See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/245411661

Soil-pipe interaction due to tunnelling:

Comparison between Winkler and elastic
continuum solutions

Article in Gotechnique January 2005

DOI: 10.1680/geot.2005.55.6.461

CITATIONS READS

56 676

4 authors, including:

Assaf Klar
Technion - Israel Institute of Technology
100 PUBLICATIONS 923 CITATIONS

SEE PROFILE

All content following this page was uploaded by Assaf Klar on 16 September 2014.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Klar, A., Vorster, T. E. B., Soga, K. & Mair, R. J. (2005). Geotechnique 55, No. 6, 461466
Soilpipe interaction due to tunnelling: comparison between Winkler
and elastic continuum solutions
A . K L A R * , T. E . B. VO R S T E R * , K . S O G A * a n d R . J. M A I R *
An elastic continuum solution and a Winkler solution of
the problem of tunnelling effects on existing pipelines are
given. A comparison is made between an elastic conti-
nuum solution and a closed-form Winkler solution with
Vesic subgrade modulus. Although applying the Vesic
expression results in the same moments and displace-
ments under external loading in a Winkler system and
the elastic continuum, it is found that its use is not
necessarily adequate for the problem of tunnelling effects
on pipelines and may not be conservative owing to
possible underestimation of bending moments. An alter-
native expression for the subgrade modulus is provided,
resulting in similar maximum bending moments in the
Winkler and elastic continuum systems.
KEYWORDS: buried structures; deformation; elasticity;
numerical modelling; tunnels
INTRODUCTION
One of the tasks facing engineers in the 21st century is the
operation and maintenance of ageing infrastructure such as
pipelines. The current paper addresses the effect of tunnel-
ling on existing buried pipelines. Fig. 1 shows a schematic
diagram of the problem, in which a new tunnel is excavated
under an existing pipe. The tunnel excavation generates soil
settlement around the pipe, causing it to deform. The pipe
suffers additional bending moment. The magnitude of pipe
deformation and the changes in bending moment depend on
the distribution of soil settlement due to tunnelling at the
pipeline level and the relative stiffness between the pipe and
the surrounding soil.
The conventional approach for obtaining a solution for
this problem utilises Winkler-based models such as that
proposed by Attewell et al. (1986). In such cases, an
appropriate subgrade modulus (spring coefficient) needs to
be assumed for both linear elastic and non-linear analyses.
In linear elastic analysis the subgrade modulus is usually
determined by means of the Vesic (1961) expression (sug-
gested for the current problem by Attewell et al., 1986).
Vesics expression essentially allows a beam on a Winkler
foundation to exhibit similar displacements and moments to
that of a beam on an elastic half-space when loaded with
the same load.
The aim of this paper is to discuss the validity of the
implementation of Vesics expression for the current problem
by comparing it with an elastic continuum solution. This aim is
achieved by comparing a Winkler system and an elastic
continuum system under the same key assumptions, as follows:
(a) A pipeline line is buried in homogeneous soil.
(b) The pipe is always in contact with the soil.
Manuscript received 17 June 2005; revised manuscript accepted 4
April 2005.
Discussion on this paper closes on 1 February 2006, for further
details see p. ii.
* Department of Engineering, University of Cambridge, UK.
461
Delivered by ICEVirtualLibrary.com to:
IP: 132.68.129.39
On: Mon, 10 Oct 2011 09:37:47
Nous fournissons une solution de continuum elastique et
une solution de Winkler au probleme des effets de la
construction de tunnels sur des pipelines existants. Nous
faisons la comparaison entre une solution de continuum
elastique et une solution Winkler de forme fermee avec
un module de couche de forme de Vesic. Bien quappli-
quant les resultats de lexpression de Vesic dans les
memes moments et deplacements sous charge externe
dans un systeme de Winkler et un systeme de continuum
elastique, nous trouvons que son utilisation nest pas
forcement adequate pour le probleme des effets de con-
struction de tunnels sur des pipelines et peut ne pas etre
conservatrice en raison dune eventuelle sous-estimation
du moment de flexion. Nous donnons une expression
alternative du module de couche de forme, ce qui donne
des moments de flexion maximum similaires dans le
systeme de Winkler et le systeme de continuum elastique.
x
Pipeline
Tunnel
z
y x
z
Fig. 1. Schematic representation of the problem
(c) The pipe does not affect the tunnel.
(d ) The soil response to loading, at pipe level, is not aware
of the tunnel. (In the elastic continuum system this
relaxing assumption allows us the use of Mindlins
(1936) Greens function for vertical load in a semi-
infinite half-space.)
(e) The pipeline is continuous.
( f ) The greenfield soil displacement at the pipe level is
described by a Gaussian curve (Peck, 1969) given as
"
#
1 x 2
Sv x Smax exp
(1)
2 i
where Smax is the maximum settlement, x is the
horizontal distance from the tunnel centreline, and i is
the distance to the inflection point of the greenfield
settlement trough profile.
462 KLAR, VORSTER, SOGA AND MAIR
As mentioned, if a VesicWinkler system is loaded with
the same load as the elastic continuum system it will exhibit
a similar bending moment to that of the elastic continuum
system. In the current problem, the tunnelling effect may be
represented as loads on the system related to the soil green-
field displacements at the pipe level, Sv (x). These loads will
generally be different in the VesicWinkler and elastic
continuum systems; hence the use of Vesics expression for
this problem will generally result in different bending mo-
ments in a Winkler system compared with those in the
elastic continuum system. The magnitude of the difference
is presented in this paper.
ELASTIC CONTINUUM SOLUTION
The pipe behaviour may be represented by the following
equation:
[S]fug fFg
where [S] is the stiffness matrix of the pipe composed of
standard beam elements, {u} is the pipe displacements, and
{F} is a force vector representing the soil loading acting on
the beam elements. The pipe is buried within the soil, and
additional external loads to that of the soil do not generally
exist, although they can be added if necessary, building
structure loading for example.
The soil continuum displacement, u C , can be represented
using a Greens function:
Xn
fuC gi f f g j G i, j (3)
j1
where {f} are the forces acting on the soil medium, and Gi, j
is the Greens function that defines the elastic soil continuum
displacement at point i due to unit loading at point j.
The summation in equation (3) can also be written as
follows:
f u CA g i
fu gi CL
f u C g i z}|{ z}|{ (4) any displacement at a point due to uniform load is equal to
X n
f f g i G i,i f f g j G i, j
j1
j6 i
where {u CL } is defined herein as local displacement, which
is the displacement at a point due solely to its loading, and
{u CA } is additional displacement at that point due to forces
acting at different points. Because of assumption (c) only
degrees of freedom of the pipe need to be considered, and
the index i can therefore be related only to the pipe. Never-
theless, {u CA } still involves quantities that result from the
tunnel (i.e. j index is still related to the tunnel degree of
freedom). However, as it was assumed that the pipe does not
affect the tunnel, these quantities (i.e. forces) can be de-
coupled as follows:
f u CAP g i
f u CL g i
f u C g i z}|{ z
}|{
X
last pipe node
f f g i G i,i
f f g j Gi, j f u CAT g i
f u CL g i
f F g i
f f g i
(6)
G i,i
The compatibility relation of fug fuC g fuCL g fuCAP g
fuCAT g is required, and by introducing this and equation
(6) into equation (2) the following relation is obtained:
S f ug K  f ug 8K  f u CAP g K  f u CAT g
h i < 1
 i j (7)
K G i,i
i, j :
0 i 6 j
where [K*] is a local soil stiffness matrix, and is diagonal.
Equation (5) shows that f u CAP g  sf f g where

