Vous êtes sur la page 1sur 8

Computers and Geotechnics 55 (2014) 240–247

Contents lists available at ScienceDirect

Computers and Geotechnics


journal homepage: www.elsevier.com/locate/compgeo

Modelling the effect of climate change induced soil settling on drinking


water distribution pipes.
B.A. Wols a,b,⇑, P. van Thienen b
a
KWR Watercycle Research Institute, Groningenhaven 7, 3430 BB Nieuwegein, The Netherlands
b
Delft University of Technology, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: In the future, the more frequent occurrence of severe heat waves and long dry periods due to climate
Received 5 June 2013 change can cause lowering of the ground water level and therefore consolidation of the soil. Conse-
Received in revised form 31 July 2013 quently, increased differential settlements are expected that may damage underground water infrastruc-
Accepted 8 September 2013
ture. Models were developed to assess the impact of differential settlements on pipe failure. The main
Available online 10 October 2013
concept of these models is that the pipe-soil system is schematized as a beam on an elastic foundation
using Winkler type springs. For climate change induced settling, a parametric function of the soil settle-
Keywords:
ment is proposed. A Monte-Carlo analysis has been applied to predict pipe failure probabilities.
Climate change
Soil settlement
Ó 2013 Elsevier Ltd. All rights reserved.
Consolidation
Pipe-soil interaction
Pipe failure
Drinking water distribution systems

1. Introduction Problems with differential soil settlements that cause damage


to (buried) infrastructure have been most widely analyzed for [8]:
The soft soils in the Netherlands make the water distribution
networks vulnerable to soil settlements. In addition, climate  Building of underground tunnels (tunnelling) causing local soil
change is expected to cause more extreme weather events in the displacements [9].
future, such as more severe heat waves and longer periods of  Excavation of building pits causing a deformation of the retain-
drought [1], that may result in increased soil settlements [2]. These ing wall of the building pit and corresponding ground deforma-
settlements, in particular differential soil settlements, associated tions [8].
with soil (property) transitions, cause additional loadings and  Ground water lowering due to excavation works (artificial low-
stresses in the drinking water distribution pipes, increasing the ering of ground water levels) or ground water extraction [10].
probability of pipe failure.
Little research is conducted to the influence of climate (change) For these types of soil settlements, expressions of the settle-
on the structural integrity of drinking water distribution pipes: ment profiles are available. Such an expression is a prerequisite
pipe failure seems to be affected by temperature and soil moisture for a proper analysis of the problem. However, for settling induced
[3], and an increased pipe breakage rate is often observed in dry by ground water lowering due to climate effects, no (differential)
(summer) periods due to soil shrinkage and settlements [4,3,5,6]. soil settlement profile is available.
Stronger effects were observed in more expansive soils [5,6]. More- The effect of tunnelling-induced ground movements on under-
over, there is a lack of models that predict the deterioration of ground infrastructure has attracted growing research attention
water distribution assets induced by climate change [7]. The cur- recently [9,11–17]. In these studies, models have been developed
rent research therefore focuses on developing models for drinking that predict bending moments in the pipe induced by tunnel build-
water distribution pipes that are affected by climate change ing. The main concept of these models is that the pipe-soil system
induced soil settlements. is schematized as a beam on an elastic foundation. So, the interac-
tion between pipeline and soil is handled using an elastic subgrade
reaction foundation model incorporating Winkler type springs [9],
⇑ Corresponding author at: KWR Watercycle Research Institute, Groningenhaven referred to as subgrade modulus. Within this modelling frame-
7, 3430 BB Nieuwegein, The Netherlands. Tel.: +31 (0)30 60 69 604; fax: +31 (0)30 work, various topics were evaluated. The subgrade modulus repre-
60 61 165.
senting the pipe-soil system is originally determined by means of a
E-mail address: bas.wols@kwrwater.nl (B.A. Wols).

0266-352X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compgeo.2013.09.003
B.A. Wols, P. van Thienen / Computers and Geotechnics 55 (2014) 240–247 241

