Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
a r t i c l e i n f o a b s t r a c t
Article history: There are various analytical, empirical and numerical methods to calculate groundwater inflow into tun-
Available online 2 May 2017 nels excavated in rocky media. Analytical methods have been widely applied in prediction of groundwa-
ter inflow to tunnels due to their simplicity and practical base theory. Investigations show that the real
Keywords: amount of water infiltrating into jointed tunnels is much less than calculated amount using analytical
Groundwater inflow methods and obtained results are very dependent on tunnel‘s geometry and environmental situations.
Analytical equation In this study, using multiple regression analysis, a new empirical model for estimation of groundwater
Multiple regression analysis
seepage into circular tunnels was introduced. Our data was acquired from field surveys and laboratory
Stepwise algorithm
Tunnel
analysis of core samples. New regression variables were defined after perusing single and two variables
relationship between groundwater seepage and other variables. Finally, an appropriate model for estima-
tion of leakage was obtained using the stepwise algorithm. Statistics like R, R2 , R2P and the histogram of
residual values in the model represent a good reputation and fitness for this model to estimate the
groundwater seepage into tunnels. The new experimental model was used for the test data and results
were satisfactory. Therefore, multiple regression analysis is an effective and efficient way to estimate
the groundwater seepage into tunnels.
Ó 2017 Published by Elsevier B.V. on behalf of China University of Mining & Technology. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
http://dx.doi.org/10.1016/j.ijmst.2017.03.005
2095-2686/Ó 2017 Published by Elsevier B.V. on behalf of China University of Mining & Technology.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
416 H. Farhadian, H. Katibeh / International Journal of Mining Science and Technology 27 (2017) 415–421
parts of them drilled in discontinuous rock masses, used for intro- The general purpose of multiple regression is to learn more
ducing an empirical equation using multiple regression analyses. about the relationship between several independent or predictor
By examining the relation between the effective variables on variables and a dependent or criterion variable. In general, multiple
groundwater flow into tunnels and considering each of the analyt- regression procedures will estimate a linear equation of the form.
ical equations’ parameters for seepage into tunnel in various media
(water table, permeability coefficient, radius of the tunnel, the Y ¼ b0 þ b1 X 1 þ b2 X 2 þ þ bn X n þ e ð1Þ
amount of overburden) and regression analysis, a new empirical
model is presented to calculate groundwater seepage into circular where e is the error rate of model. The regression coefficients or ‘b
tunnels. terms’ represent the independent contributions of each indepen-
dent variable to the prediction of the dependent variable. The
regression line expresses the best prediction of the dependent vari-
2. Analytical relations of groundwater seepage flow into tunnel able ðYÞ, given the independent variables ðXÞ. However, there is
and their validity limits usually substantial variation of the observed points around the fit-
ted regression line. The deviation of a particular point from the
Groundwater inflow depends on a number of factors, such as regression line (its predicted value) is called the residual value
the permeability of the rock mass, the groundwater table, the aper- [42,43].
ture of rock fractures, and the size of the excavation. Groundwater
inflow equations are based on Darcy’s Law and conservation of
mass [37]. Analytical methods considering the parameters such 3.2. Fitness criteria
as rock mass permeability, water table height above tunnel axis,
and tunnel radius, is used for estimation of groundwater seepage One of the points which should be considered about regression
rate into tunnels. Table 1 shows the analytical equations used to models is the assumptions that are taken into account errors. These
estimate seepage flow to tunnels. Fig. 1 shows the applied param- assumptions are: (1) the mean of errors ðeÞ is equal to zero, (2) the
eters in the equations presented in Table 1. The presented analyt- error variances ðr2 Þ is constant, (3) errors are uncorrelated, and (4)
ical equations are not valid under these conditions: the vertical errors are normally distributed. The last one is essential for testing
seepage flow towards tunnel, bedding variation in rock around the assumptions and obtaining the confidence limit. It should be
the tunnel, and inexact determination of the rock mass permeabil- noted that for assessing the fitness of the final inflow model, valid-
ity [38]. ity of the assumptions should be evaluated and analysis must be
directed to testing of the model well-fitness. Extreme deviations
from the assumptions can lead to unstable models. Normally, these
3. Materials and methods deviations are not obvious with the standard tabloid statistics such
as T or the F-statistics and R2 . These are general properties which
3.1. Multiple regression analysis
do not guarantee a good final model [44].
