geng1300047h.pdf

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


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




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




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




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


XSMP number

Figure 7.XSMP data collected during investigation relative to 2

May 2012

7




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


2 May 2012

19 June 2012

20 12June 20

04 12July 20

06 12July 20

11 July 2012

Hyde Park North CarriageDrive


YSMP number

Figure 8.YSMP data collected during the investigation relative to

2 May 2012, using a far-off benchmark

8




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


2 May 2012

19 12June 20

20 June 2012

4 12July 20

6 July 2012

11 July 2012

Hyde Park North CarriageDrive


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




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




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




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




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.




geng1300047h.pdf

Documents