You are on page 1of 18

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

net/publication/323098891

Deformation Modulus of Rock Masses: An Assessment of the Existing


Empirical Equations

Article  in  Geotechnical and Geological Engineering · February 2018


DOI: 10.1007/s10706-018-0491-1

CITATIONS READS

3 230

2 authors:

Ali Kayabasi Candan Gokceoglu


Eskisehir Osmangazi University Hacettepe University
20 PUBLICATIONS   659 CITATIONS    188 PUBLICATIONS   8,417 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

“Spatial Modeling in GIS and R for Earth and Environmental Science” View project

Interest in bimrocks View project

All content following this page was uploaded by Candan Gokceoglu on 17 December 2019.

The user has requested enhancement of the downloaded file.


Geotech Geol Eng (2018) 36:2683–2699
https://doi.org/10.1007/s10706-018-0491-1

ORIGINAL PAPER

Deformation Modulus of Rock Masses: An Assessment


of the Existing Empirical Equations
A. Kayabasi . C. Gokceoglu

Received: 2 November 2017 / Accepted: 5 February 2018 / Published online: 9 February 2018
Ó Springer International Publishing AG, part of Springer Nature 2018

Abstract Rock mass deformation modulus is an modulus and the empirically calculated rock mass
important parameter for all geotechnical applications. modulus values were performed by simple regression
However, the determination of rock mass deformation analyses. The empirical equations with higher predic-
modulus with in situ tests are highly expensive and tion capacity were also examined with root mean
time consuming. For this reason, rock engineers and square error, values account for and prediction error
engineering geologists have proposed numerous evaluations. Among the empirical equations compared
empirical equations based on various rock mass and in this study, two empirical equation giving best
intact rock properties to estimate the deformation performance and other four empirical equations pro-
modulus of rock masses. In the present study, an viding acceptable results were determined.
assessment of the existing empirical equations was
undertaken. For the purpose of the study, the data Keywords Deformation modulus  Rock mass 
obtained from four investigation galleries opened Regression  Empirical equation  Prediction error
during a dam construction (Artvin Dam, Turkey) were
used. A total of 34 plate loading tests were employed
in these galleries. The tested rock mass is poor quality
tuff. Rock mass rating (RMR89), rock tunnelling 1 Introduction
quality index (Q) and geological strength index of
each test levels were determined. The empirical Deformation modulus of rock masses in geotechnical
deformation modulus values of rock mass were projects, such as underground openings, dams and
calculated by the 26 most cited empirical equations slopes, is an important input parameter in the project
proposed by various researchers. The cross-checks design stage. This modulus is determined by various
between the measured rock mass deformation in situ tests (i.e., plate loading, dilatometer, flat jack
test and pressuremeter tests). However, all in situ test
equipment and hardware are highly expensive, and the
A. Kayabasi
application of these in situ tests are time-consuming
Department of Geological Engineering, Eskisehir
Osmangazi University, 26480 Eskisehir, Turkey and require experienced technical staff and difficult
e-mail: akayabasi@ogu.edu.tr installation processes, such as the excavation of long
galleries. Test procedure differences and blasting
C. Gokceoglu (&)
damage are also disadvantageous sides of in situ
Department of Geological Engineering, Hacettepe
University, Beytepe, 06800 Ankara, Turkey measurements (Palmström and Singh 2001). For this
e-mail: cgokce@hacettepe.edu.tr reason, many researchers have proposed empirical

123
2684 Geotech Geol Eng (2018) 36:2683–2699

equations for estimating the modulus of deformation it is often difficult or even impossible to decide which
in the last two decades. These empirical equations for method is the most accurate.
estimating deformation modulus are simple and Consequently, these equations are commonly used,
cheaper to perform than in situ tests, but the perfor- but they require special attention because almost each
mance of these equations is not always sufficient. empirical equation has certain limitations due to the
Several of these empirical equations were produced volume of data set used when developing such
from intact rock and rock mass properties, such as parameters as the empirical equation and/or type of
uniaxial strength, rock quality designation, and weath- rock mass. As a result of this evaluation and the
ering degree (Kayabasi et al. 2003; Gokceoglu et al. increase in the number of empirical equations, certain
2003; Zhang and Einstein 2004). Several others were assessments of the existing empirical equations are
based on rock mass classification systems (i.e., indispensable. For this reason, the main purpose of this
Bieniawski 1973; Grimstad and Barton 1993; Sonmez paper is to assess the performances of these empirical
et al. 2006). Aksoy et al. (2012) compared of different equations by in situ test results and to draw conclu-
empirical equations of deformation modulus of rock sions on the existing empirical equations. The 26 most
mass suggested by different researchers by using field cited empirical equations were collected from the
displacement measurements and numerical modeling literature. Thirty-four in situ plate loading test results
results. They stated that rock mass deformation which were performed at the Artvin dam project of
modulus obtained from Barton (2002) empirical Turkey were used in this paper to determine the
equation were higher than other empirical equations, performance of these empirical equations.
rock deformation modules obtained from empirical
equations suggested by Sonmez et al. (2004) and Hoek
and Diederichs (2006) are very close to each other. 2 Intact Rock and Rock Mass Properties
However, rock mass deformation values obtained of the Artvin Dam
from the equation suggested by Sonmez et al. (2004)
were lower in hard and big block-sized rock masses The Artvin dam is one of the dam projects on the
and higher in weak and little block sized rock masses Çoruh river in northwestern Turkey (see Fig. 1). The
than that suggested by Hoek and Diederichs (2006). data used in this study were provided from Artvin dam
Lastly they stated that the performance of the equation site. All data employed were collected from plate
suggested by Sonmez et al. (2004) for determining the loading tests performed by EIEI (General Directorate
rock mass deformation modulus in numerical mod- of Electrical Power Research Survey and Develop-
elling was more realistic. Feng and Jimenez (2015) ment Administration) in research galleries (Gürsoy
presented a procedure, within the Bayesian framework and Uysal 1988). The galleries were opened by
for model updating, to systematically update the blasting and excavation method with an utmost care
predictive distribution of deformation modulus of of undisturbance of rock masses. However disturbance
rock mass, and its associated predictive uncertainty, of the rock masses is inevitable with the blasting
when new ‘‘project-specific’’ data are available. method. The effect of blasting vary with rock prop-
Finally, they used an example case of an assumed erties, amount of explosive used and the distance of
circular rock tunnel design to show that the reduction the blast holes (Palmström and Singh 2001). One of
of predictive uncertainties achieved with the Bayesian the disadvantageous point of study is the damage from
updating has a significant influence on the computed blasting is insignificant in this study.
reliability estimates of a rock engineering design. The Artvin dam site is mainly covered by greenish
Zhang (2017) provided a comprehensive review of the grey coloured, coarse to fine grained tuff and dark grey
different empirical methods for determining the coloured, laminated phyllitic lithologies. Both the
deformation modulus of rock masses, and a compar- bedding of the tuff lithology and the lamination of
ative analysis and discussion was subsequently con- phyllite rock masses are NE striking and SE dipping.
ducted on the accuracy and main issues of these The discontinuity spacings are generally medium
methods. Zhang (2017) concluded that the estimated spaced to very closely spaced according to classifica-
values from the various empirical methods could be tion recommended by ISRM (1981), while apertures
very different for certain of the rock masses; therefore, are wide to very narrow. The discontinuity surfaces are

