8/11/2019 geng1300047h.pdf
1/13
Proceedings of the Institution of Civil Engineers
http://dx.doi.org/10.1680/geng.13.00047
Paper 1300047
Received 19/04/2013 Accepted 08/11/2013
Keywords: excavation/field monitoring & testing/land surveying
ICE Publishing: All rights reserved
Geotechnical Engineering
Deep excavations: monitoring mechanisms
of ground displacement
Fearnhead, Maniscalco, Standing and Wan
Deep excavations: monitoringmechanisms of grounddisplacementNick Fearnhead MScAssistant Geotechnical Engineer, Atkins, Epsom, Surrey, UK; formerlyImperial College London, UK
Kleaven Maniscalco MScGeotechnical Engineer, Rockcut Limited, Malta; formerly Imperial CollegeLondon, UK
Jamie R. Standing PhDSenior Lecturer, Imperial College London, UK
Michael S. P. Wan MScGeotechnical Engineer, Crossrail Limited, UK; also postgraduateresearcher; Imperial College London, UK
As construction methods become more advanced and design more refined, better control and smaller magnitudes of
ground displacements resulting from deep excavation can be achieved. Mechanisms of ground movements expected
from deep excavation and methods for their prediction are briefly reviewed. Such predictions, whether empirical or
numerical, are critical when the observational method is implemented during the construction process. Their accuracy
can only be verified by correspondingly accurate monitoring of actual vertical and horizontal surface and subsurface
ground displacements throughout construction. It is therefore relevant to investigate whether instruments routinely
used to determine such displacements are sufficiently accurate. Monitoring techniques involving precise levelling,
rod extensometers and torpedo inclinometers were examined in detail over a 3-week investigation period at a
greenfield site to quantify the short-term precision and accuracy of the measured displacements obtained from them.
Where possible, the influence of external factors (e.g. temperature and weather) on measurements is quantified and
recommendations for best practice are made. Providing that the monitoring techniques discussed are performed to a
high standard, the data indicate that ground displacements around deep excavations can be monitored to a
sufficient accuracy for calibration and validation of numerical analyses and ultimately safe engineering design and
construction.
1. IntroductionDeep excavations within the urban environment are frequently
necessary as part of infrastructure projects or for the construction
of buildings with deep basements. Ground removal usually
necessitates some form of support, the extent of which depends
on various factors such as soil and groundwater conditions andthe excavation/construction method. The combined response be-
tween the ground and the supporting structure constitutes a
complex soil structure interaction boundary value problem. As
with most geotechnical engineering structures, there is a much
greater degree of confidence in assessing resulting forces, stresses
and stability compared with ground and structural displacements.
In assessing the latter, frequent recourse is made to empirical
approaches or numerical analyses. Often the observational meth-
od is adopted (Nicholson et al., 1999; Peck, 1969) where
predicted responses (primarily displacements) are checked by
field monitoring during the works and contingency measures
implemented as and when necessary. Additionally numerical
analyses are often refined and updated taking account of meas-
ured ground and structural responses to make improved predic-
tions of subsequent phases of work.
As shown in Figure 1, the magnitude of strains relating to
retaining walls for deep excavations is at the small strain end of
the scale. Improved methods of design and construction have
resulted in even smaller observed displacements. However,
controlling displacements is particularly important in the urban
environment, where often structures adjacent to an excavation
require safeguarding against damage.
Region 1Very small strain
Region 2Small strain
Region 3Large strain
Retaining walls
Foundations
Tunnels
00001 0001 001 01 0 10
Shear strain, : %s
Stiffne
ss,
G
G0
Figure 1.Ranges of strains pertaining to different construction
activities (afterAtkinson and Salfors, 1990)
1
8/11/2019 geng1300047h.pdf
2/13
In this paper, first the mechanisms of ground response to deepexcavations are briefly reviewed along with typical ranges of
displacements that might be expected. The results from a series
of measurements performed as part of field research being run in
conjunction with the Crossrail project are then presented. The
measurements discussed formed part of the base readings and the
specific intention was to investigate potential sources of error and
find ways of eliminating/minimising them.
2. Typical mechanisms of ground movementaround deep excavations
Typical profiles of vertical and horizontal ground and wall
displacements (
v and
h) for a deep excavation are shown inFigure 2. The form and magnitude of the displacements depend
on many factors, including the dimensions, as well as the
geometry of the excavation, the ground and groundwater condi-
tions, the methods, sequence and duration of construction, typeand stiffness of the wall, and the quality and control of workman-
ship (Gaba et al., 2003). Many of the above factors, apart from
ground conditions, are within human control and can be carefully
selected and varied according to budget, timescale and factor of
safety required.
