Structures 8 (2016) 94–110

Contents lists available at ScienceDirect

Structures

journal homepage: www.elsevier.com/locate/structures

Construction aerodynamics of cable-stayed bridges for record spans:

Stonecutters Bridge
S.H. Robin Sham BSc PhD DIC FCGI FRSA CEng FICE FIStructE FHKIE a,⁎, Tom A. Wyatt PhD DIC FREng CEng FICE b
a
Global Long Span and Specialty Bridge Director, AECOM, USA
b
Imperial College London, UK

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

Article history: This paper presents the extensive aerodynamic investigations conducted for the construction of the 1018 m-
Received 27 August 2014 main span Stonecutters Bridge. The emphasis is on the behaviour of the bridge structure during the erection
Received in revised form 26 June 2016 stages, notably its response to vortex shedding, buffeting and divergent amplitude effects. A description is
Accepted 29 August 2016
given of the principal aspects of the wind tunnel testing carried out, including section model testing, aeroelastic
Available online 4 September 2016
freestanding tower model testing on different erection heights, and aeroelastic bridge model testing to investi-
Keywords:
gate the bridge erection stages. The engineering implications of the results are discussed. Wind tunnel testing
Cable-stayed bridge on stay cables has also been carried out, with special reference to determining cable drag and the susceptibility
Wind tunnel testing to rain-wind induced oscillations. These investigations advanced general understanding of the behaviour of
Vortex shedding cable-stayed bridges during construction, and contributed significantly to the development of the erection meth-
Buffeting od and sequence for the Stonecutters Bridge leading to the expeditious completion of this work.
Divergent amplitude © 2016 Published by Elsevier Ltd. on behalf of Institution of Structural Engineers.
Drag measurements
Rain-wind induced oscillations

1. Introduction of comprehensive wind tunnel testing; together with the evaluation

and use of the test results for construction engineering.
1.1. Context of this study The planning of the wind tunnel investigations on construction aero-
dynamics drew on detailed study of the eigen-modes of the bridge
The Stonecutters Bridge crosses the Rambler Channel, Hong Kong. throughout the different stages of erection [4], together with an assess-
This is a major shipping artery, requiring a clear span of 1018 m and a ment of the wind data captured by a high mast structure installed at the
navigation clearance of 73.5 m (Figs 1.1 and 1.2). The deck comprises bridge site.
a twin-box girder, (Figs. 1.3 and 1.4) of 53.30 m overall width with a The wind tunnel test programme on construction aerodynamics was
central gap of 14.3 m, supported by single-shaft bridge towers 298 m comprehensive. It included section model investigations of the basic
high. The cantilever erection of the main span constituted one of the aerodynamic parameters and of the risk of vortex excitations. The
largest operations of its type in recent history (Fig. 1.5). The bridge work was conducted at the largest practicable model scale and it also
was opened to traffic in December 2009. took advantage of the greater flexibility in selection of test wind speed
The Government of Hong Kong Special Administrative Region had scale. The performance of guide vanes (Figs. 1.4 and 1.6) in reducing ex-
invited tenders for construction in August 2003. The Contract was citation was studied in the section model testing (in Section 2 below).
awarded to the Maeda-Hitachi-Yokogawa-Hsin Chong JV (MHYH JV) Aeroelastic tower models were used to investigate wind action on the
in April 2004. AECOM Asia Company Limited (AECOM) (formerly bridge towers during erection, including construction equipment (in
Maunsell Consultants Asia Limited) was appointed by the MHYH JV Section 3 below). Aeroelastic bridge model testing was used to examine
for advice during the tender period, and as its consultant for the con- response to the turbulent wind. Results from aeroelastic bridge models
struction process. One aspect of the consultancy was the advice on at selected stages in the cantilever construction, including temporary
bridge aerodynamics, and the planning, management and supervision erection conditions and construction equipment, were matched to
full dynamic analysis of the bridge structure for the whole erection
sequence, stage-by-stage (in Section 4 below). Specialist studies on
⁎ Corresponding author at: AECOM 8/F Grand Central Plaza, Tower Two, 138 Shatin
full-scale stay cable models were also conducted, to examine the
Rural Committee Road, Shatin, Hong Kong. combined effects of rain and wind and their mitigation (in Section 5
E-mail address: [email protected] (S.H.R. Sham). below).

http://dx.doi.org/10.1016/j.istruc.2016.08.010
2352-0124/© 2016 Published by Elsevier Ltd. on behalf of Institution of Structural Engineers.
 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110 95

Fig. 1.1. Stonecutters Bridge – elevation.

Fig. 1.2. Stonecutters Bridge – plan.

Fig. 1.3. Stonecutters Bridge – typical cross section.

A comprehensive account of the Stonecutters Bridge construction aerodynamics of the Stonecutters Bridge by [9–12,16]; and of the
has been given in [6]. Reference is also made to [3] for the main span Sutong Bridge by [13].
erection, to [17] for the back span construction, and to [14,15] for fur-
ther specific features of the construction engineering. 1.2. Overview of the problem
Wind tunnel tests focussed primarily on the completed cable-
stayed bridge system of the Stonecutter Bridge have been reported There has been much experience of the aerodynamic behaviour of
by [2,5]. [7] commented on the investigations on construction long-span suspension bridges, leading to extensive design guidance,
96 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110

Fig. 1.4. Stonecutters Bridge – cross section with details of guide vanes.

