Steel Innovations Conference 2013
Christchurch, New Zealand
21-22 February 2013
HARBOUR CITY CENTRE WELLINGTON: SEISMIC STRENGTHENING
A.W. Thornton1, M.E. Davies2, A.G. Cattanach3
ABSTRACT
The existing Harbour city Centre structure was designed and constructed circa 1929 and consists of concrete
encased riveted steel frames (beams and columns). It has a high ground floor space, 6 suspended floors
and a concrete roof.
In 1979 concentric chevron bracing was added, with 4 bays of chevrons in each direction (longitudinal and
transverse). At elastic stress levels these braces only provided approximately 30% of NZS 1170.5
requirements beyond which brittle compression brace buckling and connection failure would occur.
Their performance has been significantly improved by the installation of axial (in-line), hysteretic dampers
which will enable the braces to yield in both compression and tension without the braces buckling.
Scale prototypes and a full mock-up damper was production tested under prolonged cyclic action. The
results from the testing were correlated with an analytical model. Mechanical and hysteretic performance was
excellent.
Non-linear time history analyses were carried out to verify the design with respect to overall building
response, to 100% NBS load levels including near field effects. The analysis/design was Peer Reviewed and
given Building Consent.
During 2011 the project was successfully completed in a staged manner with much of the building remaining
occupied during the retrofit. The strengthening has assisted the client in securing significant, high-value lease
agreements for the building.
Introduction
The Harbour City Centre project initiated from an IEP review of the building’s seismic capacity by Wellington
City Council as part of their Earthquake Prone Building policy resulting from the 2004 Building Act.
Dunning Thornton carried out a review of WCC IEP, and in examining the building and its history identified
the existing bracing retrofit in 1979 as the critical elements requiring assessment and retrofit Arranged in a
concentric manner the bracing was prone to brittle, compression failure. Additionally, the arrangement of the
end connection where the braces attached to the weak axis of the existing riveted plate columns was
identified as having a tension capacity of approximately half of the compression buckling load capacity of the
brace. Following an assessment of these weaknesses, Dunning Thornton reported to the owner their
concurrence that the building was indeed likely to be Earthquake Prone.
1
Director, Dunning Thornton Consultants Ltd., Wellington N.Z.
2
Design Engineer, Dunning Thornton Consultants Ltd., Wellington N.Z.
3
Director, Dunning Thornton Consultants Ltd., Wellington N.Z.
A study of possible strengthening options was commissioned. The
building lies on a key part of Wellington retail’s high retail value
“Golden Mile” and as such, prolonged disruption to the ground floor
would have a significant effect on the income stream from the
building. The upper floors were of more mixed income quality: the
key tenant were about to embark on a refit of their open plan
tenancy, but the floors above this comprised many smaller office
spaces often still using the original breezeblock partition walls.
The existing braces, retrofitted in 1979, were positioned throughout
the floor plate with many located in the existing central light well to
minimize intrusion into the floors. Preliminary analysis indicated that
if ductility could be introduced into the brace performance, they could
meet a significant proportion of 1170.5 demand. While the existing
braces compromise the space to some extent, their presence had
been accepted and the addition of alternative or further new
elements was considered undesirable and potentially more
disruptive.
The focus of the subsequent study was therefore how the performance of the existing braces could be
improved. Elastic strengthening of the braces by plating up the UC sections would overload the existing end
connections, columns and foundations. Options to add energy dissipation to the braces and to minimise the
overstrength actions from them was identified as most desirable.
Replacement with new proprietary viscous or buckling restrained braces would provide this performance
objective, but would involve significant disruption to the (live) building, and have a significant cost. The
strategy of cutting the existing braces and inserting new specifically designed energy absorbing devices was
proposed. This paper focusses on the development of these devices, their testing and the analysis of the
retrofit building.