123
Geotech Geol Eng (2018) 36:2683–2699 2685

Fig. 1 Location map of Artvin dam site

generally moderately and highly weathered, but static elasticity modulus of intact rock is 20.03 GPa.
rarely, slightly weathered and completely weathered Four investigation galleries were opened in the Artvin
surfaces are observed (Table 1). dam site, and 34 plate loading test were performed in
As seen in Table 1, the uniaxial compressive these research galleries by EIEI. The in situ deforma-
strength values of the tuffs in the Artvin dam site tion moduli of rock masses were determined from
vary between 38.7 and 187.5 MPa, while the average these plate loading test data. When these data were

123
2686 Geotech Geol Eng (2018) 36:2683–2699

Table 1 Rock mass properties of testing levels


Rock mass property Gallery T1 Gallery T2 Gallery T3 Gallery T4
Min Max Avg Min Max Avg Min Max Avg Min Max Avg

UCS (MPa) 38.7 138.6 93.4 112.4 181.8 169.4 42.0 141.6 114.8 46.0 187.5 136.2
Ei (GPa) 20.01 20.04 20.03 20.03 20.09 20.04 20.01 20.04 20.03 20.01 20.05 20.04
Em (GPa) 2.5 12.64 7.97 3.38 20.13 11.39 1.52 12.05 4.8 3.8 19.15 10.03
RQD (%) 25 50 50 45 100 66 40 100 61 40 100 71
RMR 29.18 59.28 48.3 36.50 71.50 51.94 38.34 49.38 40.77 29.40 67.20 52.84
Q 0.19 5.46 1.55 0.43 21.23 10.56 0.21 1.80 7.21 0.20 13.20 3.83
GSI 17 43 28 18 48 28 16 28 24 16 55 39
Structural rating (SR) 28 68 31.2 30 68 43 30 68 42 30 68 43
Joint surface condition (SCR) 1 7 5 1 12 4 1 5 3 1 13 6
Spacing (mm) 600 to \ 50 mm 600 to \ 50 mm 600 to \ 50 mm 50–400 mm
Persistency (m) [ 20 m [ 20 m [ 20 m [ 20
Roughness Slickensided-rough Slickensided-rough Slickensided-rough Slickensided-rough
Aperture (mm) 1–30 1–5 1–5 1–15 cm
Weathering UW to HW SW to HW MW to CW UW–HW
Infilling [ 5 mm 2–5 mm 3–5 mm 2–5 cm
Groundwater condition Moisted Moisted-dry Moisted Moisted

evaluated together, the minimum and maximum Twenty-six empirical equations were collected from
deformation modulus values were determined as the literature for this study. Many of these equations
1.52 and 20.13 GPa, respectively. Farmer and use Rock Rass Rating (RMR) (Bieniawski 1973;
Kemeny (1992) determined that the intact rock Serafim and Pereira 1983; Nicholson and Bieniawski
modulus is in the order of 5–20 times higher than 1990; Mitri et al. 1994; Aydan et al. 1997; Read et al.
in situ measured values. In this study lowest in situ 1999; Gokceoglu et al. 2003; Ramamurty 2004;
measured deformation modulus is approximately 13 Galera et al. 2005; Sonmez et al. 2006; Chun et al.
times lower than laboratory determined intact rock 2006; Işık et al. 2008; Mohammadi and Rahmannejad
modulus. This is arising from jointing systems, 2010; Shen et al. 2012; Alemdag et al. 2015). Several
weathering degree, aperture thickness differences. equations employ Tunnelling Quality Index
The rock mass rating (RMR), Q system and the (Q) (Grimstad and Barton 1993; Barton 2002) and
geological strength index (GSI) values were calcu- the Geological Strength Index (GSI) (Hoek and Brown
lated for each test location. The rock mass classifica- 1997; Hoek and Diederichs 2006; Beiki et al. 2010;
tions RMR (Bieniawski 1989), Q system (Barton et al. Ghamgosar et al. 2010) and others considered rock
1980), GSI (Hoek and Brown 1997) and the quanti- mass properties, such as rock quality designation and
tative GSI chart, which was proposed by Sonmez and weathering degree (Kayabasi et al. 2003; Gokceoglu
Ulusay (1999), were used for the determination of GSI et al. 2003; Zhang and Einstein 2004; Sonmez et al.
values. The average RMR, Q and GSI values and the 2006; Alemdag et al. 2016). Empirical equations used
plate loading test results of rock mass properties are in this study are given in Table 2. Each empirical
presented in Table 1. equation is numbered as seen in Table 2. The Bieni-
awski’s (1973) equation is numbered as Ee1 (estimated
equation 1). In situ plate loading test results are also
3 Empirical Equations symbolized as Em (measured deformation modulus).
There are several limitations to this approach:
Various empirical equations for determining defor- Bieniawski (1973) equations are for RMR [ 50; the
mation modulus of rock masses were proposed. Grimstad and Barton (1993) equation is proposed for

