hz.
(d ) The ‘Arup’ and ‘FE’ prediction use the same modal
properties and comparable modal response theory. The
difference in response arises from the Arup assessment of 220 mm FLAT SLAB
harmonic force being about 30% lower than the SCI’s 75 N
for a floor of this frequency. Both are broadly consistent
with the SCI P076 prediction.
(e) It is intuitively reasonable that as a floor is increased in
size the magnitude of response will reduce, and four of the
methods show this to be the case. The SCI P076, Arup and
Mode : 1 Mode : 2
the FE results show that this tendency is asymptotic Freq 6·57 Hz Freq 6·65 Hz
towards a certain level, and that there comes a point where
there is no benefit in adding further bays increasingly
Z_DISP
distant from the point at which the floor is excited. This
⫺1·00
too is intuitively reasonable. SCI P331, however, shows the ⫺0·85
⫺0·69
response to be asymptotic to zero, implying that if a floor ⫺0·54
plate is big enough, there will be virtually zero vibration in ⫺0·39
⫺0·23
any bay – a result which is clearly counterintuitive. ⫺0·08
0·08
0·23
Mode : 3 0·39
4.2. Example 2: A post-tensioned 220 mm concrete floor Freq 6·77 Hz 0·54
0·69
on a 7.5 m 3 7.5 m grid 0·85
Three of the methods considered here are written specifically 1·00
for steel structures. Only the Arup method claims to be able to Fig. 6. The first three modes of a 16 3 3 bay post-tensioned
predict vibration of both steel and concrete structures reliably, concrete floor (example 2)
thereby permitting a valid comparison to be made between

Grid 7.5 m 3 7.5 m

Concrete 220 mm post-tensioned normal weight flat slab
Floor size: 16 bays in one direction, varying numbers of bays in the orthogonal direction up to 16 bays. (Identical with
layouts considered for Slimdek floor above)

Table 2. Floor summary

70 Structures & Buildings 160 Issue SB2 Predicting footfall-induced vibration: Part 1 Willford et al.
 5·0 considered here only the Arup method statistically
FE–75 N harmonic force
4·5 quantifies this variation or indicates the statistical
Arup 750 N Walker
4·0
significance of the chosen loads.
(d ) The AISC method incorporates forces that are clearly
3·5
Response factor

inconsistent in trend with measured data. In addition, the

3·0
reason for the difference in loading for floors and for
2·5
bridges is not given and is difficult to justify intuitively.
2·0
(e) It is reasonable to expect floor size to have some effect on
1·5
total response, and all methods except that due to the AISC
1·0
show this to be the case. For the example considered, the
0·5 AISC is at least conservative, as it assumes a (small) limited
0·0 floor size and hence predicts higher response.
2 by 16 3 by 16 4 by 16 6 by 16 10 by 16 16 by 16
Bay arrangement (f) The two methods published by the SCI (P076 and P331)
predict responses for the same floor which can differ by
Fig. 7. Response factors for a 220 mm post-tensioned slab of more than a factor of 2 in either direction.
varying size (g) The basis of SCI P331 has not been published,
independently peer reviewed or made available for peer
review. It predicts responses that are inconsistent and
potentially very un-conservative in comparison with all
other methods. It can yield results that are very different
PT concrete/Slimdek comparison, from an equivalent FEA of the same floor. It includes a
varying floor size calculation for participating mass which shows response to
6·0
be asymptotic to zero for an increasing number of bays,
5·0 Arup 220 PT slab
which is both counter-intuitive and inconsistent with
Arup, Slimdek measurements. These features lead to very significant
Response factor

