Вы находитесь на странице: 1из 15

Engineering Geology 185 (2015) 116130

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Rockburst laboratory tests database Application of data


mining techniques
Manchao He a, L. Ribeiro e Sousa a,b,, Tiago Miranda c, Gualong Zhu a
a
b
c

State Key Laboratory for Geomechanics and Deep Underground Engineering, Beijing, China
University of Porto, Porto, Portugal
University of Minho, Guimares, Portugal

a r t i c l e

i n f o

Article history:
Received 13 March 2014
Received in revised form 22 September 2014
Accepted 13 December 2014
Available online 18 December 2014
Keywords:
Rockburst
Experimental tests
Data mining
Rockburst index

a b s t r a c t
Rockburst is characterized by a violent explosion of a block causing a sudden rupture in the rock and is quite common in deep tunnels. It is critical to understand the phenomenon of rockburst, focusing on the patterns of occurrence so these events can be avoided and/or managed saving costs and possibly lives. The failure mechanism of
rockburst needs to be better understood. Laboratory experiments are undergoing at the Laboratory for
Geomechanics and Deep Underground Engineering (SKLGDUE) of Beijing and the system is described. A large
number of rockburst tests were performed and their information collected, stored in a database and analyzed.
Data Mining (DM) techniques were applied to the database in order to develop predictive models for the
rockburst maximum stress (RB) and rockburst risk index (IRB) that need the results of such tests to be determined. With the developed models it is possible to predict these parameters with high accuracy levels using
data from the rock mass and specic project.
2014 Elsevier B.V. All rights reserved.

1. Introduction
A large number of accidents and other associated problems occur
during construction and exploration of underground structures, and
are very often related to uncertainties concerning ground conditions.
Many researchers have collected, analyzed and published reports on accident cases in tunnels during construction and exploration (HSE, 1996; Vlasov et al., 2001; Sousa, 2006, 2010). In the
study conducted by Sousa (2010), data on accidents were collected
from the technical literature, newspapers and correspondence
with experts in the underground domain. The data were stored in
a database and the accidents were classied into different categories, providing an evaluation on their causes and their consequences. The main goal was to determine the major undesirable
events that may occur during tunnel construction, their causes
and consequences and ultimately present mitigation measures to
avoid accidents on tunnels during the construction. Different
types of events were identied and classied (Rock Fall, Collapse
and Daylight Collapse, Rockburst, Excessive Deformation, Water
Inow, Fires and Explosions). The accidents can cause loss of
lives, equipment damage and damage to the tunnel structure that
may lead to collapse.
In deep underground structures, rockburst is a frequent type of
event caused by the overstressing of rock mass or intact brittle rock,
Corresponding author at: University of Porto, Porto, Portugal. Tel.: +351 966012385.
E-mail address: sousa-scu@hotmail.com (L.R. e Sousa).

http://dx.doi.org/10.1016/j.enggeo.2014.12.008
0013-7952/ 2014 Elsevier B.V. All rights reserved.

when stresses exceed the local compressive strength of the material. It


can cause spalling or in the worst cases, sudden and violent failure of
the rock mass. Rockbursts can cause serious, and often fatal, injuries.
They are mainly dependent on the stress exerted on the rock, which increases with depth. Rockburst is also common in deep underground
mines. This phenomenon can also occur in tunnels for transportation
systems and hydroelectric projects (Sousa, 2012a).
For deep underground engineering rockburst is one of the most important accidents. They are not easy to predict. Rockburst hazard assessment is therefore a very important task and the major topic of this
paper. Laboratory tests are one way to analyze the rockburst phenomenon. In this paper, and after a comprehensive presentation of rockburst
occurrences, a description of a unique laboratory system developed at
the SKLGDUE, Beijing, of China University of Mining and Technology is
presented. In this work, the information regarding 139 of these tests
was gathered in a database. Then Data Mining techniques (Multiple
Regression, Articial Neural Networks and Support Vector Machines)
were applied to the database in order to develop predictive models for
rockburst parameters associated to these tests, namely the rockburst
maximum stress (RB) and rockburst risk index (IRB). The developed
models are based on data concerning the rock mass and specic project
and can be used to predict these rockburst indexes when it is not possible to carry out these rockburst tests or in preliminary stages of design
of underground works to obtain an approximate prediction of their
values. Different sets of input parameters were considered so that the
models can adapt to the available data. Also, the importance of each
input parameter was assessed.

M. He et al. / Engineering Geology 185 (2015) 116130

117

Table 1
Statistics of rockburst occurrence at Jinping II.
Tunnel

No rockburst
(%)

Level I
(%)

Level II
(%)

Level III
(%)

Level IV
(%)

Auxiliary A
Auxiliary B
HP no. 1
HP no. 2
HP no. 3
HP no. 4
Total HP

81.01
83.72
92.62
88.39
92.48
87.44
90.10

12.54
10.32
6.76
7.62
7.20
10.52
8.04

4.13
4.67
0.62
3.36
0.32
1.78
1.62

1.73
1.12
0.00
0.50
0.00
0.26
0.21

0.09
0.17
0.00
0.12
0.00
0.00
0.04

HP High Pressure tunnel.

Fig. 1. Jinping II hydropower scheme.

2. Rockburst occurrences
Rockburst is characterized by a violent explosion of a block causing a
sudden rupture in the rock mass and can be common in very deep tunnels. This phenomenon can occur in tunnels for transportation systems,
hydroelectric projects and mining operations, and has been more associated with mine excavations for a long time. Therefore it is critical to
understand rockburst, focusing on the patterns of occurrence so these
events can be avoided and/or managed saving costs and possibly lives
(Camiro, 1995; Kaiser, 2009; Tang et al., 2010).
All rockbursts produce seismic waves and seismic disturbances.
These seismic events have been recorded at seismological stations
many times hundreds of miles away from the origin (Tang et al.,
2010). Rockburst rst became a recognized problem in the Kolar Fold
Field of India in 1898 and by the end of 1903, 75 rockbursts had occurred with fatalities and serious injuries at a depth of about 450 m
below the surface. Investigations concluded that rockbursts were
caused by great pressures on the mine pillars (Blake, 1972). Rockburst
accidents were also reported in gold mines of Witwaterstrand, South
Africa, in the early 1900s at lower depths damaging pillars and other
mine workings. Rockbursts occur frequently in South Africa, therefore
long-term researches have been carried out systematically in the country on the mechanisms of rockbursts. In 1975, 680 accidents took place
in 31 gold mines which claimed a large number of fatalities and loss of
production (Tang et al., 2010).
Also in other types of underground structures rockburst have been
reported for instance during the construction of the Simplon hydraulic
tunnel in the Alpes region at depths greater than 2200 m, in the Shimizu
tunnel in Japan, for depths between 1000 m and 1300 m and in the
Kanestu tunnel, constructed mainly in quartz diorite, rockburst occurred at an overburden depth between 730 m and 1050 m (Tang
et al., 2010). Norwegian tunneling experience includes a signicant