123
Geotech Geol Eng (2018) 36:2683–2699 2687

Table 2 Summary of empirical equations considered in the study


No Author Equation Limitations Litology

Ee1 Bieniawski (1973) Em = 2RMR - 100 RMR [ 50 Mudstone, siltstone, sandstone,


shale, dolerite (hard rock masses)
Ee2 Serafim and Pereira (1983) Em = 100(RMR-10)/40 RMR \ 50 Granite, gneis, Mudstone, siltstone,
sandstone, shale, dolerite. (Soft
rock masses)
Ee3 Nicholson and Bieniawsk Em = Ei[0.0028 RMR2 Argillaceous, arenaceous,
(1990) ? 0.9 exp(RMR/22.92)] crystalline rocks
Ee4 Grimstad and Barton (1993) Em = 25 log Q For Q [ 1 Wide range of rock masses: chalk
marls, sandstones, shales,
gneisses, granites, ignimbrites,
tuffs
Ee5 Mitri et al. (1994) Em = Ei[0.5(1 - (Cos(p*RMR/100)))]
Ee6 Hoek and Brown (1997) Em = (rc/100)0.5*10(GSI-10)/40 rc \ 100 Slates, schist, phyllites, weak rocks
(MPa)
Ee7 Aydan et al. (1997) Em = 0.1(RMR/10)3 Gneiss, granite, sandstone
Ee8 Read et al. (1999) Em = 0.1(RMR/10)3 Graywacke sandstones, mudstones
Ee9 Barton (2002) Em = 10(Qrc/100)1/3 Schist, quartz, marble, phyllilte,
dolamite quartzite
Ee10 Kayabasi et al. (2003) Em = 0.135[(Ei(1 ? RQD/ Quatzdiorite, limestone
100))WD]1.811
Ee11 Gokceoglu et al. (2003) Em = 0.001[((Ei/UCS) Quartzdiorite, limestone, shale
(1 ? RQD/100))/WD]1.5528
Ee12 Gokceoglu et al. (2003) Em = 0.0736e(0.0755RMR) Quartzdiorite, limestone, shale
0.0186RQD-1.91)
Ee13 Zhang and Einstein (2004) Em = Ei10( Mudstone, siltstone, sandstone,
shale, dolerite, granite, limestone,
greywacke, gneiss
Ee14 Ramamurty (2004) Em = Eie(-0.035(5(100-RMR)) Wide range of rock masses and
intact rocks
Ee15 Galera et al. (2005) Em = Eie(RMR-100)/36 Wide range of igneous,
metamorphic, detritic and
carbonate sedimantary rock
masses
Ee16 Sonmez et al. (2004) Em = Ei(sa)0.4 Quartzdiorite, limestone, shale
Ee17 Sonmez et al. (2006) Em = Ei*10[((RMR - 100) Quartzdiorite, limestone, shale
(100 - RMR))/
(4000 exp(- RMR/100))]
Ee18 Hoek and Diederichs (2006) Em = Ei[0.02 ? (1 - (D/2))/ D=1 Sedimantary, igneous,
(1 ? e((60?15D-GSI)/11)] metamorphic rocks
Ee19 Chun et al. (2006) Em = 0.003228e(0.0495RMR) Granite, gneiss, andesite, tuff,
sandstone, shale
Ee20 Isık et al. (2008) Em = (6.7*RMR - 103.6)/1000 RMR C 27 Graywacke
Ee21 Mohammadi and Rahmannejad Em = 0.0003RMR3 - 0.0193RMR2 Limestone, marble
(2010) ? 0.315RMR ? 3.4065
Ee22 Beiki et al. (2010) Em = tan(1.56 ? (ln(GSI)2)1/2r1/3 Shale, sandstone-quartzite,
Limestone, marl-lime stone with
silica-veins, Sandstone, siltstone
and mudstone

123
2688 Geotech Geol Eng (2018) 36:2683–2699

Table 2 continued
No Author Equation Limitations Litology

Ee23 Ghamgosar et al. (2010) Em = 0.0912e0.0866GSI Limestone, schist slate


Ee24 Shen et al. (2012) Em = 1.14Eie-((RMR-116)/41)^2 Mudstone, siltstone, sandstone,
shale, dolerite (hard rock
masses), Granite, gneis,
Mudstone, siltstone, sandstone,
shale, dolerite. (Soft rock masses)
Ee25 Alemdag et al. (2015) Em = 0.058e0.0785RMR Basalt, tuffites, diabases, sitstones
Ee26 Alemdag et al. (2016) Em = 0.00067RQD2 ? 0.00067RQDr Siltstones
? (0.00067RQDr ? 0.00067r2)/
(RQD ? 99.5)
*Ee Emprical equation, Em Measured deformation modulus, Ei intact rock modulus, rc Intact rock strength, GSI Geological Strength
Index and s and a Rock mass constants for GSI