Vesic expression [9]. However, Klar et al. [12] incorporated a more The boundary conditions at both ends of the beam are the soil
rigorous elastic continuum solution and proposed a modified displacement and zero rotation. The mechanical model shown in
subgrade modulus of the soil that is more in-line with the more Eq. (1) in combination with a settlement profile Sv(x) is solved
accurate elastic continuum solution. Wang et al. [16] used different numerically by means of Matlab scripts, using the built-in function
subgrade moduli for pipe uplift and downward movements of bvp4c.m to solve boundary value problems for ordinary differential
pipes. The effect of local yielding of the soil (plastic deformations) equations.
has been assessed by Klar et al. [18], where the local yielding can
be incorporated in a calculation as a (small) correction to the
2.2. Settlement profile
elastic bending moments. The difference between jointed and con-
tinuous pipelines has been evaluated by Klar et al. [13], showing
In tunnelling problems, often the soil displacement is schema-
that a jointed pipeline generally experiences lower bending
tized by a Gaussian displacement curve [12], which is however
moments. The assumption of treating the pipe as a beam rather
not applicable for climate change induced settling. The climate
than a shell has been evaluated by Klar and Marshall [14], which
change induced settling is mainly caused by lowering of the
is allowed if the pipe to soil stiffness is large. Zhang et al. [19]
ground water level, which results in higher grain stresses and
extend the problem to multi-layered soils of both continuous and
therefore a settlement (consolidation). In peat layers, lowering of
jointed pipelines, showing that in layered soils with substantial
the ground water may also result in oxidation of the peat. The
differences in elasticity, treating the soil as homogeneous can
latter will mainly occur at the top layer above the buried water
result in significant errors. Concerning the soil displacement profile
infrastructure and is therefore not considered here. The settling
by the tunnel building, in [11], a modified Gaussian settlement
caused by higher grain stresses can be estimated by consolidation
profile is proposed that better fits experimental results. Validation
theory, for example the formula of Koppejan [23]. More recently,
has been performed in several studies by means of experiments
the a–b–c-isotache model has been developed to calculate soil set-
[11,15,19,20] or by means of FEM models [14,16].
tlement [24]. This formula calculates the soil strain (in 1D over the
Wang et al. [16] developed an empirical estimation for the max-
depth) over time (creep) induced by additional stresses in the soil.
imum bending strain in a pipe induced by a ground settlement.
However, these formulae do not provide analytical expressions for
Pipe stresses can then directly be estimated when settlements,
the lateral distribution of settlements, which are required to ana-
pipe dimensions, soil and material properties and burial depth
lyze the effect of differential settlements along the axial direction
are known.
of the pipe. Differential settlements (i.e. laterally varying) as a re-
In this paper a method is developed for predicting pipe stresses
sult of consolidation may be caused by:
resulting from differential soil settlements, e.g. induced by climate
change. This method is based upon the tunnelling method. A para-
 Lateral variations in ground water levels.
metric function of the soil settlement profile is proposed. More-
 Lateral variations in permeability resulting in different time
over, analogous to Wang et al. [16], an empirical expression is
scales of consolidation.
derived for the climate change induced settlements. This approach
 Lateral variations in soil stiffness.
is further extended towards a Monte Carlo analysis allowing the
prediction of failure probabilities of a pipe subject to differential
Since the soil may consist of many different layers with possibly
settlements.
different soil properties, specific software is required to obtain
realistic settlement profiles. Such software (e.g., Plaxis, D-Settle-
2. Methods ment, etc.) uses a 2D or 3D soil profile known from maps (com-
posed of different soil layers with different stiffnesses, etc.) and a
2.1. Mechanical model ground water profile. Such a profile for a specific situation can be
used as an input for Eq. (1). However, we are also interested in
The following assumptions were made: an analytical expression of the soil displacement profile to perform
parametric studies. Since we did not find any representative
1. Pipe is schematized by a beam. analytical expression in the literature, a simplified displacement
2. Pipe remains in contact with the soil. profile is proposed that assumes a certain transition length
3. Elastic behaviour of pipe material. between soil layers of different stiffness, permeability or ground
4. No internal or external loading on the pipe (q = 0). water level:
5. No joints.   
x
Sv ðxÞ ¼ 0:5Smax 1 þ erf pffiffiffi ; ð2Þ
The schematization of the pipeline behaviour as a beam on elas- i 2
tic grounds is given by (assuming Euler beam theory):
! which describes a smooth transition of differential settlement Smax
2 2
d d uðxÞ over a transition length i. The value of these two parameters may
2
EI 2
þ K  ðuðxÞ  Sv ðxÞÞ ¼ qðxÞ; ð1Þ differ depending on soil composition. By choosing the appropriate
dx dx
values any curvature and transition width of the profile can be
where q(x) represents an external load on the pipe [N/m], which can obtained. The profile is shown in Fig. 1. We also applied another
be traffic loads and ground loads or internal loads. Sv(x) represents settlement profile (based on a hyperbolic tangent function), for
the soil displacement profile and u(x) the pipe displacement [m]. which similar type of results were obtained (results not shown
The parameter EI represents the bending stiffness of the pipe here). In this work, values of i were chosen in the range of
[N m2]. The spring constant of the soil K [N/m2], referred to as sub- 0.5–15 m, which corresponds to a total transition length of
grade modulus, may be a function of x, accounting for transitions in 2–60 m. Larger values of i will not induce any significant bending
soil composition. The external load is set to zero, because an infinite moments in the pipe. The small values of i may typically occur at
long pipe supported by an elastic soil without settlements that is a transition between soil layers or shallow zones of different soil
loaded equally will induce no bending moments. In other types of types (for example shallow clay lenses in a sandy soil). An arbitrary
problems, where erosion causes loss of support, loadings need to soil displacement can also be constructed by adding the individual
be imposed on the beam [21,22]. soil displacement profiles of Eq. (2):
242 B.A. Wols, P. van Thienen / Computers and Geotechnics 55 (2014) 240–247