number of tunnels subjected to high rock stresses. The majority of the


problems are associated with spalling due to anisotropic stresses
below steep valleys. This is found normally in road tunnels along or between the fjords under high overburden. An example is the 24.5 km
long Laerdal tunnel where moderately intense spalling and slabbing
was encountered most of the times. In some areas heavy rockbursts
caused violent ejection of sharp edged rock plates (Sousa, 2010).
Rockburst also occurred at the Gothard base railway tunnel in
Switzerland.
In China the rst rockburst was recorded in 1933 at Shengli Mine in
Fushun. Based on the available data it is estimated that over 2,000 coal
bursts occurred in 33 mines in China during 19491997. In the period
from 2001 to 2007 rockburst in deep metal mines caused more than
13,000 accidents and at least 16,000 casualties (Adoko et al., 2013).
In China many rockbursts occurred also during excavation of high
pressure, drainage and auxiliary tunnels of the Jinping II hydropower
scheme (Figure 1), (Wu et al., 2010; Feng and Hudson, 2011; Feng
et al., 2012a). This scheme is composed by four high pressure tunnels,
each with 16.67 km in length, 60 m spacing between them, two parallel
auxiliary tunnels A and B, 17.5 km long with a span of 6 m excavated by
Drill & Blast (D&B) and a drainage tunnel of about 16.73 km, with a diameter of 7.2 m excavated by a Robbins TBM and by D&B. The high pressure tunnel nos. 1 and 3 were excavated, respectively, by a Robbin TBM
and by a Herrenknecht TBM both 12.4 m in diameter. The high pressure
tunnel nos. 2 and 4 were excavated by D&B method with an equivalent
diameter of about 13 m. They were excavated in marble, sandstone and
slate strata, with a maximum overburden up to 2500 m (Figure 2), (Wu
et al., 2010; Wang et al., 2012).
A consulting Workshop took place in 2009 organized by the Chinese
Society for Rock Mechanics and Engineering about the Jinping II long
tunnels. Several reports were elaborated mainly focusing on rockburst
problem (He, 2009; Hudson, 2009; Kaiser, 2009; Qian, 2009; Sousa,
2009). Conclusions were established and suggestions were made
about TBM problems, the estimation of in situ stresses, the geometry
of the excavations when using D&B method, and modeling specially
for rockburst prediction. For rockburst the suggestions included the

Fig. 2. Simplied geological prole of the high pressure tunnel no. 1 of Jinping II project (Wang et al., 2012).

118

M. He et al. / Engineering Geology 185 (2015) 116130

Table 2
Classication of rockburst as used at Jinping II.
Rockburst
level

Description

Type of sound

Characteristics of duration

Depth of the block (m)

Impact in excavation

I
II
III
IV

Light
Moderate (mild)
Intensive (strong)
Very strong (excess of loads)

Sound of cracking
Clear sound of cracking
Sound of a strong explosion
Sound of an intensive explosion

Sporadic explosion
Long duration and not progressive with time
Fast with increase of overburden
Sudden with increase of overburden

b0.5
0.5 1.0
1.02.0
N2.0

Small
A certain impact
Reasonable impact
With large impact

establishment of the Rockburst Vulnerability Index (RVI) to be calibrated with the rockburst experience (Hudson, 2009), the establishment of
a database containing information about rockburst and the description
of the events, application of DM techniques and development of
Bayesian Networks (BN) for predicting the probability of occurrence of
rockburst, its location, depth and width, and time delay for the different
types of rockburst (Sousa, 2009).
During construction different types of rockbursts were observed at
Jinping II which permitted to describe the mechanism and to settle
criteria for rockburst (Feng et al., 2012b; Wang et al., 2012; Yan et al.,
2012). According to Wang et al. (2012), the statistics of rockburst events
is represented in Table 1. The drainage tunnel was partly excavated by
the TBM until around 6 km, and after by D&B due to the occurrence of
very strong rockbursts (Liu et al., 2011). The classication of the
rockburst is presented in Table 2, corresponding level I to light
rockburst, level II to moderate, level III to intensive and level IV to
very strong (Peixoto et al., 2011).
Rockbursts along auxiliary tunnels mainly occurred in the strata T2z
and T2b (marbles). The most intensive rockbursts in T2b are very strong,
intensive in T3 (sandstones), and moderate in the other strata. Most
rockbursts occurred within 612 m from the face and 520 h after
excavation. For the high pressure tunnels, the number and intensity
increase in spite of a higher percentage of no rockburst length as it is
presented in Table 1. The fractured face was rough or dome-shaped.
Tensile and shear failure resulted in wedge or dome-shaped fracture
surfaces with a depth that sometimes reached 1.6 m. Since the start of
the high pressure tunnels, 77 rockbursts of several levels occurred at
tunnel no. 1; about 200 happened at tunnel no. 2; about 100 at tunnel
no. 3; and more than 100 at tunnel no. 4. Since the tunnels were excavated in marbles, rockburst of levels III and IV occurred with increasing
depth. The maximum ejection distance of level IV rockbursts reached 5 m
and the depth of the crater ranged from 3 m to 5 m. Most intensive
rockburst occurred within 1030 m from the working face. The majority

Fig. 3. Schematic representation of potential for rockbursts and the effect of connement
(Castro et al., 2012).

of the rockbursts occurred on the north spandrel (13 clock position)


and at the south arch corner (710 clock position), mainly due to the principal stress directions and the geological structures (Wang et al., 2012).
There are several mechanisms by which the rock fails, originating
rockburst. The main source mechanisms are usually associated with
local underground geometry of the cavities, structural elements like pillars and the existing geology (Ortlepp and Stacey, 1994; Camiro, 1995;
M. He et al., 2012). The representation of potential rockburst phenomena
is indicated in Fig. 3. The rockburst is usually classied as a strainburst,
pillar burst or fault slip burst (Castro et al., 2012; M. He et al., 2012).
These topologies normally occur in large scale mining operations, but,
in civil works the most common phenomenon is strainbursting, although
buckling and face crushing may also occur. Also impact-induced
rockburst has to be considered for less stressed and deformed rock formations, created by blasting, caving and adjacent tunneling.
Rockburst phenomena have been extensively investigated by many
researchers based on in situ and laboratory tests and also by theoretical
approaches. Laboratory tests play an important role in understanding
rockburst mechanisms, calibration of numerical models, evaluation of
mechanical parameters, and identication of the stress state where a dynamical event may be initiated. This event can be classied according to
the potential damage, scale and violence (M. He et al., 2012; Wang
et al., 2012). In terms of damage the classication referred in Table 2
can be used. In terms of scale, rockbursts can be divided into sparse
(length of rockburst L 10 m), large-area (10 b L 20 m), and continuous rockbursts (L 20 m). According to the failure pattern, it can be divided in slabby spalling, bending failure, dome/wedge-shaped failure, and in
cavern collapse. In terms of severity of rockburst damage as a function of
failure depth rockburst can be classied as represented in Fig. 4.
The construction method seems to have an inuence on the
behavior of the excavation in what regards to rockburst. Not only the
existence of a support system that stops the violent ejection of fragments of rock is essential to guarantee safety, but also the type of construction process seems to have an effect on the severity of the
rockburst. According to experience, for the same type of conditions
and for the same rock, strain bursting is more likely to occur in a TBM
tunnel than in a D&B tunnel, as happened at the Jinping II hydroelectric
scheme (Feng et al., 2012b). However, for the surrounding rocks with
lower stress levels, external disturbances, such blasting, caving and adjacent tunneling, can also trigger rockbursts. Rockbursts can be classied in strainbursts and impact-induced bursts as indicated in Fig. 5
(M. He et al., 2012). Strainbursts can be divided into the following
sub-types: instantaneous burst, delayed burst and pillar bursts. The
impact-induced burst can also be divided into the three sub-types,
like rockburst induced by blasting or excavation, by roof collapse and
by fault slip, according to their formation mechanisms. An inuence diagram containing the factors that affect the occurrence of a rockburst is
presented in Fig. 6. The type and the strength of the rock are other important factors affecting rockburst and its severity. Rockburst occurs
more likely and with greater severity in brittle rocks. Geometry of the
cavity, particularly the shape, the existence of faults and the in situ
state of stress are also important parameters to be considered in the inuence diagram of rockburst.
Many indexes and predictive models related to rockburst can be
found in literature (Adoko et al., 2013). As stated, in the Workshop

M. He et al. / Engineering Geology 185 (2015) 116130

119

Fig. 4. Severity of rockburst damage as a function of depth of failure.