The two types of settlement profile outside the excavation, shown
in Figure 2, depend on the magnitude and shape of the wall
deflection, which in turn depends on many factors including wall
type, system stiffness and construction sequence (Ou, 1993). Maxi-
mum surface settlement, vm, occurs very close to the wall in the
case of the spandrel profile, associated with large wall deflections
in the initial phases of excavation and the wall deforming as a
cantilever. This is more likely to occur with bottom-up construction
without stiff propping. A concave profile, with maximum settle-
ments at some distance from the wall, is more likely to develop
when stiff props are installed soon after excavation starts and with
top-down construction. Combinations of both cantilever and sub-
sequent deep-seated wall deflections can lead to surface settlements
with features of both spandrel and concave settlement profiles
(Hsieh and Ou, 1998).Roscoe and Twine (2010) usefully compare
observed modes of wall deflection for top-down and bottom-up
construction. They found that for top-down construction, maximum
horizontal wall displacements, hm 0.15%H(andhm , 0.2%H)
where H is the retained height, while for similar bottom-upexcavations, hm , 0.4%H.
3. Quantifying expected ground movementsThere are two primary approaches to estimating ground and
retaining system displacements resulting from deep excavations.
Historically empirical approaches were developed from numerous
case histories (e.g. Peck, 1969), providing correlations between
horizontal and vertical displacements and excavation depth, for
various factors, especially soil and wall types. This approach is
still often used and more refined correlations have subsequently
been developed (e.g. Clough and ORourke, 1990). Numerical
analysis is the approach most commonly used in modern times,especially with the availability of much greater computing power,
modelling capabilities and refined soil parameters from advanced
laboratory testing.
Both approaches are frequently used with success and both have
drawbacks. The empirical approach relies on the accuracy of the
instruments with which case study ground or wall movements
were measured, and caution is also required in checking that the
ground conditions relevant to the project under consideration
compare closely with the case studies. Checks should be made
continuously during the works using appropriate field monitoring.
Similarly, as the accuracy of numerical analyses is strongly
dependent on the quality of the analysis and the input data
(particularly soil parameters) it is essential to verify the analysis
from the earliest stages of construction. As with empirical design,
in situ monitoring during construction is the primary way to
verify such methods of analysis, thus highlighting the need for
vm
hm
Wall
Temporaryprops
Stiff, permanentbase slab
(a)
(b)
WallStiff, permanentroof slab
vm
hm
Temporarypropping asexcavationprogresses
Castingof stiff,permanentbase slab
Figure 2.Typical wall and ground displacements associated with
deep excavations (afterHsieh and Ou, 1998): (a) concave
settlement profile (typical of top-down construction); (b) spandrel
settlement profile (typical of bottom-up construction)
2
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
3/13
highly accurate monitoring methods. This is particularly the casewhen the observational approach is being implemented.
In order to assess the applicability of various monitoring meth-
ods, their accuracy and instrument resolution need to be assessed
in relation to the expected magnitudes of ground movement
around a deep excavation. An impression of the range of
displacement magnitudes to be expected can be gained from
Clough and ORourke (1990) or Gaba et al. (2003) who provide
comprehensive guidelines categorised in terms of different wall
and soil types. Clough and ORourke concluded that average h
and v values are about 0.2%H and 0.15%H respectively (the
latter is significantly lower than the 1%H estimated by Pecks
method). They also produced a set of dimensionless settlement
profile design charts categorised according to soil type, derived
from a large collection of case study data encompassing different
wall types (Figure 3).
These dimensionless profiles have been used to estimate surface
settlements behind a wall based on maximum settlements, deflec-
tion ratios and the soil type classifications given in three well-
documented case studies: the underground car park at the Palace
of Westminster (Burland and Hancock, 1977); Victoria Embank-
ment, London (St John et al., 1993) and the Lion Yard, Cam-
bridge (Ng et al., 2004). Table 1 shows a summary of the
measured movements and calculated deflection ratios from thecase studies. In all cases the deflection ratio is less than the upper
limit of 0.5%Hpredicted byClough and ORourke (1990), falling
very close to the average values h , 0.2%H and v , 0.15%H.
Therefore, these values have been used to calculate expected
horizontal wall deflection and vertical settlement for various
excavation depths, as given in Table 2.
Table 2 indicates that total displacements of the magnitude of
tens of millimetres will typically develop by the end of excava-
tion. Such movements can be readily measured with a variety of
instrumentation; however, it is the displacements that develop
incrementally over the duration of the works that are required,especially for the validation and updating of numerical analyses.
Realistically, measurements need to be made to sub-millimetre
accuracy.