Q [ 1; Mitri et al. (1994) is for rc \ 100 MPa. Beiki Bieniawski 1990), Ee16 (Galera et al. 2005), Ee11
et al. (2010) provides negative values for extreme GSI (Gokceoglu et al. 2003), Ee14 (Ramamurthy 2004),
values (Tahir and Mohammad 2014). Due to these Ee17 (Sonmez et al. 2006), Ee5 (Mitri et al. 1994).
limitations, the number of regression input parameters Meanwhile, these equations produce higher estimated
are decreased, and negative values were discarded in deformation values: Ee21 (Mohammadi and Rahman-
regression analysis. nejad 2010), Ee2 (Serafim and Pereira 1983), Ee19
(Chun et al. 2006), Ee7 (Aydan et al. 1997), Ee1
(Bieniawski 1973) and Ee8 (Read et al. 1999). enerally,
4 Cross-Checks Between Measured Deformation all of the equations’ estimates of the deformation
Modulus and Empirical Equations Derived modulus value are close to the measured deformation
from RMR modulus. The estimated deformation modulus values
tend to scatter with increasing RMR values.
The collected empirical equations were grouped in this Figure 4 shows both estimated and deformation
study according to the classification systems on which modulus data plotted by test location. The overesti-
they are based. The first group was the RMR based mating equations were excluded for clarity. The Ee5
equations. Sixteen of 26 empirical equations estimate (Mitri et al. 1994), Ee12 (Gokceoglu et al. 2003), Ee15
the deformation modulus with RMR values of the rock (Galera et al. 2005), Ee17 (Sonmez et al. 2006) and Ee19
mass. Testing locality RMR values were calculated (Chun et al. 2006) tend to be closer to the measured
from testing locality rock mass properties according to deformation values (Em) (see Fig. 4).
Bieniawski (1989).
In situ measured deformation modulus and empir-
ical estimated deformation modulus based on RMR 5 Cross-Checks Between Measured Deformation
plotted with RMR values of testing localities are Modulus and Empirical Equations Derived
shown in Figs. 2 and 3. Figure 2 shows 14 empirical from the Q Index
equations and measured deformation modulus values
plotted against RMR values of plate loading testing The second group of equations were based on the Q
localities. Ee24 (Shen et al. 2012) and Ee25 (Alemdag index classification. There are two empirical equations
et al. 2016) equations overestimated deformation that use Q data to evaluate the deformation modulus of
values; therefore, in the interest of clarity, these rock masses. One of them is the Grimstad and Barton
equations were plotted separately as Fig. 3. (1993) equation. The limitation of this equation is for
The following equations produce lower estimated Q [ 1. Therefore, 24 of the total in situ test sites were
deformation values than measured deformation val- correlated with the empirical equation of Grimstad and
ues: Ee21 (Işık et al. 2008), Ee3 (Nicholson and Barton (1993). The other equation is the Barton

123
Geotech Geol Eng (2018) 36:2683–2699 2689

Fig. 2 Estimated
deformation modulus values
from RMR

Fig. 3 Outlier estimation of


deformation modulus by
Ee25 and Ee26 equations from
RMR

(2002). Figure 5 shows the plots of Q index based on Figure 6 shows the measured deformation modulus
empirical deformation modulus and in situ measure- and estimated deformation moduli at each test loca-
ments of the plate loading tests. Both empirical tion. The Grimstad and Barton (1993) equation is valid
equations give estimated deformation modulus closer for Q [ 1. Broken lines show that this equation is not
to in situ test results for Q \ 5. With higher Q index valid along these testing localities. The estimated
values, estimated deformation values are higher than deformation modulus values from Ee4 (Grimstad and
in situ determined deformation test results. Barton Barton 1993) and Ee9 (Barton 2002) are in closer
(2002) specifically estimates a higher deformation agreement with measured deformation values. The
modulus. Grimstad et al. (2002) equation overestimated

123
2690 Geotech Geol Eng (2018) 36:2683–2699

Fig. 4 Estimated
deformation modulus from
RMR and measured
deformation modulus at
each test location

Fig. 5 Relationship
between measured
deformation modulus and
estimated deformation
modulus from Q index

deformation values at the T17, T18 and T28, T29 6 Cross-Checks Between Measured Deformation
locations because the rock masses along these test Modulus and Empirical Equations Derived
locations have higher Q index values. A slight from GSI Index
overestimation is also observed with the Barton
(2002) equation. This overestimation is The third group was the Geological Strength Index
attributable to the high intact rock strength value of (GSI)-based empirical equations. A qualitative GSI
the rock mass. Both Grimstad and Barton (1993) and chart based on visual inspection has been available
Barton (2002) provide higher estimated deformation since 1994. Sonmez and Ulusay (1999) suggested a
values than measured deformation values under good quantitative numerical basis for evaluating the GSI
rock mass conditions. system. Later, the quantitative GSI chart was slightly