Adapted by Kaiser, 2009.

held for Jinping II, Hudson (2009) proposed the index RVI which can
provide an indication of the likelihood of rockburst occurrence. As in
other empirical systems, like RMR and Q systems, for the key variables
certain values are allocated and the index is established from these
values. The RVI index is calibrated with experience and can be rened
with future knowledge. For the Jinping II hydroelectric scheme the
parameters proposed were the following: height of overburden;
compressive strength of the intact rock; intact rock brittleness (brittle,
semi-brittle, semi-ductile, ductile); presence of folding; presence of
bedding plane separations or faults; and inputs from numerical models
(when available). Also, it is important for the calibration of the RVI indexes to take into consideration the scale effect since rockburst is stronger when the equivalent dimension of the tunnel increases. This fact
was also evident at Jinping II where the rockburst was stronger in the
high pressure tunnels than in the access tunnels (Hudson, 2009). The
depth of the rockburst failure and the volume of rock involved can be
correlated with the RVI value. Adoko et al. (2013) proposed accurate
predictive models for rockburst intensity (RBI) based on fuzzy inference
system and adaptive neuro-fuzzy inference systems (ANFIS), and eld
measurements data. Based on a literature review a database was gathered composed by 174 rockbursts. The input parameters for the models
were: the maximum tangential stress, the uniaxial compressive
strength, the uniaxial tensile strength of the rock and the elastic strain
energy index and in some cases the stress coefcient and the rock brittleness coefcient, while the RBI was the output. The ANFIS model indicates the best performance. This work used a data-driven approach i.e.
the models were derived from real data whereas many other models
present in literature use only knowledge from experts to derive rules
and models. In the present work this data-driven approach was also

used since a database was gathered and the models were developed applying specic intelligent algorithms.
As shown, brittleness of the rock is one of the major issues
concerning rockburst. Its characterization and description is essential
for the understanding of the rockburst phenomenon. However, the definition and the measurement of brittleness are not yet standardized and
many empirical relations with different approaches can be found in literature to predict rock brittleness (Gong and Zhao, 2007). Normally, it is
measured indirectly as a function of rock strength parameters like the
uniaxial compressive and the Brazilian tensile strength (Yagiz and
Candan, 2010). Yagiz (2009) proposed a brittleness index (BI) based
on the punch penetration test. Later, Yagiz and Candan (2010) proposed
two models to predict BI based on a fuzzy inference system and nonlinear regression analysis. To develop the models, which presented a relevant accuracy level, the authors gathered a database with 48 samples
using as input parameters the uniaxial compressive strength, the
Brazilian tensile strength and the unit weight of the rock.
Rockbursts are not easy to predict. Investigations using Acoustic
Emission (AE) monitoring are sometimes recommended since they
allow one to monitor the accumulation of cracking and to evaluate the
tendency for the rock to suffer rockburst (Tang et al., 2010; Wang
et al., 2012). Rockburst is affected by several factors. The stress release
is one of the factors affecting the level of these events. Therefore, the
characteristics of blasting vibration and effects should be investigated
in detail, which happened in the case of Jinping II. By optimizing the
blasting parameters and the network used in the blasting, the risk of
rockburst can be reduced (Wang et al., 2012). It is important to develop
appropriate methodologies for the support design in the tunnel sections
characterized by potential rockburst. Special reference is made for the

Fig. 5. Laboratory experiments methods based on rockburst classication (He et al., 2012a).

120

M. He et al. / Engineering Geology 185 (2015) 116130

Type and rock strength

Geometry
(Shape and equivalent
diameter)

Faults
(Folding)

Rockburst

Stress state
(Overburden & K=h/v)

Construction method
(Support & advanced
rate)

Severity
(Time delay)

Dimensions of burst
(Location)

Damage of tunnel

Fatalities & injuries

Fig. 6. Inuence diagram of rockburst.


Adapted from Sousa, 2010.

principles of rockburst prevention followed at Jinping II (Wang et al.,


2012) and also to the work developed in Canada by Kaiser and Cai
(2012) in implementing an interactive design tool for conducting
rockburst support design in tunnels under burst-prone ground. Finally
special emphasis is made for a new Constant-Resistance and LargeDeformation (CRLD) bolts developed at SKLGDUE in order to mitigate
the disturbance impacts and control rockburst (M. He et al., 2012).
These bolts can bear many abrupt loads maintaining a good supporting
performance. They have an ideally elastoplastic behavior compared
with a traditional bolt which was corroborated by experimental tests
(He et al., submitted for publication).
Risk analysis and risk management for rockburst should follow the
guidelines established by ITA (Eskesen et al., 2004), and proposed by
other authors (Sousa, 2010; Brown, 2012). One process suggested for
DUSEL (Deep Underground Science and Engineering Laboratory) cavities by Popielak and Weining (2010), involves the establishment of conceptual models, to perform numerical analysis in order to study the
potential impacts of rockburst and risk management plans as well as
strategies for its implementation. For instance, during the excavations
of high pressure tunnels of the Jinping II scheme by TBMs some construction processes were implemented for the purpose of rockburst
control namely to avoid the occurrence of very strong rockbursts and
to prevent negative impacts on safety (Wang et al., 2012). The main
idea of the plan was to release the high in situ stresses with pilot tunnels
and simultaneously these pilot tunnels served as geological exploratory
pits and the working faces were used for micro-seismic monitoring. The
pilot tests were classied in three types as represented in Fig. 7. This

approach allowed reducing the risk of occurrence of strong rockbursts.


For the high pressure D&B tunnels preventive measures were also
taken including blasting control, use of shotcrete and bolt support
(Wang et al., 2012).
The risk of rockburst can be also analyzed by using BN probabilistic
tools. The methodology for risk assessment and decision making proposed by Sousa (2012b) for tunnel projects during the design and construction phases can be implemented for the different hazards including
of course the rockburst scenario. The decision support framework can
consist of two models, namely: a rockburst (geologic) prediction
model and a decision model. This methodology was successfully applied
to tunnels of Porto Metro in Northern Portugal where three collapses
occurred during the construction (Sousa and Einstein, 2012).
3. Rockburst laboratory tests
3.1. True triaxial system
Laboratory experiments on rockburst have been carried out by researchers using several integrated static and dynamic loading and uniaxial and biaxial testing machines. However, among these laboratory
tests, no one has simulated physically the rockburst process and reproduce the circumstances of the occurrence of the rockburst. A modied
true-triaxial experimental system was developed at SKLGDUE, for
modeling in a more realistic way the rockburst process (He et al.,
2012a, 2012b). It is a system that includes the main machine, hydraulic
pressure controlling unit and data acquisition for forces, displacements,

Fig. 7. Pilot tunnels used at Jinping II (Wang et al., 2012).

M. He et al. / Engineering Geology 185 (2015) 116130

121

Fig. 8. Rockburst testing system.

acoustic emission (AE), high speed digital recording and also infrared
(IR) thermal monitoring system (Figure 8). During the test procedure,
one surface is unloaded immediately from the triaxial condition,
which simulates in an approximate way the stress path in the excavation face during tunneling. The evolution model for a rockburst can be
translated by three phases as illustrated in Fig. 9, with cracking in a vertical plan, followed by a buckling deformation and the nal rockburst
ejection (He et al., 2010b). Fig. 10 illustrates the dropping system used
in the system for the unloading of a face which is implemented through
a change of a piston movement. The stress paths used by the testing system simulate the different types of rockburst that can occur (He et al.,
2010b, 2012c).
The development of the triaxial rock test machine makes it possible
to have a better understanding of the rockburst phenomena. To illustrate the rockburst loading path, the results of a rockburst test for a
sandstone sample is presented (Figure 11). The sample was rst loaded
with three principal stresses at the same loading rate step by step to
simulate the in situ original stress state (51.8; 39.3 and 29.5 MPa). At
each loading level, the interval of loading time was 5 min. Then the minimum horizontal stress on one surface of the sample was unloaded.
Therefore the simulation of the rockburst started. A detailed analysis
of the results of this test is presented in the publication of He et al.

(2012a), regarding rockburst time, classication and typical results of


the failure process.
3.2. Measuring systems
The measuring system for the rockburst experiments includes
(M. He et al., 2012):
3.2.1. Data acquisition system
The changes in forces and displacements that occurred during the
rockburst process are registered by a dynamic strain amplier and a
portable data acquisition instrument, which consists of sensors, ampliers, a data acquisition instrument, a computer and the appropriate
processing software. The dynamic data collection device can achieve a
data acquisition speed as high as 100 kps. Fig. 12 presents the data
acquisition apparatus.
3.2.2. High-speed image recording system
The system is equipped with a high-speed digital camera in order to
record the kinetic characteristics of the rock fragments ejected during
the rockburst event, providing excellent material for the analysis of

Fig. 9. Evolution model for a rockburst (He et al., 2012a).