4. Investigation into the accuracy ofmonitoring methods
Having outlined the main mechanisms of ground movement
around deep excavations and the ways in which the form andmagnitude of displacements can be predicted, suitable methods to
measure them are now discussed. The emphasis is on conven-
tional monitoring techniques, requiring manual readings. Baseline
10
05
00 05 10 15 20
d H/
v
vm
/
(a)
d
H
vm
vm
v
v
10
05
00 05 10 15 20 25 30
d H/
v
vm
/
(b)
Settlementenvelope
Settlementenvelope
Settlement
envelope
H
d
10
05
00 05 10 15 20
d H/
v
vm
/
(c)
075
Figure 3.Dimensionless settlement profiles adjacent to deep
excavations (Clough and ORourke, 1990): (a) sands; (b) stiff to
very hard clays; (c) soft to medium clays
Project Excavation
depth,H: m
hm: mm vm: mm Lateral extent of
movements: m
hm/H: % vm/H: %
Victoria Embankment
(St John et al., 1993)
19 30 25 50 0.158 0.132
Palace of Westminster
(Burland and Hancock, 1977)
18.5 29 20 80 0.157 0.108
Lion Yard (Ng et al., 2004) 10 22 10 55 0.220 0.100
Table 1.Summary of deflection ratios from case studies
considered
3
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
4/13
measurements at a greenfield research monitoring site in Hyde
Park, London (Wan and Standing, 2014) were used to assess the
accuracy of precise levelling, rod extensometers and torpedo
inclinometers (for surface and subsurface vertical displacements
and subsurface horizontal displacements respectively).
The three methods of measurement are commonly used inpractice and were chosen as being the most prone to inaccuracy
due to the inherent necessity for human involvement and
exposure to external factors such as weather. Other instrumenta-
tion used on the research site, that might equally be used for
monitoring ground and wall responses around a deep excavation
under construction, are total stations (especially robotic devices),
vibrating wire piezometers and spade pressure cells and in-place
inclinometers. These are not considered here as they are judged
to be relatively unaffected by human or environmental errors.
At and close by the Hyde Park research site three arrays of
surface monitoring points (SMPs) for precise levelling, 15 rod
extensometer boreholes of varying depths up to 50 m and
torpedo inclinometer casings to a depth of 40 m were installed,
as shown in Figure 4. Various investigations into the factors
affecting accuracy for the three monitoring methods were made
using these points and boreholes. The findings from the
practical field assessment of each method are now described
and discussed.
H: m hm/H: % vm/H: % hm: mm vm: mm
5 0.2 0.15 10 7.5
10 0.2 0.15 20 15
15 0.2 0.15 30 22.5
20 0.2 0.15 40 30
25 0.2 0.15 50 37.5
30 0.2 0.15 60 45
35 0.2 0.15 70 52.5
40 0.2 0.15 80 60
Table 2.Estimated vertical and horizontal displacements for
various excavation depths
Wall benchmark
Surface monitoring pointRod extensometer
Inclinometer
N
25 m 0 25 m 50 m
CrossraileastboundtunnelCrossrailwestboundtunnel
XSMP30
XSMP25
XSMP23
XSMP10
YSMP1
YSMP12
YSMP21YSMP25
YSMP20
ZSMP17ZSMP9
ZSMP5
ZSMP1
HP10
YSMP24
YSMP13
YSMP14
YSMP12A
YSMP14A
XSMP1
XSMP11XSMP10A
NorthCarria
geDrive
HP21RX
HP18RX
BayswaterRoad
Clarendon
P
lace
Figure 4.Plan showing layout of instrumentation for research
monitoring at Hyde Park
4
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
5/13
Resolution, precision and accuracy are three terms frequentlyused when discussing monitoring data (along with various forms
of error). Straightforward definitions provided by Dunnicliff
(1988) are given in abbreviated form here: resolution is the
smallest division that can be read on a device; precision is the
closeness of a reading to the arithmetic mean of a number of
readings; and accuracy is the closeness of a reading to the true
value.
5. Precise levellingTwo of the three lines of SMPs were installed within the confines
of the Hyde Park greenfield site, parallel to one another and
200 m apart (XSMP and YSMP lines with 30 and 31 SMPs
respectively) while the ZSMP line (with 17 SMPs) ran along the
pavement of Bayswater Road and so was influenced by traffic
vibration. Details concerning the form of the XSMP and YSMP
points are given inFigure 5(note that ZSMP points were survey
nails). An important feature, as discussed later, is the use of the
carefully designed BRE socket (Building Research Establishment
Digest 386 provides more details (BRE, 1993)). A benchmark
installed in a wall 30 m outside the greenfield site was used as a
short-term reference point and its stability was checked periodi-
cally using a nearby 80 m deep datum. The precise levelling
surveys were performed using a Leica DNA03 digital level, witha resolution of 0.01 mm, used with an invar bar-coded staff.
Two-peg tests were performed to check for collimation errors
within the precise level and corrections were made if necessary.
Foresight and backsight distances were kept similar and limited
to < 30 m to improve accuracy. The Leica DNA03 was pro-
grammed to take three repeat readings in quick succession for
each measurement, displaying the range and standard deviation.
Measurements were only accepted if the range of these three
readings was
8/11/2019 geng1300047h.pdf
6/13
behaviour as observed for the XSMP line is evident, but results
are evenly spread over a much larger range.
The maximum variation of any one YSMP over the 3-week
investigation period was 1.70 mm, with the average variation
being 1.31 mm: significantly greater than the 0.26 mm average
for the XSMP line. It can be seen from examination of the
profiles that a significant proportion of the total variation between
surveys occurs at all SMPs including the first point, YSMP29.