Table 1
Analytical equations for prediction of groundwater seepage into tunnels (H0 = distance between tunnel center and water table; h = hydraulic head into the tunnel; z = overburden;
r = tunnel radius; K = hydraulic conductivity; D = hydraulic load above load surface and Q = water inflow to tunnel).
Karlsrud [16] Q ¼ 2pK ln h A combination of above mentioned equations, according to field observations, is edited for reducing error in deep and
ð2hr1Þ
shallow tunnels (under water table)
Lombardi [19] Q ¼ 2pK h In this equation, Karlsrud method has been corrected with application of exact conditions
r ð1þ0:4h Þ
r2
ln2h
El-Tani [21] Q ¼ 2pK kk2 1 In this equation El-Tani has applied Mobius transformation method and fourier series and presented a new analytical
2
h
þ1 lnk
2 1=2
solution for flow calculation, in which k ¼ ðh=rÞ ððh =r2 Þ 1Þ
2pKLðAþDÞ
Park et al. [25] Q¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi In this equation: Water table is above the land surface. Hydrostatic load along the tunnel border depending on the
ðHDÞ ðHDÞ2
ln r þ r2
1 stage is varying
Li et al. [39] Q ¼ KðS þ C:HÞ In this equation, S and C are coefficient related tot he tunnel‘s shape and depth
H. Farhadian, H. Katibeh / International Journal of Mining Science and Technology 27 (2017) 415–421 417
variable selection subject include: construction of subset models (1) forward, (2) backward and (3) stepwise regression which is a
and selection of the best one. combination of the two previous methods [46].
Appropriate benchmark for a regression model that has been
widely used is the coefficient of multiple determinations ðR2 Þ. Sup-
4. Analysis and discussion
pose that R2p is the coefficient of multiple determinations for a sub-
set regression model of p term, namely p 1 regression variables 4.1. Introducing the variables
and the intercept is b0 . An alternate formula for R2 is
In this study, the data and site characteristics of Amirkabir,
SSR ðpÞ SSE ðpÞ
R2p ¼ ¼1 ð2Þ Ghomroud and Nosoud Tunnels are used (Fig. 2). At different sta-
Syy Syy tions along the tunnels, the borehole cores had been obtained
The SSR ðpÞ, SSE ðpÞ and Syy represent the sum of squares regres- and data from laboratory measurements were collected. Also,
sion, the residual sum of squares and the total sum of squares for raw and processed data from joints survey and lugeon tests were
a subset model with p term respectively. This equation shows that available. In addition, groundwater inflow from these tunnels
had been recorded exactly. Table 2 presents the main parameters
adding predictors to a regression will always increase R2 as the p
of different sections in the studied tunnels for estimation of
will decrease. Therefore, analyst adds regression variables to the
groundwater inflow.
model until when by adding a new variable to the model, R2p
Amirkabir water conveyance tunnel is detailed as follows: the
increases only slightly [45]. 30 km Amirkabir tunnel is located in Alborz mountain range in
Iran. The tunnel is located in the Karaj Formation consisting of
3.3.1. Adjusted coefficient of multiple determinations complex sedimentary and volcanic rocks such as monzodiorite,
To avoid misinterpretation from R2 , some analysts prefer to use gabro, tuff, sandstone and conglomerate limestone. Amirkabir tun-
the adjusted R2 ðR2p Þ statistic which for a subset p, is defined as nel was operated to transfer water from Amirkabir dam to Tehran
[47–49].
follows:
Ghomroud water conveyance tunnel is demonstrated as
Table 2
Main parameters of the different sections in the studied tunnels for estimation of the water flow using multiple regression analyses.
Fig. 4. Histogram of residual values for estimated training data from the Eq. (4).