122

M. He et al. / Engineering Geology 185 (2015) 116130

Fig. 10. Illustration of the dropping system for load bar and loading plate (He et al., 2010b).

Fig. 12. Data acquisition system (He et al., 2012a).

the mechanism of rockburst. It consists of high-speed cameras, capture


cards, disk for storage of the images and the control software.
3.2.3. AE monitoring system
The system is equipped with an AE monitoring instrument in order to
obtain the features of the development and changes in microcracks and
the characteristics of energy release during the rockburst process. The
AE includes an acquisition card, a continuous current source, a preamplier and an AE sensor (Figure 13) (He and Zhao, 2013). The AE characteristics are very important and can be used and analyzed in order to
understand the crack propagation phenomena in rocks, as represented
in Fig. 14 for the instantaneous rockburst of the Carrara marble. The signal
and images can be investigated from the tests. The evolution and frequency is important to investigate the failure characteristics of the samples. By
signal processing using fast Fourier transformation analysis, the time frequency spectrum for the AE signals can be obtained (Jia, 2013).
3.2.4. IR thermal monitoring system
The IR thermography system shows the changes in the surface
temperature of a sample which lead to uneven heat transfer. The
changes of temperature can be analyzed using an IR thermal monitoring system providing data for the changes in the characteristics
of the sample and consequently to better understand the mechanism
of rockburst. IR thermography has been extensively used as a nondestructive and non-contact technique to inspect cracks or defects inside the materials (Gong et al., 2009; He et al., 2009, 2010a). IR
thermography, together with such image processing procedures as
feature extraction and frequency-spectral analysis, were used for visualization and characterization of the mechanical and structural responses of rock masses (He et al., 2010a). The IR detection scheme
used in the rockburst tests is illustrated in Fig. 15.

4. Database with laboratory test results


4.1. Organization of the database
From 2004 to 2012 a large number of rockburst laboratory tests
were performed at SKLGDUE. A survey was conducted through several
reports containing results from rockburst tests processed during these
years and a database was elaborated with 139 tests from 4 countries,
mainly from China (88.5%), and also from Italy (5.0%) with samples
from Carrara quarries, from Canada (5.0%) with granite samples from
Creighton mine, and from Iran (1.5%) from samples obtained in petroleum wells. The tested samples were mainly coal (43), sandstone (28)
and granite (25), but other rocks were also tested. Fig. 16 illustrates
the relative distribution of the tests for the different types of rocks.
A form was elaborated and a great number of elds were considered,
namely: (1) location of the test; (2) dimensions of the sample in length,
width and height; (3) rock material; (4) main minerals and cracks;
(5) stresses before loading; (6) stresses during the test; (7) characteristics of the rockburst test; and (8) critical depth. In Table 3, the different
information inserted for each eld is referred.
The samples had a prismatic shape with an average length of 59 mm,
with a minimum of 39 mm and a maximum of 111 mm; the width
corresponding to the face to be unloaded had 33 mm in average,
61 mm as maximum and as 19 mm minimum; the height was 148 mm
in average, 229 mm in maximum and 98 mm in minimum. The volume
was 312 cm3 in average, 1186 cm3 in maximum and 72 cm3 in minimum.
The main minerals present in the samples were clay, carbon,
quartz, calcite and feldspar. The information about the percentage
of these minerals in the samples existed for almost all the cases.
For the existence of cracks the following designation was adopted:
0 no information; 1 few cracks; 1.5 cracks in some parts;
2 a lot of cracks.
All the tests were of the strainburst type. The rockburst critical
depth (He) was calculated assuming a simplied circular shape in
the crown of the tunnel, a concentration factor for the stresses
equal to 2 and a specic weight of 27 kN/m 3 for the overburden
rock mass, by the following expression:
He 18:52 RB

where RB is the rockburst maximum stress obtained in the test.


A rockburst risk index (IRB) was also calculated following the formula (He, 2009):

Fig. 11. Sandstone rockburst loading path (He et al., 2012b).

IRB

H
H
0:054
:
He
RB

M. He et al. / Engineering Geology 185 (2015) 116130

123

Fig. 13. AE detection system (Jia, 2013).

4.2. Statistical results and classications


Some of the most relevant statistics concerning the database parameters are presented in Table 4, namely the mean values of some of the
main parameters for different rock types.
UCS values have an average of 65.5 MPa for all samples, with a minimum average value of 8.0 MPa for shale and a maximum of 234.1 MPa
for peridotite. The average values for the most representative rocks
were 11.9 MPa for coal, 113.9 MPa for granite, and 83.4 MPa for
sandstone.
The deformability modulus (E) has an average of 22.3 GPa
for all samples, with a minimum average value of 2.0 GPa for mudstone and a maximum of 74.1 GPa for peridotite. The average values

for the most representative rocks were 2.4 GPa for coal, 33.8 GPa for
granite, and 24.5 GPa for sandstone. In relation to the Poisson ratio,
the average value is equal to 0.25, with a minimum average value
of 0.18 for marble and a maximum of 0.37 for mudstone and for
shale.
Regarding the depth (H) of the samples, the average value is
equivalent to 678 m, with a minimum average value of 250 m for dolomite and a maximum of 3375 m for limestone. For the critical depth He,
the average value is equal to 1529 m, with a minimum average value of
343 m for shale and a maximum of 2989 m for granite. For average
values and the most representative rocks H and He were equal,
respectively, to 507 m and 352 m for coal, 700 and 2989 m for granite,
and 854 and 1878 m for sandstone.

Fig. 14. The AE spectral on points and micro-crack structure characteristics after an instantaneous rockburst test (He et al., 2012c).

124

M. He et al. / Engineering Geology 185 (2015) 116130


Table 3
Fields considered in the database.
Field

Topics

Location of the test


Dimensions of sample
Rock material

Location sample; depth (m); country; date.


Code, length; width; height (mm); volume (cm3)
Type of rock; RQD; UCS (MPa); specic weight
(g/cm3); E (GPa) elastic modulus; Poisson ratio.
% clay; % feldspar; % calcite; % carbon; existence of cracks
v vertical in situ stress; h1 horizontal in situ
stress; h2 horizontal in situ stress in the face to be
unloaded; I1 (rst invariant of the stresses); I2
(second invariant of the stresses); I3 (third invariant
of the stresses).
Rockburst maximum stress (RB); maximum stress
axis; loading rate in MPa/s; unloading rate for
vertical stresses in MPa/s; unloading times.
Type of burst position; duration of the test in
minutes; time of burst delay (minutes); mainly
shape of fragments.
Critical depth; rockburst risk index.

Main minerals and cracks


Stresses before loading
(MPa)

Stresses during tests


(MPa)
Fig. 15. Scheme of an IR detection system on a rockburst tests (He et al., 2012a).

