Professional Documents
Culture Documents
ranged from 0.254s to 0.272s) at the base of the frame is 2.347g for the final phase of
testing, stage 10. The spectral acceleration for the corresponding period of the original
unscaled Hollister record is 0.323g. Consequently, despite scaling the PGA by a factor of
22 from a value of 0.12g to 2.7g, the effective scaling factor for the periods of interest is
closer to 7.3. Figure 5.9 also shows that for the intermediate levels (stage 2) of excitation,
the filtering has had an even greater effect on the spectral accelerations.
5.3.5 Conclusions for the dynamic shake table investigation of a steel frame
This fully dynamic investigation of a steel frame has been successful in achieving its
three aims as outlined by the following points.
1. A shake table test has been performed using a real earthquake ground motion.
2. A level of excitation has been obtained experimentally that causes the steel frame
to reach a displacement ductility of five, which may be considered as collapse.
3. An analytical model of the frame has been developed and compared with the test
response.
Unfortunately however, the test did encounter difficulties related to load application.
It has been shown that the shake table filtered the earthquake motion so that much lower
accelerations were applied to the frame than intended. Consequently, the results appeared
illogical during performance of the test and only became reasonable when the actual
applied base motion was examined. It is postulated that this filtering was caused by the
bolted connection of the frame to the shake table being too loose.
In the context of this report, the investigation has served two purposes. Firstly, it has
demonstrated how a shake table test may be carried out in practice. Secondly, the
investigation has served to demonstrate how sensitive a shake-table test can be to
excitation filtering caused by inadequate test set-up. This emphasises the points made
within this report regarding the value of experimental set-up and also the highlights the
importance of placing instrumentation that will allow assessment of the actual applied
input excitation.
shown in Figure 5.10. A hollow R.C. pier is studied because it is considered that hollow
piers may behave differently than the solid piers on which the development of most
current assessment procedures have been based. In addition, the inelastic shear behaviour
of R.C. members is still not well understood, even for new solid R.C. members.
Consequently, the aims of this case study are:
To achieve these aims, a cyclic quasi-static test up until failure will be performed on
the hollow R.C. pier. It is noted that several similar investigations would be necessary
before any strong conclusions could be made about the non-linear behaviour of such
piers. However, the results for the single pier shown in Figure 5.10 will emphasise the
importance of careful experimental set-up. This section includes excerpts from [115]
where similar tests on a number of hollow R.C. piers are reported. For detailed results
and a comprehensive discussion on the behaviour of these other hollow rectangular piers,
readers are referred to [115].
The concrete cylinder compressive strength of the pier was found to be 30.3 MPa.
Reinforcing steel yield strengths were also obtained through testing; 520MPa for
longitudinal reinforcing and 710MPa for transverse steel. An axial load of 250kN,
corresponding to an axial load ratio of 0.07, was applied to the pier through the post
tensioned bar shown in Figure 5.10. The test set-up is discussed further in Section 5.4.5.
116 Timothy J. Sullivan, Rui Pinho & Alberto Pavese
1200 mm
200 mm
375 mm 450 mm 375 mm
450 mm
Load
line 300 mm
Post-tensioned
bar
1350 mm
1250 mm
450 mm
300 mm
Long. Bars: 40 φ8
600 mm
The pier was discretized into ten elements for which 200 monitoring points were
introduced. As can be seen from Figure 5.11, a fine mesh was adopted for the base of the
pier that receives the highest curvature demands and is consequently where the inelastic
behaviour is concentrated.
A concrete crushing strain value of 0.004 was adopted, as recommended in [68] for
design and assessment. Note that it is suggested in [68] that spalling may be expected
between strains of 0.006 and 0.010, therefore an even higher value of crushing strain may
have been appropriate due to the confining nature of the support. A concrete
confinement factor of 1.11 was determined in accordance with the Mander [80] model.
It was assumed that the tensile strength of concrete was zero for the pushover analysis
since concrete will not contribute greatly to flexural strength. It is acknowledged that
some tensile strength is important for reinforcement bonding to concrete yet such
behaviour was not explicitly modelled in the pushover.
Reinforcement strain hardening was incorporated linearly within the program and
strain penetration into the footing was allowed for by an increased pier height. This
effective length was set equal to the clear height of the pier plus the strain penetration
118 Timothy J. Sullivan, Rui Pinho & Alberto Pavese
Note that this effective length is only the appropriate height for flexure since the shear
will develop between the clear faces of the pier. However, this length was considered
reasonable for the pushover model since the shear capacity is determined independently
and superimposed on the flexural capacity curve.
200000
150000
Base Shear (N)
100000
Yield displacement = 7.9mm
50000
0
0 10 20 30 40 50
Top Displacement (mm)
Figure 5.12. Pushover curve for the pier modelled in SeismoStruct [69].
