Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Young 322.00

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

6th International Conference on Earthquake Geotechnical Engineering

1-4 November 2015


Christchurch, New Zealand

Seismic Performance of Christchurch Wastewater Treatment Plant Pond


Bunds

R. A. Young 1, M. F. L. Gibson 2, D. V. Toan 3, G. J. Alexander 4

ABSTRACT

The Canterbury earthquake sequence (CES) in New Zealand caused significant damage to the
Christchurch City Council’s oxidation pond bunds at the Wastewater Treatment Plant. Bund
damage was predominantly associated with seismic shaking, liquefaction and lateral spreading
along the 14 km of embankments. This paper describes seismic assessment of the bunds before
and after each of the significant CES events in 4 September 2010, 22 February and 13 June 2011.
The seismic response at cross sections through the embankments was back analysed and calibrated
against field deformation measurements, using simple pseudo static limit equilibrium (LE)
methods. Back analysis is supported by detailed walkover reconnaissance and comprehensive cone
penetration testing (CPT) performed before and after the significant events. Factors of safety
against limited flow failure (the material solidifies after finite deformation) and Newmark sliding
block ground deformations, collectively referred to as lateral spreading, were evaluated to assess
their relative contribution to the observed ground movements.

Introduction
The Christchurch Wastewater Treatment Plant (CWTP) oxidation pond bunds are located in
Bromley, to the east of Christchurch’s CBD. The ponds are extensive, covering 208 hectares, and
are contained by 14km of soil bunds. As a result of the earthquakes on 4 September 2010, 22
February 2011 and 13 June 2011 the bunds suffered variable, and in places significant damage.
Christison et al (2012) provide a comprehensive discussion on the observed damage and inferred
failure mechanisms, the risks associated with potential future movement and the design of
remedial solutions. This paper discusses a back analysis approach that was adopted to assess the
seismic performance of the bunds and used to predict the future performance of mitigation
design options.

Cone Penetrometer Testing (CPT) carried out before and after each earthquake indicated that
there was a degradation in the cone resistance (q c ) of the soils, the extent of which varied and
could be closely correlated to the observed ground damage. It was concluded that the seismic
induced liquefaction and associated lateral movement could be modeled using simplified Limit

1
Technical Director – Geotechnical Engineering, Beca Ltd, CEng, CPEng, MICE, Christchurch, New Zealand,
richard.young@beca.com
2
Associate – Geotechnical Engineering, Beca Ltd, CPEng, MIPENZ, IntPE(NZ), Christchurch, New Zealand,
marcus.gibson@beca.com
3
Chief Geotechnical Engineer, Beca Ltd, CPEng, FIPENZ Auckland, New Zealand, dovan.toan@beca.com
4
Technical Fellow – Geotechnical Engineering, Beca Ltd, CPEng, FIPENZ, CEng, MICE, Auckland, New Zealand,
gavin.alexander@beca.com
Equilibrium (LE) methods. Firstly non-liquefied Newmark sliding block (kinematic) analyses
were adopted to define a zone of deformation and magnitude of displacement. Secondly the
potential for post-liquefaction static limited flow failure was assessed and, if predicted, the extent
and magnitude of lateral displacement recorded. Adding the displacements together (lateral
spreading) indicated good correlation with the observed extent of deformations and provided a
suitable method for predicting the bunds’ performance, particularly where moderate to severe
damage was observed.

The concurrency of liquefaction and ongoing ground accelerations are considered. We concluded
that, based on the short duration of shaking experienced at CWTP, adding the effects of this to
the calculated kinematic and post-liquefaction limited flow failure displacements would over-
predict the slope movement observed.

Site Description

The ponds are underlain by sand of fixed and semi-fixed dunes and beaches (Christchurch
Formation) which investigations had proved to in excess of 20m depth. These deposits are
characterised by fine to medium sand with minor silt horizons that can be laterally extensive but
do not form continuous layers beneath all of CWTP. The consistency of the sand varies both
vertically and horizontally; typically it is medium dense but does range between loose to dense.