but the recent advance of cable-stayed bridges to spans in excess of typically be similar, especially for the crucial torsional natural frequency
1000 m has presented a new frontier in terms of dynamics and aerody- required to ensure avoidance of divergent-amplitude oscillation (flut-
namics, particularly during bridge construction. The past decade pre- ter). The greater vertical stiffness of the cable-stayed structural form
sented significant opportunities for exploring the behaviour of ultra- raises the natural frequencies of vertical motion, which tends to reduce
long span cable-stayed bridges during erection, including statics, dy- sensitivity to buffeting by the vertical component of gusts.
namics and aerodynamics. Notable examples include the 1018 m main The potential dynamics and aerodynamic problems during erec-
span Stonecutters Bridge and the 1088 m main span Sutong Bridge tion are, however, drastically different between the two bridge sys-
[14,15], the latter being the longest cable-stayed bridge span upon its tems. The classical suspension bridge may not develop its torsional
completion. The suspension bridge has hitherto been unchallenged for stiffness prior to closure of the span, most obviously in the
such spans. commonly-adopted method of starting deck erection from within
Once the erection stages have been completed the structural dynam- the span, typically working from the mid-span incrementally to-
ics of the cable-stayed and suspension structural forms in service would wards the towers. In this circumstance, the risk of flutter is likely to

Fig. 1.5. Stonecutters Bridge – large cantilever erection.

 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110 97

Fig. 2.2. Test configuration with deck lifting gantry.

Fig. 1.6. Stonecutters Bridge erection – guide vanes installed prior to girder segment lifting.

be a crucial consideration during suspension bridge construction. The solution to this system under the action of localised loading com-
This problem is avoided by the incremental cantilever erection of prises decay functions of the form e–x/λsinλx and e–x/λcosλx (in which
cable-stayed bridges. However, incremental cantilever erection cul- x is the distance from the point of perturbation), the decay length pa-
minates with two very slender cantilever structures, before their rameter being given by λ = (4EI/kv)1/4. In general, the value of kv is
connection at mid-span. closely constrained by the design presumption that each stay supports
In the vertical plane, bending caused by the vertical component of the weight of the local segment of deck at a consistent dead-load stress
wind gust action is superimposed on the action of the weight of the and has the result that with slender girders the decay length will be
erection equipment, including the lifting of a further deck section at small compared to the free span. In the case of the Stonecutters Bridge,
reaching out beyond the current outermost stays. The additional mass the decay length based on the stiffness of the stays towards mid-span is
as well as the absence of the deck flexural continuity means that the around λ = 80 m.
natural frequencies are lower than in the in-service condition, increas- The BEF solution for a load P applied at the end of a long (semi-
ing the sensitivity to gust action. infinite) member gives the deflection y0 = Pλ3/2EI, and the maximum
Valuable insight into structural behaviour in the interaction of girder moment M = 0.322Pλ = 0.64(EI/λ2)y0 occurring at a distance
the stay system with beam flexure in the vicinity of the cantilever-tip (π/4)λ from the load. In bridge cantilever erection, the deflections are
is given by the beam-on-elastic-foundation ‘BEF’ analogy [18], in increased by deflection of the pylon permitted by the extension of the
which the deck girder (characterised by its flexural rigidity EI kNm2) stays in the side span, but the moment is unaffected. The vertical-
is regarded as supported by a continuous elastic support derived plane natural modes reflect this focus on behaviour in the vicinity of
by ‘smearing’ the vertical stiffness of the discrete stays (stiffness kv the free end. For Stonecutters Bridge the critical dynamic moments
kN/m2, the vertical reaction per unit length per unit vertical deflection). were found to occur around 100 m from the end; in the first vertical

Fig. 2.1. Section model Test Configuration-1.

98 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110

Fig. 2.3. Static force coefficients in wind axis system, for Test Configuration-1].

mode, the value of this moment was about M = 0.64(EI/λ2)y0. The fre- piers permits significant rotation of the deck beam in plan at the main
quency of the second vertical mode was only about 50% higher than the towers. The natural frequency is thus much lower than in-service and
fundamental, so this mode was also expected to make a significant con- the lateral response to wind gust buffeting becomes a major problem,
tribution to the stressing. especially in the context of the twin deck configuration of Stonecutters
In the horizontal plane, the structure is basically a flexural cantilever, Bridge, introducing an element of ‘vierendeel ladder’ to the structural
but flexure of the anchor arm with lateral compliance of the supporting compliance.

Fig. 2.4. Drag coefficients of selected test configurations.

 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110 99

Fig. 2.7. (a) Vortex shedding response: Full scale bending oscillation amplitude at 0° wind
incidence.

Fig. 2.5. Hand rail system with 60% porosity toeboard.

This complexity of buffeting behaviour emphasised the prime im-

portance of an aeroelastic model of the erection cantilever at construc-
tion stages approaching full length; this is described in Section 4,
below. Two scenarios for modeling the turbulent wind were investigat-
ed. ‘Ocean fetch’ (wind broadly from the south-west) gave the highest
mean wind speed, but the more turbulent ‘land fetch’ proved to give
maximum values for some criteria sensitive to dynamic response, de-
spite the lower mean wind speed. The criterion for freedom from
divergent-amplitude oscillation was 95 m/s sustained speed at deck
level. Fig. 2.7. (b) Vortex shedding response: Full scale torsional oscillation amplitude at 0 wind
incidence.