0
Displacement (m)

−0.02
−0.04 i
−0.06
−0.08
−0.1
−10 −8 −6 −4 −2 0 2 4 6 8 10
x (m)

Fig. 1. Soil displacement profiles used for the suggested settlement profile
(i = 2.5 m, Smax=0.1 m).

N 
X   
ðx  x0;n Þ
Stot ¼ 0:5Smax;n 1 þ erf pffiffiffi : ð3Þ
n¼1 in 2
Fig. 2. Behaviour of soil. Stresses versus strain of pipe material.
where x0,n is the position where the transition occurs. The parame-
ters Smax,n, x0,n and in can be varied to obtain an arbitrary profile.

2.3. Soil and pipe material properties


For the proposed settlement profile (Eq. (2)) this results in (using
the bending moments obtained in Eq. (11) and (12)):
The properties of soils and pipe materials are shown in the Sup-  
porting information. In this work, soils are classified into sand, clay exp  1 Smax D
rp;max ¼ EMnorm pffiffiffiffiffiffiffi 2 2
; ð7Þ
and peat, which is a rough classification based upon the grain size. 8p i
Pipe materials AC and PVC are considered here, which together
where Mnorm (normalized to the maximum bending moment for a
represent roughly 80% of the Dutch drinking water distribution
situation where the pipe follows the ground displacement exactly)
network. The pipe material is modelled as elastic, which is a safe
may vary between 0 and 1 depending on the pipe-soil (interaction)
approach for materials that may also undergo plastic deformation,
stiffness. A structural reliability function is defined by:
such as PVC. Calculation of the thickness of the pipes is also pro-
vided in the Supporting information. Z ¼ rf  rp;max : ð8Þ
where rf represents the yield stress of the pipe material. If Z < 0 fail-
2.4. Subgrade modulus
ure occurs (pipe material starts yielding).
The spring coefficient K representing the pipe-soil interaction is
called subgrade modulus, and is a property of the pipe-soil system. 3. Results
The subgrade modulus is determined by the method described in
Wang et al. [16]. It is a function of burial depth, pipe diameter, soil 3.1. Bending moments with similar pipe and soil displacement
type and properties (see Supporting information). An upward
spring is defined to account for pipe uplift and a downward spring If the soil is very stiff compared to the pipe, the pipe follows the
is determined from the bearing capacity equations for strip foun- ground displacement exactly. The bending moments can be di-
dations. The forces of the pipe acting on the ground may result rectly determined from those displacements:
in collapsing of the ground, as given by the plastic deformation 2
d Sv ðxÞ
part in Fig. 2. Values of subgrade moduli for four different pipe MðxÞ ¼ EI 2
; ð9Þ
diameters are shown in the Supporting information. Softer soils re- dx
sult in lower subgrade moduli. There is a small influence of pipe where Sv represents the soil displacement profile. This will be an
diameter: larger pipe diameter result in larger subgrade moduli. upper bound for the bending moments in the pipe, because in real-
ity the pipe-soil interaction will reduce the pipe curvature. Analyt-
2.5. Failure mechanisms ical expressions of bending moments can be obtained by solving Eq.
(9) for a given soil displacements profile.
The pipe material starts yielding when the maximum stress in For the displacement profile induced by consolidation (Eq. (2)),
the pipe exceeds the yield stress. This is regarded as failure. The the bending moments with equal pipe and soil displacement
stresses in the pipe are calculated in the following manner: First, become:
the curvature (j) can be calculated from the bending moment in  
Mg ðxÞ Smax x x2
the pipe: ¼ 3 pffiffiffiffiffiffiffi exp  2 ; ð10Þ
EI i 2p 2i
Mp
jp ¼ : ð4Þ which will be largest at x/i = ±1:
EI
 