Characteristics of
rockburst test

The rockburst maximum stresses (RB) obtained in the tests have an


average value for all samples of 82.6 MPa, with a minimum of 16.5 MPa
for mudstone and a maximum of 161.4 MPa for granite. The average
value of coal was equal to 19.0 MPa and for sandstone 101.4 MPa.
Fig. 17 shows the distribution of the rockburst stresses obtained in all
samples. A large number of samples for soft rocks are recognized with
values lower than 40 MPa.
The percentage of some minerals namely clay, calcite, feldspar,
quartz and carbon were obtained from the tested samples. The average
values for all samples and for the highest average values were: clay
11.9% and 31.1% (mudstone); calcite 8.7% and 99.5% (limestone);
feldspar 15.4% and 78.7% (peridotite); quartz 22.1% and 51.6%
(sandstone); carbon no value and 93.2% (coal).
The rockburst risk index (IRB) was calculated for all the samples in
accordance with Eq. (2). The average value for all the samples is around
1.08. Very different values were calculated for the existent rock types,
with low values for basalt, granite and slate, and high values for shale,
coal, mudstone and limestone, varying from 1.5 to 3.5. For the most representative rocks the following average values were obtained: 1.65
(coal); 0.25 (granite); and 0.62 (sandstone). Taking these values into
consideration the rockburst index was classied in four classes as indicated in Table 5. The distribution for the classes for all tests and for the
most representative rock formations (coal, granite and sandstone) are
indicated in Table 6. Low IRB values were obtained in 56% of the samples,
and very high values were 13% of the total. Fig. 18 shows a plot with the
number of samples distributed by IRB.
Some representative graphs were also obtained relating the IRB with
other parameters. Fig. 19 represents the relation between IRB with RB. A

Critical depth (m)

Basalt
Cloud Schist

4%
4% 9%

3%

Coal
Dolomite
Granite

20%

Limestone
31%

1%
3%

Marble
Mudstone

5%

Peridotite
1%

18%

Sandstone
1%

Shale
Slate

Fig. 16. Rockburst tests in different rock types.

relation between IRB and K (ratio between average horizontal stresses


and vertical stress due to overburden) is represented in Fig. 20. The following relation was achieved, with a correlation coefcient of 0.904:
logK 0:1260:780logIRB

5. Application of DM techniques
5.1. General
Nowadays, due to the advances in information and communication
technologies, there is an extraordinary expansion of data generation
that needs to be stored. The data can hold valuable information, such
as trends and patterns that can be used to improve decision making
and optimize processes. Due to the great potential of this subject
there has been an increasing interest in the Knowledge Discovery
from Databases (KDD) and DM elds that led to the fast development
of electronic data management methods (Fig. 21).
DM is a relatively new area of computer science that lies at the intersection of statistics, machine learning, data management and databases,
pattern recognition, Articial Intelligence (AI) and other areas. DM
consists in the searching and inference of patterns or models in the
data which can represent useful knowledge. There are several DM
techniques, each one with its own purposes and capabilities, namely
Decision Trees and Rule Induction, Neural Networks, Support Vector
Machines and BN, Learning Classier Systems and Instance-Based Algorithms (Berthold and Hand, 2003; Miranda et al., 2011). The increasing
interest on DM led to the necessity of dening standard procedures to
carry out this task. In this context, the two most used methodologies
in DM are the CRISP-DM (Cross-Industry Standard Process for Data Mining) and the SEMMA (Sample, Explore, Modify, Model, and Assess),
(Bulkley et al., 1999; Chapman et al., 2000; Miranda, 2007).
In the eld of geotechnics many advances have been observed, like
in constitutive models, test equipment and instruments, probabilistic
methodologies to deal with uncertainty, numerical tools and many
others. However, the tools and techniques normally used to analyze
geotechnical data did not undergo signicant development during the
last decades. The establishment of standard organization and representation methods of geotechnical data in an electronic format is a subject
whose importance was recognized by the geotechnical community,
particularly with the creation of Joint Committee JTC2 from the Societies
ISRM, ISSMGE and IAEG. In the events organized by this Committee
some developments were presented and the relevance of the application of DM techniques was emphasized.

M. He et al. / Engineering Geology 185 (2015) 116130

125

Table 4
Statistical values for different relevant parameters. 1 Basalt; 2 Schist; 3 Coal; 4 Dolomite; 5 Granite; 6 Limestone; 7 Marble; 8 Mudstone; 9 Peridotite; 10 Sandstone;
11 Shale; 12 Slate.
Parameters

UCS (MPa)
E (GPa)

H (m)
He (m)
RB (MPa)
IRB

All rocks

Mean
Mean
Mean
Mean
Mean
Mean
Mean

65.5
22.3
0.25
678
1529
82.6
1.08

Rocks
1

10

11

12

143.9
62.2
0.21
400
2585
139.6
0.16

1000
1413
76.3
0.81

11.9
2.4
0.28
507
352
19.0
1.65

250
2315
125.0
1.11

113.8
33.8
0.24
700
2989
161.4
0.27

24.1
9.7
0.24
3375
1278
69.0
3.49

58.4
43.1
0.18
500
1561
84.3
0.36

11.5
2.0
0.37
910
306
16.5
3.02

234.1
74.1
0.20
1554
2317
125.1
0.68

83.4
24.5
0.24
854
1878
101.4
0.62

8.0
3.0
0.37
500
343
18.5
1.54

58.8
13.5

500
1587
85.7
0.34

However studies concerning the application of formal KDD processes are rare in geotechnical engineering activities. The results of KDD
processes were recently presented concerning geotechnical data gathered in two important underground works in predominantly granite
rock masses, Venda Nova II and Bemposta II hydroelectric schemes
(Miranda and Sousa, 2012). New alternative regression models were
developed using several DM techniques for the analytical calculation
of strength and deformability parameters and geomechanical indexes.
These models were built up considering different sets of input data,
allowing their application in different scenarios of data availability.
Most of the models use less information than the original formulations
but maintain a high predictive accuracy, which can be useful in the preliminary design stages in any case where geological/geotechnical information is limited. The application of DM also provided insight to the
most inuential parameters for the behavior of the rock mass of interest.
An important application was performed for the DUSEL laboratory,
located at the former Homestake gold mine (McPherson et al., 2003).
The laboratory is seen as a multi-discipline facility with particle physics
providing the lead but other disciplines being a signicant part of the
facility, including geomicrobiology, geosciences, and geoengineering.
At DUSEL, a large database of geotechnical data was already produced.
The geotechnical database was analyzed using these innovative DM
tools, including BN models, and new and useful models were developed
for the prediction of geomechanical parameters (Sousa et al., 2012).
The main issue of the DM task is building a model to represent data.
In this step of the KDD process, learning occurs by adopting a search
algorithm for training. This process occurs over a training set until a
given criteria is met. After training, the model is built and its quality is
normally evaluated over a test set not used for training. There are
several different models but there is no universal one to efciently
solve all the problems. Each one presents specic characteristics
(advantages and drawbacks) which make them better suited in a certain case. The modeling techniques applied in this study were: Multiple
Regression MR, Articial Neural Networks ANN and Support Vector
Machines SVM.

In regression problems, the goal is to estimate the model which minimizes an error measurement between real and predicted values considering N examples. In this work the used error measures were the
mean absolute error (MAE) and the root mean squared error (RMSE),
as dened by equations:
N
X
jei j

Mean Absolute Error : MAE

i1

Root Mean Squared Error : RMSE

v
u N
uX 2
u
ei
u
t i1

where N is the total number of samples and ei is the difference between


the real value and the estimated value by the model.
The MR is similar to the simple regression being the main difference
the number of independent variables involved. The simple regression
involves only one independent variable whereas the MR involves
several independent variables and establishes a relationship among
them and the dependent variable (Berry and Linoff, 2000). ANN was
conceived to imitate the biological networks of neurons found in the
brain. They are formed by groups of connected articial neurons in a
simplied but very similar structure to the brain neurons. Like the biological structures, these networks can be trained and learn from a set
of examples to nd solutions to complex problems, recognize patterns
and predict future events. The acquired knowledge can then be generalized to solve new problems. This means that they are self-adaptive systems. Multi-layer networks are the most common type of network and
are composed by different parallel layers of neurons (Haykin, 1999).
Fig. 22 shows the scheme of a multi-layer network used in this work.
The rst is the input and the last the output layer. Intermediate ones
are called hidden layers. There are several architectures or topologies
for the network, each one with its own potentialities, but the most
used is the multilayer feed-forward. In this type of network connections
are unidirectional (from input to output) and there are no connections
between neurons in the same layer forming an acyclic network.
The SVM (Cortes and Vapnik, 1995) were originally developed to be
used in classication problems i.e., to model discrete labeled outputs.
After the introduction of the -insensitive loss function, it was possible
to apply SVM to regression problems. The basic idea of the SVM is to
transform the input data into a high-dimensional feature space by
using a nonlinear mapping, which is normally unknown, using a set of
Table 5
Classication in accordance with the rockburst index.