2. Section model wind tunnel testing (Fig. 2.2), as well as that of equipment for supporting possible
tuned mass dampers.
2.1. Scope The maximum test wind speed was 24 m/s (corresponding to over
100 m/s full-scale), giving a Reynolds Number of 1.8 × 106 referred to
Classical section model testing of the deck section was undertaken to the overall deck width; only marginal changes in coefficients were re-
determine force coefficients for the erection configurations, to make a corded above a Reynolds Number of 0.8 × 106. For example, in Test
detailed assessment of mitigation of vortex-shedding excitation by Configuration-1 (Fig. 2.1), the difference in the drag coefficients mea-
guide vanes (Fig. 1.4), and to explore the risk of divergent-amplitude in- sured in smooth flow at these two Reynolds Numbers was around 2%.
stability (flutter), including the sensitivity of stability at high wind Tests were performed in smooth flow and in grid-generated turbulence.
speeds to mass, damping and frequency ratios. The work also examined In recognition of the inescapable severe mis-match of turbulence scale
potential flutter sensitivity to inclination of the mean wind to the hori- to the model linear scale, a reduced total intensity of turbulence was
zontal. The model was fabricated largely from carbon fibre, to optimize generated, designed to simulate the specified turbulence spectrum
stiffness, to a linear scale of 1:50. over the range of frequency (f) defined by 0.3 b fb/V b 3.0 for both nom-
In addition to full validation of dynamic performance, static wind inal ‘land fetch’ and ‘ocean fetch’ winds.
load coefficients were obtained from force-balance measurements.
The basic configuration included the guide vanes and all temporary
equipment including rail beams and temporary safety handrails
with netting and toeboard (Fig. 2.1). These measurements revealed
sensitivity of the lateral wind drag loading to handrail details
which were therefore subject to a number of changes (Section 2.2).
The effect of adding the deck lifting gantry was also measured

Fig. 2.6. Divergent amplitude response. Fig. 2.8. Variation of bending total damping with reduced wind speed.
100 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110

Fig. 2.9. Variation of torsional total damping with reduced wind speed.

All force coefficients and nominal Reynolds Number were based on

the deck width (B = 53.3 m full-scale, 1060 mm model scale). The
test results for the basic configuration (Test Configuration-1) as a func- Fig. 3.2. Temporary work and construction equipment at 1:200 model scale used in the
tion of the angle of inclination of the wind in the vertical plane are aeroelastic tower model testing.
shown in Fig. 2.3.
The drag and lift coefficients measured for the basic section (Config-
uration-1) at 0° wind incidence in smooth flow were 0.098 and 0.106 than the section with temporary safety handrail and solid toeboard
respectively. Turbulence generally increased the drag force by 15% for (Fig. 2.5). Additional tests were therefore formulated, to assist in the de-
sea fetch conditions and by 30% for land fetch conditions. sign optimisation of the temporary safety handrail system, with a view
to reducing the lateral wind forces on the structure during construction.
A 40% increase in wind drag, relative to the configuration of a bare
2.2. Assessment of the effect of the temporary safety handrail system
deck without temporary safety handrail, was recorded for the configu-
(Fig. 2.4)
ration with a handrail system with safety net and a solid toeboard.
This increase was reduced to 25% using a porous toeboard of similar po-
The static load measurements for the basic deck section without
rosity to that of the safety net. The flow pattern around the deck section
handrails revealed that the basic section had a significantly lower drag
was believed to be significantly different for the two temporary safety
handrail systems tested. The system with solid toeboard was likely to
induce flow separation at the upstream edge while the system with
the porous toeboard was unlikely to cause the same effect.
For the deck configuration with an open “post-and-rail” handrail
system, without safety net and porous toeboard, the increase in drag
relative to the basic section was in the range 17% to 20%. The edge rail
beams for the under-deck access gantry also contributed to the increase
in the overall drag of the basic bridge deck cross section.
The drag coefficients measured in selected test configurations are
shown in Fig. 2.4.

2.3. Divergent amplitude response

The basic test configuration, including guide vanes, cable anchorages

and temporary handrails, represented a full-scale mass of m = 24.4 t/m,
with effective mass radius of gyration r = 17.5 m. An extensive study
was made of the development of the eigen-modes throughout the erec-
tion sequence [4], to identify potential mode combinations for
divergent-amplitude response and to ensure robust coverage of the
very high wind speed specified as the flutter criterion for this project.
The study included eigen-analyses of the bridge erection stages in sce-
narios with and without construction plant and equipment, to probe
the sensitivity of the modal response to plant/ equipment mass. The
presence of construction plant/equipment reduces the fundamental fre-
quency by 7% at the maximum erection cantilever condition. The princi-
pal contribution to plant/equipment mass was due to the deck erection
gantry each weighing 743 t.
The lowest natural frequency for a torsion mode, arising prior to con-
nection at mid-span while still supporting erection equipment, was
ft. = 0.43 Hz. This corresponds to the maximum cantilever erection con-
dition. (See Section 4 for an account of the eigen-analyses for the erec-
tion stages). The combination of box girders with inclined stay cable
Fig. 3.1. Stonecutters Bridge - temporary work and construction equipment used for tower planes led to a high value of the ratio (ft/fb) of the fundamental frequen-
erection. cies in torsion and in bending, typically ft./fb = 2.06. As these modes had
 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110 101

Fig. 3.3. Aeroelastic tower model wind tunnel investigations on different erection heights.