This implies that the majority of the error associated with the
YSMP survey was generated in traversing 150 m from the wall
benchmark, requiring two change points, to reach the first point
YSMP29 (note that for the XSMP survey, a backsight was made
directly to the benchmark without change points). In order to
assess this possibility, the reduced levels of all the points were
related to a deep anchor at 50 m depth within rod extensometer
borehole HP21RX, whose reference head was also surveyed with
the YSMPs (Figure 4). Adjusting the YSMP reduced levels
relative to the 50 m deep anchor, as opposed to the wall
benchmark 150 m away, resulted in a vast improvement, as is
evident fromFigure 9. The maximum variation of any one YSMP
is 0.88 mm and the average variation 0.44 mm, agreeing closely
with corresponding values of 0.87 mm and 0.26 mm for the
XSMP survey (for which the same wall benchmark was used). In
contrast to the heave observed using the original data, the
adjusted data in Figure 9 indicate very little change in level
between May and July. This illustrates the potential for misinter-pretation of data if the magnitude of the precise levelling error is
similar to that of the ground displacements: careful selection of
an alternative benchmark reduced errors by more than 50%. In
the case of deep excavation monitoring, a deep datum could have
the same benefit and also be used for the assessment of base
heave.
5.2.3 ZSMP survey loop closing error and effect of
traffic
Closing error is often used as a measure of errors accumulated
over backsight and foresight readings taken during a survey.
However, this is not a definitive measure of the accuracy of the
survey, as positive and negative incremental errors of any size
can cancel each other out to give a closing error close to zero.
The ZSMP surveys had the smallest average closing error of all
three loops, while having the largest range of readings, with a
maximum range of reduced level being 1.07 mm and the average
12 October 2011
11 November 2011
25 January 2012
14 March 2012
2 May 2012
July (2012)average
4
2
0
2
4
6
8
10
30 29 28 27 26 25 23 22 21 20 19 18 17 16 15 14 13 12 11 10A 10 9 8 7 6 5 4 3 2 1
Changeinreducedlevelrelativetosurveyof12October2011:mm
Hyde ParkXSMP number
Hyde ParkNorth CarriageDrive
(Trees schematic only)
Figure 6.June/July data compared to long-term variation in
XSMP reduced level
6
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
7/13
range of readings 0.51 mm. Despite the small closing errors, large
errors on individual points could occur, as were detected when
returning to the start point of the survey when control checks
were made on a number of points; for example, for one survey, a
control check on one of the ZSMPs indicated a difference from
the initial reading of 1.42 mm, despite the final closing error
being only 0.45 mm. In this case the closing error could be
misrepresentative of potential errors.
The inaccuracies associated with the ZSMP survey were due to
two main factors. First, their location on the kerb of a busy main
road resulted in considerable vibration and also the close
proximity of passing vehicles caused buffeting, making it difficult
for the staff holder to maintain a steady, level position. Second,
installing BRE sockets was not possible in the kerbside and so
flat-topped nails, which barely protruded above the kerb surface,
were used, making it difficult for the staff to be adequately
positioned with its centre over the nail.
5.3 Repeatability of screwing together BRE plugs and
sockets
The BRE points used for the XSMP and YSMP lines consist of a
socket grouted in place and a removable plug as shown in Figure
5. In order to check the repeatability of screwing the BRE plug
into its socket a number of repeat readings were taken in quick
succession, removing and replacing the BRE plug each time
without moving the tripod. Repeat readings were always within
0.05 mm of the original reading and in most cases only differing
by 0.02 mm, confirming that the BRE plug can be relocated into
the socket with high precision. Similar repeat readings were taken
on the socket used for the wall benchmark, which was not of
BRE design (there are several alternatives commercially avail-
able), for which the readings were within a 0.17 mm range. The
BRE design involves precision machined mating faces and spigot
that cause the plug and socket to be drawn together co-centrically
(other advantages are given by BRE (1993) and Standing et al.
(2001)). The magnitude of the error caused by relocating the
survey sockets was deemed insignificant compared with the
survey errors discussed earlier.
5.4 Effect of distance/range on the accuracy of staff
readingThe Leica DNA03 has a quoted operating range of 1 .8110 m,
although for high-accuracy readings it is good practice to keep
the staff distance between 2 m and 20 m, with a limit of 30 m. In
order to assess the effect of staff distance on readings, two sets of
0
1
2
3
4
5
30 29 28 27 26 25 23 22 21 20 19 18 17 16 15 14 13 12 11 10A 10 9 8 7 6 5 4 3 2 1
Changeinreducedlevelrelativetosurveyof2May2012:mm
2 May 2012
18 June 2012
19 June 2012
20 June 2012
22 June 2012
22 June 2012(set 2)
5 July 2012
Hyde ParkNorth CarriageDrive
(Trees schematic only)
XSMP number
Figure 7.XSMP data collected during investigation relative to 2
May 2012
7
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
8/13
measurements were made in which the staff was held on a change
plate at varying distances from the tripod station, with five repeat
readings being taken each time. Tests were performed both on
site and indoors to investigate environmental effects. A strong
correlation between increasing staff distance and range of read-
ings is evident from the data inFigure 10. The effect was reducedwhen external factors such as wind and traffic were removed in
the case of the indoor tests. The results indicate that if the staff is
more than about 20 m away from the precise level, an error of
about 0.2 mm could occur for outdoor conditions (and this could
become much worse with adverse weather).
