You are on page 1of 8

Predicting the Settlements above Twin Tunnels Constructed in

Soft Ground
D. N. Chapman, C. D. F. Rogers, D. V. L. Hunt
School of Engineering, Civil Engineering, The University of Birmingham, U.K

ABSTRACT
This paper presents a simplified method of adapting the settlement curve above the second tunnel to
allow a more realistic prediction of the overall settlements above multiple tunnel constructions. The
method is still based on the use of the error function curve and so can easily be incorporated into
current software packages or spreadsheets. The method applies a modification factor to the soil within
the zone of overlap between the predicted areas of disturbed soil above the tunnels. In clay soils it is
shown that the ground settlements need to be modified by approximately 60% to take into account the
prior disturbance from a first tunnelling operation.
Although there is only limited good quality published case history data related to multiple tunnel constructions,
the paper describes the good agreement obtained when the method is applied to several published results.

1. INTRODUCTION
Many semi-empirical equations have been introduced by various authors for predicting the movements
above single tunnels (e.g. Peck, 1969, O, Reilly and New, 1982). The equations have been shown over
many years, when compared to case history data of single tunnels, to make accurate predictions of the
vertical and horizontal displacements.
Based on the accuracy of these existing equations OReilly and New (1991) extended their use for
predicting the movements above twin tunnels. These equations were based on the assumption that the
first and second tunnels would have identical settlement profiles. The individual profile above each
tunnel was predicted by the existing single tunnel equations, the total profile was merely an addition of
each profile. However, when using the equations of OReilly and New (1991) for predicting the
movements above twin tunnels with a delay between drives several differences occurred when
compared to the field data. Good quality twin tunnel case history data is one way of improving these
empirical equations for twin tunnel prediction. However, many twin tunnel case histories do not
usually have sufficient data points in critical positions and are usually given as a total profile and do
not show the individual contributions made by each tunnel. Quality data that does fit this description
has been reported by Cording and Hansmire (1975), Nyren (1998) and Cooper et al. (2002) and for
this reason are included in this paper. Due to the lack of data and through advancements in computing
power researchers have been aiding understanding of movements above twin tunnels by use of
numerical modelling using numerical methods (e.g. Addenbrooke, 1996; Chapman et al., 2003).
Unfortunately, the resulting magnitudes of settlements produced using these numerical methods do not
accurately reproduce the actual magnitudes of displacement found at full scale. Hence they cannot be
used directly to improve the empirical equations. However, the results from such studies have aided
the understanding of the movements above twin tunnels and provided a starting point for possible
modifications to be made to existing empirical methods.

C33

This paper consists of two parts. The first part describes the currently available empirical equations for
predicting movements above single tunnels. The effect of parameter selection in connection to the
prediction of horizontal and vertical displacements is highlighted for each equation. The second part of
the paper considers how one might adapt these equations in order to more precisely incorporate the
ground movements that are found above twin tunnels. Possible modified equations are given and
trialled for comparison with field data.
2. PREDICTIVE METHODS FOR MOVEMENTS ABOVE A SINGLE TUNNEL
For most soft ground tunnelling problems the ground movements are caused by movements of clay
occurring at the construction level due to the tunnelling operations and the removal of ground for
tunnel liner placement. These movements are incorporated into a volume loss parameter, Vl (usually
expressed as a percentage).
Vl =

4 Vs
D 2

(1)

where, Vs is the theoretical volume of the settlement trough per metre length of excavation found by
integrating Equation 2 (Vs = i (2)0.5) and D is the diameter of the tunnel. For most tunnelling
operations the value of the volume loss can be expected to range from 0.5% to 2.0% depending on the
type of tunnelling method being employed. The volume Vs is assumed to be the same as the losses that
occur at the tunnel Vt.
It is now well established and accepted that the surface settlement profile above a tunnel in soft ground
may be represented by a Gaussian distribution curve of the form shown in Equation 1 (Peck, 1969 and
OReilly and New, 1982).
W (x ) = W

max

x2
exp
2
2i

(2)

where, W is the vertical displacement and x is the distance from the tunnel centreline in the transverse
direction. Wmax is maximum settlement, which occurs at the point x = 0 and i is the distance to the
point of inflexion on the curve (described later). By combining Equations 1 and 2, the maximum
settlement at the surface can be found from the volume losses expected during tunnelling operations
and is given by Equation 3.
W max =