An Introduction to Structural Testing Techniques in Earthquake Engineering 119
The second phase of the analytical assessment is to predict the shear capacity of the
pier. Experimental evidence produced in the last ten years has made clear that traditional
code formulations to predict shear strength tend to be overly conservative at low levels of
ductility and increasingly non conservative at high levels of displacement ductility. In
general, most models follow the traditional approach of computing the shear strength as
the sum of a concrete and a steel contribution (see, for example the ATC-40 formulation
described below). More recently, revised models have been proposed that include a
reduction of the concrete contribution as a function of the flexural ductility demand as
presented in [74].
The following shear models (refer [81] for details of all models) have been considered
and compared in this investigation:
The ATC-40 model computes the shear strength (Vn) as the sum of a concrete (Vc)
and a steel (Vs) contribution as shown in Eq. (5.5).
Vn = Vc + Vs (5.5)
where Vc and Vs are given in SI units by Eq.s (5.6) and (5.7) respectively.
ª Pe º ' (5.6)
Vc = 0.29 « k1 + » f c (0.8 Ag )
¬« 13 . 8 A »
g ¼
Av f yh D '
Vs = (5.7)
0. 6 s
In the first of these equations, the factor k1 equals 1 for a displacement ductility of
two or less, and zero for a displacement ductility greater than two. Pe is the axial load, Ag
is the gross section area of the pier and f ‘c is the concrete cylinder compressive strength.
In the second equation, Av is the cross sectional area of the transverse reinforcing, with
120 Timothy J. Sullivan, Rui Pinho & Alberto Pavese
yield strength fyh, and hoop spacing or spiral pitch of s. Finally, D’ is the core diameter
measure from centre to centre of the transverse reinforcing. It should be selected for a
hollow pier by considering that the ratio D’/s is used to determined how many transverse
bars are effective in resisting the development of a diagonal shear crack.
The nominal shear strength for the Caltrans formulation also uses the formulation
presented in Eq. (5.5). In the Caltrans model, the steel component is found using Eq.
(5.7) and the concrete component is determined as shown in Eq. (5.8).
It is seen that these formulations are very similar to the ATC formulation, with the
main differences being the higher concrete contribution at ductility levels greater than
two, and no strength when the pier is in tension.
Both UCSD models recognise that there is a component of the shear strength that
does not degrade with ductility and which is not associated with the steel contribution.
The proposed expressions are consequently composed of three terms as shown in Eq.
(5.11), where an axial force component is added to the concrete and steel contributions.
Vn = Vc + Vs + V p (5.11)
Av f y (D − 2 cov+ d bs ) Av f y (D − c − cov )
VsO = cot θ VsM = cot θ (5.13)
s s
V pO = V pM = P
(D − c ) (5.14)
2L
An Introduction to Structural Testing Techniques in Earthquake Engineering 121
In Eq. (5.12) the effective shear area, Ash, can be taken as 0.8 Ag for solid sections,
while for hollow section the suggested values are given by Eq. (5.15).
0.35
0.30
UCSD-Mod
UCSD-Orig
0.25
γ (MPa)
0.20
0.15
0.10
0.05
0.00
0 2 4 6 8 10
Displacement ductility
The α and β terms shown in the UCSD-MOD version of Eq. (5.12) account for the
aspect ratio and longitudinal reinforcement respectively. The aspect ratio factor can be
calculated using Eq. (5.16) and the factor to account for longitudinal reinforcement
through Eq. (5.17).
1 ≤ α = 3 − H / D ≤ 1 .5 (5.16)
β = 0.5 + 20 ρ l ≤ 1 (5.17)
122 Timothy J. Sullivan, Rui Pinho & Alberto Pavese
where ρl=Asl / Ac is the geometrical reinforcement ratio for which Ac is the confined
concrete core area. Finally, it is noted that the UCSD-MOD equations have been derived
from results partially obtained through testing of hollow piers.
These four models have been used to determine the shear capacity of the section at
different values of displacement ductility. The next stage is to superimpose these shear
capacities onto the pushover curve obtained in Section 5.4.3.1.
400
ATC-40
300 Caltrans
UCSD-Orig.
200
Lateral Force (kN)
UCSD-Mod.
100 Flexural response
0
-100
-200
-300
-400
-80.0 -60.0 -40.0 -20.0 0.0 20.0 40.0 60.0 80.0
Displacement (mm)
Figure 5.14. Predicted non-linear response envelope for the hollow RC pier.
An Introduction to Structural Testing Techniques in Earthquake Engineering 123
Another important point is that the predicted capacity curves can be used to set
appropriate displacement limits for the actuator. From Section 5.4.3.1 it was shown that
the outer reinforcing bars should start yielding at a top displacement of 5.6mm. This
corresponds to a drift of 0.41%. Similarly the yield displacement for the pier of 7.9mm,
converts to a yield drift of 0.59%. Appropriate test drift levels might progress from 0.4%
initially, for which an elastic response should be expected, through to 3.0% which
corresponds to a displacement ductility of five. It is expected that at a displacement
ductility of five the pier will either have reached or be on the point of collapse. Note that
these ductility demands do not account for bar slippage or movements associated with
shear cracking.