5.5 Errors induced by tripod location changes
It is common practice when precise levelling to keep the number
of tripod location changes to a minimum. It is thought that the
error induced in moving the tripod, setting up at a new location
and relevelling the device is greater than taking a longer-range
measurement to avoid this change. To quantify this potential
error, two loops (one on grass, one on concrete) involving eight
tripod location changes (total distances of 363 m for grass and
437 m for concrete with overall cumulative foresight/backsight
distances between 15 and 25 m) were performed in which the
reduced level of a single SMP was repeatedly measured five
times from each location. Assuming the SMP remains stable
implies that any variation in reduced level during each loop was
due to relocation of the tripod. The results shown in Figure 11
indicate a total error of 0.91 mm accumulated over the eight
tripod moves, for the concrete positions, compared with 0.28 mm
on grass. In both cases the first five tripod location changesresulted in ,0.2 mm variation in reduced level. This result
supports the idea that tripod moves should be kept to a minimum
when surveying.
5.6 Effect of thermal expansion
Thermal expansion of the staff, tripod, SMP or near-surface
ground can cause problems in the short term if temperatures
change noticeably during a survey. Quantifying such errors
often can be difficult, but on one occasion a dramatic change
in weather conditions from overcast and rain to direct sunlight
resulted in a noticeable rise in temperature while the tripod
remained at one location. Three SMPs were remeasured after
being in direct sunlight for about 20 min and the readings were
compared with those taken in overcast conditions, showing
increases in reduced level of 0.16 mm, 0.23 mm and 0.22 mm.
The similar increase in reduced level of each of the three
points possibly indicates the thermal expansion of the tripod
10
05
0
05
10
15
20
25
29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 14A 13 12A 12 11 10 9 8 7 6 5 4 3 2 1
Changeinreducedlevelrelativetosurveyof2May2012:mm
2 May 2012
19 June 2012
20 12June 20
04 12July 20
06 12July 20
11 July 2012
Hyde Park North CarriageDrive
(Trees schematic only)
YSMP number
Figure 8.YSMP data collected during the investigation relative to
2 May 2012, using a far-off benchmark
8
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
9/13
relative to the staff and SMPs. Although this does not offer a
quantitative evaluation, the result implies that the effect of
temperature change can be of similar magnitude to other
errors.
5.7 Accuracy of the levelling bubbles on the staff and
precise level
Prior to readings being taken, levelling bubbles on both the
precise level instrument and the staff should be positioned within
10
05
0
05
10
15
20
25
29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 14A 13 12A 12 11 10 9 8 7 6 5 4 3 2 1
Changeinreducedlevelrelativetosurveyof2May2012:mm
2 May 2012
19 12June 20
20 June 2012
4 12July 20
6 July 2012
11 July 2012
Hyde Park North CarriageDrive
(Trees schematic only)
YSMP number
Figure 9.YSMP reduced levels adjusted relative to the 50 m deep
datum (HP21RX) within the park
0
005
010
015
020
025
030
035
0 5 10 15 20 25 30 35 40
Rangeofreadingstaken:mm
Distance of staff from station: m
Laboratory
Hyde Park
Figure 10.Effect of distance of staff from station on repeatability
of the reading
04
02
0
02
04
06
08
10
0 1 2 3 4 5 6 7 8
ChangeinreducedlevelofX
SMP29relativeto
firstreading:m
m
GrassConcrete
Number of tripod moves
Figure 11.Effect of number of tripod moves on accuracy of
precise levelling
9
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
10/13
the prescribed circle so that they are at the correct orientation.When precise levelling, both of these elements are susceptible to
human judgement and therefore error. In order to quantify
potential associated errors, tests were performed indoors with the
staff securely clamped and repeat measurements were taken to
establish the range of readings with the staff/instrument levelling
bubble still within the prescribed circle (which would be deemed
acceptable in the field). The maximum error that could be caused
by the staff or instrument bubble not being centred was found to
be 0.1 mm and 0.07 mm respectively (over a sight distance of
,6 m). This is a reassuring result given the difficulty in
maintaining staff verticality in windy conditions.
5.8 Effect of using different operators
Using the same instrument operator and staff holder at all times
might be considered good practice when precise levelling to
avoid personal idiosyncrasies causing overall errors. To test this,
four loops involving three tripod moves and measurement of the
reduced level of seven SMPs were performed using different
combinations of staff holder and instrument operator each time.