0 . 313 V l D 2
i

(3)

Subsurface settlement profiles are found by assuming Gaussian profiles predicted by Equation 2. Wmax
and hence W are found assuming no volume change with depth (i.e. by assuming that each subsurface
settlement profile has the same volume/unit length as the surface profile) and using subsurface values
of i. As i decreases with depth both Wmax and hence W must increase. The effects of doubling or
halving the volume loss will double or halve the W, U and Wmax and Umax.
On the basis of field data OReilly and New (1982) suggested that the distance to the point of
inflection, defined as i = KZo, where Zo is the depth of the tunnel and K is a dimensionless trough
width parameter. OReilly and New (1982) also suggested a value of K to be around 0.5 for London
Clay when considering surface settlements. Extending the assumptions of Equation 4 to subsurface
regions leads to an underestimate of i (which in turn leads to an overestimate in W) with depth
especially in regions close to the tunnel. Mair et al. (1993) on the basis of centrifuge and field data
improved the subsurface prediction of i by using Equation 4.

C33

i/ Zo = 0.175 + 0.325 (1 - Z/ Zo)

(4)

In the special case where Z=0 (i.e. ground surface level) the equation yields i = 0.5. The equation gives
a linear decrease in i with depth. Many other non-linear empirical equations exist for subsurface values
of i and have been presented by Heath and West (1996). Grant and Taylor (2000) also postulated
possible theories as to how i should change with depth when considering the movements above a
single tunnel although no formulation was given by the authors.
The horizontal movements shown in Figure 1 can be found by assuming a direct relationship with the
vertical movements as shown in Equation 5.

Wx
U(x ) =

Zo

(5)

where U is the horizontal movement. The equation is based upon the assumption of a position of
vector focus (in this case the bottom term in the bracket). OReilly and New (1982) reported this
vector focus to be situated at the exact centre of the tunnel (highlighted as position 2 in Figure 3). The
maximum horizontal displacement, Umax, occurs at the point of inflection, i. By combining Equation 2
and Equation 5, Umax can be given in terms of Wmax. When assuming a vector focus at depth Z0, Uma x =
0.202 Wmax as presented in Figure 2. The location of the point for vector focus is highly important and
directly effects the magnitudes of horizontal movements. Based on the work of Mair et al. (1993)
improved predictions have been found assuming a vector focus at 0.175Zo/0.325 below the tunnel axis,
(highlighted as position 3 in Figure 3). This yields the simple relationship between the horizontal (U)
and vertical (W) movements with depth given by Equation 6:

Wx

U( x ) =
(1 + 0.175 / 0.325) Zo

(6)

Using this new point of focus Umax = 0.131 Wmax. The magnitudes of horizontal displacements
predicted by Equation 6 are lower than those predicted by Equation 5 as shown in Figure 2. If a point
of vector focus were assumed at depth 0.5Zo (highlighted as position 1 in Figure 3), Umax = 0.404
Wmax. The magnitudes are higher than those predicted by Equation 5 or 6. The changes in W and U
and their corresponding maximums can be seen in Figure 2.
x

0.404Wmax
0.202Wmax
0.131Wmax

Free surface

Z
U

0.5 Zo

Zo

1.0Wmax

Vector focus 1
Vector focus 2
Vector focus 3
(see Figure 3)
3

Figure 2 Variation in horizontal displacement due


to changes in position of vector focus

C33

0.175
Zo +
Zo
0.325
Point of vector focus

Figure 3 Variation in position of vector focus

1.2 Umax
1.0 Umax
0.8 Umax

0.8i
1.0i
1.2i

Umax

0.8i
1.0i
1.2 i
U

0.8Vl
1.0Vl
1.2Vl
W max

0.8Wmax
Wmax

K=0.6
K=0.5
K=0.4

1.2 Wmax

Figure 4 Variation in displacements due to trough


width and volume loss increases (Wmax constant)