The intent of the works was to improve the seismic resistance of Building One such that satisfactory
performance occurs during a seismic event with a 1/500 annual probability of exceedance. In other words,
the strengthened structure shall comply with current building code seismic requirements.
Description of Existing Structure
The existing structure (Building One) was designed and constructed circa 1929. It has a high ground floor
space, 6 suspended floors and a concrete roof. It abuts an older, timber floored building constructed around
1900 that is 3 stories high (Building Two). The two buildings are currently tied together but Building Two is
scheduled for demolition and redevelopment.
The structure of the 1929 building consists of concrete encased riveted steel frames (beams and columns).
The floors are reinforced concrete supported on a grid of secondary and primary concrete-encased steel
beams.
On three sides (north, south and west) there are concrete facade frames: the beams are again concrete-
encased steel but the columns are further stiffened by the extension of the encasement to form wing walls.
At the east side are two stair/lift towers. These are essentially steel-framed with concrete infilled panels. The
remainder of the structure is an open framework.
The columns are all founded on concrete pads/piers sitting on rock, 1-3m below Lambton Quay.
In 1979 concentric chevron bracing was added, with 4 bays of chevrons in each direction (longitudinal and
transverse). At elastic stress levels these braces only provide approximately 30% of NZS 1170.5
requirements. Beyond the elastic limit compression brace buckling and connection failure would occur. More
fundamentally, a critical weakness in one of the chevron connection types was also identified; this would likely
lead to premature failure even before the brace capacity was fully developed.
Proposed Strengthening
Working within their elastic limit, the braces provide approximately 30% of NZS1170.5 requirements in
compression. During 20011 their performance was improved by the installation of axial (in-line), hysteretic
dampers which will enable the braces to yield in both compression and tension without the braces buckling.
The resulting dampers utilise steel pins fitted through the cut braces, put into “two-point” bending by the
addition of flange plates across the cut. This would retain the high initial stiffness of the dampers, but allow
significant post-yield capacity, hence effective ductility, through curvature of the pins over a long length.
Figure 1. Elevation of a typical damper.
Figure 2. Yielded pin after testing.
The design challenges were around the connection points of the pins; how to allow the high rotations around
the bearing pints without inducing double-curvature to the pin or unwanted additional friction forces. The final
design employs a combination of brass bearing seatings, and proprietary SKF shaft bearings.
Figure 3. Section through pin showing bearing arrangement.
Other strengthening elements for the building included:
- The addition of an additional bay of conventional k-bracing to the north side to balance horizontal
irregularity (resisting torsion).
- The addition of vertical anchors to resist uplift and to increase the compression capacity of the
foundations.
- Re-jacketing (encasing) of some columns to improve axial capacity and to provide additional
resistance to soft-storey inclinations in the high ground floor space (and associated bracing of the
mezzanine floor).
- Strengthening of selected brace connections to the existing beams and columns (in doing so
eliminating the critical weaknesses).
- Conventional compression strengthening of some of the upper floor braces.
Blockwalls are provided between the mezzanine floor and the ground floor. These walls provide acoustic
separation and, in some locations, support gravity loads from the mezzanine floor. To mitigate “short
column” effects the walls have been de-stiffened in-plane as much as practicable by the introduction of
vertical joints/sawcuts.
Interaction with Building Two
The structural work detailed will improve the performance of Building One, though the facades are built
together with those from Building Two. The full performance benefit from the upgrade will not take effect until
Building One is seismically isolated from Building Two, or Building Two is removed. In the interim Building
One will not meet 100% NBS however it would still remove the building from earthquake prone status and
achieve an estimated 80%NBS.
Future Additions
An allowance has been made for a possible future high-rise structure on the site of Building Two that would
partially bear on the existing Building One. The future structure was assumed to extend nine storeys above
the existing structure. Lightweight (steel framed) construction was assumed and the additional vertical loads
to be imposed on the existing structure would be of the order 6000kN.