Closing errors varied from 0.40 mm to 0.83 mm and the maxi-
mum range in reduced level measurement of any one SMP was
0.43 mm. This error is similar in magnitude to that of other
surveys completed, suggesting that changing operator and staff
holder does not have a detrimental effect on the accuracy of
precise levelling providing good practice is followed.
5.9 Summary of potential errors incurred with precise
levelling
The results from the various investigations discussed above are
summarised in Table 3. Many of the errors considered are
influenced by environmental or human factors, which are difficult
to quantify: although the results are not absolute, they offer a
guide as to the most influential factors causing errors. It is
concluded that measurement of SMP reduced levels can beperformed with sub-millimetre accuracy, typically about
0.3 mm. This corroborates precise levelling accuracies given by
Standing et al. (2001), who also provide practical tips for good
practice. Of the many factors investigated in this study, the effects
of traffic, levelling socket design and number of tripod moves
affected the overall accuracy the most. These factors can be
minimised by careful planning of the survey layout to minimise
the number of tripod moves. Using a nearby reference point, for
example a deep datum, would facilitate this and offers other
advantages.
6. Rod extensometersRod extensometer anchors were installed at 11 different depths
down to 50 m below ground level in the boreholes at the research
site (Figure 4). A dial gauge was used to measure changes
between the reference head and the top of rods connected to the
anchors, as shown inFigure 12. Repeat dial gauge readings were
taken for each rod until three consecutive readings within a
0.05 mm range were obtained. Although the resolution of the dial
gauge readings was 0.01 mm and the quoted accuracy 0.01 mm,
this only relates to relative displacements between anchors within
one borehole. The overall rod extensometer monitoring system
accuracy is estimated to be 0.3 mm as it is controlled by the
inherent accuracy of precise levelling of the reference heads,
which is necessary to obtain absolute reduced levels of theanchors (in conjunction with the dial gauge readings). Five sets of
rod extensometer readings were taken over the 3-week investiga-
tion period to provide an indication of the short-term accuracy of
rod extensometer measurements and influencing factors.
Initially results were processed relative to the wall benchmark
referred to in Section 5. As expected, following the discussion
above, errors occurred as a consequence, with an unlikely
Factor Range of readings: mm
Maximum range Average range
Reduced level of a single SMP within: XSMP loop 0.87 0.26
YSMP loop 0.88 0.44
ZSMP loop 1.07 0.51
Moving between Hyde Park and wall benchmark 1.02 0.50
Wall benchmark movement relative to deep datum 1.57 ,1.57
Replacement of BRE SMP plug into socket 0.05 0.02
Replacement of wall plug into socket 0.17 0.12
Measurement over a distance of about 20 m 0.20 0.10
Moving of tripod location 0.91 ,0.20
Thermal expansion 0.23 ,0.20
Staff bubble not centred 0.10 ,0.10Instrument bubble not centred 0.07 ,0.07
Table 3.Summary of factors affecting accuracy of precise
levelling
10
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
11/13
apparent average range of movements of 12 mm at the deep
anchor positions (3540 m), at which depths the ground should
be unaffected by seasonal variations. The data were then
processed relative to the deep datum at 50 m depth in the same
way as for the precise levelling data. The maximum range in
reduced level recorded for any one anchor was within 1.02 mm
with the average range of values recorded being within only
0.45 mm. In order to assess the precision of the dial gauge
readings alone, five sets of readings were taken over a short
period when reference head movements should have been negli-
gible. The maximum range of dial gauge readings obtained for
any one rod was 0.48 mm and the average range was 0.09 mm.
This is greater than might be expected given the quoted accuracy,
but in situ accuracy is affected by factors such as fine debris on
the top of the rods or guides. Despite these factors the overall
absolute accuracy of the rod extensometer system is still judged
to be about 0.3 mm (i.e. the accuracy of the precise levelling).
The change in anchor reduced level with depth was also plotted
for each individual borehole so that relative movement between
anchors within a single borehole could be identified. Processing
the data in this way enabled easier identification of erroneous
readings and overall precise levelling error. In most cases the five
data sets followed almost identical profiles, offset from oneanother according to variation in precise levelling results, as
shown in the example from one of the borehole extensometers in
Figure 13. It can be seen that the scatter in data between surveys
is about 0.2 mm and that erroneous readings are easily spotted
by their deviation from these profiles, for example the spike
between 20 and 25 m for the survey of 11 July 2012.
One way of improving the absolute accuracy of rod extens-
ometers readings (i.e. 0.3 mm) is to install a deep anchor in
each borehole. Provided that the deep anchor is unaffected by
ground movements, absolute vertical displacements of the an-
chors above it can be determined relative to it without the need
for precise levelling. This would improve the accuracy to
correspond to that of the dial gauge readings (i.e. 0.05 mm,
based on the average measured range within 0.09 mm). However,
in most cases this would be prohibitively expensive.