123
Geotech Geol Eng (2018) 36:2683–2699 2691

Fig. 6 Deformation moduli


estimated from Q index and
measured deformation
modulus at each test location

modified by Sönmez and Ulusay (2002). This latest empirical equation. The broken line of Ee22 (Beiki
version of quantitative GSI chart was used for the et al. 2010) arises from the negative value estimation
determination of GSI values of rock masses in this of this empirical equation. The empirical equations
study. proposed with GSI produce estimations of the defor-
There are four equations that use the GSI value to mation modulus that are lower than the measured
estimate the deformation modulus. Hoek and Brown deformation values in general (Fig. 8).
(1997) was the first to use GSI value and intact rock
strength of the rock masses for estimation of the
deformation modulus of rock masses. The intact rock 7 Cross-Checks Between Measured Deformation
strength must be higher than 100 MPa for this Modulus and Empirical Equations Derived
equation. Due to this limitation, 10 test locations are from Different Rock Mass and Intact Rock
discarded from calculations. Later, Hoek and Dieder- Parameters
ichs (2006) proposed an equation. These researchers
used a rock materials laboratory elasticity modulus The fourth group were the empirical equations based
with GSI values for estimating deformation value of on the rock mass and intact rock properties (i.e. RQD,
rock masses. Beiki et al. (2010) estimates deformation WD, rci, s and a). This group consists of five empirical
modulus with GSI and intact rock strength (r) values. equations that estimate the deformation modulus using
Ghamgosar et al. (2010) produced an equation with rock mass parameters, such as rock quality designation
only the GSI value. Figure 7 shows plots of the (RQD), weathering degree (WD), intact rock strength
measured deformation modulus values and estimated (rci), and rock mass constants (s, a).
deformation modulus values versus GSI values by According to Fig. 9, the empirical equation pro-
testing location. All of the equations estimate the posed by Kayabasi et al. (2003) is overestimating the
deformation modulus to be lower than the measured deformation modulus at four testing localities (TT9,
deformation modulus. The Beiki et al. (2010) equation T10, T18 and T29). The properties that these testing
provided negative deformation value when the GSI localities have in common are high RQD, high
value of rock mass was under 18.85, i.e., this equation weathering degree and high intact rock strength.
provided a negative deformation value at six testing Kayabasi et al.’s (2003) equation overestimates the
locations. deformation modulus for good rock conditions. Other
Both measured and estimated deformation are empirical equations Gokceoglu et al. (2003), Zhang
plotted by test location. The broken line of Ee6 (Hoek and Einstein (2004), Sonmez et al. (2004), Alemdag
and Brown 1997) arises from the limitation of this et al. (2016) estimate the deformation modulus within

123
2692 Geotech Geol Eng (2018) 36:2683–2699

Fig. 7 Relationship
between measured
deformation modulus and
estimated deformation
modulus from GSI

Fig. 8 Estimated
deformation moduli from
GSI and measured
deformation modulus at
each test location

a range of 20 GPa over or under the measured performed between estimated deformation values and
deformation modulus values. the measured deformation values. The graphs of
simple regression analyses are shown in Fig. 10.
A number of equations, such as those of Bieniawski
8 Correlation Measured Deformation Modulus (1978), Grimstad and Barton (1993) and Beiki et al.
with Estimated Deformation Modulus (2010), have limitations. Due to these limitations, the
deformation modulus for some testing locations was
As mentioned in the previous sections, many empirical not estimated. RMR values of 23 plate loading testing
equations have been proposed in the literature. The 26 locations are lower than 50; therefore, the number of
most cited empirical equations were selected for this the estimated deformation values are 11 from Bieni-
study. The estimated deformation modulus values at awski’s equation. Grimstad and Barton (1993) is for
plate loading test locations were calculated from these the typical range of Q values, and it produces a
empirical equations. Simple regression analyses were negative value of deformation modulus for Q \ 1. Q

123
Geotech Geol Eng (2018) 36:2683–2699 2693

Fig. 9 Estimated
deformation moduli from
RQD, WD, rci, s–a and
measured deformation
modulus at each test location

values of 10 of the plate loading testing locations are (PE) and cumulative frequency chart was drawn to
lower than 1, and the number of the deformation determine prediction performance of empirical equa-
modulus values successfully estimated with Grimstad tions and is shown in Fig. 10.
and Barton (1993) equation was 24. The suggested The empirical equations performances are con-
equation by Beiki et al. (2010) provides negative rock verging mainly at - 400 (%) and ? 200 (%) predic-
mass deformation values for values of GSI under tion values. Negative values of PE mean that the
18.85, and the estimated number of deformation empirical equation is overestimating the measured
values was 28 with this equation. deformation values. Accumulation of the negative side
The data placement along the 1:1 line is indicative of the horizontal axes show the overestimation of the
of the estimation performance for the equations. The empirical equation. To improve upon the clarity of
better placements along the 1:1 line are observed with Fig. 11a, the equations with higher VAF and RMSE
the equations of Hoek and Brown (1997), Gokceoglu values were removed and the optimal empirical
et al. (2003), Hoek and Diederichs (2006), and equations were found and displayed in Fig. 11b.
Ghamgosar et al. (2010). According to Fig. 1b, the equations of Nicholson
The root mean square error (RMSE) indices and and Bieniawski (1990), Mitri et al. (1994), Hoek and
values accounted for (VAF) are calculated to quantify Brown (1997), Gokceoglu et al. (2003), Zhang and
the prediction performance of the equations for simple Einstein (2004), Galera et al. (2005), Sonmez et al.
regression analysis. A perfect prediction is represented (2004), Hoek and Diederichs (2006) and Ghamgosar
with 0 in RMSE values and 100% in VAF values. The et al. (2010) produce good prediction performances.
coefficient of determination (R2), RMSE and VAF The RMSE and VAF revealed that the highest
indices for each equation are tabulated in Table 3. predictive capability was provided by the Galera
The VAF value of the empirical equation of Barton et al. (2005) equation among the empirical equations
(2002) (Ee9) is the lowest with a value of – 6404.50%, compared in this study. Gokceoglu et al.’s (2003)
whereas the highest value of VAF is 70% with the equation is the second best when considering all data.
empirical equation of Sonmez et al. (2004) (Ee16). The Hoek and Diederich’s (2006), Ghamgosar et al.
best RMSE performances are observed with the (2010), Nicholson and Bieniawski’s (1990) and Hoek
empirical equations of Mitri et al. (1994) (Ee5) and and Brown’s (1997) equations also exhibited good
Alemdag et al. (2016) (Ee27). performances. Mitri et al.’s (1994) and Sonmez et al.’s
Regression analysis and coefficient of determina- (2004) equations yielded more inconsistent results
tion values, RMSE and VAF values are not enough for among the equations having good prediction
prediction performance. Prediction performance of performances.
empirical equation is calculated with the prediction
error value in statistics. The calculated prediction error