The data acquisition system included global measurements of horizontal, vertical and
diagonal displacements, measurements of longitudinal deformation (curvature) in the
plastic hinge zones, some strain gauges on stirrups and longitudinal bars and verification
of possible base translation and rotation. In total, data was recorded at more than 80
points over the pier as indicated by Figure 5.15 and Figure 5.16 which present the
transducer locations for two sides of the pier. A similar number and arrangement of
transducers were also provided on the other sides of the pier.
The data for each transducer was collected and stored on a central computer that
monitored the results as the test proceeded. As shown in the photo of Figure 5.17, the
wiring associated with so many transducers is substantial and careful planning and
organisation is required simply to keep track of each transducer. The photo also clearly
shows the actuator used to apply the horizontal displacements and part of the top
actuator used to control the axial load.
124 Timothy J. Sullivan, Rui Pinho & Alberto Pavese
SIDE C
24 60
0 13
25
load cell
1 5 9 14 LP
26 59
LVDT
2 16
6 10
27 58
3 17
7 11
28
(a) 4
8 12
18
52 51 29 30 54 55
Key to pier-side names
53 56
57
(b)
Figure 5.15. (a) Key to names assigned to pier faces. (b) Locations of transducers on face
A of the pier and the positions of the load cells. The LP devices are resistive transducers,
whereas the LVDT are linear variable differential transformers or inductive transducers.
C D
Figure 5.16. Locations of transducers on reinforcing bars on sides C and D of the pier.
An Introduction to Structural Testing Techniques in Earthquake Engineering 125
Figure 5.17. Photo of pier ready for testing. Note the actuators at the top of the pier,
some transducers on the face of the pier, and the large amount of wiring that connects to
the computer.
Cyclic displacements were applied to the top of the pier to drift levels of 0.4%, 1.2%,
1.8% and 3.0%. Three cycles were performed at drift levels of 0.4% and 1.2%, with two
cycles performed at 1.8% and 3.0%. The global response, in terms of actuator force
(shear) versus tip displacement, was monitored as the test proceeded. After the
displacement cycles at 1.2% drift it was observed that the axial load was not being
maintained at 250kN as intended and had instead risen to over 300kN. This was because
the actuator responsible for controlling the axial load was not being operated correctly.
Consequently, the test was paused and the vertical actuator was released and re-set at
250kN. The actual cyclic time-history of imposed displacements is shown in Figure 5.18.
Included in this figure is the effective displacement of the pier with a rigid base obtained
by allowing for the footing rotations. Included in this figure are the displacements of the
pier modified with account for the footing rotations so that the values are equivalent to a
pier with a rigid base. Although there is little difference between the raw and adjusted
displacement values, this modification is made as it is desirable to consider the rigid
footing case for comparison with the analytical prediction.
126 Timothy J. Sullivan, Rui Pinho & Alberto Pavese
The fact that the test had to be paused and the axial load adjusted implies that the
experimental set-up was not carried out correctly. Ideally, the changing axial load would
have been noted within the first set of cyclic displacements to the 0.4% drift level. This
level is below the yield drift of the pier and is performed to enable checks of the system
set-up, with the aim of preventing exactly this type of mistake. The impact of the axial
load adjustment on the non-linear behaviour of the pier will be considered in the review
of the results.
50
40
Test paused and axial load
30
adjusted to be constant 250kN
20
Displacement (mm)
10
0
1
-10
-20
-30
Displacement History
-40 Disp w/o footing rotation
Disp due to footing rotation
-50
Time
Figure 5.18. Cyclic displacement history as applied to the pier, adjusted for foundation
rotation.
Apart from the adjustment of axial load, the test proceeded smoothly. It was found
that the pier began to lose significant shear capacity at 3.0% drift indicating that failure
had been reached. An unexpected crack pattern developed with vertical cracks on the
sides of the piers, as presented in the following results section.
essentially elastic close to the flexural prediction, but with very narrow hysteresis loops. It
is believed that these loops are associated with the cracking that was predicted at a drift
less than 0.1%.
400
ATC
300 Caltrans
UCSD-Orig.
200
Lateral Force (kN)
UCSD-Mod.