Figure 5 Variation in displacements due to


trough width changes (Volume loss constant)
Grant and Taylor (2000) suggested that further improvements could be made to the prediction of U for
subsurface regions if three separate points of vector focus were assumed for three distinct subsurface
regions. The focal points were found as the intersection of the tangents to i with depth. In conclusion,
using a vector focus below or above the tunnel centre line will decrease or increase the horizontal
movements above the tunnel respectively.

The value of K chosen will also effect W and U. Figures 4 and 5 consider three values for K of 0.4, 0.5
and 0.6. Figure 4 considers changing K while maintaining a constant value of Wmax, hence increasing
the volume loss. Figure 5 considers the effect on W when using a constant volume loss and changing
values of K. When using a constant value of Wmax, the increase in the trough width from 0.4 to 0.6
causes a similar increase in the volume loss and the magnitude and position of U max. Although, when
keeping the volume loss constant, by increasing K from 0.4 to 0.6, W decreases. Wmax can be seen to
have an inverse relationship with K. It should also be noted that the position of Umax increases relative
to K, but the magnitude of Umax remains unchanged provided Vl is kept constant.
3. PREDICTING MOVEMENTS ABOVE TWIN TUNNELS
The simplest assumption when predicting movements above a second tunnel is for the settlement
profile to match that above the first tunnel (i.e it has a greenfield profile). When considering side by
side tunnels of the same size, the displacements above the second tunnel are found based on the same
assumptions made for the first tunnel. The total settlement profile is found by summation of the
separate profiles calculated using Equation 3 and 4, due to each tunnel. However, many authors
reported that the method is not accurate for predicting settlements if a time delay is expected between
the construction of each tunnel. The inaccuracies that occur in the settlement profile have been related
to the following changes in: Soil stiffness (pre failure), volume loss, magnitude of W and position of
Wmax , trough width parameter, K (hence i), near and remote limb lengths (related to K), magnitude of
U, position of U=0 and Umax,, position of the vector focus for movements. The various changes to
these parameters when the second of twin tunnels is constructed have been highlighted by various
authors using the results of field data and numerical modelling. Mair and Taylor (1997) explained the
changes in settlement above the second tunnel by stating that the soil will have been previously
strained by tunnel 1 and bigger losses would be expected for tunnel 2. Chapman et al. (2002) defined
the area of previously strained soil as the overlapping zone found from the overlapping bounds to

C33

movement for each tunnel. The soil inside this zone was highlighted as being vulnerable to changes in
stiffness due to straining and hence would account for the changes in behaviour that occurred. The
details of these changes in behaviour and their incorporation into current empirical models are detailed
in this section.
The effect of pre-failure stiffness in relation to twin tunnels was first highlighted by Addenbrooke
(1996). The results of numerical modelling using the finite element method and non-linear small strain
soil models showed increased horizontal and vertical displacements on the side of the first tunnel
driven. Similar results were found by Chapman et al. (2003) when considering twin tunnels.
Peck (1969), Cording and Hansmire (1975), Addenbrooke (1996) and Cooper et al. (2002) reported
increases in volume loss for the second tunnel constructed. Cording and Hansmire (1975) reported as
much as a 50% increase in volume loss for a second 6.4 m diameter tunnel at a centre to centre spacing
of 11m (d/2R of 0.5). Similar increases were reported by Addenbrooke (1996), who showed that the
volume loss increase decreased with tunnel spacing. Trend lines for the prediction of volume loss in
relation to tunnel spacing have been set by Cording and Hansmire (1975), with improvements to the
trend line based on a wider range of field data being suggested by Cooper et al. (2002). Chapman et al.
(2003) found that similar increases in volume loss could be achieved by using Equation 7 on the
settlement profile for a second tunnel
Cording and Hansmire (1975) Addenbrooke (1996) Cooper et al. (2002) and Chapman et al. (2003)
showed that the magnitude of Wmax is increased with its position being eccentrically placed towards
tunnel 1. The eccentricity and relative increase in Wmax were found to decrease with increased tunnel
spacing and also for subsurface regions.
As highlighted in Figure 1 an increase in settlement can be caused by an increase in volume loss for
the second tunnel, which is easy to implement into the empirical model. However, this still gives a
symmetrical curve with no eccentricity in Wmax. Field data and numerical modelling have shown an
asymmetry in the curves with more settlement occurring on the side nearest the first tunnel. A method
of incorporating all of these differences was proposed by Chapman et al. (2003) based on a
modification factor shown in Equation 7.