The correlation coefficients between groundwater flow and the rates in the tunnels for which their domain is presented in Table 2),
other variables show a relatively weak relationship. It is empha- approximately 30% was selected randomly for testing the model
sized that a simple model cannot be a good solution for analysis and the rest were used for network training.
of groundwater seepage into tunnel. Therefore it is inevitable to
use multivariate models. From the available data (measured inflow 4.2. Determination of single-and two-variable relationship between
groundwater seepage and other variables
At this stage, all the relations obtained from the previous stage
were defined as new variables. So, using the stepwise algorithm,
the appropriate regression equation between the four primary
variables and several new variables can be obtained. F IN statistics
for input variables in the model of the 90% confidence level is 0.1
and F OUT to discard it from the equation, because it was smaller
Fig. 5. Trend of predicted water inflow (Q p ) versus observed water inflow (Q o ) for than the F IN is 0.15 was considered. The final regression model is
the test data from the Eq. (4). obtained as follow:
Fig. 6. Comparison between observed and predicted groundwater inflow for test data.
420 H. Farhadian, H. Katibeh / International Journal of Mining Science and Technology 27 (2017) 415–421
!
1:217 0:741 0:739r model represent a good correlation to observed groundwater
Q ¼ 0:004 þ 200Khp 2z þ 2
ð4Þ seepage into tunnels.
ln 2h
r
ln r ðlnzÞ
(4) The proposed empirical model for estimating groundwater
seepage into tunnel is used for the test data and results
R ¼ 0:963; R2 ¼ 0:927; R2p ¼ 0:924 are satisfactory. Therefore, the above-mentioned
The unit of the proposed model (Q ) is (L=s=m). methodology is an effective and efficient way to estimate
the groundwater seepage into tunnel.
The statistic, R2 and in the model Eq. (4) are in an acceptable
range and from this perspective, the above-model is considered
valid. Trend of predicted water inflow (Q p ) versus observed water
inflow (Q o ) in Fig. 3 indicates the validation of the model. In Fig. 3, Acknowledgements
the training data (predicted) are placed in front of their actual
values. The authors would like to thank Institute of SCE and
In Fig. 4, histogram of residual values for estimating training Mahab Ghods Consultant Engineering Company, especially
data in this model has been plotted. As can be seen, the histogram Dr. J. Hasanpour, Dr. A. AaliAnvari and his colleagues in the
shows a zero mean, variance close to one and the normal distribu- Tunneling Division for their help in the collection of required data.
tion which firmly insists a proper fit of the model.
References
4.4. Model validation
[1] Ma R, Li G, Zhang N, Liu C, Wei Y, Zhang M. Analysis on mechanism and key
factors of surrounding rock instability in deeply inclined roadway affected by
After acquiring the model of teaching, the accuracy of this model argillation and water seepage. Int J Min Sci Technol 2015;25(3):465–71.
was analyzed using the 30% test data which were selected ran- [2] Schmidt B. Potential topics for research and development in underground
engineering. Tunn Undergr Sp Tech 1991;6(3):283–6.
domly. Fig. 5 shows the results of predicted water inflow (Q p ) ver- [3] Sievanen U. Leakage and groutability. Working report; 2001.
sus observed water inflow (Q o ) for test data. As can be seen, the R2 [4] Coli M, Pinzani A. Tunnelling and hydrogeological issues: a short review of the
current state of the art. Rock Mech Rock Eng 2014;47(3):839–51.
have an acceptable range and therefore, the above model is consid- [5] Nikvar Hassani A, Katibeh H, Farhadian H. Numerical analysis of steady-state
ered valid. Fig. 6 shows comparison between the estimated values groundwater inflow into Tabriz line 2 metro tunnel, northwestern Iran, with
of groundwater seepage with flow rate observed on the test data. special consideration of model dimensions. Bull Eng Geol Environ
2015;31:161–71.
As can be seen, in sections where groundwater seepage into the
[6] Farhadian H, Katibeh H. Effect of model dimension in numerical simulation on
tunnel is high, there are significant differences between these two assessment of water inflow to tunnel in discontinues rock. World Acad Sci, Eng
values and this is probably due to the differences in scale between Technol, Int J Environ, Ecol, Geol Geophys Eng 2015;9(4):257–60.
[7] Chen R, Tonon F. Rock fracture simulation for groundwater prediction in rock
the joints has been measured in laboratory and real media. But
tunnels. In: Proceedings of rapid excavation and tunneling conference (RETC).
trends change is considerably identical between predicted seepage San Francisco: USA; 2011.
into tunnel using this model and observed flow. [8] Muskat M. The flow of homogeneous fluids through porous media. New
The proposed empirical model can be used for prediction of York: McGraw Hill; 1937.