The maximum strain in a cross-section occurs at the outer diameter Mg;max Smax 1
¼  2 pffiffiffiffiffiffiffi exp  : ð11Þ
of the pipe and can be calculated from the curvature: EI i 2p 2
ep ¼ jp D=2: ð5Þ
3.2. Bending moments with pipe-soil interaction
The largest (flexural) stresses occur at the position of the maximum
bending moment. Assuming a linear stress–strain relation (Hooke’s
In all relevant cases, the pipe has some non-zero stiffness, so
law), the maximum stress can now be written as:
that the pipe displacement will only partly follow the ground dis-
M p;max D placement. The pipe-soil interaction will therefore reduce the
rp;max ¼ E : ð6Þ
EI 2 bending moments in the pipe. The calculated bending moments
B.A. Wols, P. van Thienen / Computers and Geotechnics 55 (2014) 240–247 243

in the pipe are normalized with the bending moments when the (Fig. 4). The bending moment can now be estimated directly from
pipe exactly follows the soil displacement: the rigidity without the need for a numerical model, and can there-
fore be used in situations where large numbers of pipe calculations
M p ðxÞ
M norm ðxÞ ¼ ; ð12Þ are required, for example to apply a probabilistic approach or to as-
Mg;max sess a large water distribution network.
where the subscript g represents the ground induced pipe displace-
ment only and p the pipe displacement with pipe-soil interaction. 3.3. Probabilistic approach
The bending moments can be computed from the displacements
2
calculated in the numerical model using MðxÞ ¼ EI d dxuðxÞ
2 . A probabilistic approach has been followed for a pipe without
Settlement profiles of the soil and pipe are shown in Fig. 3. The joints using the empirical expression given by Eq. (14). In the prob-
soil settlement profile exhibits point symmetry around the origin. abilistic approach, the uncertainty in all the parameters that define
Due to the pipe-soil interaction, the pipe displacement has a smal- the reliability function (Eq. (8)) is taken into account, so that the
ler curvature than the soil settlement. The pipe displacement is probability of failure can be calculated:
asymmetric, since the soil subgrade modulus is smaller in upward
Pf ¼ PrðZ < 0Þ: ð15Þ
direction. The bending moments (Fig. 3, lower panel) are zero near
the ends and in the middle. The maximum bending moments in A Monte Carlo method is employed to calculate the probability of
the pipe occur at the positions of maximum curvature where the failure. In this method, the values of each relevant parameter are
pipe moves in downward direction (relative to the soil). Since drawn from a probability density function. The reliability function
the pipe displacement is more gradual than the soil displacement, Z is calculated many times. The probability of failure is then calcu-
the maximum bending moments are lower and occur further away lated from the number of results with Z < 0 divided by the total
from the origin than for the case in which the pipe would follow number of repetitions. The probability density functions, coefficient
the soil displacement exactly. of variation and mean for the relevant parameters are listed in
Analogous to the tunnelling settlement approach [16], a rigidity Table 2. The mean value and coefficient of variation may differ
parameter is introduced that is representative of the relative pipe- per situation and information availability. When no information
soil stiffness: on soil settlement is available, the values in Table 2 in combination
EI with known pipe material, soil type and pipe dimensions can be
R3 ¼ 4
; ð13Þ used. The variation in pipe parameters may vary upon pipe mate-
Ki
rial, pipe diameter and pipe age. The values shown in Table 2 are
where i is the characteristic length of the differential settlement. For obtained from common engineering practices for steel pipes
K the average value of upward and downward subgrade modulus is (elasticity and yield stress for steel were taken from ([25])). Since
taken. A clear relation between R3 and the normalized bending mo- the soil parameters are often unknown or difficult to obtain, a large
ment can be found that includes the relevant physical parameters: variance is chosen for these parameters. A normal distribution is
pipe diameter, burial depth, soil subgrade modulus, pipe elasticity used for the pipe parameters, whereas a lognormal distribution is
and maximum displacement of the soil. In Fig. 4, the model has used for the soil parameters to account for the large variability in
been calculated 10 000 times while changing these parameters. soil parameters.
For each calculation, each of the five parameters was given a value As an example, a Monte Carlo analysis is conducted for an AC
drawn from a log normal probability density function with mean pipe in sand with a diameter of 0.5 m. The random values of each
and standard deviation given in Table 1. A good relation was found parameter drawn from probability density function are shown in
between R3 and the maximum normalized bending moment, which Fig. 5. The resulting distribution of the structural reliability func-
is fitted by the following empirical expression (analogous to Wang tion is shown in Fig. 6. The probability of failure is 0.046. Note that
et al. [16]): this is only due to bending moments from differential settlements,
no other external or internal loadings on the pipe are considered.
1 þ a1 Rb31
M norm ¼ ; ð14Þ A point of attention in a Monte Carlo analysis is the number of
1 þ a2 Rb32 repetitions. Especially for low failure probabilities, a sufficiently
large number of repetitions need to be chosen that cover the failure
where a1, a2, b1 and b2 are fit parameters. Since the ratio of the up-
space of Z. As a rule of thumb, the number of repetitions n should
ward subgrade modulus and downward subgrade modulus as well
be [26]:
as the yielding point vary as a function of soil type (see Supporting
 
information), separate curve fits were made for sand, clay and peat 1
n > 400 1 : ð16Þ
Pf

An example of convergence is shown in the supporting information.