Fig. 17. Histogram with the distribution of the rockburst maximum stress RB.

Value of IRB

Classication

IRB b 0.6
0.6 b IRB 1.2
1.2 b IRB 2.0
IRB 2.0

Low
Moderate
High
Very high

126

M. He et al. / Engineering Geology 185 (2015) 116130

Classication for IRB

All samples

Coal

Granite

Sandstone

Low
Moderate
High
Very high

77
17
26
18

3
10
18
12

25
0
0
0

22
3
2
1

RB / MPa

Table 6
Distribution of samples for the different classes of rockburst index.

functions known as kernels. Then, the SVM nds the best hyperplane of
linear separation within the feature space (Miranda et al., 2011; Sousa
et al., 2012).
5.2. Modeling and evaluation

0.6

Among the DM algorithms used in this study, only MR provides an


equation relating to the output and the input variables. The modeling
software was the R program environment (R Development Core Team,
2010) which is an open source freeware statistical package. Within
this framework a specic library RMiner (Cortez, 2010) was used
which allows applying several algorithms and evaluating their behavior
under a different set of metrics. In this work the goal was to predict the
RB and the IRB using different groups of input variables. Hence, the
models were developed considering two groups of input parameters
namely a main and a secondary group. The main group contains the variables that may have a signicant inuence on the prediction capacity of
the models and in the secondary group the ones with expected marginal
inuence. In Table 7 the four data groups considered as input parameters are presented.
Before tting the ANN and SVM models, the data were normalized to
a zero mean and one standard deviation and the outputs were postprocessed with the inverse transformation (Hastie et al., 2009).
The performance of the models was accessed by using 20 runs under
20-fold cross validation approach (Hastie et al., 2009). Under this
scheme, the data are divided into 20 different subsets, being one used
to test the model and the remaining to t it, which means that all data
are used for training and testing. The mean and condence intervals
for the error measures are then computed considering the results of
all the runs and a 95% condence interval of a T-student distribution.
These statistical measures dene the range of expected errors for future
predictions of the nal model, which is estimated using all the data for
training.
Often, complex models, such as ANN and SVM, are viewed as blackboxes which is their main drawback. However, it is possible to have
an insight on how they work by applying a sensitivity analysis and
studying the most important input variables on the prediction of the
target variable (Kewley et al., 2000; Cortez and Embrechts, 2011).
Such procedure is carried out analyzing the model responses when a
given input is changed. Such quantication is determined by keeping
all the inputs constant, except one that is varied through its range of

1.2

2.0

Fig. 19. Distribution of IRB vs. RB.

values. A parameter with a strong inuence induces a high variance in


the model output whereas a parameter with low importance induces
a low variance.
5.3. Model for RB
The results for group G1, with the main variables, show good performances for all the developed models with error measures uctuating
approximately between 20 MPa and 32 MPa for a variable that ranges
from 10.6 MPa to 255.5 MPa (with a mean value of 82.6 MPa). Also
the values of the correlation coefcient are rather high and near 0.9.
All the models present similar results as shown in Table 8, where the
metrics for each model are presented for the groups of variables G1
and G2, considering the parameters MAE and RMSE, and R as the correlation coefcient. However, the model based on the SVM slightly outperforms the remaining. Fig. 23 presents the plot of experimental vs.
predicted values for the SVM model. A fairly good distribution of values
around the 45 slope line can be observed pointing out for a good performance of the model.
The most important parameters in the prediction of RB are UCS and
h1, followed by Depth and v and nally by E and h2 with relative importance levels below 10%. Fig. 24 presents the importance of the variables for the SVM model. The evaluation of RB for group G1 is
presented in Table 9.
Eq. (6) translates the obtained multiple regression model for RB.
RB 9:1320:013  H 0:381  UCS 0:364  E 1:211
 h1 0:069  h2 0:365  v

Considering also the secondary set of input parameters (group G2)


the results slightly improve and the SVM model continues to have the

Number of Samples

80

60

Coal
Others

40

20

0
<0.6

0.6-1.2

1.2-2

>2

IRB

Fig. 18. Number of samples distributed by rockburst index IRB.

Fig. 20. Relation between the rockburst indices IRB and K.

M. He et al. / Engineering Geology 185 (2015) 116130

127

Fig. 21. Phases of a KDD process (Fayyad et al., 1996).

best performance. In terms of importance, the main parameters are almost the same as in the previous case, namely UCS and h1, followed
by Depth, v and Q. As it can be stated, in this case the importance of
E and h2 drops and Q appears with a signicant relative importance.
The results are presented at Table 10.

5.4. Model for IRB


For Group G3 and considering all the tests, the model based in the
ANN presents excellent results (Table 11). The SVM model also has
very good results and the MR model presents the worst results. It is important to emphasize that the relation between the input variables and
IRB is highly non-linear which explains the excellent prediction capacity
of the ANN model. Fig. 25 presents the plot of experimental versus predicted values for the ANN model. An excellent distribution of values
around the 45 slope line can be observed pointing out for a good performance of the model. The most important variables are H and RB
followed by E, K and UCS. The importance of the variables in the ANN
model is illustrated in Fig. 26.
The evaluation of IRB for group G3 is presented at Table 12.

Eq. (7) translates the obtained multiple regression model for IRB
group G3.
4

IRB 1:432 8:035  10  H8:429  10


 E0:007  RB 0:074  K

 UCS0:009

Considering additional secondary variables (group G4) the same tendency is observed in terms of the performance of the models. Table 13
illustrates for this group the evaluation of IRB. The most important variables are also H and RB, followed by the volume of the samples, UCS
and K. Comparatively with the previous group E loses importance and
in the other hand the secondary variable volume of the sample presents
a signicant importance.
Concerning group G3 the performance of the models is quite
different. The MR model presents the worst performance with considerably high error values, between 0.458 and 0.602 for MAE and RMSE,
respectively for a parameter with a mean value of 0.954 and ranging
from 0.046 to 5.207. On the other hand, the ANN model presents
excellent results translated by very low error values and an R coefcient
very near to the unity. The SVM model also presents good results but it
is clearly outperformed by the ANN model. For group G4, that considers
also a secondary set of parameters, the results are similar in relation
to the previous case, although a slight decrease on the model's perforTable 8
Metrics for the evaluation of RB.
DM tech.

RB
G1

MR
ANN
SVM

G2

MAE

RMSE

MAE

RMSE

21.20
21.32
20.49

31.08
32.54
30.51

0.873
0.864
0.879

20.46
21.21
19.64

30.43
31.48
29.51

0.880
0.636
0.888

Fig. 22. Scheme of the multi-layer ANN.

Table 7
Generated groups for evaluation of RB and IRB.
Parameter

Depth (m)
Uniaxial compressive strength (MPa)
Deformability modulus (GPa)
Horizontal in situ stress (loading face) (MPa)
Horizontal in situ stress (unloading face) (MPa)
Vertical stress due to overburden
Percentage of clay (%)
Percentage of quartz (%)
Percentage of feldspar (%)
Percentage of calcite (%)
Percentage of carbon (%)
Volume of the sample
Rockburst maximum stress (MPa)
Ratio between average horizontal stresses and v

Symbol

H
UCS
E
h1
h2
v
Cl
Q
F
Ca
Cb
Vol
RB
K

RB

IRB

G1

G2

G3

G4

Y
Y
Y
Y
Y
Y

Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y

Y
Y
Y

Y
Y

Y
Y
Y

Y
Y

Y Yes; No; G1 Group 1; G2 Group 2; G3 Group 3; G4 Group 4.

Fig. 23. Experimental versus predicted RB values for SVM model.

128

M. He et al. / Engineering Geology 185 (2015) 116130


Table 11
Metrics for the evaluation of IRB.
DM tech.

IRB
G3

MR
ANN
SVM

Fig. 24. Importance of variables for predicting RB using SVM algorithm.

mance is observed. In fact, the input parameters that were added to the
process could not improve the model's performance due to their low
importance in relation to IRB. The only exception is the input parameter
volume of the sample that replaced E in the order of importance.