s i, j f G i, j for i 6 j, 0 for i jg, and with
f f g
f F g
S f ug equation (7) becomes
[[S] [K  ] [K  ][ s][S]]fug [K  ]fu CAT g (8)
(2)
which is solved numerically to obtain the elastic continuum
solution. Following assumption (c) {u CAT } is the greenfield
displacement.
It is worth noting that omitting [K*][*s][S] in the above
equation results in a Winkler-like model, where the soil
reaction acting on the pipe is not affected by the soil
response at different locations along the pipe. The term
[K*][*s][S] can thus be regarded as an additional term that
takes account of continuum effects. This, however, does not
mean that the solution obtained by omitting this term is the
Winkler solution, because the components of [K*] are differ-
ent from those that will be constructed using the Vesic
(1961) subgrade modulus.
In this study, the Mindlin (1936) solution (Greens func-
tion) for a point load is used to construct the components of
equation (8). However, as Mindlins solution does not satisfy
displacement at the point of loading, a reference displace-
ment value for that point was considered to be the average
displacement around the circumference of the pipe. This is
identical to assuming a barrel load around the pipe. Actually
the average displacement, over the same area (or volume) as
that of the uniform load, due to an equivalent concentrated
load at that point, as illustrated in Fig. 2. This is due to the
reciprocity property of the Greens function, which states
that the response at x9 due to a delta function at x is equal
to the response at x due to a delta function at x9.
To enable a general solution corresponding to different
soil and pipe characteristics the results were normalised. The
proposed normalisation covers all independent parameters,
and was used to describe bending moment. Fig. 3 shows the
computed normalised maximum sagging bending moments
Concentrated load system Uniform load system
ux 5 G(x; x)P p 5 P/A
(5) x
ux 5 eG(x; x)pdA
P
A