The bunds were constructed by excavating the upper horizons of local sand and compacting them
into a generally dense state. The batters are typically formed at 2.5H:1V, although there is some
variation in the inclination of these side slopes. Crest widths vary between 5m and 40m with the
external faces of the bunds being up to 5m high and internal faces, up to 3m high. The
surrounding groundwater is high, typically within 1m of ground level, and the water level in the
ponds is typically 0.5m below the bund crests.

Damage to Oxidation Pond Bunds

Ground shaking induced extensive liquefaction and damage to the bunds, becoming
progressively worse after each event with high peak ground accelerations (PGA). Detailed
inspections of the bunds revealed that the extent and severity of damage varied between
“insignificant” or barely observable (L0) through to “very severe damage” i.e. several metres of
lateral spreading (L5) [New Zealand Geotechnical Society (NZGS, 2010)]. Concentrated damage
to the bunds after the September 2010 event was limited to the east, around ponds 1, 2 and 3.
Following the February 2011 event these movements were reactivated and additional local
damage to all the remaining bunds were noted. This pattern of damage was replicated following
the June 2011 event, with the level of damage slightly increased as discussed in Christison et al
(2012).

Geotechnical Investigation

Before 4 September 2010 and then between the subsequent two major events, CPT testing was
carried out at (broadly) coincident locations. The testing indicated that the degree of loosening of
the bund fill and underlying soils varied spatially and between the events. Prior to the
earthquakes the q c of the well compacted bund fill was typically 20MPa +5MPa, falling to
approximately 10MPa +5MPa following the earthquakes. CPT indications of loosening of the
bund fill (Figures 1 & 2) correlated well with the observed surface ground damage following the
various events, generally associated with damage exceeding category L3 (NZGS, 2010).

CWTP Ponds

Locality Plan Pond 1

Location CWTP Pond Pond 4


4 CPT – Figure 2
Pond 2A Pond 2B

Location CWTP Pond


2A CPT – Figure 2

Pond 3

Pond 5 L4
L3
L2
Pond 6 L1

Collapse over pipe


Location of section
N Figure 5 Ejecta sand boils

Figure 1. Plan of CWTP oxidation pond bunds with damage (13 June 2011)

Canterbury Earthquake Sequence

Details of the three earthquakes considered in this paper, including the distance from the
epicentres to the CWTP and the induced horizontal PGA, are given in Table 1. Pages Road Pump
Station 1 (PS1) seismograph is located 2.5km to the west of the CWTP on a comparable soil
profile and examination of the strong ground motion data indicates:

• Reduction in the horizontal (but not vertical) ground acceleration after a period of
shaking, refer Figure 3
• Irregularity in the horizontal acceleration waveform after a period of shaking
• Moderation in the magnitude of horizontal compared to vertical acceleration
• These phenomena are only noticeable for the two events having epicenters closest to
CWTP.
CWTP Pond 4 CPT CWTP Pond 2A CPT

Pre CES
Post 22 Feb 2011

Post 4 Sep 2010


Post 4 Sep 2010
Elevation (mRL)

Elevation (mRL)
Post 22 Feb 2011
Pre CES

Cone Resistance qc (MPa) Cone Resistance qc (MPa)

Figure 2. Reduction in q c observed between pre CES and post earthquake CPT

Table 1. Earthquake Characteristics at CWTP Oxidation Ponds

Earthquake Epicentre Magnitude Duration Horizontal Reduced Horizontal


Distance PGA 1 Acceleration
Duration PGA
4 Sep 2010 40km M w 7.1 ~50 s 0.18 g - -
22 Feb 2011 6km M w 6.2 13 s 0.46 g 3–4s ~0.1 g
13 Jun 2011 2km M w 6.0 10 s 0.26 g 4–5s ~0.1 g
1
Conditional Peak Ground Accelerations (PGA) for liquefaction assessment Bradley and Hughes (2012)

For the two closer events the reduction in horizontal ground acceleration typically occurred after
four to six cycles of significant ground shaking. The reduced PGA occurred over approximately
three cycles of shaking for a total period of approximately 5 to 10 seconds i.e. at a much reduced
frequency. After approximately 10 seconds of shaking the horizontal ground accelerations had
reduced to less than 0.05g.
Post-liquefaction ground motion
Acceleration (g)

Pre-liquefaction ground motion

Time (s)