6. Final considerations

G4

MAE

RMSE

MAE

RMSE

0.458
0.069
0.191

0.602
0.114
0.408

0.789
0.989
0.991

0.446
0.094
0.246

0.606
0.207
0.479

0.786
0.976
0.872

With the available data composed by 139 laboratory rockburst tests,


DM techniques were applied using MR, ANN and SVM algorithms, in
order to obtain predictive models that can improve the knowledge
about rockburst through the laboratory rockburst tests. The developed
models can be used to infer the rupture stress RB and also the rockburst
index IRB. Two sets of input parameters were used considering the most
important variables and considering additional secondary ones. The results for RB emphasized the importance of UCS and the horizontal in
situ stresses in the face not unloaded, followed by the depth of the
sample and the vertical estimated in situ stresses. Using a multiple regression algorithm an equation was determined. All the developed
models presented excellent results, however the model based on the
SVM algorithm presents the best performance. The models developed
for IRB presented excellent results when ANN algorithm was used,
which translates the highly non-linear relation with the input variables.

Failure mechanism of rockburst needs to be well understood. The


laboratory tests developed at SKLGDUE provide important information
on the subject and were described in detail. A large number of tests
were performed and it was decided to create a database with the obtained information. Special reference was made to the introduction of
a rockburst index IRB and a classication for the rockburst in accordance
with this index. Important relations were achieved with the index and
the rockburst maximum stresses obtained in the tests RB, as well as
with the ratio K between the average in situ horizontal stresses and
the vertical stresses due to overburden.

Table 9
Evaluation of RB for group G1 of variables.
Parameters

DM technique
MR

ANN

SVM

H
UCS
E
h1
h2
v

0.130
0.276
0.084
0.354
0.025
0.134

0.140
0.212
0.109
0.256
0.098
0.183

0.117
0.254
0.094
0.272
0.126
0.136

MR

ANN

SVM

0.144
0.173
0.029
0.230
0.028
0.079
0.042
0.121
0.044
0.076
0.017
0.016

0.122
0.162
0.042
0.187
0.040
0.093
0.040
0.098
0.058
0.065
0.038
0.056

0.104
0.210
0.046
0.189
0.078
0.073
0.049
0.067
0.054
0.024
0.035
0.070

Fig. 25. Experimental versus predicted IRB values for ANN model.

Table 10
Evaluation of RB for group G2 of variables.
Parameters

H
UCS
E
h1
h2
v
Cl
Q
F
Ca
Cb
Vol

DM technique

Fig. 26. Importance of variables for predicting IRB using ANN model.

M. He et al. / Engineering Geology 185 (2015) 116130

Acknowledgments

Table 12
Evaluation of IRB for group G3 of variables.
Parameters

H
UCS
E
RB
K

129

DM technique
MR

ANN

SVM

0.449
0.034
0.120
0.320
0.077

0.321
0.089
0.057
0.430
0.104

0.261
0.148
0.212
0.285
0.094

The most important variables are the depth of the samples H and RB,
followed by UCS and K. The volume of the samples is also important
when considering the secondary variables which can be related with
the scale effect.
AE results obtained during laboratory rockburst tests were not yet
analyzed using these DM techniques. It is planned to perform in the future in order to bring new approaches to the study of the rockburst
phenomenon.
Due to the nature of the rock and the discontinuity nature of the rock
masses, the evaluation of the mechanical strength characteristics is inuenced by the dimensions of the tested volumes. Although rockburst
is characterized by a violent explosion of a block of rock and consequently this phenomenon depends more of the nature of the rock, a
study was already initiated with selected cases of rockburst that
occurred during the construction of several tunnels covering several hydroelectric projects and mines. A rst step is intended to complete the
database with more cases that occurred at the megaproject of Jinping
II and then DM technique will be applied to this particular project in
order to identify the more relevant parameters and establish models
for some of the parameters. Several rockburst tests were also performed
for Jinping II site particularly in marbles. In these formations a large
number of in situ rockburst events occurred. Using probabilistic approaches a method for extrapolating laboratory data to eld scale will
be evaluated.
The application of various AI techniques on existing in situ data, in
particular DM and BN, identied the importance of various parameters
involved in the rockburst phenomenon. Special attention was given to
the development of inuence diagrams and the use of BN, which
allow replacing with advantage, other techniques to better address the
uncertainties involved in the phenomenon of rockburst. With the enlarged database more complex studies will be implemented regarding
the assessment of rockburst and its consequences. The studies already
performed using laboratory rockburst tests can be correlated with the
rockbursts that occurred in situ and the scale effect can be analyzed in
detail. The particular situation that occurred at Jinping II hydroelectric
scheme will be emphasized due to the existence of important information obtained with laboratory rockburst tests and the occurrence of
rockburst in situ. The establishment of a new rockburst coefcient is
intended to be calibrated with the obtained results.

Table 13
Evaluation of IRB for group G4 of variables.
Parameters

DM technique
MR

ANN

SVM

H
UCS
E
RB
K
Cl
Cb
Vol

0.422
0.023
0.065
0.249
0.074
0.020
0.022
0.125

0.219
0.101
0.059
0.346
0.071
0.058
0.079
0.072

0.201
0.110
0.053
0.246
0.028
0.089
0.042
0.130

The authors want to express their acknowledgments to the support


from the State Key Laboratory for Geomechanics and Deep Underground
Engineering (Beijing) and China University of Mining and Technology to
the project entitled Risk Assessment Activities Applied to Slope Stability,
Rockburst and Soft Rocks at the State Key Laboratory for Geomechanics
and Deep Underground Engineering (Beijing) of China University of
Mining and Technology. Also the authors acknowledge the support
of the State National Basic Research Program of China (973 Project
no. 2010CB226800).