Displacement (m)

0
−0.02
−0.04
−0.06 4. Discussion
−0.08
−0.1 pipe
soil 4.1. Scenario study
0.4
0.3
0.2
A scenario study for typical pipes that are encountered in Dutch
Mnorm

0.1 soils is performed using the numerical model given by Eq. (1). Val-
0
−0.1 ues of normalized bending moments and normalized pipe stresses
−0.2
−0.3
(normalized with the yield stress of the pipe material) are shown
pipe
in Table 3 for (1) different soil types, (2) different pipe materials
−15 −10 −5 0 5 10 15
and (3) different pipe diameters. The pipes are buried at a depth
x (m)
of 1 m. The pipe stresses also increase with pipe diameter, but
Fig. 3. Settlement profiles and bending moments with pipe-soil interaction. Sand, reach a maximum after which they decrease. The more flexible
AC pipe, D = 0.5 m, i = 1.5 m, Smax = 0.1 m, Kd = 6.35 MN/m2, Ku = 1.80 MN/m2. PVC pipes have higher normalized bending moments than AC
244 B.A. Wols, P. van Thienen / Computers and Geotechnics 55 (2014) 240–247

Fig. 4. Normalized maximum bending moments as a function of rigidity R3 while varying pipe diameter, settlement length, burial depth, pipe elasticity and soil displacement.

Table 1 trends as the pipe stresses calculated by the deterministic


Variation of pipe and soil parameters. approach (compare for example the maximum stresses and the
Parameter Symbol Distribution cV l probability of failure for pipes in peat). An explanation is that the
+
mean value of the structural reliability calculated from all the
Diameter D Log normal 0.2 0.4 m Monte Carlo repetitions differs from the value of the structural reli-
Elasticity pipe material E Log normal 0.5 1.05  1011 Pa ability function where all the mean values of each parameter are
Burial depth H Log normal 0.2 1.0 m applied (which is basically the deterministic approach). The prob-
Transition length i Log normal 0.5 5m
ability of failure that is solely induced by differential settlements is
Maximum differential Smax Log normal 0.1 0.25 m
settlement between 0.003% for PVC in the softest soil up to 52% for AC in the
stiffest soil.
+
Coefficient of variation (cV ¼ rl with standard deviation r and mean l).

4.2. Sensitivity study

The sensitivity of the model to soil settlement parameters and


Table 2
pipe diameter has been investigated using the empirical fitted
Parameters used in Monte Carlo analysis.
model (expression (14)). The pipe stresses are linearly related to
Parameter Symbol Distribution cV+ l the differential settlement (Smax), since Smax has no influence on
Diameter D Normal 0.004 Var Mnorm, and Mg,max is linearly dependent on Smax (Eq. (11)). Other
Thickness wall t Normal 0.04 Var parameters do change Mnorm: the effect of pipe diameter and set-
Elasticity pipe material E Normal 0.04 See SI
tlement transition length is shown for various pipe materials and
Yield stress pipe material rf Normal 0.08 See SI
Soil subgrade modulus K Log normal 0.5 see SI soil types in Fig. 8. Increasing the diameter makes the pipe stiffer.
Transition length i Log normal 0.5 5m As a result, the maximum bending moment as the pipe would
Maximum differential settlement Smax Log normal 0.25 0.1 m follow exactly the ground displacement increases (Fig. 8a), since
Emperical fit parameters a, b Normal see Fig. 4 the stiffer pipe requires a larger force to follow the displacement
+
Coefficient of variation (cV ¼ rl with standard deviation r and mean l). of the soil. However, the pipe will also become stiffer in relation
to the soil stiffness (increase of rigidity), so that the pipe is less
affected by the soil movement resulting in lower normalized bend-
pipes, because the soil displacement is more easily followed. ing moments (Fig. 8b). The net result is shown in the maximum
However, the normalized stresses are lower than for AC due to stresses in the pipe that depend on both the normalized bending
the lower elasticity of PVC. Also, for the softer soils the bending moment and pipe diameter (Fig. 8c). When increasing the pipe
moments and stresses are lower, because the pipe curvature is diameter, stresses first increase, then reach a maximum, after
smaller. Note that the imposed displacement is the same for all soil which they decrease. The position of this maximum depend on soil
types, whereas in reality the softer soils are more likely to settle. type and material type.
For these typical drinking water distribution pipes buried in the Increasing the transition length of the soil displacement results
soil, stresses caused by a differential settlement of 10 cm over a in lower maximum bending moments for a situation when the pipe
length of 10 m may increase by 5–135% of the material’s yield would follow exactly the soil displacement (Fig. 8d), since the dis-
stress. placements are more gradual. However, since the pipe can follow a
In Fig. 7, the normalized bending moments are plotted as a lower soil curvature more easily, the normalized bending moments
function of rigidity R3 for the different soil types, pipe materials will increase (Fig. 8e). The largest pipe stresses will occur at the
and pipe diameters. The fitted relation between bending moment lowest values of i (Fig. 8f). This maximum also depends on pipe
and rigidity (Eq. (14)) is also shown, confirming that the empirical material and soil type.
fit works for different pipe materials and soil types. Therefore, the
empirical function has been used in the Monte Carlo analysis to 4.3. Arbitrary soil displacement
calculate the failure probability (Table 3). The Monte Carlo analy-
ses use the probability density functions as presented in Table 2, Instead of Eq. (3), any soil displacement profile can be used, for
where the mean and coefficient of variation for i is changed to example a displacement profile obtained from field studies, or
2.5 m and 0.1 for this specific scenario. The failure probabilities cal- advanced soil deformation models. In Fig. 9, an example of an
culated by the Monte Carlo analysis do not always follow the same arbitrary soil displacement profile is shown, which has been
B.A. Wols, P. van Thienen / Computers and Geotechnics 55 (2014) 240–247 245