Figure 3. Horizontal ground acceleration - PS1 N-S [PRPC] (22 Feb 2011)
Raw data sourced from www.geonet.org.nz

Ground Deformation

The damage to oxidation pond bunds was summarised as follows (Christison et al, 2012):
• All bunds deformed, with lateral displacements of 10mm up to approximately 1500mm
• In areas of L3 and L4 damage lateral spreading was approximately 300-700mm per event
• Settlement of bund crests was 50mm to 200mm and in isolated locations up to 500mm
• Distribution of damage was variable with change from L3 to L1 damage often occurring
over short distances (<10m)
• CPT testing indicated loosening of bund fill to typical depth of 3m to 4m
• Both circular and planar failure mechanisms were observed - varied with geometry
• Seepage through bunds increased after each significant event
• Catastrophic instability or piping failure of bunds did not occur
• Highest and steepest embankments exhibited greatest deformations

Modeling of Ground Deformation by Pseudo Static Limit Equilibrium (LE) Methods

Pseudo static LE methods of analysis were adopted as a simple, efficient means to develop an
understanding of how the bunds performed. Observed surface expressions of ground deformation
suggested that the movement could be modeled as classical circular (typically followed by toe
heave) or planar type slip surfaces (with no toe heave); these ground displacements being
approximated by a combination of Newmark sliding block and limited lateral flow
displacements. The failure mechanisms varied from location to location depending on the ground
conditions and geometry.

Back Analysis Methodology

The back analysis methodology is given in Figure 4 with 2D plane strain models developed at
intervals along the bunds where there were variations in stratigraphy and geometry. The
locations were selected based on the availability of pre-earthquake investigation data and
variations in ground deformation, which enabled failure mechanisms and factors affecting the
severity of damage to be investigated. Free faces within 200m of the site, typically with a height
ratio of less than 20% were modeled using Geo-Slope SLOPE/W software. Seismic loading was
taken as the peak horizontal ground acceleration for the three events recorded at the PS1
seismograph. Zones of liquefaction together with residual normalized undrained shear strength
ratios (τ/σ vo ’) were estimated in accordance with the methods of Idriss and Boulanger (2008)
and Olsen and Stark (2002) – both gave similar results. Special attention was paid to areas where
there were abrupt changes in the performance of the bunds over short distances.

Back Analysis Discussion


As each of the three events had different, but defined characteristics (e.g. extent of liquefaction,
liquefied strengths & PGA) it was possible to obtain consistency between the back analysis
parameters and the variable level of damage observed. Where this correlation was not good
sensitivity assessments were undertake to investigate the effects of soil parameters, liquefied
condition, geometry and loading on the model. The safety map feature within Slope/W allows all
slip surfaces with a FoS less than 1 to be drawn, which was used to model the extent of
deformation. A detailed review of the outputs is required, applying an appropriate level of
judgment to assess the reasonableness of the results.

Kinematic Deformation
Kinematic (Newmark sliding block type) displacements, calculated for the bunds based on the
back analysed critical accelerations and integration of the acceleration records at the PS1,
indicated displacements of <20mm for the 4 September 2010 earthquake and 50 – 150mm for the
22 February 2011 event. Empirical estimates such as Jibson (2007) under predicted
displacements compared to the analysed and observed displacements, with empirical estimates of
<10mm displacement for 95% confidence. The kinematic deformation (non-liquefied)
corresponds with a L0 to L2 damage severity observed at the ponds. The extent of deformation is
consistent with that observed at the ground surface.
Limited Flow Failure
Comparisons between the back analysis models and field observations identified that the major
cause of bund damage was limited flow failure corresponding to the L3 and L4 damage
observed. Abrupt increases in damage could be modeled by factors of safety for the liquefied
static case falling just below unity, as a result of minor changes in ground conditions,
liquefaction extent and geometry. The extent of limited flow failure deformation was, on
occasion, considered to be too deep. The CPTs where limited flow failure was observed
indicated that the reduction in q c of the soil beneath the centre of the embankment was apparent
to a depth of up to 3m below the external toe level of the bund (Christison et al, 2012).
Elsewhere the extent of limited flow failure indicated by the back analysis was consistent with
the field observations (ref Figs 1 & 5).
Figure 4. Pseudo Static LE Back Analysis Methodology Assessment
Location of head scarp and
instability for Pond 6 L3 damage Loosening of soil profile observed from pre
and post 22 Feb 2011 earthquake CPT