broadly matching mode shapes, this combination was judged likely to critical speed as well as increase in absolute dimension, as well as a
be the critical case for verification of divergent amplitude (flutter) be- stiffer model supporting a higher tuning frequency of the mounting.
haviour. Lower values of ft./fb = 1.60 and 1.20 were also checked, to en- The Reynolds Number at potential vortex resonance speed, referred to
sure that sensitivity to frequency ratio would not give concern the duct width, was thus about 3500. The outcome in validation of
regarding possible coupling of modes with inferior shape match. As a guide vane performance was regarded as fully satisfactory.
further check on sensitivity, some tests were repeated with an arbitrari- Vortex shedding response was investigated for the basic configura-
ly increased mass, corresponding to 36.6 t/m. This was considered to tions as described above (‘Configuration 1’) and also with the guide
cover the observation that although there was generally little relevant vanes omitted (‘Configuration 1E’). The measurements were carried
dynamic interaction between deck and tower, there were some out in smooth flow for angles of incidence varying from −5° to 5° in in-
modes in which this did cause significant increase in the effective crements of 2.5° and at a level of structural damping much below the
mass as referred to the deck. For the above basic dynamic properties, anticipated full-scale structure levels (typically of the order of 0.013 log-
the flutter normalised mass parameter (taking the reference dimension arithmic decrement for bending and 0.023 logarithmic decrement for
B = 53.3 m, ignoring any effect of the central gap, and mass radius of gy- torsion). The frequency ratio ft./fb was set at 1.2. Vertical and torsional
ration r = 16.0 m) is μ = mr/ρB3 = 2.30. responses appeared independent with near-constant amplitude at the
For the Stonecutters Bridge, the design wind speed criteria against critical wind speed. Results are shown in Fig. 2.7 a and b, with wind
flutter instability stipulated 95 m/s at deck level, in the “without traffic” speeds given in terms of the normalised reduced velocities V/fbB and
condition, for 0° of wind incidence. The Selberg [8] analytic approxima- V/ftB respectively. Strong beneficial effects by the guide vanes were ob-
tion to the classical flutter speed VF is then VF = 3.7ftB{μ(1–f2b/ f2t )}1/2 = served. The guide vanes were found to effectively reduce the vortex
112 m/s, offering confidence that the effect of the central gap would suf- shedding responses for all wind incidences examined, substantially
ficiently counteract any shortfall of the flutter aerodynamics of this eliminating the resonant vortex shedding at 0° wind incidence.
well-streamlined deck section to meet the specification requirement.
The responses recorded in the testing on approach to the wind speeds 2.5. Aerodynamic damping
set for given wind angles of attack in the vertical plane are summarized
in Fig. 2.6, expressed in terms of reduced wind speeds. No critical insta- The aerodynamic damping in bending and in torsion of the bridge
bility was found within the range of design wind speeds for any of the deck were measured for the basic bridge deck configuration (Configura-
cases tested. tion-1).
The variations of the bending and torsion total damping (structural
and aerodynamic damping) with reduced wind speeds, and for smooth
2.4. Vortex shedding response flow, sea fetch and land fetch conditions, are depicted in Figs. 2.8
and 2.9. The results demonstrated that the aerodynamic damping of
The choice of linear scale for a section model calls for compromise the bridge deck in vertical bending and torsion was likely to be large
between representation of details of shape and representation of the compared to the structural damping at reduced wind speeds higher
overall flow pattern, the latter calling for a sufficiently high Aspect than V/fbB = 1 and V/ftB = 1.5 in bending and torsion respectively, ris-
Ratio (AR), defined as the ratio of model length between tunnel walls ing at design wind speeds to as much as twenty times the inherent
to the deck width. The large scale 1:50, resulting in the unusually low structural damping anticipated for a structure of this type (for welded
value AR = 2.7, took into account difficulty reported by other investiga- steel bridges, the structural damping is typically 0.02 logarithmic decre-
tors in validation of the performance of guide vanes [1] that was attrib- ment). This suggested that for these modes the buffeting response of the
uted to Reynolds Number effects referred to the width of the duct bridge deck would be largely governed by the aerodynamic damping;
created by the vane, which is only 600 mm full-scale. This was accepted, this was consistent with the findings of the aeroelastic bridge model
recognising the double advantage in Reynolds Number from increased testing (Section 4).

Fig. 3.4. Tower response without construction plant/equipment. Fig. 3.5. Tower response with construction plant/equipment.
102 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110

Fig. 4.1. Aeroelastic bridge model wind tunnel investigations at 1:200 scale.