jfirst pipe node x

j6 i x

where {u CAP } is the additional displacement due to forces

resulting from soil pipe interaction, and {u CAT } is the By definition G(x; x) 5 G(x; x ), hence
additional displacement due to the existence of the tunnel.
This decomposition is valid as long as assumption (c) holds. eux d A eG(x; x)PdA P
Remembering that the force acting on the soil is the uav 5
A
5
A
5 eG(x; x)dA 5 eG(x; x)pdA 5 ux
A A A A A
reaction for the pipe, one can define the soil reaction on the
pipe from the above equation: Fig. 2 Explanation of the barrel load identity

Delivered by ICEVirtualLibrary.com to:

IP: 132.68.129.39
On: Mon, 10 Oct 2011 09:37:47
 SOILPIPE INTERACTION DUE TO TUNNELLING 463
12 2060

Sagging 2040
10
Z/r0 5 3 2020
Z/r0 5 7
08 Z/r0 5 15 000

Mi 2/ElSmax
Mi2/EISmax

020
06
040

04 060 R 5 El/i 3r0Es

080 R 5 80
02 R 5 20
100 R 5 05
R 5 00
0 120
001 01 1 10 100 25 24 23 22 21 0 1 2 3 4 5
R 5 El/i3r0Es x/i

Fig. 3. Maximum sagging moments Fig. 4. Normalised bending moment along the pipe, Z/r0 7

@4 u
occurring at the tunnel centreline in relation to a rigidity 44 u 44 Sv x (9)
factor R, defined as R EI/Es r0 i3 , where EI is the bending @x 4
p

stiffness of the pipe, Es is Youngs modulus of the soil, and where 4 K=4EI , EI is the bending stiffness of the pipe,
r0 is the radius of the pipe. The normalised bending moment K is the subgrade modulus, u is the vertical pipe displace-
is defined as Mi2 /EISmax . It should be noted that the rigidity ment, and Sv is the greenfield soil settlement (i.e the soil
factor and the normalised bending moment depend on i and settlement at the pipe level if it did not exist). An alternative
Smax . Hence any uncertainty in their determination will have mechanical system to equation (9) is shown in Fig. 5.
an impact on the estimation of the bending moment. In For an infinite Winkler beam, a concentrated load P
practice, effort should be made to estimate or minimise that creates a bending moment M of the following magnitude at
uncertainty (e.g. parametric studies and intensive field meas- a distance t from the location of the load (after Hetenyi,
urements). 1946):
The embedment depth may significantly affect the bending
moments experienced by a pipe, because the values of Smax P
M exp
t cos t
sin t (10)
and i are functions of depth. However, the normalised 4
bending moment for a given R is not highly affected by the
embedment depth, as seen from the comparison of different The continuous loading due to the soil trough settlement can
depth ratios in Fig. 3. Hence for later comparison with the be replaced by an infinite number of infinitesimal concen-
Winkler solution Z/r0 7 is chosen. It should also be noted trated loads dP(x), the magnitudes of which depend on the
that the results, in the range plotted, were found to be distance from the tunnel centreline, x:
practically independent of the ratio i/r0 when R was chosen dP x KSv xdx (11)
as a non-dimensional controlling parameter. Furthermore, all
elastic analysis values presented in this paper assume a The maximum bending moment in the pipe occurs above the
Poissons ratio
of 0.25, which is an acceptable value for tunnel centreline, and is referred to as the maximum sagging
soil under drained conditions. Nevertheless, it was found that moment. Using equation (10), each of the mentioned con-
the response is not sensitive to the value of Poissons ratio centrated loads contributes the following amount to the
using the current normalisation with Es (e.g. difference in bending moment at x 0:
bending moments of less than 1.5% between
0.25 and