W mod = 1 +


d ' + x
M 1
W

AK 1 Z *

(7)

where M is a Modification factor (typically 0.6), A is the multiple of i to make a half trough width
(typically 2.5-3), Z* is Zo Z, d is the distance between tunnel axes (-ve for distances on the left hand
side of tunnel 2) and K1 is the value of K for Tunnel 1.The other symbols have their usual meanings.
The equation is only applied to soil inside the overlapping bounds to movement for the first and
second tunnel. The factor was based on the results of finite element analyses which highlighted that
the maximum relative increase in settlement for a second tunnel always occurred over the centreline of
the first tunnel driven. The bracketed value refers to the modification used, i.e. 60% refers to an M
factor of 0.6.
Cording and Hansmire (1975) suggested that the trough width could possibly be increasing for the
second tunnel on the side nearest to the first tunnel driven. Cooper et al. (2002) provided methods of
estimating the relative increases in trough width based on an increase in the volume of the near limb,
V2n, relative to the remote limb, V2r. A trend line for the ratio of V2r/V2n was given for varying
tunnel spacing. The volume and hence the half trough width of the near limb were assumed to be
greater than that of the remote limb, i.e. Kn > Kr, where Kn and Kr are the trough width parameters for
the near and remote limbs respectively. Changes to K have been noted in this paper. When considering
the settlements above a second tunnel any increases made to the trough width while keeping Wmax
constant would achieve an increase in the total volume/ unit length of the settlement profile and hence

C33

would assume an increase in volume loss. However, the changes in trough width alone do not account
for the increases in Wmax that occur or its eccentric placement towards tunnel 1. Chapman et al. (2003)
found that similar patterns of V2r/V2n could be found by using Equation 7 when assuming no changes
to K for near and remote limbs, (i.e. Kn = Kr). If the changes in K could be exactly quantified their
inclusion into equation 8 could easily be facilitated by assuming different values of K when
considering movements on the near and remote limbs as shown in Equation 8, K2 represents the value
of K for tunnel 2 with separate values being used for Kn and Kr:

x2

W = Wmax exp
2
2 (K 2 Z*)

(8)

Cording and Hansmire (1975), Addenbrooke (1996) and Chapman et al. (2003) have all reported
higher horizontal displacements on the near limb compared to the remote limb. As far as the authors
are aware no methods are currently available for predicting these horizontal movements. This paper
suggests that if improvements are made to the prediction of W such as that given by Equation 7 and 8
and if the relationship given by Equation 2 for horizontal movements is still valid, it should be
possible to improve predictions of U. One possible formulation is shown by Equation 9:
W x
U ( x ) = mod f
Zf

(9)