123
2694 Geotech Geol Eng (2018) 36:2683–2699

Fig. 10 Graphs showing the relation between measured deformation values (Em) and the estimated deformation values (Ee)

123
Geotech Geol Eng (2018) 36:2683–2699 2695

Fig. 10 continued

9 Results and Conclusions equation between rock mass classification systems


were not used in this study due to the rudeness of
There are considerable numbers of empirical equation method and may cause wrong values as stated by
that have been proposed in the rock mechanics Palmström and Singh (2001).
literature. However, the prediction performance of The 26 most cited equations were compiled, and
these empirical equations have not been widely tested. these equations were grouped according to classifica-
As mentioned in the introduction, the purpose of this tion systems on which they are based. The first group
study was to determine the performance of these was the RMR based equations. Sixteen of the 26
empirical equations. For this reason, 34 plate loading empirical equations estimate the deformation modulus
tests results which were performed in research with RMR values of the rock mass. The simple
galleries of Artvin dam sites, loading test location regression analysis between RMR values and mea-
RMR, Q and GSI values were used as data in this sured-estimated deformation values showed that Shen
study. RMi rock mass classification system was not et al. (2012) and Alemdag et al. (2015), Mohammadi
performed in this study because RMi values were not and Rahmannejad (2010), Serafim and Pereira (1983),
determined directly. In other words, the transiton Chun et al. (2006), Aydan et al. (1997), Bieniawski

123
2696 Geotech Geol Eng (2018) 36:2683–2699

Table 3 Coefficient of determination (R2), the values of root mean square errors (RMSE) and values account for (VAF) indices of
the estimated modulus of deformation from the empirical equations
Equation no Equation type The coefficient of Value account Root mean square
determination R2 for VAF (%) error RMSE

Ee1 Linear 0.7441 - 393.52 10.71


Ee2 Linear 0.6261 15.46 4.59
Ee3 Linear 0.6504 38.13 6.85
Ee4 Linear 0.4117 - 404.12 8.87
Ee5 Linear 0.6378 61.69 2.93
Ee6 Power 0.5082 45.01 5.44
Ee7 Linear 0.6326 - 11.19 5.13
Ee8 Linear 0.6433 - 12.47 6.07
Ee9 Logarithmic 0.6376 - 6404.50 38.76
Ee10 Linear 0.2307 - 1240.20 19.76
Ee11 Power 0.2645 - 269.38 9.39
Ee12 Linear 0.5981 59.44 5.87
Ee13 Power 0.1432 - 65.07 6.77
Ee14 Linear 0.6532 41.48 3.63
Ee15 Linear 0.6530 99.44 3.22
Ee16 Linear 0.4219 70.07 3.78
Ee17 Linear 0.6426 - 61.75 1.94
Ee18 Logarithmic 0.4819 55.29 5.62
Ee19 Linear 0.6364 66.47 3.45
Ee20 Linear 0.6394 3.97 8.14
Ee21 Logarithmic 0.6336 0.09 4.59
Ee22 Power 0.2426 72.93 5.21
Ee23 Logarithmic 0.4806 52.69 6.24
Ee24 Logarithmic 0.6492 - 2994.40 39.28
Ee25 Logarithmic 0.6395 - 24371 73.49
Ee26 Linear 0.3475 22.45 3.11

(1973) and Read et al. (1999) overestimate deforma- modulus value of rock massess. One of these equations
tion values where as Isık et al. (2008), Nicholson and is the Grimstad and Barton (1993) equation, and the
Bieniawski (1990), Galera et al. (2005), Gokceoglu other is the Barton (2002) equation. When Q \ 5, both
et al. (2003), Ramamurthy (2004), Sonmez et al. equations give estimated deformation modulus closer
(2006), Mitri et al. (1994) produce lower estimates of to in situ test results. At higher Q index values,
deformation values. 20 testing location RMR values estimated deformation values with these equations
are lower than 50; therefore, Bieniawski (1973) yield higher values than in situ determined deforma-
equation was evaluated with 14 data points. This tion test results.
equation tends to overestimate those with high RMR The third group was the geological strength index
values (RMR [ 60) and underestimate those with low (GSI) based empirical equations. Four empirical
RMR values (RMR \ 60). equations use GSI for estimation of deformation
The second group was the Q index classification modulus, Hoek and Brown (1997), Hoek and Dieder-
based empirical equations. There are two empirical ichs (2006), Beiki et al. (2010) and Ghamgosar et al.
equations that use Q data to evaluate the deformation (2010). These empirical equations’ estimates of the