0.5, for the complete range plotted). dP(x)
dM(x) exp (
jxj)[ cos (jxj)
sin (jxj)] (12)
It can be shown that when R approaches zero the bending 4
moment values tend to those obtained by forcing the pipe to
follow the soil (i.e. M
EI  @ 2 Sv =@x 2 ). This can also be
`
`
seen in Fig. 4, which provides bending moments along a
pipe for different values of R. As R increases, the normal-
ised bending moment decreases and highlights the potential
overestimation of bending moments when assuming the pipe
follows the curvature of the soil.
In the following section a closed-form solution for the
Winkler problem under the same assumptions described ear- 1 x 2
sv(x) 5 Smax exp 2
lier is obtained. This solution is later compared with the 2 i
Equal to:
above elastic continuum solution. 1 x 2
P(x) 5 KSmax exp 2
2 i

dP 5 Ksv(x)dx

CLOSED-FORM SOLUTION OF THE WINKLER

PROBLEM dx
The work by Attewell et al. (1986) is often used to
` x `
analyse tunnelpipe interaction. They obtained a numerical
solution for the Winkler problem under the aforementioned
assumptions and used the following differential equation to
represent the pipeline behaviour: Fig. 5. Mechanical representation of equation (9)

Delivered by ICEVirtualLibrary.com to:

IP: 132.68.129.39
On: Mon, 10 Oct 2011 09:37:47
464 KLAR, VORSTER, SOGA AND MAIR
The influence of all infinitesimal concentrated loads is there- THE VALIDITY OF VESICS SUBGRADE MODULUS
fore FOR SOILPIPE INTERACTION
1 The solution for the Winkler system (equations (15) or
KSv (x)
M max exp (
jxj)[ cos (jxj) (16)) requires knowledge of the subgrade modulus K.

1 4 Attewell et al. (1986) suggest the use of the Vesic (1961)

sin (jxj)]dx equation for the subgrade modulus, which is given by
r

1 "
2 # 12 Es B
4
Es
x :
K 1 0 65 (17)
2EI Smax3
exp
x
0:5 EI 1

2
0 i
where B is the width of a beam (in our case 2r0 ). This
3 cos x
sin xdx (13) equation refers to a beam resting on the surface of an
infinite half-space.
Rewriting equation (13), a normalised maximum sagging The physical meaning of this subgrade modulus is as
moment can be defined: follows. If this subgrade modulus is used to define the
1 "
2 # maximum moment in an infinite Winkler beam under a
M max i2 x concentrated load, the moment is computed as
23 i2 exp
x
0:5
EISmax i
:
0
P EI 0 27 0:25
M 0:37Pb 4
1

2 (18)
3 cos x
sin xdx (14) 4 Es b

A closed-form solution for the above equation is feasible where 2b B.

and is equal to As a reference, the Biot (1937) solution for the same
n  h  pi conditions (i.e. concentrated load on an infinite beam) but
M max i2 p 3
2 cos  2 1
2C 2= 2 for elastic continuum is:
EISmax
:
 h EI 0 277 0:277
p io M 0:37Pb 1

2 (19)
sin  2 1
2S(2= 2) : Es b 4

u   u  
The two expressions are practically the same, and equation
2t dt; S(u)  sin 12t 2 dt
1 2
C u  cos (15)
0 0
(19) provides the physical meaning of the Vesic (1961)
equation, which is simply an analogue, essentially allowing a
where  i, and C(u) and S(u) are Fresnel integrals. beam on a Winkler foundation to exhibit similar displace-
Alternatively, the following relation may be used: ments and moments to those of a beam on an elastic
"p foundation when loaded with concentrated loads.
M max i2 X 1
j 4j 2 3 42 j1 2 j! 4 Because the above analogue refers to a beam resting on