One line of columns was strengthened to increase their gravity load capacity for this demand. This allowance
was solely for gravity loads and does not account for the additional P-delta seismic demands that are inherent
with increases in vertical load. Analysis of the combined structures would be required should any
redevelopment occur, and the combined structure would need to perform to full new building standard.
Testing
The dampers went through a two-stage testing process. Dunning Thornton particularly thank Chris Gannon
from Robinson Seismic for assistance with organizing the testing and with assistance in brainstorming ideas
of how to achieve robust detailing around the pin connections.
The first stage involved prototype testing of a single dowel and led to development of the seating/bearing
details at the end of the dowel. The successful hysteretic performance of the dowel was confirmed.
Figure 4. Prototype damper setup.
In the second stage a full mock-up damper/brace arrangement was production tested under prolonged cyclic
action. The results from the testing were correlated with an analytical model. Mechanical and hysteretic
performance was very satisfactory.
As all properties of the brace system were theoretically derivable from “code” physics, the testing was
determined to be “production testing” in accordance with AS/NZS 1170.0 provided that the results correlated
with the theoretical predictions. A single test would therefore be sufficient as the design was not dependent
on achieving a 95% certainty from the “scatter” of results that would arise from a system that was not
theoretically predictable.
During developed design the damper was tested beyond the limit of its flexural (hysteretic) displacement to
determine its ultimate capacity. The capacity of the damper was found to exceed the rated capacity of the
testing equipment. Hence, the ultimate failure capacities of the dampers are unknown. The ultimate demand
on the dampers in this application is within the tested load.
Typical results from testing are shown in the graph below.
Figure 5. Production testing results.
.
Structural Modelling
The structure was recreated as a three-dimensional model in ETABS for the purpose of undertaking non-
linear time history analyses and verifying the design. Non-linear elements are confined to the strengthened
“brace” elements and a rocking wall each side of the east-side stairwells.
In order to allow for the effect of P-Δ, the approximate average horizontal centre of mass was determined and
a column member was created in this location. The column member was pinned at the top and bottom of
each storey so that it did not provide any lateral resistance, and had a very high axial stiffness so that its
natural period was less than that of the lateral modes used in the analysis. A vertical point load representing
the weight of each floor was applied at the relevant height.
The rotational inertia was calculated manually and input for each floor, and extra mass was included at first
floor level as an allowance for the mezzanine. The horizontal masses assigned to each (rigid) diaphragm
were sourced (by ETABS) from the vertical loads and mass moments of inertia applied to the “column” in the
model. The structure’s effective vertical centre of mass was found to be close to level four.
Some members of the existing structure were de-stiffened from their gross concrete stiffness such that their
contribution to resisting base shear was limited to their capacity. The additional ductility demand on these
elements was also considered.
Actions were checked on the existing plated steel girders that contribute to frame action. The contribution of
the girders was limited to the nominal capacity of the typical riveted beam-column connection. Flexural
yielding of the steel angles fixed to the flanges was found to be the governing mechanism (note that this is a
ductile yield mechanism).
The concrete façade, which has a significant effect on the shape of the first mode, was also de-stiffened in
the model. As the short beam members will be subject to high shear demands, any concrete encasement
would crack. Accordingly the beam elements were assigned the stiffness of the bare steel beam section
only. The stiffness of the columns in the façade was reduced from that of the gross concrete section to
reflect the reinforcing in the wing walls.
To re-create the observed behaviour of the dampers, three non-linear link elements were used in parallel.
The link elements were as follows:
A “plastic” link – a bi-linear hysteresis with properties such that the initial stiffness, yield force and
post-yield stiffness agreed with tested and theoretical results.
A “hook” link – a link element with no stiffness except for when its positive (tensile) displacement
exceeds a specified value. The value used was the displacement at which the damper ran out of
free travel and began to stiffen. The stiffness of this link was such that, when combined in parallel
with the “plastic” link, it agreed with experimental results
A “gap” link – as for the “hook” link, except using negative (compressive) displacement.