3. Aeroelastic tower model wind tunnel investigations 4. Aeroelastic bridge model wind tunnel investigations

The Stonecutters Bridge towers are 298 m high. The lower A linear scale of 1:200 was chosen for the aeroelastic model
reinforced concrete sections of the towers have a complex (Fig. 4.1), to permit good representation of both turbulence and struc-
geometry varying from an extended oval shape of 24 m × 18 m, to a tural detail. To achieve this scale within the wind tunnel while main-
14 m-diameter hollow circle at deck girder level (+70mPD) and taper- taining adequate clearance of the cantilever tip from the tunnel wall,
ing again to an 11 m-diameter circle at + 175mPD. The upper towers the land span was slightly curtailed, with substituted spring restraint
have a circular shape narrowing from 11 m to 7 m in diameter. to retain correct mode shapes affecting the cantilever response.
The temporary work and construction equipment used for the tower The analytical investigations earlier supported the supposition that
erection consisted primarily of a tower crane, a jib crane, working plat- the full half-main span, maximum cantilever condition, denoted by
forms, bracings and ancillary items (Fig. 3.1) and they were modelled at Erection Stage ES34, Mode Number 6, was the most onerous case,
1:200 scale for the aeroelastic tower model testing (Fig. 3.2). with dynamic response to buffeting placing severe demands on the
The wind tunnel investigations for the bridge tower construction in- structure. The modal data of the first six eigen-modes of Erection
cluded an assessment of the vortex shedding resonant response and Stage ES34 are given in Table 4.1. The mode shapes of the first six
aerodynamic stability of the tower during erection, together with an as- eigen-modes of Erection Stage ES34, and their respective descriptions
sessment of the peak wind-induced overturning moments, based on a are given in Figs. 4.2a to f. However, examination of the changes of dy-
series of 1:200 scale aeroelastic model tests. Four erection heights namic modal patterns as a function of cantilever length had identified
were selected for the investigations: 119 m, 174 m, 246 m and 293 m one other stage of interest, at an erection cantilever some 100 m shorter
(full-height freestanding tower). As part of the wind tunnel studies, than the maximum cantilever condition denoted by Erection Stage ES40
the aerodynamic interference effects of construction plant and equip- (see also Section 1.2). The modal data of the first six eigen-modes of
ment for the respective construction stages on the mean wind loading Erection Stage ES40 are given in Table 4.2. The test programme was
and the vortex shedding response of the tower were examined therefore repeated with the aeroelastic model shortened accordingly.
(Fig. 3.3). Aeroelastic modeling of cable-stayed bridges is not subject to the
The aeroelastic tower model wind tunnel investigations verified the same limitation in suspension bridge modelling. In suspension
structural adequacy and aerodynamic performance of the tower during bridge aeroelastic modelling dependence on gravitational stiffness pre-
the erection stages. A search was conducted to detect possible adverse vents free choice of wind speed scaling. Denoting the scaling factors
effects in certain wind directions, in which the wake from the tower (model value/full size value) for length as RL and for wind speed as RV,
erection crane might influence flow separation from the main body of conservation of aerodynamic similarity is satisfied by frequency scale
the tower. No such adverse effect was identified. Rn = RV/RL and mass scale Rm = R2L . Response deflections are then di-
The tests concluded that damping is effective in suppressing vortex rectly scaled, Ry = RL, but this implies that accelerations are increased
shedding response and that such response was generally reduced by Ra = R2V./RL. The wind speed scale RV = 1/8 was chosen to give the
the presence of construction plant/equipment. best balance between model robustness and wind speed range in the
The tower responses recorded in the testing are given in Figs. 3.4 tunnel, which with the chosen linear scale RL = 1/200, gave model fre-
and 3.5. quencies Rn = 200/8 = 25 times full-scale. The mass scale can normally

Table 4.1
Modal data of the first six eigen-modes of Erection Stage ES34.

Erection Stage ES34

Stage: East cantilever (leading segment: ES-34)

Model reference: 5029 - G1B backward eigen ES-33.mcb

Mode no. Circular frequency (rad/s) Frequency (Hz) Maximum displacement Node of maximum displacement Degree-of-freedom Generalized mass Remarks

1 0.708651 0.112785 0.000443 410,096 Y 0.001000302 Lateral mode

2 1,173,462 0.186762 0.0004 150,077 Y 0.000999012 Lateral mode
3 1.264947 0.201323 0.000638 430,096 Z 0.001000098 Vertical mode
4 1.947137 0.309897 0.000393 410,096 Z 0.000999726 Vertical mode
5 2.315941 0.368593 0.000284 420,111 z 0.001000403 Vertical mode
6 2.719029 0.432747 0.000692 350,027 z 0.000949684 Torsional mode
 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110 103

Fig. 4.2a. Erection Stage ES34 Mode Shape 1 - first lateral mode.

Fig. 4.2b. Erection Stage ES34 Mode Shape 2 - second lateral mode.
104 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110

Fig. 4.2c. Erection Stage ES34 Mode Shape 3 - first vertical mode.

be satisfied by careful weight control in selection of components for the adjuster. For the deck girder, the usual solution to this problem is to rep-
model, but the frequency scale requires that the elastic properties of el- licate the aerodynamic form with lightweight segmented non-
ement cross sections are R2V = 1/64 times the value that would be given structural cladding, concealing a slim elastic spine. For the twin-box
by direct linear modelling to scale RL. deck, concealment of the corresponding longitudinal spine members
In the case of the stays, this was met by introduction of small coil enforced their location much further apart than otherwise desirable.
tension springs at the upper end of each stay, together with a turnbuckle Considerable design effort was needed to achieve an acceptable

Fig. 4.2d. Erection Stage ES34 Mode Shape 4 - second vertical mode.
 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110 105

Fig. 4.2e. Erection Stage ES34 Mode Shape 5 - third vertical mode.

compromise for vierendeel bending in plan, while also matching the re- screw fixing at one end, and spring-borne from the spine at the other
quired vertical and torsional properties. The resulting spine was based end to prevent structural participation. Details such as railings photo-
on photo-etched lattice trusses 9 mm deep as longitudinal members etched in brass and much erection equipment were added on
within each box, with stiffer cross-members, fabricated in brass. The the decks and the towers. Two options for the deck lifting gantries,
cladding was fabricated largely by vacuum-forming thermoplastic and two access gantries below the decks, were modelled in detail
sheet, forming 270 mm (54 m full-scale) units each with adjusting- (Fig. 4.3).

Fig. 4.2f. Erection Stage ES34 Mode Shape 5 - first torsional mode.
106 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110

Table 4.2
Modal data of the first six eigen-modes of Erection Stage ES40.