[9] Polubarinova KPA. Theory of ground water movement. New Jersy: Princeton
groundwater flow into tunnel where the following conditions University; 1962.
exist: [10] Goodman R, Moye D, Schalkwyk A, Javendel I. Groundwater inflow during
tunnel driving. Eng Geol 1965;1:150–62.
[11] Zhang L, Franklin JA. Prediction of water flow into rock tunnels: an analytical
(1) Tunnel is excavated in rocky media and predominant rock solution assuming an hydraulic conductivity gradient. Int J Rock Mech Min Sci
layers are tuff, limestone and sandstone. Geomech Abstr 1993;30(1):37–46.
(2) The radius of the tunnel is in the range of 2.3–2.8 m. [12] Fernández G, Alvarez T. Seepage-induced effective stresses and water
pressures around pressure tunnels. J Geotech Eng 1994;120(1):108–28.
(3) Rock mass permeability of the tunnel site is between [13] Heuer RE. Estimating rock-tunnel water inflow. In: Proceeding of the rapid
5 108 m/s and 2.36 106 m/s. In other words, applica- excavation and tunneling conference.
tion of this model for tunnels which are drilled in faults [14] Heuer RE. Estimating rock tunnel water inflow II. In: Proceedings of rapid
excavation and tunnelling conference.
and crashed zones with high level of fracture density is not
[15] Lei S. An analytical solution for steady flow into a tunnel. Ground Water
proposed. 1999;37:23–6.
(4) Water inflow regime is laminar. [16] Karlsrud K. Water control when tunneling under urban areas in the Olso
region. NFF Publ 2001;12(4):27–33.
[17] Raymer JH. Predicting groundwater inflow into hard-rock tunnels: estimating
5. Conclusions the high-end of the permeability distribution. In: RETC proceedings, vol. 83;
2001. p. 1027–38.
Based on the results obtained from this study, it is possible to [18] Raymer JH. Groundwater inflow into hard-rock tunnels: a new look at inflow
equations. In: Rapid excavation and tunneling conference. Seattle; 2005. p.
make the following conclusions: 457–68.
[19] Lombardi G. Private communication; 2002.
(1) Many equations have been used to estimate groundwater [20] El Tani M. Water inflow into tunnels. In: Proceedings of the world tunnel
congress ITA-AITES. Balkema: Oslo; 1999. p. 61–70.
seepage into tunnel. Most of these models are theoretical- [21] El Tani M. Circular tunnel in a semi-infinite aquifer. Tunn Undergr Sp Tech
experimental and have been used in water transfer tunnels. 2003;18:49–55.
In this study, the amount of groundwater seepage into rock [22] El Tani M. Helmholtz evolution of a semi-infinite aquifer drained by a circular
tunnel. Tunn Undergr Sp Tech 2010;25:54–62.
tunnels is obtained using multivariate regression analysis. [23] Cesano D, Bagtzoglou AC, Olofsson B. Quantifying fractured rock hydraulic
(2) A relationship between groundwater seepage into tunnel heterogeneity and groundwater inflow prediction in underground
and hydrogeological data using multiple regression analyses excavations: the heterogeneity index. Tunn Undergr Sp Tech 2003;18:19–34.
[24] Perrochet P, Dematteis A. Modeling transient discharge into a tunnel drilled in
was developed.
heterogeneous formation. Ground Water 2007;45(6):786–90.
(3) Perusing one and two variable relationship between ground- [25] Park K, Owatsiriwong A, Lee JG. Analytical solution for steady-state
water seepage and other variables, new regression variables groundwater inflow into a drained circular tunnel in a semi-infinite aquifer:
a revisit. Tunn Undergr Sp Tech 2008;23:206–9.
were defined. An appropriate model Eq. (4) is obtained for
[26] Farhadian H, Katibeh H, Huggenberger P. Empirical model for estimating
groundwater seepage using stepwise algorithm. Statistics groundwater flow into tunnel in discontinuous rock masses. Environ Earth Sci
of R and R2, and the histogram of residual values in the 2016;75(6):1–16.