1  

EISmax 2 j! 1 4 j! the surface of an infinite half-space, Attewell et al. (1986)
j0
suggested taking twice the value of the Vesic expression, as
p the pipe is buried in the soil, K 2K1 . This value of K
2
5 (16) corresponds to the case where the pipe is buried at infinite
1 2 j! depth, and is not strictly true. However, for a pipeline
The
p limit of the above relations when i tends to zero is affected by tunnelling, it results in a higher estimation of
2 3 (can easily be seen from equation (15)), which is bending moment compared with a more suitable value be-
exactly the solution obtained under the assumption of a tween K1 and 2K1 for the embedment depth in question;
single concentrated load of magnitude P KVL,p where VL hence the use of K 2K1 is considered conservative.
is the volume loss at the pipe level, equal to 2 Smax i. This The basis for creating any analogue is that it should have
means that as the pipe rigidity increases it feels the soil the same input. Vesic (1961) derived equation (17) as an
loading increasingly as a localised loading. analogue on the basis that the two systems (i.e. Winkler and
Figure 6 shows a comparison of the above solution with elastic continuum) are loaded by the same external loads.
the numerical values derived by Attewell et al. (1986) to fit However, in the case of the effect of tunnelling on existing
their suggested solution. General agreement exists between pipelines, the basis for creating an analogue should be an
the numerical values and the current closed-form solution. identical input of greenfield settlement profile. As shown in
Fig. 5, the tunnel effect may be represented by a force
distribution along the pipe that relates to the greenfield
10 Current closed-form solution settlement. Only if this force distribution in both systems
(i.e. Winkler and continuum) is equal will the use of Vesics
08 expression result in identical bending moments in the two
Mmaxi 2/ElSmax

systems. This force distribution, f (x), at the level of the

06
pipe, is equal to that which will cause a greenfield settle-
ment in a pipeless system. In a Winkler system this force
04
distribution is equal to f (x) KSv (x), whereas in the
Attewell et al.: 24 load elements continuum solution, presented in matrix form, it is {f}
02
Attewell et al.: 12 load elements [s]
1 {Sv }, where s is the flexibility matrix of the soil
[s] i, j Gi, j (note: this is different from [*s]). These two
0 05 10 15 20 force distributions are not generally the same, and hence the
1/i Vesic expression might not necessarily be adequate for the
current problem.
Fig. 6. Comparison between closed-form solution and previous For comparison purposes two non-dimensional controlling
solution parameters ( EI=Es r40 and i/r0 ) are considered and varied. In

Delivered by ICEVirtualLibrary.com to:

IP: 132.68.129.39
On: Mon, 10 Oct 2011 09:37:47
 SOILPIPE INTERACTION DUE TO TUNNELLING
3
the elastic continuum solution R r0 =i EI=Es r40 and is a
function of these two parameters. For the Winkler closed-
form solution, it can be shown
 that when using K 2K1