Erection Stage ES40

Stage: East cantilever (leading segment: ES-40)

Model reference: 5029 - G1B backward eigen ES-27.mcb

Mode no. Circular frequency (rad/s) Frequency (Hz) Maximum displacement Node of maximum displacement Degree-of-freedom Generalized mass Remarks

1 1.03955 0.165449 0.000448 410,108 Y 0.000999524 Lateral mode

2 1.232441 0.196149 0.000396 150,077 Y 0.000999226 Lateral mode
3 1.490479 0.237217 0.000625 430,108 Z 0.000999853 Vertical mode
4 2.209547 0.35166 0.000358 410,108 Z 0.00100103 Vertical mode
5 2.691459 0.428359 0.000432 420,123 z 0.000999881 Vertical mode
6 3.15015 0.501362 0.000181 430,191 Y 0.00099867 Lateral mode
7 3.353757 0.533767 0.000765 340,021 z 0.000945609 Torsional mode

The tower mass and elastic properties were included in the dynamic Fig. 4.5 depicts the aeroelastic bridge model testing on Erection Stage
system, together with temporary work items carried on the tower. To ES34 in smooth flow, and Fig. 4.6 depicts the aeroelastic bridge model
examine sensitivity to damping, with special relevance to the possible testing on Erection Stage E40 in sea fetch.
provision of tuned mass dampers during erection, electro-magnetic The test results were largely within the bounds of anticipation, giv-
dampers giving controlled additional damping for each principal direc- ing high confidence into the incorporation with the computer model.
tion (lateral, vertical, torsional) were also provided (Fig. 4.4). The vertical responses were generally somewhat less than the ‘a-priori’
With this scaling, model motion accelerations were higher than full- analysis. The effect of the free end of the cantilever appeared only mod-
size, Ra = R2V/RL = 3.1. Vertical acceleration of the deck, acting with the est, and introduction of the end of the adjacent cantilever in close prox-
mass supported, clearly contributes to the stay tension. As the model imity had little consistent effect on critical responses (Fig. 4.7).
was clearly only valid in the near-linear regime of taut stays, pre-
tension was imposed on one or two pairs of stays towards the cantilever 5. Stay cable wind tunnel investigation
tip that were at risk of going slack. This was achieved by a connection
from the deck to the floor of the wind tunnel, acting through springs For the stay cable wind tunnel investigation, cable models were
which were very soft by comparison with the ‘true’ stays, to ensure manufactured at 1:1 scale to ensure Reynolds Number similitude be-
that there was no significant constraint of motion; the links were tween model and prototype; and using the same material as the full-
co-linear with the associated stays to minimise transmission of scale stay cable tube in order to reproduce the same characteristics of
horizontal forces to the deck. Accelerations substantially in excess rain water adhesion. Each cable model was supported on two pairs of
of ‘g’ were recorded in the exploration of behaviour beyond the de- linear springs installed in the test rig. The supporting springs were de-
sign wind speeds. signed to be representative of the typical full-scale stay cable support
The nature of the model elastic deck structure made it impractical to conditions.
measure stress resultants directly by strain gauging, and substantial The first series of stay cable wind tunnel testing sought to determine
computational studies were involved in evaluation of the required the static wind forces on the parallel wire strand stay cables of the
values for design verification from measurements by accelerometers Stonecutters Bridge (Fig. 5.1). For two selected cable diameters which
and non-contact displacement transducers. This was, however, compat- were exemplary of the actual cables used, with the surface texture
ible with the primary requirement of the physical testing to validate a (dimpled pattern) proposed for construction, static load measurements
comprehensive computer model which was then utilized for checking were taken. Such measurements were also taken on corresponding ca-
the full erection sequence, stage-by-stage, and for investigation of the bles with a smooth surface texture for comparison (Fig. 5.2).
sensitivity of buffeting behaviour to changes in site conditions and to The test matrix comprised measuring drag forces on two stay cable
construction loads. diameters, 139 mm and 169 mm, with two cable surface textures (one
smooth and one dimpled), at different inclinations relative to the

Fig. 4.3. Construction plant/equipment on the aeroelastic bridge model. Fig. 4.4. Electromagnetic dampers used in the testing.
 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110 107

Fig. 4.7. Aeroelastic bridge model wind tunnel testing - investigations of proximity effects.

Fig. 4.5. Aeroelastic bridge model wind tunnel testing on Erection Stage ES34 in smooth
flow.

longitudinal and transverse axes of the bridge. A total of twelve tests

were conducted. Fig. 5.3 shows the results of the drag coefficients mea-
sured from the twelve tests.
It was observed that the drag coefficients corresponding to incident
wind along the longitudinal axis of the bridge (at azimuth angles 8.84°
and 10.45°) were usually lower in magnitude than those due to incident
wind perpendicular to the longitudinal bridge axis (at azimuth angles
98.84° and higher). The test results also showed that, compared to a
smooth surface, a dimpled surface texture has a drag characteristic
that is less sensitive to changes in the other test conditions and param-
eters. For example for the 139 mm-diameter cable tested at an azimuth
angle of 10.4° and through the range of wind speeds used in the testing Fig. 5.1. Static load measurements on full-scale stay cable.
(20 m/s–50 m/s), the variation in drag coefficient measured on a dim-
pled surface texture is 0.03; while the variation in drag coefficient mea-
sured on a smooth surface under the same test conditions is 0.22.
The second series of stay cable wind tunnel testing sought to inves- The rain-wind induced vibration tests were conducted in an open
tigate the behaviour of the stay cables in respect of rain-wind induced circuit of a down-blowing type with an open test section. The test sec-
oscillations. In particular the objectives were to establish tion was 2.3 m wide by 1.5 m high, with wind speeds ranging from
1 m/s to 22 m/s. It was equipped with facilities which can reproduce
• the efficiency of the dimpled surface texture of the cable sheathing any rain and wind conditions required for this type of testing. The dis-
which was proposed for construction, in mitigating rain-wind in- placements of the cable model were measured by using force transduc-
duced oscillation, and ers installed at the top of the supporting springs in the test rig.
• the efficiency of mechanical damping in mitigating rain-wind induced Prior to the rain-wind induced vibration test, a cable model was ex-
oscillations. cited into a free-decay motion with an initial velocity. The time history
of the cable model's displacements was recorded, with which data the
damping level and frequency were then determined. In all the tests,
the damping levels of the model both at the commencement and at
the end of the test were verified to ensure that no significant change
of the model damping had occurred during the test.