Wmod is the settlement of the modified profile (From Equation 7), xf is the horizontal distance from
vector focus and Zf is the depth to the point of vector focus. As far as the authors are aware little
attention has been given to changes in the position of vector focus for a second tunnel driven. The
depth of the vector focus has been highlighted in this paper as being a significant influence on the
magnitude of U. The horizontal positioning of the vector focus is equally important as it controls the
position of U=0. By consideration of field data in Section 4 possible deviations in horizontal position
of vector focus may be occurring, which can be taken into account by using Equation 11. Chapman et
al (2002) reported an offset in the position of U=0 when considering numerical modelling. Cording
and Hansmire (1975) also reported an eccentric placement of U=0 towards tunnel 1. Equation 9
incorporates this shift in U=0 by offsetting the vector focus from the centreline of the tunnel axis.
4. PREDICTING CASE HISTORY MOVEMENTS
Three published case histories are considered in this Section: the Heathrow Express (U.K.), Lafayette
Park (U.S.A.), and St James Park (U.K.). Table 1 shows the details of each twin tunnel case history
being considered. By employing the currently available techniques for predicting movements above
single tunnels and considering the various methods for improving settlement prediction above twin
tunnels, it can be shown how it is possible to match quite accurately the recorded settlement profiles
for a second tunnel driven. Figure 7 shows the subsurface displacements of the Piccadilly line tunnels
on the Heathrow Express (Cooper et al., 1998). The subsurface settlements were measured 13 m below
the ground surface inside existing twin running 4 m diameter tunnels that were at a 700 skew to the
new 9 m diameter tunnels being constructed 13 m below. A settlement profile based on Equations 1, 2,
3, 6 and 8 is shown for tunnel 1. At a depth of 13 m below the ground surface and 13m above the
tunnel axis, K should be equal to 0.675, however, the skew elongates the profile and a value of 0.8 was
found more appropriate. This profile was superimposed over tunnel 2 and shows the greenfield
settlement that would most likely occur for the tunnel 2 if tunnel 1 had not been present. The modified
profile was determined by increasing K on the near limb whilst keeping Wmax constant. The resulting
settlement profile was then multiplied by Equation 9 in the overlapping zone. The new profile closely
matches the field data. Figures 7 and 8 are shown in order to highlight how improvement to the
prediction of vertical displacements combined with the use of Equation 9 can lead to improved

C33

Location

Lafayette Park

St James Park

HeathrowExpress

Tunnel

Green

Green

Green

D(m)

6.4

6.4

4.8

4.8

Zo (m)

14.6

14.6

20.5

20.5

26

26

Z (m)

13

13

d(m)

-11

22.5

23

V(%)

3.63

2.7

1.2

1.8

Wmax (mm)

52.7

69.6

18.48

22.6

29.3

35

Umax (mm)

10.39

38.52

3.64

11.6

Kn

0.5

0.6

0.4

0.43

0.8

1.2

Kr

0.5

0.35

0.4

0.4

0.8

0.8

Focal Pt (x,z)

(-11,22.5)

(-9,9)

(22.5,31.5)

(22.5,12)

N/A

N/A

0.6

0.6

0.6

-40

-30

-20

-10

Chainage (m)
0
10

20

30

40

50
0

-10

-20

-30

Greenfield W tunnel 2
Modified W tunnel 2
Greenfield W tunnel 1
W actual (tunnel 1)
W actual (tunnel 2)

Displacement (m)

Table 1 Twin tunnel details

-40

-50

Figure 6 Heathrow Express after driving


Upline tunnel (Cooper and Chapman, 1998)

-40

-30

-20

Chainage (m)
-10

Chainage (m)
0

10

20

40

10

20

30

40

50

10

Displacement (mm)

Displacement (mm)

20

-10
20

-20
-40
-60
-80
-100

Greenfield W tunnel 2
Modified W tunnel 2
Greenfield U tunnel 2
Modified U tunnel 2
Actual W
Actual U

-10

-20

-30

Greenfield W tunnel 2
Modified W tunnel 2
Greenfield U tunnel 2
Modified U tunnel 2
U actual
W actual

-40

Figure 7 Lafayette
Hansmire, 1975)