−7
x 10

2
150
300
1.5

100
200
φ

φ
1

100 50
0.5

0 0 0
1 2 3 4 7
0.01 0.02 0.03 0.04 0.05 0.2 0.4 0.6 0.8
x 10
σf (Pa) t (m) D (m)

−10 −7
x 10 x 10

4 30

6
3
20
4
φ

φ
2

10
1 2

0 0 0
1 2 3 4 2 4 6 8 10 12 0.05 0.1 0.15 0.2
10
x 10 6
E (Pa) K (Pa) x 10 Smax (m)

0.6

0.4
φ

0.2

0
5 10 15
i (m)

Fig. 5. Probability density functions of parameters used in Monte Carlo analysis, number of repetitions: 1  108. Sand, AC pipe.

x 10
−8 4.4. Validation
8
The Winkler type of model that has been used for tunnelling
7 problems is extensively validated by measurements [11,15,19,20]
and FEM calculations [14,16]. Therefore, it is expected that the
6
method proposed in this paper, which is also a Winkler type of
5 model, will be accurate. The main assumption is the differential
settlement profile, which has not been validated by measured data.
φ

4 Despite the lack of validation, the settlement profile is chosen


parametrically, resulting in a generic profile that can be adjusted
3
with parameters Smax and i. Moreover, the probabilistic approach
2 also accounts for the uncertainty associated with this problem by
introducing probability density functions. Wider distributions can
1
be chosen for parameters that are uncertain, such as settlements
0
and/or soil parameters. When there is a large degree of uncertainty
−2 −1 0 1 2 3 4 in the parameters, the method is able to provide a first rough esti-
7
Z (Pa) x 10 mation of the stresses in a buried pipeline. For more accurate pre-
dictions, detailed information on site specific parameters, such as
Fig. 6. Probability density function of structural reliability function Z, number of
soil composition and expected settlements, is required.
repetitions: 1  108. Sand, AC pipe.

5. Conclusions
obtained from 50 random individual displacement profiles. The
parameters were drawn from uniform probability density In the future, the more frequent occurrence of severe heat
functions. waves and long dry periods due to climate change can cause
246 B.A. Wols, P. van Thienen / Computers and Geotechnics 55 (2014) 240–247

Table 3
Bending moments, stresses and probability of failure for a buried pipe. Maximum differential settlement is 0.1 m over a characteristic distance i of 2.5 m.