Fig. 4.6. Aeroelastic bridge model wind tunnel testing on Erection Stage ES40 in sea fetch. Fig. 5.2. Dimpled pattern on HDPE cable sheathing.
108 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110

Fig. 5.3. Results of drag measurements on stay cables.

Fig. 5.4. Effect of damping on rain-wind induced vibration of 169mm-diameter cable.

Exploratory tests were conducted on dimpled cable models of 2 dif- +60° in increments of 5°, and at wind speeds of 5 m/s–10 m/s in incre-
ferent diameters, to determine the critical azimuth angles. Each cable ments of 1 m/s, and then in the range of 10 m/s–16 m/s in increments of
was tested at a medium rain intensity of around 80 mm/h, at 0.006 1 m/s but adjusted to fine increments of 0.5 m/s near the peak model
log dec cable damping, with the azimuth angle varying from −60° to response.

Fig. 5.5. Effects of damping on rain-wind induced vibration of 139mm-diameter cable.

 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110 109

Production tests on the two dimpled cables were conducted at the meant that relatively little benefit would be obtained through addition-
critical azimuth angles and wind speeds identified in the exploratory al (mechanical) damping; only a 10% reduction of the resonant contri-
tests, at rain intensities varying from 15 mm/h to 120 mm/h, in incre- bution to response in the first vertical mode was obtained by adding
ments of 5 mm/h or 10 mm/h, at 0.006 logarithmic decrement cable damping up to 0.15 logarithmic decrement. Given the difficulty of creat-
damping. The study continued at increased damping levels of 0.02, ing passive mechanical devices for damping of vertical motion at such
0.04, and 0.06 logarithmic decrement, at the critical azimuth angles low frequencies, mitigation of buffeting response in these modes
identified in the earlier testing and for a rain intensity of 80 mm/h. would be very unpromising. The aerodynamic damping of lateral re-
As a control, comparative tests were carried out on two cable models sponse was relatively modest, so that in this case additional damping
(2 different diameters) with smooth surfaces, at one azimuth angle could significantly reduce the dynamic displacements, although in the
identified in the exploratory tests, at 0.006 logarithmic decrement event this was not found necessary. Reduction by up to 40% was obtain-
cable damping, for a rain intensity of 80 mm/h and at wind speeds in ed by increasing damping to 0.2 logarithmic decrement.
the range of 5 m/s–16 m/s. The rain-wind induced oscillation tests on stay cables investigated
From the test results of the rain-wind induced vibration testing, the the effects of dimpled pattern on the cable sheathing and of increased
influence of rain intensity on rain-wind induced vibrations of dimple ca- damping on the dynamic behaviour of the stay cables. The dimpled pat-
bles was observed. For the 169 mm-diameter cable model, the most se- tern could suppress or alleviate rain-wind induced vibrations. Increased
vere oscillations occurred at an azimuth angle of 35° and at a wind damping was also effective in suppressing such oscillations. The drag
speed of 10.0 m/s. For the 139 mm-diameter cable model, the most se- coefficients measured on cables with dimpled surface texture were
vere oscillations occurred at an azimuth angle of 40° and at a wind within the permissible design values of 0.8 for transverse wind and
speed of 12.5 m/s. Using these sets of azimuth angles and wind speeds, 1.2 for longitudinal wind.
further investigations showed that for the 169 mm-diameter dimpled The work on the aerodynamic investigations for the Stonecutters
cable model, the most severe oscillations occurred at a rain intensity Bridge construction has accumulated a significant body of knowledge
of 80 mm/h at amplitude of around 100 mm which is 0.59 times of which has enabled the successful erection of the 1018 m main span
the cable diameter. The corresponding parameters for the 139 mm- cable-stayed bridge. It has also advanced the understanding of the aero-
diameter cable were a rain intensity of 80 mm/h and an amplitude of dynamic behaviour of ultra-long span cable-stayed bridges, and in par-
35 mm, which is only 0.25 times of the cable diameter. ticular some of the fundamental characteristics of a vented, twin-box
Using the sets of governing azimuth angles and wind speeds, the ef- type of superstructure.
fects of increased damping, at 0.019, 0.038 and 0.063 logarithmic decre-
ment were examined.
Figs. 5.4 and 5.5 depict the test results for the 169 mm-diameter Acknowledgements
cable and 139 mm-diameter cable respectively.
The dimpled pattern could suppress or alleviate rain-wind induced The authors wish to thank the Maeda-Hitachi-Yokogawa-Hsin
vibrations. Increased damping was also effective in suppressing such Chong JV and the Highways Department of Hong Kong Special Adminis-
oscillations. trative Region, for permission to publish this paper.
The section model and aeroelastic model wind tunnel tests, includ-
6. Conclusions ing model design and fabrication, were undertaken by BMT Fluid Me-
chanics Ltd., UK and the stay cable wind tunnel tests were carried out