123
Geotech Geol Eng (2018) 36:2683–2699 2697

Fig. 11 Graphs showing the relationships between prediction error and cumulative frequency for the empirical equations evaluated in
this study a the evaluation all empirical equations b the evaluation for the empirical equations between - 80 and ? 100%

deformation modulus are lower than measured defor- Hoek and Diederichs (2006) and Ghamgosar et al.
mation modulus. The Beiki et al.’s (2010) equation (2010).
gives a negative deformation value for GSI values The RMSE, VAF and the prediction error (%) and
under 18.85, so negative deformation value estimates the cumulative frequency of data graph revealed the
were produced for 6 of the testing locations in this best equations for prediction of the deformation
study. modulus. The highest predictive capability was pro-
The fourth group was the rock mass and intact rock vided by the Galera et al.’s (2005) equation among the
properties (i.e. RQD, WD, rci,s and a) based empirical empirical equations compared in this study. Gokceo-
equations. There are five empirical equations estimat- glu et al.’s (2003) equation is the second best after
ing deformation modulus with RQD, WD, rci,s and a. Galera et al.’s (2005) equation when considering all
Kayabasi et al. (2003) overestimates the deformation data. Hoek and Diederichs (2006), Ghamgosar et al.
modulus of rock masses whereas other empirical (2010), Nicholson and Bieniawski’s (1990) and Hoek
equations suggested by Gokceoglu et al. (2003), and Brown’s (1997) equations also displayed good
Zhang and Einstein (2004), Sonmez et al. (2004), performances. The modulus of deformation of rock
Alemdag et al. (2016) estimate the deformation masses is governed by discontinuity characteristics
modulus within the range of 20 GPa over or under such as aperture, spacing, infilling etc., weathering
the measured deformation modulus values. degree, strength and deformability of intact rock
Estimated deformation and measured deformation pieces in rock mass, mineralogy and texture of intact
modulus values were compared for each of the rock pieces, in situ stress conditions etc. Due to this
empirical equations. The best data accumulations complexity, to estimate and/or to determine modulus
along the 1:1 line were observed with the equations of deformation accurately is very difficult. On the
Hoek and Brown (1997), Gokceoglu et al. (2003), other hand, an empirical equation is based on the

123
2698 Geotech Geol Eng (2018) 36:2683–2699

dataset employed. For this reason, when assessing or Barton N, Loset F, Lien R, Lune J (1980) Application of
using an empirical equation, this complex situation Q-system in design decisions concerning dimensions and
appropriate support for underground installations. Sub-
should be considered. Additionally, the type of rock surface Space, Pergamon, pp 553–561
mass and the number of the in situ tests used in the Beiki M, Bashari A, Majdi A (2010) Genetic Programming
present study are limited. Consequently, it is not approach for estimating the deformation modulus of rock
forgotten that the findings obtained from the present mass using sensitivity analysis by neural network. Int J
Rock Mech Min Sci 47:1091–1103
study based on a dataset representing one rock type. Bieniawski Z (1973) Engineering classification of rock masses.
Otherwise, the results and conclusions presented here Trans S Afr Inst Civ Eng 15:335–344
is open to speculation. Additionally, depending on Bieniawski ZT (1978) Determining rock mass deformability:
various factors, in situ measurements of deformation experience fromcase histories. Int J Rock Mech Min Sci
Geomech Abstr 15:237–247
modulus of rock masses have sometimes some uncer- Bieniawski ZT (1989) Engineering rock mass classifications.
tainties. These uncertainties were summarized by Wiley, New York
Palmström and Singh (2001). During the present Chun B, Lee Y, Seo D, Lim B (2006) Correlation of deformation
study, to minimize these uncertainties, testing condi- modulus by PMT with RMR and rock mass condition.
Tunn Undergr Space Technol 21(3–4):231–232
tions were assessed and rock mass parameters were Farmer IW, Kemeny JM (1992) Deficiencies in rock test data.
selected carefully. In: Proceedings of international conference on Eurock
Finally, this paper reviewed and evaluated the 1992. Thomas Telford, London. pp 298–303
performance of empirical equations for estimating the Feng X, Jimenez R (2015) Estimation of deformation modulus
of rock masses based on Bayesian model selection and
rock mass deformation modulus (Ee). However, these Bayesian updating approach. Eng Geol 199:19–27
type studies should be supported and continued by Galera JM, Alvarez Z, Bieniawski ZT (2005) Evaluation of the
additional data and studies for other rock masses. Even deformation modulus of rock masses: comparison between
under ideal conditions, it is difficult or impossible to pressure meter and dilatometer tests with RMR predictions.
In: Gambin M, Mestat P, Baguelin F (eds) Proceedings of
decide which method is the most accurate (Zhang ISP5-PRESSIO 2005. LCPC publication Paris
2017). For these reasons, a crosscheck should be Ghamgosar M, Fahimifar A, Rasouli V (2010) Estimation of
performed before using these empirical equations for rock mass deformation modulus from laboratory experi-
an important project. ments in Karun dam. In: Zhao, Laboise, Dudt, Mathier
(eds) Proceedings of the international symposium of the
international society for rock mechanics. Taylor & Francis
Acknowledgements The authors thank Geological Engineers Group, pp 805–808
N. Gürsoy and B. Uysal for permission to use the data and the Gokceoglu C, Sonmez H, Kayabasi A (2003) Predicting the
General Directorate of Electrical Power Research Survey and deformation moduli of rock masses. Int J Rock Mech Min
Development Administration for performing the tests. Sci 40:701–710
Grimstad E, Barton N (1993) Updating the Q-System for NMT.
In: Proceedings of the international symposium on sprayed
References concrete-modern use of wet mix sprayed concrete for
underground support, Oslo, Norwegian Concrete
Association
Aksoy OC, Geniş M, Aldaş UC, Özacar V, Özer CS, Yılmaz Ö
Grimstad E, Kankes K, Bhasin R, Magnussen AW, Kaynia A
(2012) A comparative study of the determination of rock
(2002) Rock Mass Q used in designing Reinforced Ribs of
mass deformation modulus by using empirical approaches.
Sprayed Concrete and Energy Absorption. 4 th Int. Symp.
Eng Geol 131–132:19–28
on Sprayed Concrete, Davos, Switzerland
Alemdag S, Gurocak Z, Gokceoglu C (2015) A simple regres-
Gürsoy N, Uysal B (1988) Orta Çoruh Havzası Artvin Baraj
sion based approach to estimate deformation modulus of
Yeri Hidrolik Kriko Yükleme Deney Sonuçları. General
rock masses. J Afr Earth Sci 110:75–80
Directorate of Electrical Power Researchers Survey and
Alemdag S, Gurocak Z, Cevik A, Cabalar AF, Gokceoğlu C
Development Administration Report (in Turkish)
(2016) Modeling deformation modulus of a stratified sed-
Hoek E, Brown ET (1997) Practical estimates of rock mass
imentary rock mass using neural network, fuzzy inference
strength. Int J Rock Mech Min Sci 34(8):1165–1186
and genetic programming. Eng Geol 203:70–82
Hoek E, Diederichs M (2006) Empirical estimation of rock mass
Aydan Ö, Ulusay R, Kawamoto T (1997) Assessment of rock
modulus. Int J Rock Mech Min Sci 43:203–215
mass strength for underground excavations. In: Proceed-
Işık NS, Ulusay R, Doyuran V (2008) Deformation modulus of
ings of the 36th US rock mechanics symposium, New
heavily jointed-sheared and blocky greywackes by pres-
York, pp 777–786
suremeter tests: numerical, experimental and empirical
Barton N (2002) Some new Q value correlations to assist in site
assessments. Eng Geol 101:269–282
characterization and tunnel design. Int J Rock Mech Min
ISRM (International Society for Rock Mechanics) (1981) Part 1.
Sci 39:185–216
Site characterization, the quantitative description of