Soil Pipe Max. Mnorm Max. norm. stresses Probability of failure


D (m) D (m) D (m)
0.10 0.20 0.50 1.00 0.10 0.20 0.50 1.00 0.10 0.20 0.50 1.00
Sand PVC 1.00 1.00 0.99 0.93 0.05 0.12 0.23 0.43 4.0E5 9.9E4 2.0E2 7.6E2
Sand AC 1.00 0.99 0.91 0.70 0.19 0.48 0.88 1.35 1.7E2 9.8E2 3.6E1 5.2E1
Clay PVC 1.00 1.00 0.94 0.72 0.05 0.12 0.22 0.33 3.0E5 5.4E4 6.1E3 1.6E2
Clay AC 1.00 0.92 0.68 0.40 0.19 0.45 0.66 0.78 1.3E2 6.1E2 1.6E1 2.0E1
Peat PVC 1.00 0.95 0.73 0.45 0.05 0.11 0.17 0.21 1.2E5 1.0E4 5.5E4 7.8E4
Peat AC 0.97 0.70 0.42 0.21 0.19 0.34 0.40 0.40 5.4E3 1.6E2 2.4E2 1.9E2

Sand Clay Peat


1.2
PVC
AC
1
fit

0.8
M norm

0.6

0.4

0.2

0
−5 −4 −3 −2 −1 0 1 2 −5 −4 −3 −2 −1 0 1 2 −5 −4 −3 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
R3 R3 R3

Fig. 7. Normalized maximum bending moments as a function of rigidity R3 for different soil types, pipe materials and pipe diameters.

(a) (b) 1.2 (c) 1.2


1 1

105 0.8 0.8


Mmax (Nm)

Mnorm

σnorm

0.6 0.6

0.4 0.4
0
10 0.2 0.2

0 0
0 0.5 1 1.5 0 0.5 1 1.5 0 0.5 1 1.5
D (m) D (m) D (m)

(d) (e) (f) AC, sand


1 1 PVC, sand
AC, clay
0.8 0.8 PVC, clay
105
Mmax (Nm)

AC, peat
Mnorm

σnorm

PVC, peat
0.6 0.6

0.4 0.4

100 0.2 0.2

0 0
0 5 10 15 0 5 10 15 0 5 10 15
i (m) i (m) i (m)

Fig. 8. Normalized bending moments and pipe stresses as a function of pipe diameter (D) and transition length (i). Sand, Smax = 0.1 m, i = 2.5 m when D is varied and D = 0.5 m
when i is varied.

lowering of the ground water level and therefore consolidation of of the complexity of the system was developed, but also analytical
the soil. Consequently, increased differential settlements are models for more simplified systems. The most simple analytical
expected that may damage underground water infrastructure. model is obtained when the pipe follows the same displacement
Several models were developed to assess the impact of differential as the soil. However, this assumption is not realistic for most bur-
settlements on pipe failure. A numerical model that contains most ied pipelines and pipe-soil interaction needs to be incorporated.
B.A. Wols, P. van Thienen / Computers and Geotechnics 55 (2014) 240–247 247

Appendix A. Supplementary data


Displacement (m)

0.03
0.02
0.01 Supplementary data associated with this article can be found,
0
in the online version, at http://dx.doi.org/10.1016/j.compgeo.
−0.01
−0.02
pipe 2013.09.003.
soil