Park

(Cording

and

Figure 8 St James Park (Nyren, 1998)

predictions of horizontal movements. Figure 7 shows the movements at Lafayette Park reported by
Cording and Hansmire (1975). By modifying a Gaussian curve, the settlement profile can be made to
quite accurately fit the case history data. The Gaussian curve highlighted as Greenfield tunnel 2, was
found using Equations 1, 2, 3, 6 and 8 assuming a 3% volume loss above a 6.4 m diameter tunnel at a
depth of 14.6 m with K=0.5. The modified W was found by increasing K on the near limb to 0.6 (from
a value of 0.5) reducing K on the remote limb to 0.3 (from a value of 0.5) and then applying Equation
8 in the overlapping zone. The horizontal movements were found using Equation 9, assuming a vector
focus eccentrically placed 2 m towards tunnel 1 and 9 m below ground level. Figure 8 shows the
movements at St James Park reported by Nyren (1998). The first tunnel was driven 31 m below
ground level and the second at 20.5 m below ground level. The Gaussian curve was found assuming a
2% volume loss on a 4.8 m diameter tunnel with a K of 0.4 (based upon a sand overlaying a clay).The
modified settlement was found by increasing the near limb (K=0.43) only and applying equation 9 in
the overlapping zone. The horizontal movements were found using Equation 9, assuming a vector
focus situated 12 m below ground level at the vertical centreline axis of tunnel.
5. CONCLUSIONS
This paper has shown several ways in which analyses of the settlement profile above a single tunnel
analyses can be affected by the parameters selected in the empirical equations. The role of these
parameters has been highlighted. This paper has described some of the results from the current

C33

research at Birmingham University investigating methods for predicting the settlements that occur
above twin tunnel constructions. While the behaviour is complicated, it has been shown that current
empirical equations combined with modification factors can be used in order to obtain similar vertical
and horizontal displacement profiles to those found from case history data when considering the short
term behaviour associated with driving tunnels side by side. As more twin tunnel case history data
becomes available it is hoped that more detailed guidelines can be provided for practitioners.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the Engineering and Physical Sciences Research Council in
the UK for their financial support of this research project.
REFERENCES
Addenbrooke, T.I., 1996. Numerical analysis of tunnelling in stiff clay. PhD Theses, Imperial College,
London: p.373.
Cooper, M.L., Chapman, D.N., Rogers, C.D.F., 2002. Prediction of settlement in an Existing Tunnel
cause by the Second of Twin Tunnels. Design of structures, 2002, Transportation research record
1814, pp.103-112.
Chapman, D.N., Rogers, C.D.F., Hunt, D.V.L., 2003. Investigating the settlement above closely
spaced multiple tunnel constructions in soft ground. Proc. of World Tunnel Congress 2003,
Amsterdam 2003, vol 2. pp.629-635.
Cording, E.J., Hansmire, W.H., 1975. Displacement around soft ground tunnels. Proc. 5th Pan
American Conf. on Soil Mech. & Foundation Eng., Buenos Aires, Vol. 4. pp.571-633.
Grant, R.J., Taylor, R.N., 2000. Tunnelling Induced ground movements in clay, Proc. Inst Civ Engrs
2000. 143, pp.43-45.
Heath, G. R., West, K. J. F., 1996. Ground movements at depth in London clay. Proc.Instn Civ. Engrs
Geotech, Engng , 119, Apr, 65-74.
Mair, R.J., Taylor, R.N., 1997. Bored tunnelling in the urban environment. Proc. Conf. on Soil
Mechanics. and Foundation Engineering, Hamburg Vol. 4. Theme lecture.
Mair, R.J., Taylor, R.N., Bracegirdle, A., 1993. Subsurface settlement profiles above tunnels in clays.
Geotechnique 43 (2). 315-320.
Nyren, R., 1998. Field measurements above twin tunnels in London Clay. PhD theses, Imperial
College
OReilly, M.P. New, B.M., 1982. Settlements above tunnels in the United Kingdom their magnitude
and prediction. Proc. of Tunnelling82 Symposium, London. pp.173-181.
OReilly, M.P., New, B.M., 1991. Tunnelling Induced Ground Movements; Predicting their
Magnitude and Effects. 4th Int. Conf. On Ground Movements & Structures. pp.671-697.
Peck, R.B., 1969. Deep Excavation and Tunnelling in Soft Ground. State of the art volume. 7th Int.
Conf. On Soil Mech. & Foundation Eng., Mexico. pp.225-290.

C33

You might also like