123
Geotech Geol Eng (2018) 36:2683–2699 2699

discontinuities in rock masses. In: Brown ET (ed) ISRM Shen J, Karakus M, Xu C (2012) A comparative study for
suggested method: rock characterization, testing and empirical equation in estimating deformation modulus of
monitoring. Pergamon Press, London rock masses. Tunn Undergr Space Technol 32:245–250
Kayabasi A, Gokceoglu C, Ercanoglu M (2003) Estimating the Sonmez H, Ulusay R (1999) Modifications to the geological
deformation modulus of rock masses: a comparative study. strength index (GSI) and their applicability to stability of
Int J Rock Mech Min Sci 40:55–63 slopes. Int J Rock Mech Min Sci 36:743–760
Mitri HS, Edrissi R, Henning J (1994) Finite element modeling Sönmez H, Ulusay R (2002) A discussion on the Hoek-Brown
of cable bolted stopes in hard rock ground mines. Presented failure criterion and suggested modification to the criterion
at the SME annual meeting, New Mexico, Albuquerque verified by slope stability case studies. Yerbilimleri
pp 94–116 (Earthsciences) 26:77–99
Mohammadi H, Rahmannejad R (2010) The estimation of Sonmez H, Gokceoglu C, Ulusay R (2004) Indirect determina-
deformation modulus using regression and artificial neural tion of the modulus of deformation of rock masses based on
network analyis. Arab J Sci Eng 15:205–217 the GSI system. Int J Rock Mech Min Sci 1:849–857
Nicholson GA, Bieniawski ZT (1990) A nonlinear deformation Sonmez H, Nefeslioglu HA, Gokceoglu C, Kayabasi A (2006)
modulus based on rock mass classification. Int J Min Geol Estimating of rock modulus: for intact rocks with an arti-
Eng 8:181–202 ficial neural network and for rock masses with a new
Palmström A, Singh R (2001) The deformation modulus of rock empirical equation. Int J Rock Mech Min Sci 43:224–235
masses—comparisons between in situ tests and indirect Tahir M, Mohammad N (2014) Prediction performance and
estimates. Tunn Undergr Space Technol 16(3):115–131 generalization of the empirical estimationof rockmass
Ramamurthy T (2004) A geo-engineering classification for deformation modulus based on rockmass classification
rocks and rock masses. Int J Rock Mech Min Sci 41:89–101 systems. Int J Sci Eng Technol 3(12):1488–1498
Read SAL, Richards LR, Perrin ND (1999) Applicability of the Zhang L (2017) Evaluation of rock mass deformability using
Hoek–Brown failure criterion to New Zealand greywacke empirical methods—a review. Undergr Space. https://doi.
rocks. In: Vouille G, Berest P (eds) Proceedings of the org/10.1016/j.undsp.2017.03.003
nineth international congress on rock mechanics, Paris, Zhang L, Einstein HH (2004) Using RQD to estimate the
August 2, pp 655–660 deformation modulus of rock masses. Int J Rock Mech Min
Serafim JL, Pereira JP (1983) Considerations on the geome- Sci 41:337–341
chanical classification of Bieniawski. In: Proceedings of
the symposium on engineering geology and underground
openings, Lisboa, Portugal, pp 1133–1144

123

View publication stats

You might also like