To verify that this combination of links produced the required behaviour, a separate model was created with a
single pair of braces and a mass. The output from this model suggested that the combination of links does
not create numerical instabilities and generates the appropriate hysteretic behaviour.
The eastern side of the building contains a stair well and elevator shaft at each end, and reinforced concrete
walls (constructed in relation to damage following 1942 earthquake) are positioned in these areas. The
capacities of these walls were assessed (refer section M) and it was determined that some would rock.
Some walls were not included in the model because there was insufficient diaphragm capacity to transfer
lateral demands. The historic repairs had only been undertaken at the first two levels, hence only these
levels were modelled. Modelling of these walls was considered necessary because their corner locations
mean they are likely to have an influence on the structure’s torsional behaviour. The historic strengthening
also provides part of the resistance to the ground floor’s inclination to otherwise form a soft storey.
These walls were modelled as rocking elements with vertical loads on top of the walls to represent the gravity
load borne by the walls. The gravity load is eccentric and the distribution of the modelled vertical load reflects
this. To create rocking behaviour, “gap” or compression-only springs were provided as vertical restraints.
The stiffness of these springs was approximated to the foundation conditions. Similar to the braces, a
separate model was created to verify that the behaviour of walls matched what was intended. Dynamic
(vertical “pounding”) effects on the column bases are not considered critical because of the low stresses in
the wall and the small contribution of the walls to the total base shear. The vertical loads on the walls did
generate some vertical modes but inspection of the participating masses shows these walls to have negligible
influence on the structure’s response.
Seismic Demands
Earthquake records were, in general, scaled using the procedure prescribed by NZS 1170.5. However, in a
departure from 1170, seven time-history ground motion records were used and the records were scaled so
that their average spectra matched the design spectra found using NZS 1170.5. The period range of interest
was taken around the effective period. Design demands were determined from the average maximum of
records.
While conducting the time-history analyses, it was observed that, during oscillations that corresponded to
velocity pulses of forward-directivity/near fault events (typically the Lucerne record), the dampers exceeded
their design displacement and their stiffness increased accordingly. The effective period of the structure
therefore shortens during this peak cycle. Comparing the NZS 1170.5 spectra and the scaled spectra at this
effective period, the acceleration of the NZS 1170.5 spectra is noticeably higher. To account for this
overestimation of demand, the magnitude of the scaling was reduced so that the average of the records’
spectra coincided with the NZS 1170.5 design acceleration at the critical effective period (i.e. when the
structural demand is the greatest).
Figure 6. Spectral trace at the point of interest.
Acceptability Criteria
An unacceptable response is considered to be failure of members or connections which results in either a
loss of gravity load bearing capacity or a significant, irrecoverable reduction in lateral stiffness. Specifically,
these were considered to be:
Buckling of columns.
Shear failure or confinement failure (i.e. “bursting”) of reinforced concrete column encasements.
Foundation failure or uplift below braced frames.
Failure of a brace connection.
Fracture of a damping device (ultimate capacity taken as the maximum tested load).
Buckling of a brace is assessed on the circumstances of the buckling. If, for example, a brace buckles
multiple times during a seismic event, this is considered unacceptable. However, if the brace only buckles
under a single extreme cycle due to, for example, a single near-field velocity pulse, this was considered
acceptable provided that:
The girder that the brace connects into can form a uni-directional plastic hinge at mid-span, and the
connections along the relevant load paths have been detailed to resist the over-strength actions (i.e.
collapse of the floor is prevented).
The brace buckles very close to the peak of the cycle and the reduction in stiffness is not significant
enough to allow a soft-storey collapse to occur due to P-Δ effects (i.e. one brace buckling at the
edge of the structure, not all braces on the floor buckling).
Inter-storey drift of the floor restrained by the brace is considerably smaller than 2.5% (i.e. there is
ample opportunity for the ground motion to reverse and “catch” the building).