H. Farhadian, H. Katibeh / International Journal of Mining Science and Technology 27 (2017) 415–421 421
[27] Farhadian H, Katibeh H, Huggenberger P, Butscher C. Optimum model extent [39] Li P, Zhang D, Zhou Y. Study on prediction methods and its influence factors of
for numerical simulation of tunnel inflow in fractured rock. Tunn Undergr Sp water inflow into tunnels. J Beij Jiaot Univ 2010;34(4):11–5.
Tech 2016;60:21–9. [40] Wu T. Permeability prediction and inage capillary pressure simulation in
[28] Farhadian H, Katibeh H, Sayadi AR. Prediction of site groundwater rating (SGR) sandstone reservoir. Texas: Texas A&M University; 2004.
for Amirkabir tunnel using artificial neural networks. Int J Min Metal Mech Eng [41] Jia TG, Liu J. Stability of mine ventilation system based on multiple regression
(IJMMME) 2013;1(1):5–9. analysis. Int J Min Sci Technol 2009;19(4):463–6.
[29] Ribacchi R, Graziani A, Boldini D. Previsione degli afflussi d’acqua in galleria e [42] Rezaee MR, Jafari A, Kazemzadeh E. Relationships between permeability,
influenza sull’ambiente. Meccanica e Ingegneria delle rocce 2002:143–99. porosity and pore throat size in carbonate rocks using regression analysis and
[30] Jacob CE, Lohman SW. Nonsteady flow to a well of constant drawdown in an neural networks. J Geophys Eng 2006;3(4):370–6.
extensive aquifer. Trans Am Geophys Union 1952;33(4):559–69. [43] Kallu RR, Keffeler ER, Watters RJ, Agharazi A. Development of a multivariate
[31] Marechal JC, Perrochet P. New analytical solution for the study of hydraulic empirical model for predicting weak rock mass modulus. Int J Min Sci Technol
interaction between Alpine tunnels and groundwater. Bull Soc Geol Fr 2015;25(4):545–52.
2003;174(5):441–8. [44] Mustafa R. Application of statistical methods for flow unit identification and
[32] Ming HF, Wang MS, Tan ZS, Wang XY. Analytical solutions for steady seepage characterization of a reservoir using well log and core data. West
into an underwater circular tunnel. Tunn Undergr Sp Tech 2010;25:391–6. Virginia: West Virginia University; 2003.
[33] Chen R. Groundwater inflow into rock tunnels. Texas: University of Texas at [45] Coli Cameron A, Windmeijer FAG. An R-squared measure of goodness of fit for
Austin; 2010. some common nonlinear regression models. J Economet 1997;77(2):329–42.
[34] Chen R, Tonon F. Rock tunnel groundwater prediction in a simulated rock [46] Montgomery DC, Peck EA. Introduction to linear regression
fracture network. In: Proceedings of North American tunneling conference. analysis. USA: Wiley; 1992.
Indianapolis (USA); 2012. p. 401–10. [47] SCE Company. Site hydrogeology reports for KRJ project; 2005.
[35] Farhadian H, Nikvar Hassani A, Katibeh H. Groundwater inflow assessment to [48] SCE Company. Geological and engineering geological report for Amirkabir
Karaj Water Conveyance tunnel, northern Iran. KSCE J Civ Eng 2016:1–10. water conveyance tunnel project (Lot1). Unpublished report; 2006.
[36] Farhadian H, Katibeh H. Groundwater seepage estimation into amirkabir [49] SCE Company. Geological and engineering geological reports for KWCT
tunnel using analytical methods and DEM and SGR method. World Acad Sci, project; 2006.
Eng Technol, Int J Civil, Struct, Construct Architect Eng 2015;9(3):255–60. [50] Mahab Ghods Consultant Engineering Company. Engineering geological for
[37] Liu F, Xu G, Huang W, Hu Sh, Hu M. The effect of grouting reinforcement on Ghomroud project; 2004.
groundwater seepage in deep tunnels. Blucher Mech Eng Proc 2012;1 [51] SCE Company hydrogeology reports for Nosoud project; 2005.
(1):4727–37.
[38] Farhadian H, Aalianvari A, Katibeh H. Optimization of analytical equations of
groundwater seepage into tunnels, a case study of Amirkabir tunnel. J Geol Soc
India 2012;80:96–100.