0:27
and
0.25, i 0:813 EI=Es r40 i=r0 , which can
then be substituted into equations (15) or (16) to compute
the maximum bending moment. Fig. 7 shows the comparison
between the normalised bending moment resulting from the
continuum elastic analysis and the Winkler solution using
Vesics expression.
For i/r0 10 the agreement between the two solutions is
quite good, and from a practical point of view they are
identical. However, as i/r0 decreases, the difference in-
creases, resulting in significant differences in values. This
suggests that the Vesic analogue is not necessarily adequate
for all cases, and if the soil is assumed to be linear elastic,
the Winkler solution may not be conservative (i.e. under-
estimation of bending moments). If a more suitable value of
K between K1 and 2K1 is used, as suggested earlier, the
underestimation of bending moment will be even more
pronounced.
It should be noted that the conditions assumed in both the
Winkler and the continuum solutions might deviate from the
exact scenario as the ratio i/r0 becomes small, for the
following three possible reasons:
(a) The simple beam theory would not be accurate for a
case where the extent of deformation is comparable to
the pipe radius.
(b) The pipe is located close to the tunnel, in which case
assumption (c) above may not be justifiable.
(c) The diameter of the pipe is similar to or bigger than
that of the tunnel, in which case assumption (c) again
may not be justifiable.
In practice, where conditions (b) and (c) are met, the ratio
i/r0 may also be smaller than 5, corresponding to the
region in which significant difference between the proposed
continuum and Winkler solutions is noted (Fig. 7). If in this
region the solutions obtained from both the proposed con-
tinuum and Winkler models already differ significantly from
the true elastic solution, it may be argued that neither
solution is satisfactory, and a different approach for solving
the problem should be applied. Furthermore, the comparison
between the two models may be considered irrelevant unless
the true solution is known. Nonetheless, it is of interest to
consider the two possibilities of the deviation trend of the
proposed continuum solution from the accurate one: as the
ratio i/r0 decreases, the proposed solution either (a) under-
estimates or (b) overestimates the true solution. In the
former case the proposed continuum solution provides a
better estimation of the true solution than the Winkler
approach, because the Winkler method underestimates the
solution even more (Fig. 7). In the case of (b) it is not clear
10
08 i/r0 5 5
Mmaxi 2/ElSmax
06
04 i/r0 5 25
02 Winkler solution
3D elastic solution
0
01 1 10 100
EI/Esr 04
Fig. 7. Comparison between continuum solution and Winkler
solution using Vesic analogue
465
which of the two solutions best estimates the true solution.
However, in this case the proposed continuum solution al-
ways provides a conservative estimation compared with the
Winkler and accurate solutions.
Considering the above, if in practice the application of a
more sophisticated approach is not undertaken for cases
matching the three conditions mentioned earlier, it is always
safer to use the proposed continuum method rather than the
Winkler solution, as it will either be closer to the accurate
solution or at least more conservative than the Winkler
solution.
In the field, there is uncertainty in estimating Smax and i.
This uncertainty may overshadow the difference between the
two models. The relative importance of estimating the settle-
ment trough parameters compared with the choice of spring
coefficient in a Winkler model may be illustrated using the
two extreme cases considered earlier for the Winkler model:
(a) As the relative pipe soil rigidity approaches zero, the
bending moment becomes independent of the model,
and is proportional to the curvature of the soil
settlement trough (i.e. / Smax /i2 ).
(b) As the relative pipe soil stiffness increases, the volume
loss and spring coefficient become more important, and
in the limit case, shown earlier, the bending moment is
proportional to both (/ KSmax i).
From these two limits, it is evident that there is a shift in
the significance of estimating the spring coefficient, ranging
from negligible importance at low relative rigidity to the
same importance as Smax and i for very high relative rigidity.
In practice there is no significant advantage in utilising the
accurate spring coefficient if the relative rigidity, R, is small.
On the other hand, if R is large the deviation resulting from
an inaccurate estimation of the spring coefficient may be of
the same order as that resulting from a similarly inaccurate
estimation of i or Smax . This also highlights the importance
of appropriate allocation of resources for correctly estimat-
ing Smax , i (or, more accurately, the shape of the settlement
trough), and the stiffness of the soil as a function of R.
In the following section a revised subgrade reaction
modulus for use in the Winkler system is suggested based
on the analogy of identical greenfield settlement input.
AN ALTERNATIVE ANALOGUE FOR WINKLER
SOLUTION
As the maximum bending moment is often a parameter
that controls the possible pipe damage due to tunnelling
underneath, it was chosen as the entity for the comparison
between the Winkler and the elastic continuum systems. A
subgrade reaction modulus that will result in similar bending
moments in the Winkler and the elastic continuum systems
is proposed here.
In the Winkler system the normalised bending moment is
a function of i, as shown in Fig. 6, whereas in the elastic
i/r0 5 10 continuum it was found to be a function of R. Strictly
speaking, there is an influence of depth Z/r0, but it is
relatively small, as described previously. Itpwas
found
that in
order for the functions to fit closely, i 4 3=R. Rearrange-
ment of this equation leads the subgrade modulus to be
equal to
12Es r0
K (20)
i
The coefficient of subgrade reaction, defined as k K/B (B
1000 10000
2r0 ), is therefore equal to k 6Es /i. Fig. 8 shows the
comparison between the Winkler solution with the subgrade
modulus of equation (20) and the continuum solution as
before. Good agreement exists between the two. Hence it is
Delivered by ICEVirtualLibrary.com to:
IP: 132.68.129.39
On: Mon, 10 Oct 2011 09:37:47
 466 KLAR, VORSTER, SOGA AND MAIR
10 For the current case of a Gaussian settlement trough an
i/r0 5 10 alternative expression for the subgrade modulus was sug-
08 gested to create similar maximum bending moment in the
Winkler and elastic continuum systems.
Mmaxi 2/ElSmax