7. Torpedo inclinometersIn order to assess short-term precision of torpedo inclinometers,
tests were performed over a 1 h period on a single 20 m deep
borehole. Three sets of readings were taken at 0.5 m intervals
from bottom to top of the 20 m inclinometer casing in both the a-
axis and b-axis keyways, orientated at 908 to one another. The
torpedo used has microelectromechanical system (MEMS) de-
vices orientated in orthogonal directions so that a- and b-axes areread simultaneously. Torpedo inclinometer measurements are
largely unaffected by external human or environmental factors
due to their location below ground and automated data-logging
system, leaving their accuracy dependent on the instrument itself.
Primary sources of error are inadequate time allowed for initial
temperature equilibration of the torpedo within the casing and
dirt on the wheels of the torpedo or in the keyways. Errors can
also occur if the keyways are twisted; this should be avoided
Reference head Dial guage Referenceholes
Ground level
Cast-ironlockable
cover
Lean-mixcement
Stainlesssteel rod Guide
bracketRodsleeving
Anchor head(approximately
035 m long)
Edge ofborehole
Note: boreholes arebackfilled with astiff bentonite-cementgrout after rodinstallation
Extendableprongs
Figure 12.Schematic diagram of rod extensometers and
reference head (Nyren, 1998)
2 May 2012
19 June 201221 12June 20
4 12July 20
5 12July 2011 12July 20
0
5
10
15
20
25
30
35
40
45
04 03 02 01 0 01 02 03 04
Depthbelow
groundlevel:m
Subsurface vertical displacements: mm
Figure 13.Results plotted on a hole-by-hole basis allows for
easier identification of erroneous readings
11
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
12/13
during installation and can be checked using specific torpedodevices with a function to determine twisting.
Cumulative differences in horizontal deflections recorded from
the three sets of measurements are shown in Figure 14. The
differences are cumulated from the base of the borehole, which is
assumed fixed, and the second and third sets are compared with
the first set (i.e. taken to be zero). In the a-axis sense the readings
indicate an increasing development of error from the second set
to the third set and as the torpedo is raised towards the surface,
with maxima of 0.7 mm and 1.1 mm. For the readings in the b-
axis sense the maximum error is 0.3 mm. Readings were
commenced in the a-axis keyway after about 5 min and it appears
from the data that the torpedo had not acclimatised sufficiently
(in terms of temperature within the borehole), affecting the first
set of results. If the third set is compared with the second set,
greatly improved readings are observed with deviations within
0.5 mm and 0.1 mm for the a- and b-axes respectively. In some
torpedo instruments there is an option that can be implemented
where equilibration checks are made within the readout unit so
that measurements cannot commence, at the start of a set of a
survey with the instrument at the base of the inclinometer tube,
until two consecutive readings are within a set tolerance (e.g.0.02 mm) over a fixed, short time span (e.g. 90 s). It is anticipated
that such a procedure would have resulted in better compatibility
between the first and second sets of measurements. During the
1 h period negligible horizontal movement of the inclinometer
casing would have occurred, indicating the short-term precision
of cumulative deflection measurement using a torpedo over a
20 m depth at 0.5 m intervals would be within 0.25 mm
provided that sufficient time is allowed for the torpedo device to
acclimatise.
Potential errors resulting from the process of raising the torpedo,
removing/replacing it from/in the casing, and reacclimatisation at
the base of the borehole were also investigated. The torpedo was
lowered to the borehole base and allowed to equilibrate before
being raised to 15 m depth and secured in position. Twenty repeat
readings were then taken at this depth before raising the torpedo
to 10 m and 5 m depths, taking a further 20 repeat readings at
each. The whole process was then repeated with the instrument
rotated by 1808. Each set of 20 readings had a range within
0.05 mm, suggesting a measure of precision of 0.025 mm for
the inclinometer instrument itself.
0
2
4
6
8
10
12
14
16
18
20
0 02 04 06 08 10 12
Depthbelow
groundlevel:m
HP10 a-axis cumulative horizontaldisplacement: mm
0
2
4
6
8
10
12
14
16
18
20
0 01 02 03
Depthbelow
groundlevel:m
HP10 b-axis cumulative horizontaldisplacement: mm
28 June 2012 at 10:58
28 June 2012 at 11:31
28 June 2012 at 11:57
Figure 14. Short-term repeatability of cumulative horizontal
deflections measured using a torpedo inclinometer for 28 June 2012
12
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan
8/11/2019 geng1300047h.pdf
13/13
8. ConclusionMechanisms of ground and wall displacements are reviewed
along with methods of predicting their magnitude and extent.
Drawing on three well-documented case studies, a range of
typical displacements corresponding to excavation depths has
been estimated to judge monitoring accuracy requirements. Maxi-
mum vertical and horizontal ground and wall displacements are
typically in the region of centimetres, which can be readily
measured using many conventional monitoring techniques. How-
ever, much greater, sub-millimetre, accuracy is required if the
development of displacements is to be carefully appraised. This is
necessary when assessing construction/excavation using (a) the
observational method and also (b) when validating and refining
numerical analyses of the works.