Flow failure Critical Slip Surface

Pond 6 Extent of slip surfaces with FoS<1

Typical Material Parameters


g =20kN/m3, f’=32o, c’ =0kPa
g =17kN/m3, f’=30o, c’ =3kPa
g =18kN/m3, t/s’=0.1
g =18kN/m3, t/s’=0.25
g =19kN/m3, t/s’=0.14
CPT Qc (MPa)
Liquefied Ground conditions – 22 Feb 2011 10 20 30

Figure 5. Example bund cross section showing limited flow failure instability [Pond 6]

Concurrency of PGA and Liquefied Shear Strength

For the closer and lower magnitude February and June 2011 events the acceleration waveform at
the PRPC seismograph (Figure 3) suggests that the horizontal PGA is reached after four seconds
or so, with the horizontal acceleration then dropping to approximately 0.1g before reducing to
less than 0.05g after 10 seconds or so. This drop in horizontal acceleration could be due to the
onset of liquefaction, with the loss of shear transfer and the much higher damping of high
frequency waves, and the liquefied ground isolating the seismograph from the source
accelerations. The modeling undertaken suggests that this may be the case. When the model was
run with liquefied shear strengths occurring concurrently with ground accelerations of 0.05 –
0.1g extensive and the predicted deep seated slip surfaces did not match field observations. One
explanation could be the delayed onset of liquefaction and limited flow failure due to the silty
nature of the fine sands. CPTs may support this phenomenon as indicated by the reduction in q c
from a high ‘non-liquefiable’ value to a lower ‘liquefiable’ value after the February 2011 event.
This conclusion may not apply when larger magnitude, longer duration events occur.

Conclusions

The extent of damage at CWTP modeled by kinematic and limited flow failure back analyses
correlates well with field observations of damage from multiple earthquakes and provided a
simple and efficient preliminary method to identify the mechanism and vertical and lateral extent
of ground deformation. The back analyses demonstrated a variation in the magnitude of
predicted displacement compared to observed damage, the interpretation of which required good
judgment. Marked changes in damage over short distances could be attributed to a transition
from Newmark sliding block deformation to limited flow failure. The sensitivity of stability and
triggering of flow failure due to minor variations in geometry and material strength parameters
was especially important when the liquefied static FoS was close to 1. At the oxidation ponds
there was a significant increase in damage when limited flow failure was triggered.

Acknowledgements
Christchurch City Council for supporting the sharing of details contained in this paper.

References

Ambraseys N. and Srbulov M., Earthquake induced displacements of slopes, Soil Dynamics and
Earthquake Engineering, 1995; 14: 59-71.
Bradley B., Hughes M., Conditional Peak Ground Accelerations in the Canterbury Earthquakes for
Conventional Liquefaction Assessment, Technical Report for the Ministry of Business, Innovation and
Employment, April 2012.
Brown L. J., Weeber J, H, Geology of the Christchurch Urban Area, Scale 1:25,000, IGNS Ltd, 1992.
Christison M., Young R., Gibson M., Seismic performance of Christchurch wastewater treatment plant
oxidation pond bunds, Ingenium 25th Annual Conference, 21-23 June 2012.
Idriss I. M., and Boulanger R. W., Soil liquefaction during earthquakes. Monograph MNO-12, Earthquake
Engineering Research Institute, Oakland, CA, 2008.
Jibson R.W., Regression models for estimating coseismic landslide displacement, Engineering Geology
2007; 91: 209-218.
New Zealand Geotechnical Society Inc (NZGS), Geotechnical earthquake engineering practice: Module 1 –
Guideline for the identification, assessment and mitigation of liquefaction hazards, July 2010.
Olson S. M., and Stark T. D. 2002. Liquefied strength ratio from liquefaction flow failure case histories.
Can. Geotech. J., 2002; 39: 629–647.
Youd T.L., Hansen C.M. and Bartlett S.F., Revised multilinear regression equations for prediction of lateral
spread displacement, J. Geotech and Geoenvironmental Eng., 2002; 128 (12): 1007-1017.

You might also like