06

i/r0 5 25
04 i/r0 5 5
ACKNOWLEDGEMENTS
02 The authors wish to express their sincere gratitude to the
Winkler solution British Technion Society for enabling the collaboration be-
3D elastic solution
0 tween the Technion Israel Institute of Technology and
01 1 10 100 1000 10000 Cambridge University. The authors also wish to thank the
EI/Esr 04 Cambridge MIT Institute for sponsoring the project.
Fig. 8. Comparison between continuum solution and Winkler
solution under the new analogue
NOTATION
proposed to use equation (15) or (16) with equation (20) to B width of Winkler beam
EI pipe bending stiffness
compute the maximum sagging bending moment of a pipe
Es Youngs modulus of soil
subjected to tunnelling underneath. It should, however, be {F} force vector representing soil loading on pipe
noted that the current analysis is based on an assumption {f} forces acting on the soil medium
that the soil is linear elastic. In reality soil non-linearity will Gi, j Greens function defining elastic soil continuum displacement
be involved, and this requires further investigation. Neverthe- at point i due to unit loading at point j
less, for small displacement, where elastic behaviour dom- i distance to inflection point of greenfield settlement trough
inates, the elastic continuum solution is still valid. k coefficient of subgrade reaction
K subgrade modulus
[K*] local soil stiffness matrix
CONCLUSIONS K1 Vesics subgrade modulus
M bending moment in pipe
The problem of tunnelling effects on existing pipelines Mmax maximum bending moment in pipe (sagging)
was solved using an elastic continuum method using Mind- P force acting on Winkler system
lins Greens function and a more simplistic Winkler system. R pipe soil rigidity factor
A closed-form solution for the Winkler problem was derived r0 radius of pipe
for a case where the greenfield settlement trough caused by [S] stiffness matrix of pipe
tunnelling is described by a Gaussian curve. Smax maximum settlement in greenfield
The elastic continuum solution was compared with the Sv vertical settlement in greenfield
Winkler solution with a subgrade modulus based on the u pipe displacement
Vesic equation, which was employed by Attewell et al. uC soil continuum displacement
u CA additional displacement at a point due to forces acting at
(1986). It was found that Vesics expression, which was
other places
originally derived to give the same moments and displace- u CAP additional displacement resulting from forces of soil -pipe
ments under a concentrated load in a Winkler system as in interaction
an elastic continuum, is not necessarily adequate for the u CAT additional displacement resulting from the tunnel
problem of tunnelling effects on pipelines, and may not be u CL local continuum displacement, defined as the displacement at
conservative. It should be noted that the significant differ- a point due to its own loading
ence between the two models was observed in a region Z pipe embedment depth
where both models may possibly become inadequate because Winkler model damping factor
of potential violation of the model assumptions, and that a [s] soil flexibility matrix
different approach should be considered for solving the [*s] equal to [s]
[K*]
1

Poissons ratio of soil
problem. In the context of the comparison between the  i, relative soil pipe stiffness
proposed continuum and Winkler models it was noted that if
in practice the application of a more sophisticated approach
is not undertaken, it is always safer to use the proposed
continuum method rather than the Winkler solution, as it REFERENCES
will either be closer to the accurate solution or at least more Attewell, P. B., Yeates, J. & Selby, A. R. (1986). Soil movements
conservative than the Winkler solution. induced by tunnelling and their effects on pipelines and struc-
Although the comparison was made based on a Gaussian tures. London: Blackie & Son.
soil settlement trough, the inherent problems identified with Biot, M. A. (1937). Bending of an infinite beam on an elastic
the use of Vesics expression would also be relevant for foundation. J. App. Mech. Trans. ASME 59, A1A7.
different shape functions. Vesics expression is based on the Hetenyi, M. (1946). Beams on elastic foundation. Ann Arbor, MI:
fact that the Winkler and the continuum systems are loaded University of Michigan Press.
Mindlin, R. D. (1936). Forces at a point in the interior of a semi-
by identical external loads, but in the current problem they
infinite solid. Physics 7, 195202.
are subjected to the same settlement trough, which does not Peck, R. B. (1969). Deep excavation and tunneling in soft ground.
necessarily result in the same loads. Therefore the use of Proc. 7th Int. Conf. Soil Mech. Found. Engng, Mexico City,
Vesics spring coefficient would not generally give identical 266290.
results to that of a continuum solution for settlement-induced Vesic, A. B. (1961). Bending of beams resting on isotropic elastic
loads. solids. J. Eng. Mech. Div., ASCE 87, 3553.

Delivered by ICEVirtualLibrary.com to:

IP: 132.68.129.39
On: Mon, 10 Oct 2011 09:37:47
View publication stats