Flexural response
100 Experimental data
0
-100
-200
-300
-400
-80.0 -60.0 -40.0 -20.0 0.0 20.0 40.0 60.0 80.0
Displacement (mm)
Figure 5.19. Cyclic behaviour, plus flexural and shear capacity predictions for the bridge
pier
The yield displacement of the pier was calculated as 7.9mm, or 0.59% drift, as
reported in Section 5.4.3.1. However, the test results indicate that both the stiffness and
strength of the pier have been over estimated for the second stage of cycling to 1.2%
drift. This lower than expected initial stiffness has also been observed from several tests
on similar hollow RC piers [115] and it has been suggested that this is principally the
result of shear cracking deformations. Consequently, it is expected that at 1.2% drift the
full flexural capacity of the pier had still not been developed, even if some longitudinal
bars had begun yielding. Given these points, the lower than expected response of the pier
up to the various drift levels might be considered reasonable. However, it is apparent that
even at displacements of 40mm, corresponding to 3.0% drift, the predicted capacity of
the section overestimates the response. In fact, a drop of strength is noted for cycles
beyond the 1.2% drift level. Some causes for this might be found through examination of
the damage pattern that developed during the test.
At the initial stages of the test a predictable cracking pattern developed that consisted
principally of diagonal shear cracks on the sides and some horizontal flexural cracks on
128 Timothy J. Sullivan, Rui Pinho & Alberto Pavese
the front and back faces of the pier. This early stage cracking is shown in the photo, part
(a) of Figure 5.20. Note that such cracking supports the argument that the stiffness was
reduced because of shear deformations. At the later stages of testing an unexpected form
of cracking was observed. Part (b) of Figure 5.20 shows a large vertical crack that
developed on the side faces of the pier. The test was stopped upon formation of a failure
mechanism that could be described as a combined shear-splitting failure.
Connects to a
diagonal shear crack
(a) (b)
Figure 5.20. Cracking patterns on the side of the pier (a) diagonal shear cracks at low drift
levels, and (b) combined vertical splitting and diagonal shear cracks near collapse.
the presence of any bending is expected to have subjected some longitudinal bars to an
axial load capable of transferring to lateral movement and generation of circumferential
tensile stresses in the concrete. In contrast to a situation where the axial load is constant
and a longitudinal bar is subject to tension and compression as part of flexural cycling,
these circumferential stresses would have developed before the full transverse confining
pressure had been developed in the concrete. The small cracks around the longitudinal
bars would then have developed into what appeared as a splitting failure upon the
ensuing cycling. This is because the initial splitting cracks would have been subject to
shears parallel to the longitudinal bars (known as horizontal shear for a cantilever beam)
which eventually split the column longitudinally. The observed failure mechanism was
formed when the diagonal shear cracks (formed during the initial cycle phases to 1.2%
drift) connected to the splitting cracks.
Alternatively, it could be that local weaknesses in the concrete caused the unexpected
failure mechanism to develop. However, through comparison of the results with those
reported in [115] it is clear that the axial unloading and reloading of the pier midway
through the test did alter the response. Figure 5.21 (from [115]) presents the response of
an identical pier subject to similar drift levels but with the axial load maintained at a
constant 250kN throughout the test.
350
UCSD-Orig.
UCSD-Mod.
UCB-Orig.
250
UCB-Mod.
Flexural response
Experimental data
150
Lateral force (kN)
50
-50
-150
-250
-350
-10 -8 -6 -4 -2 0 2 4 6 8 10
Displacement ductility
Figure 5.21. Experimental tests results from [115] compared with numerical predictions.
130 Timothy J. Sullivan, Rui Pinho & Alberto Pavese
It is evident that the full predicted strength has been developed and maintained for
some cycles at displacement ductility demands greater than four (drift levels of 2.4%). A
combined flexure-shear failure eventually developed at displacement ductility demands
greater than five. Further support is given to the argument that the test procedure
affected the results, because the hysteretic shapes for the initial stages of the two tests are
very similar. Readers are again referred to [115] for more details and the performance of
other shear models in predicting the results of hollow reinforced concrete piers.
Through comparison of results from [115] for a similar test, it appears that the
behaviour of the test pier at low drift levels is reasonable. Interesting observations include
a lower than expected initial stiffness. It is proposed that this low stiffness is due to
formation of shear cracks causing reduced shear stiffness, noting that the pushover curve
incorporated the effects of flexural cracking. The comparison in results suggests that at
higher levels of drift the test pier has responded unusually.
A failure sequence has been proposed to account for the unexpected non-linear
response. The hypothesized response sequence stipulates that the response was altered by
axial unloading and reloading during the testing. This axial adjustment was necessary
because of a mistake made in the experimental set-up phase. Fortunately for the purposes
of this report, this event has highlighted the importance of the experimental set-up phase
in a test.
Consequently, despite being unable to fulfil the aims of this investigation, several
interesting observations have been made about hollow reinforced concrete piers and the
non-linear behaviour of reinforced concrete. More importantly in the context of this
report, the importance of the experimental set-up phase has been highlighted, reinforcing
comments made earlier within this document.