4·0 concerns about the method’s general suitability as a design

tool. The current authors suggest that the SCI withdraw the
3·0
guide and that designers use it with caution.
2·0
(h) The Arup and FE methods give consistent results, this is
because they both incorporate standard modal response
1·0 theory. They give results which are broadly consistent with
the well-established SCI P076 method.
0·0
2 by 16 3 by 16 4 by 16 6 by 16 10 by 16 16 by 16
Bay arrangement REFERENCES
1. WYATT T. A. Design guide on the vibration of floors. The
Fig. 8. A comparison between steel and concrete designs Steel Construction Institute Publication 076, SCI, Ascot,
1989.
2. HICKS S. J. and DEVINE P. J. Design guide on the vibration
The objective of the methods is similar in that each endeavours of floors in hospitals. The Steel Construction Institute
to predict the resonant response that could be induced by a Publication 331, SCI, Ascot, 2004.
pedestrian walking at the most critical walking speed. In spite 3. MURRAY T. M., ALLEN D. E. and UNGAR, E. E. Floor
of this, very different results are obtained for the same vibration due to human activity. AISC Design Guide, Series
structure, depending on the chosen predictive method. The No. 11, American Institute of Steel Construction, Chicago,
need for a consistent, robust and generally applicable method 1997.
for predicting footfall vibration is clear. 4. BRITISH STANDARDS INSTITUTION. Appendix C Vibration
Serviceability Requirements for Foot and Cycle Track
The inconsistencies and limitations of the methods considered Bridges. BSI, London, 1978, BS 5400, Part 2.
are summarised below. 5. WILLFORD M. and YOUNG P. A Design Guide For Footfall
Induced Vibration of Structures. CCIP-016. The Concrete
(a) Of the methods considered, only one (Arup method) is Centre, Camberley, 2006.
suitable for assessing a broad range of structures of 6. ELLIS B. R. On the response of long-span floors to walking
varying form (i.e. steel composite structures, reinforced and loads generated by individuals and crowds. The Structural
prestressed concrete structures, and structures with unusual Engineer, 2000, 78, No. 10, 17–25.
or complex geometry.) 7. WILLFORD M., YOUNG P. and FIELD C. Predicting footfall-
(b) The threshold frequency below which resonant excitation induced vibration: Part 2. Proceedings of the Institution of
of a floor is considered possible varies considerably Civil Engineers, Structures and Buildings, 2007, 160, No. 2,
between 5 Hz3 and at least 19 Hz. 11 The authors consider 73–79.
10 Hz (no harmonic excitation beyond the fourth 8. SMITH J. W. The Vibration of Highway Bridges and the
harmonic) to be reasonable for design. Effects on Human Comfort. PhD thesis, University of
(c) There is considerable variation in the assumed harmonic Bristol, 1969.
loading between the various methods. There is also 9. The Structural Engineer, 16 May 2006, pp. 36–37.
considerable scatter in measured data, but of those 10. YOUNG, P. Improved floor vibration prediction

Structures & Buildings 160 Issue SB2 Predicting footfall-induced vibration: Part 1 Willford et al. 71
 methodologies. Arup Vibration Seminar, Institute of 13. BRITISH STANDARDS INSTITUTION. Guide to Evaluation of
Mechanical Engineers, London, 2001. Human Exposure to Vibration in Buildings (1 Hz to 80 Hz).
11. THE CONCRETE SOCIETY. Post-tensioned Concrete Floors – BSI, London, 1992, BS 6472.
Design Handbook. The Concrete Society Camberley, 2005, 14. NHS ESTATES. Acoustics, Design Considerations. Health
Technical report 43. Technical Memorandum 2045. Her Majesty’s Stationery
12. MSC Nastran. See www.mscsoftware.com. Office, Norwich, 1996.

What do you think?

To comment on this paper, please email up to 500 words to the editor at [email protected]
Proceedings journals rely entirely on contributions sent in by civil engineers and related professionals, academics and students. Papers
should be 2000–5000 words long, with adequate illustrations and references. Please visit www.thomastelford.com/journals for author
guidelines and further details.

72 Structures & Buildings 160 Issue SB2 Predicting footfall-induced vibration: Part 1 Willford et al.