This paper has described the extensive aerodynamics investigations at China Aerodynamics Research and Development Centre, China.
conducted for the Stonecutters Bridge construction. A series of theoret-
ical studies of the bridge behaviour during construction was conducted References
throughout the planning of the wind tunnel investigation, determina-
tion of test parameters and evaluation of test results. [1] Hui MCH, Larsen A. Aerodynamic investigation for the deck of Stonecutters Bridge
emphasising Reynolds number effects. Proc. Symp. Advances in wind and structures,
The section model testing verified the stability of the bridge against Busan, Korea; 2002.
divergent amplitude response during construction. Validation of the [2] Hui MCH. Full-bridge aeroelastic model wind tunnel tests for the Stonecutters
guide vane performance to control vortex shedding response led to Bridge. Hong Kong Inst Eng Trans 2013;20(2):109–23.
[3] Ishikura Y, Kuwabara K, Sham R, O’ Grady R. Stonecutters Bridge Hong Kong: main
their installation prior to deck lifting. The section model demonstrated
span erection. Proc. instn. civil engineers, bridge engineering 165 BE1; 2012.
the significance of the vertical and torsional aerodynamic damping p. 67–76.
compared with intrinsic structural damping, and hence the dominance [4] Kwok SCT, Sham SHR. Contract HY/2002/26 Stonecutters Bridge technical memo-
of aerodynamic damping in limiting vertical and torsional buffeting re- randum no. TM023 eigenvalue analysis of bridge erection stages. Maunsell Consul-
tants Asia Ltd.; 2005[93504/TM/023].
sponse. The static wind loading measurements on the section model [5] Larsen A, Yeung N, Carter M. Stonecutters Bridge, Hong Kong: wind tunnel tests and
identified the significance of the temporary handrail system on the studies. Proc Inst Civ Eng Bridge Eng 2012;165(2):91–104.
overall drag of the bridge deck. Additional tests validated a form and [6] Maeda-Hitachi-Yokogawa-Hsin Chong JV. Stonecutters Bridge gateway to Hong
Kong's port, Hong Kong. [ISBN: 9881778891] Bonham Kasyan Media Limited; 2010.
configuration of temporary safety handrail system to limit the lateral [7] Morgenthal G, Yamasaki Y. Behaviour of very long cable-stayed bridges during erec-
wind forces on the structure during construction to acceptable levels. tion. Proc Inst Civ Eng Bridge Eng 2010;163(BE4):213–24.
The aeroelastic tower model wind tunnel investigations verified the [8] Selberg A. Oscillation and aerodynamic stability of suspension bridges. Civil engi-
neering and construction, 13. ACTA Polytechnica Scandinavica; 1961.
aerodynamic performance and structural integrity of the tower during [9] Sham SHR. Contract HY/2002/26 Stonecutters Bridge wind tunnel test brief section
the erection stages including the full-height freestanding conditions. model testing, issue 0. Maunsell Consultants Asia Limited; 2003.
Damping was effective in mitigating vortex shedding response and [10] Sham SHR. Contract HY/2002/26 Stonecutters Bridge wind tunnel test brief full
aeroelastic bridge model testing, issue 0. Maunsell Consultants Asia Limited; 2003.
such response was in general reduced by the presence of construction [11] Sham SHR. Contract HY/2002/26 Stonecutters Bridge wind tunnel test brief full
plant and equipment. aeroelastic freestanding tower model testing, issue 0. Maunsell Consultants Asia
In the aeroelastic bridge model testing, the vertical responses re- Limited; 2003.
[12] Sham SHR. Contract HY/2002/26 Stonecutters Bridge wind tunnel test brief stay
corded were generally less than the “a-priori” analysis. The buffeting re-
cable testing, issue 0. Maunsell Consultants Asia Limited; 2003.
sponses recorded in the aeroelastic model testing also corroborated the [13] Sham SHR. Sutong bridge contract C3 wind tunnel test briefs for section model,
observations made in the section model tests, in that control of the ver- aeroelastic freestanding tower model and aeroelastic bridge model testing. Maunsell
tical and torsional responses to buffeting was dominated by aerody- Consultant Asia: Limited; 2005.
[14] Sham SHR. Construction engineering for the world's longest cable-stayed bridges.
namic damping. The very large aerodynamic damping of vertical Proc. 24th annual international bridge conference, Engineers Society of Western
motion, up to 0.6 logarithmic decrement at design wind speeds, Pennsylvania, 4–6 June 2007, Pittsburgh; 2007. p. 249–59.
110 S.H.R. Sham, T.A. Wyatt / Structures 8 (2016) 94–110

[15] Sham SHR. Frontiers of large cable-stayed bridge construction, McGraw-Hill year- [17] West BW, Sham SHR. Concrete engineering in Stonecutters Bridge Hong Kong. 2007
book of science & technology 2013. [ISBN 978–0–07-180140-9] New York: ASBI convention; 2007.
McGraw-Hill; 2013 158–70. [18] Wyatt TA. The dynamic behaviour of cable-stayed bridges: fundamentals and para-
[16] Sham SHR, Wyatt TA. Wind tunnel investigations for Stonecutters Bridge construc- metric studies. In: Ito, et al., editors. Cable-stayed bridges, 40. Elsevier developments
tion. International conference on bridge engineering challenges in the 21st century in civil engineering; 1991. p. 151–70 [ISBN 0 444 89055 6].
1–3 November, Hong Kong; 2006.