Surveying techniques using precise levelling, rod extensometers
and torpedo inclinometers have been investigated by means of field
trials at a greenfield research site. The three methods are routinely
used and involve various potential human and environmental
factors that affect their accuracy. A variety of these factors have
been assessed and where possible quantified with a primary
emphasis on precise levelling (Table 3). Applying good practice to
each technique should enable surface and subsurface displace-
ments to be measured to sub-millimetre accuracy. Precise levelling
accuracy can be achieved to 0.3 mm, which also applies to
absolute measurements of rod extensometer anchors (much greateraccuracy of0.05 mm is possible for relative movements between
anchors in individual boreholes). The main factor controlling
precise levelling accuracy is the number of change points required
in the survey. These can be minimised and greater accuracy
achieved if a deep datum, for example a rod extensometer, is
installed on the site. The overall accuracy of torpedo inclinometer
measurements is judged to be 0.25 mm over a 20 m depth of
measurement, provided that absolute displacements at one end of
the case are known (e.g. if it can be assumed that the base is fixed).
REFERENCES
Atkinson JH and Salfors G (1990) Experimental determination ofstress-strain-time characteristics in laboratory and in situ
tests. Proceedings of the 10th European Conference on Soil
Mechanics and Geotechnical Engineering, Florence, Italy,
vol. 3, pp. 915956.
BRE(1993) Monitoring Building and Ground Movement by
Precise Levelling. IHS BRE Press, Garston, UK, Building
Research Establishment Digest 386.
Burland JB and Hancock RJR (1977) Underground car park at the
House of Commons: geotechnical aspects.The Structural
Engineer55(2): 87100.
Clough GW and ORourke TD (1990) Construction induced
movements of insitu walls. In Proceedings of Design and
Performance of Earth Retaining Structures. Cornell
University, Ithaca, NY, USA, ASCE Geotechnical Special
Publication No. 15, pp. 439470.
Dunnicliff J(1988) Geotechnical Instrumentation for Monitoring
Field Performance. Wiley, New York, NY, USA.
Gaba AR, Simpson B, Powrie W and Beadman DR (2003)Embedded Retaining Walls Guidance for Economic Design.
Ciria, London, UK, Report C580.
Hsieh P and Ou CY (1998) Shape of ground surface settlement
profiles caused by excavation.Canadian Geotechnical
Journal35(6): 10041017.
Ng CWW, Menzies BK and Simons NE (2004)A Short Course in
SoilStructure Engineering of Deep Foundations,
Excavations and Tunnels. Thomas Telford, London, UK.
Nicholson D, Tse C and Penny C (1999) Observational Method in
Ground Engineering: Principles and Applications. Ciria,
London, UK, Report R185.
Nyren RJ(1998) Field Measurements Above Twin Tunnels in
London Clay. PhD thesis, Imperial College, London, UK.
Ou CY(1993) Characteristics of ground surface settlement during
excavation.Canadian Geotechnical Journal30(5): 758767.
Peck RB (1969) Deep excavations and tunnelling in soft ground.
Mexico City. Proceedings of the 7th International Conference
on Soil Mechanics and Foundation Engineering, Mexico City,
Mexico, pp. 225290.
Roscoe H and Twine D (2010) Design and performance of
retaining walls. Proceedings of the Institution of Civil
Engineers Geotechnical Engineering163(5): 279290.
St John HD, Potts DM, Jardine RJ and Higgins KG (1993)
Prediction and performance of ground response due to
construction of a deep basement at 60 Victoria Embankment.Proceedings of the Wroth Memorial Symposium, Oxford, UK,
pp. 581 608.
Standing JR, Withers AD and Nyren RJ (2001) Measuring
techniques and their accuracy. In Building Response to
Tunnelling. Case Studies from the Jubilee Line Extension,
London, Volume 1: Projects and Methods(Burland JB,
Standing JR and Jardine FM (eds)). Thomas Telford, London,
UK, pp. 273299.
Wan MSP and Standing JR (2014) Lessons learnt from
installation of field instrumentation. Proceedings of the
Institution of Civil Engineers Geotechnical Engineering,
http://dx.doi.org/10.1680/geng.13.00047.
WHAT DO YOU THINK?
To discuss this paper, please email up to 500 words to the
editor at [email protected]. Your contribution will be
forwarded to the author(s) for a reply and, if considered
appropriate by the editorial panel, will be published as a
discussion in a future issue of the journal.
Proceedingsjournals rely entirely on contributions sent in
by civil engineering professionals, academics and students.
Papers should be 20005000 words long (briefing papersshould be 10002000 words long), with adequate illustra-
tions and references. You can submit your paper online via
www.icevirtuallibrary.com/content/journals, where you
will also find detailed author guidelines.
Geotechnical Engineering Deep excavations: monitoring
mechanisms of ground displacement
Fearnhead, Maniscalco, Standing and Wan