Results and Design Checks
Axial capacities of columns subject to brace loads were checked. Peak brace loads were found by summing
the plastic link and the largest absolute value of the hook or gap links. These were checked against the axial
capacity of the critical braces and the maximum tested capacity of the dampers. Upper level (elastic) braces
were found to buckle prematurely and accordingly conventional strengthening is to be provided in these
locations.
Maximum storey displacements were within the drift limits specified by NZS1170.5. Due to there being no
adjacent structures, there is no issue with pounding (Refer to introduction regarding separation of Building
Two).
Notably, under records with near-field characteristics the dampers exceeded their intended maximum
displacement (i.e. “bottomed out”). The tangent stiffness and secant stiffness of the dampers increased, and
effective damping therefore decreased accordingly.
Inspection of the backbone of a base shear vs. centre of mass displacement plot shows an idealised ductility
of 1.35. Using this ductility and the effective period at the peak displacement to derive the equivalent static
base shear results in a value similar to that observed in the time-history output.
F
Idealised
yield point
Observed
backbone
ΔY 0.35ΔY
Δ
Figure 7. Backbone curve.
The near-field characteristics tended to induce greater displacements and, in combination with a higher
effective stiffness, this reduces the hysteretic damping considerably. If the intensity of the near-field events is
reduced by 10-20%, the dampers are far more effective. The table below presents the base shears (in kN)
as reported by ETABS and as calculated using a displacement-based design (DBD) approach. The
noticeable discrepancy in values for near-fault events occurs because one record causes a large, one-off
spike (~30,000 kN) in base shear. This record possesses forward-directivity characteristics.
Excitation Characteristics DBD Calculation ETABS Output
Near-fault 20,890 23,130
Far-field 15,660 16,800
Torsional Response
To assess the torsional response, the structure’s centre of mass was shifted to induce a torsional response.
The mass was offset by 10% of the structure’s width towards the north-western corner and also by the same
distance towards the south-eastern corner. The time-history analyses were then re-run (without adjusting
record scaling). The output from this can be seen on page EP80 of section I and is reproduced below. The
displacements shown are the average of the peaks from seven earthquake records. For a north-south
excitation the structure displays considerable torsional robustness. However, under east-west excitation the
structure’s behaviour becomes more torsional.
Variation of Level Four Displacements with Horizontal
Centre of Mass Location
250
North-west mass
Level Four Displacement (mm)
200
Central mass
in the Principal Direction
South-east mass
150
100
50
0
South-east corner; North-west corner; South-east corner; North-west corner;
East-west excitation East-west excitation North-south excitation North-south excitation
Location of Displacement Measurement; Principal Direction of
Excitation
Figure 8. Outcome of torsional response studies.
It should be noted, due to the nature of the structure, that greater than 75% of the mass is dead load which
has been explicitly located at their respective centres of mass. Therefore actual eccentricity is likely to be
considerably less than 10%.
Brace Design
To allow the structure to withstand the demands of a 1/500 probability of exceedance event, strengthening of
braces, brace connections and foundations has been detailed. Demands were determined from the non-
linear time history output, and the mean load in the element under consideration was found for the seven
records. This average inherently includes some over-strength actions as the link elements (representing the
dampers) exhibit “hardening” so it is therefore acceptable to apply a strength reduction factor of unity, also
the steel frame has some natural nominal ductility. However, when designing ground anchors, as they are
non-ductile, a strength reduction factor of 0.8 was applied.
Steel plating was added to the upper level braces (without dampers) to increase their compression capacity.
Although the compression capacity was increased, the expected failure mechanism is still buckling (i.e. the
plating is not significant enough for yielding of the non-plated cross-sections to govern). Accordingly, the
capacity of these braces’ connections needed to be evaluated. Generally, the existing bolted connections
had sufficient capacity to transfer the brace loads. However, the welded connections of the north-south
braces (a critical structural weakness) were determined to be ineffective and are to be strengthened by
adding steel gusset plates above and below the braces.