0.05 References
0.02
M norm

−0.01 [1] van den Hurk B, Tank AK, Lenderink G, van Ulden A, van Oldenborgh GJ,
−0.04 Katsman C, et al. Knmi climate change scenarios 2006 for the netherlands. Tech
−0.07 rep WR 2006-01; KNMI; 2006.
pipe
[2] Arnold G, Bos H, Doef R, Goud R, Kielen N, Luijn Fv. Water management in the
−25 −20 −15 −10 −5 0 5 10 15 20 25 netherlands. Tech rep WD0111VV007B; Ministry of Infrastructure and
x (m) Environment; 2011.
[3] Kleiner Y, Rajani B. Forecasting variations and trends in water-main breaks. J
Infrastruct Syst 2002;8(4):122–31.
Fig. 9. Effect of arbitrary soil displacement. Sand, D = 0.5 m, Kd = 6.35 MN/m2,
[4] Newport R. Factors influencing the occurrence of bursts in iron water mains.
Ku = 1.80 MN/m2. Water Supply Manage 1981;3:274–8.
[5] Hu Y, Hubble DW. Factors contributing to the failure of asbestos cement water
mains. Can J Civil Eng 2007;34(5):608–21.
[6] Gould SJF, Boulaire FA, Burn S, Zhao XL, Kodikara JK. Seasonal factors
For a continuous pipeline in a homogeneous soil, the bending influencing the failure of buried water reticulation pipes. Water Sci Technol
moments in the pipe can be accurately determined by a fitted func- 2011;63(11):2692–9.
[7] UKWIR. Impact of climate change on asset management planning. Tech rep 12/
tion, so that the pipe stresses induced by a soil settlement can be CL/01/16; UKWIR; 2012.
calculated via an analytical model. Using this analytical model, a [8] Netzel H. Building response due to ground movements. PhD thesis; Delft
probabilistic approach can be followed to obtain a probability of University of Technology; Delft; 2009.
[9] Attewell P, Yeates J, Selby A. Soil movements induced by tunnelling and their
pipe failure, while accounting for the uncertainty in the relevant
effects on pipelines and structures. Glasglow (UK): Blackie and Son Ltd.; 1986.
soil and pipe parameters. For more complex situations, when the [10] Preene M. Assessment of settlements caused by groundwater control. Proc Inst
soil deformation becomes plastic; when a lateral transition Civil Eng: Geotech Eng 2000;143(4):177–90.
[11] Vorster TEB, Klar A, Soga K, Mair RJ. Estimating the effects of tunneling on
between soils occur; or when an arbitrary soil displacement is
existing pipelines. J Geotech Geoenviron Eng 2005;131(11):1399–410.
imposed, the numerical model needs to be used. [12] Klar A, Vorster TEB, Soga K, Mair RJ. Soil-pipe interaction due to tunnelling:
Jointed pipelines were not considered in this work, however comparison between Winkler and elastic continuum solutions. Geotechnique
joints can be easily added to the numerical model by locally reduc- 2005;55(6):461–6.
[13] Klar A, Marshall AM, Soga K, Mair RJ. Tunneling effects on jointed pipelines.
ing the bending stiffness of the pipe, as suggested by [19]. Using an Can Geotech J 2008;45(1):131–9.
analytical (fitted) model and probabilistic approach for jointed [14] Klar A, Marshall AM. Shell versus beam representation of pipes in the
pipelines will be a topic of future research. Pipe failure predictions evaluation of tunneling effects on pipelines. Tunnel Underground Space
Technol 2008;23(4):431–7.
may also improve when more representative soil displacement [15] Marshall AM, Klar A, Mair RJ. Tunneling beneath buried pipes: view of soil
profiles are found (e.g. from soil mechanical models) and applied. strain and its effect on pipeline behavior. J Geotech Geoenviron Eng
Such a displacement function can be easily implemented in the 2010;136(12):1664–72.
[16] Wang Y, Shi J, Ng CWW. Numerical modeling of tunneling effect on buried
current modelling framework. pipelines. Can Geotech J 2011;48(7):1125–37.
A sensitivity study shows that there is a typical pipe diameter [17] Wang Y, Wang Q, Zhang KY. An analytical model for pipe-soil-tunneling
for which the pipe stresses are maximal, depending on pipe mate- interaction. Proc Eng 2011;14:3127–35.
[18] Klar A, Vorster TEB, Soga K, Mair RJ. Elastoplastic solution for soil-pipe-tunnel
rial and soil type. A similar observation has been found for the
interaction. J Geotech Geoenviron Eng 2007;133(7):782–92.
distance over which the differential settlements occur. [19] Zhang C, Yu J, Huang M. Effects of tunnelling on existing pipelines in layered
For a displacement of 0.10 m over a length of 10 m, pipe stresses soils. Comput Geotech 2012;43(2):12–25.
[20] Shi J, Wang Y, Ng CWW. Buried pipeline responses to ground displacements
ranging from 5% to 135% of the yield stress may occur for AC and
induced by adjacent static pipe bursting. Can Geotech J 2013;50(5):481–92.
PVC pipes that are placed in typically Dutch soils, resulting in a [21] Rajani B, Zhan C, Kuraoka S. Pipe-soil interaction analysis of jointed water
probability of failure ranging from 0.003% to 52%. mains. Can Geotech J 1996;33(3):393–404.
[22] Rajani B, Tesfamariam S. Uncoupled axial, flexural, and circumferential pipe-
soil interaction analyses of partially supported jointed water mains. Can
Geotech J 2004;41(6):997–1010.
Acknowledgements [23] Koppejan A. A formula combining the Terzaghi load compression relationship
and the Buisman secular time effect. In: 2nd International conference on soil
mechanics and foundation engineering; 1948. p. 32–7.
This study was carried out in the framework of both of the
[24] Den Haan E. Vertical compression of soils. PhD thesis; Delft University of
Dutch national research program Knowledge for Climate (http:// Technology; Delft, The Netherlands; 1994.
knowledgeforclimate.climateresearchnetherlands.nl/) and the [25] Vrouwenvelder ACWM, Siemes AJM. Probabilistic calibration procedure for the
joint research program of the Dutch Water Utility sector (http:// derivation of partial safety factors for the Netherlands building codes. Heron
1987;32(4):9–29.
www.kwrwater.nl/BTO/). The authors would also like to thank [26] CUR. Kansen in de civiele techniek, deel 1: probabilistisch ontwerpen in
Henk Kruse for fruitful discussions. theorie. Tech rep CUR report 190; Stichting CUR; 1997.

Vous aimerez peut-être aussi