The lower level braces (with dampers) do not require plating. To improve robustness, the braces’
connections were strengthened such that if the compression brace was to buckle, the beam at the apex of
the braces can yield. The connection strengthening was designed to develop the over-strength capacity of
this beam.
Installation
The methodology for step-by step installation of the dampers to the braces was developed by Dunning
Thornton. The structural steel subcontractor (MJH Ltd) developed this further to include their new clip-on,
computer-controlled cutting torch, which was faster and more accurate than cutting the holes by hand. The
simple methodology meant that installation of the dampers into the existing braces got quicker as the project
progressed. The methodology can be summarised as follows:
Cut out the flanges of the UC brace, leaving the web in place to keep the brace in alignment during
the installation.
Form the holes through the UC flanges for the pins using the cutting torch.
Fit up the pins, bushes, bearings and plates.
Complete the site welds and cut out the remaining section (web) of the UC brace.
As the on-site work was deliberately kept simple and minimal, the installation time for the dampers was
shorter than anticipated. This was beneficial to the contractor, considering they were managing multiple work
faces within a live office building.
The strengthening was completed in a staged manner during which much of the remainder of the building
remained live. In this was loss of rental income for the owner was minimised, and decanting of occupants
could occur so that desirable tenants would not seek alternative leases elsewhere.
Figure 9. Insitu completed damper.
Conclusion
Rigorous analysis was carried out using non-linear time history techniques and good correlation was
achieved with both the force-based and the displacement-based approximate analyses. This encourages the
belief that the model is a fair representation of the structure (provided that the fundamental assumptions
common to all three methods are correct).
The completed structure would be expected to perform in excess of code requirements under far-field events
(such as a rupture of the Wairarapa Fault). The axial dampers limit the demand on the upper levels and
provide reasonable damping.
Similarly, the completed structure would be expected to perform well under a full-code near-fault event (such
as a rupture of the Wellington Fault). Again, the axial dampers – although not as effectively – limit the
demand on the upper levels and provide damping. Under an event with a 1/500 probability of exceedance,
the dampers would likely exceed their “damping” design displacement (i.e. they bottom out) but buckling of
the braces would not be expected to occur. Therefore, the structure has sufficient capacity to perform in a
predictable, dependable and repairable manner under an event of this magnitude.
When subject to a near-fault event which contains a forward-directivity characteristic (such as a rupture that
initiates in the Hutt Valley and propagates towards Wellington), the structure would not perform in an entirely
optimal manner. The “pulse” which characterises this type of event causes a single significant excursion
outside of the structure’s design displacement. This one-off excursion is likely to cause some lower level
braces to buckle in compression. However, the strengthening details provided have been designed for a
secondary mechanism (pull-down of the steel girder above) to develop during the single cycle of the pulse.
This mechanism ensures that lateral resistance is not lost entirely and hence avoids collapse in this low-
probability event. It is of note that a forward-directivity pulse of 70% of code does not cause brace buckling.
It should also be noted that NZS 1170.5 states “the inclusion of the forward-directivity component will result in
actions more consistent with much lower annual probabilities of exceedance” (refer clause C7.5 of the
commentary.
The seismic retrofit was completed successfully within an occupied building. Key tenants were retained,
including a major anchor tenant whose nature of business set a high priority on the seismic capacity of their
building. With the subsequent Canterbury series of earthquakes, Wellington lessees have become very
sensitive to buildings’, seismic performance. By embarking on this bold and innovative retrofit project, the
owners have secured high-quality, desirable tenants in a key CBD location.
Acknowledgments
Chris Cannon – Robinson Seismic: Details ideas and testing
Geoff Sidwell – Aurecon: Peer Review
John Milford – Kirkcaldie & Stains: Client
LT McGuinness: Compliant Contractor