References
Adoko, A., Gokceoglu, C., Wu, L., Zuo, Q., 2013. Knowledge-based and data-driven fuzzy
modeling for rockburst prediction. Int. J. Rock Mech. Min. Sci. 61, 8695.
Berry, M., Linoff, G., 2000. Mastering data mining: the art and science of customer relationships management. John Wiley & Sons, Inc., USA.
Berthold, M., Hand, D., 2003. Intelligent data analysis: an introduction. Second edition.
Springer.
Blake, W., 1972. Rock-burst mechanics. Q. Colorado Sch. Min. 67 (1), 64.
Brown, E.T., 2012. Risk assessment and management in underground rock engineering
an overview. J. Rock Mech. Geotech. Eng. 4 (3), 193204.
Bulkley, J., Gayle, S., Hicks, B., Stephens, R., 1999. Adding the where to the who. 24th
SUGI SAS Users Group International conference, paper 173, Miami, p. 3.
Camiro, 1995. Rockburst research handbook. CAMIRO mining division, Canadian
Rockburst Research Programe, 19901995 6 vol. p. 977 (Sudbury).
Castro, L.M., Bewick, R.P., Carter, T.G., 2012. An overview of numerical modelling applied
to deep mining. In: Sousa, Vargas, Fernandes, Azevedo (Eds.), Innovative Numerical
Modeling in Geomechanics. CRC Press, London, pp. 393414.
Chapman, P., Clinton, J., Kerber, R., Khabaza, T., Reinartz, T., Shearer, C., Wirth, R., 2000.
CRISP-DM 1.0. Step-by-step data mining guide. SPSS Inc., p. 73.
Cortes, C., Vapnik, V., 1995. Support vector networks. Machine Learning 20(3). Kluwer
Academic Publishers, pp. 273297.
Cortez, P., 2010. Data mining with neural networks and support vector machines using
the R/rminer tool. In: Perner, P. (Ed.), Advances in Data Mining. Applications and
theoretical aspects. Proc. of 10th Industrial Conference on Data Mining, Berlin,
Germany, Lecture Notes in Computer Science. Springer, pp. 572583.
Cortez, P., Embrechts, M., 2011. Opening black box data mining models using sensitivity
analysis. 2011 IEEE Symposium on Computational Intelligence and Data Mining
(CIDM 2011), pp. 341348.
Eskesen, S., Tengborg, P., Kampmann, J., Veicherts, T., 2004. Guidelines for tunnelling risk
management: ITA, working group no. 2. Tunn. Undergr. Space Technol. 19, 217237.
Fayyad, U., Piatesky-Shapiro, G., Smyth, P., 1996. From data mining to knowledge
discovery: an overview. In: Fayyad (Ed.), Advances in Knowledge Discovery and
Data Mining. AAAI Press, The MIT Press, Cambridge MA, pp. 471493.
Feng, X., Hudson, J., 2011. Rock engineering design. Taylor & Francis, London, p. 468.
Feng, X., Jiang, Q., Sousa, L.R., Miranda, T., 2012a. Underground hydroelectric
powerschemes. In: Sousa, Vargas, Fernandes, Azevedo (Eds.), Innovative Numerical
Modeling in Geomechanics. CRC Press, London, pp. 1350.
Feng, X., Chen, B., Li, S., Zhang, C., Xiao, Y., Feng, G., Zhou, H., Qiu, S., Zho, Z., Chen, D., Ming,
H., 2012b. Studies on the evolution process of rockbursts in deep tunnels. J. Rock
Mech. Geotech. Eng. 4 (4), 289295.
Gong, M., Zhao, J., 2007. Inuence of rock brittleness on TBM penetration rate in
Singapore granite. Tunn. Undergr. Space Technol. 22, 317324.
Gong, W.L., Zhao, H.Y., An, L.Q., Mao, L.T., 2009. Temporal and spatial analysis of infrared
images from water jet in frequency domain based on DFT. J. Beijing Univ. Aeronaut.
Astronaut. 34 (6), 690694 (2008).
Hastie, T., Tibshirani, R., Friedman, J., 2009. The elements of statistical learning: data mining, inference, and prediction. 2nd edition. Springer-Verlag.
Haykin, S., 1999. Neural networks a comprehensive foundation. 2nd edition. PrenticeHall, New Jersey.
He, M.C., 2009. The mechanism of rockburst and its countermeasure of support. Int. Consultation Report for the Key Technology of Safe and Rapid Construction for Jinping II
Hydropower Station High Overburden and Long Tunnels, Jinping II, pp. 2328.
He, M.C., Zhao, F., 2013. Laboratory study of unloading rate effects on rockburst. J. Disaster
Adv. 6 (9), 1118 (September).
He, M.C., Gong, W.L., Li, D.J., Zhai, H.M., 2009. Physical modeling of failure process of the
excavation in horizontal strata based on IR thermography. Int. J. Rock Mech. Min.
Sci. Technol. 19 (6), 689698.
He, M.C., Jia, X., Gong, W.L., Faramarzi, L., 2010a. Physical modeling of an underground
roadway excavation in vertically stratied rock using infrared thermography. Int.
J. Rock Mech. Min. Sci. 47, 12121221.
He, M.C., Miao, J.L., Feng, J.L., 2010b. Rockburst process of limestone and its acoustic emission characteristics under true-triaxial unloading conditions. Int. J. Rock Mech. Min.
Sci. 47, 286298.
He, M., Xia, H., Jia, X., Gong, W., Zhao, F., Liang, K., 2012a. Studies on classication, criteria
and control of rockbursts. J. Rock Mech. Geotech. Eng. 4 (2), 97114.
He, M.C., Jia, X.N., Gong, G.J., Zhao, F., 2012b. A modied true triaxial test system that allows a specimen to be unloaded on one surface. In: Kwasniewski, Li, Takahashi (Eds.),
True Triaxial Testing of Rocks. CRC Press, London, pp. 251266.

130

M. He et al. / Engineering Geology 185 (2015) 116130

He, M.C., Jia, X.N., Coli, M., Livi, E., Sousa, L.R., 2012c. Experimental study on rockbursts in
underground quarrying of Carrara marble. Int. J. Rock Mech. Min. Sci. 52, 18.
He, M.C., Gong, W., Wang, J., Qi, P., Tao, Z., Du, S., 2014. Development of a novel energyabsorbing bolt with extraordinarily large elongation and constant resistance. Int.
J. Rock Mech. Min. Sci. (submitted for publication).
HSE, 1996. Safety of New Austrian Tunnelling Method (NATM) tunnels. Health & Safety
Executive, London, p. 86.
Hudson, J., 2009. Prediction rockburst occurrence and development of the rockburst vulnerability index (RVI). Int. Consultation Report for the Key Technology of Safe and
Rapid Construction for Jinping II Hydropower Station High Overburden and Long
Tunnels, Jinping II, pp. 2531.
Jia, X.N., 2013. Experimental study on acoustic emission Eigen-frequency spectrum of
strainbursts (in Chinese). PhD Thesis. China University of Mining and Technology,
Beijing, p. 208.
Kaiser, P.K., 2009. Failure mechanisms and rock support aspects. Int. Consultation Report
for the Key Technology of Safe and Rapid Construction for Jinping II Hydropower Station High Overburden and Long Tunnels, Jinping II, pp. 6271.
Kaiser, P., Cai, M., 2012. Design of rock support system under rockburst condition. J. Rock
Mech. Geotech. Eng. 4 (3), 215227.
Kewley, R., Embrechts, M., Breneman, C., 2000. Data strip mining for the virtual design of
pharmaceuticals with neural networks. IEEE Trans. Neural Networks 11 (3), 668679.
Liu, L., Wang, X., Zhang, Y., Jia, Z., Duan, Q., 2011. Tempo-spatial characteristics and inuential factors of rockburst: a case study of transportation and drainage tunnels in
Jinping II hydropower station. J. Rock Mech. Geotech. Eng. 3 (2), 179185.
McPherson, B., Elsworth, D., Fairhurst, C., Kelsler, S., Onstott, T., Roggenthen, W., Wang, H.,
2003. EarthLab: A subterranean laboratory and observatory to study microbial life,
uid ow, and rock deformation. A Report the National Science Foundation. NSF,
Washington, DC, p. 60.
Miranda, T., 2007. Geomechanism parameters evaluation in underground structures:
articial intelligence, Bayesian probabilities and inverse methods. PhD thesis.
University of Minho, p. 317.
Miranda, T., Sousa, L.R., 2012. Application of data mining techniques for the development
of geomechanical characterization models for rock masses. In: Sousa, Vargas,
Fernandes, Azevedo (Eds.), Innovative Numerical Modeling in Geomechanics. CRC
Press, London, pp. 245264.
Miranda, T., Correia, A.G., Santos, M., Sousa, L.R., Cortez, P., 2011. New models for strength
and deformability parameters calculation in rock masses using data mining techniques. Int. J. Geomech. 11 (44), 4458.
Ortlepp, W.D., Stacey, T.R., 1994. Rockburst mechanisms in tunnels and shafts. Tunn.
Undergr. Space Technol. 9 (1), 5965.
Peixoto, A., Sousa, L.R., Sousa, R.L., Feng, X.T., Miranda, T., Martins, F., 2011. Prediction of
rockburst based on an accident data base. 11th ISRM Congress, Beijing, pp. 12471252.

Popielak, R., Weining, W., 2010. Engineering and design services for excavation DUSEL.
Preliminary Design, Preliminary Report #2, Contract D10-04, Lakewood, Coloradop. 104.
Qian, Q., 2009. The strategy for controlling water inow. Int. Consultation Report for the
Key Technology of Safe and Rapid Construction for Jinping II Hydropower Station
High Overburden and Long Tunnels, Jinping II, pp. 1518.
R Development Core Team, 2010. R: a language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria (http://www.R-project.org).
Sousa, L.R., 2006. Learning with accidents and damage associated to underground works.
In: Matos, Sousa, Kleberger, Pinto (Eds.), Geotechnical Risks in Rock Tunnels. CRC
Press, London, pp. 739.
Sousa, L.R., 2009. Continuing site investigation and risk assessment. Int. Consultation Report for the Key Technology of Safe and Rapid Construction for Jinping II Hydropower
Station High Overburden and Long Tunnels, Jinping II, pp. 17.
Sousa, R.L., 2010. Risk analysis for tunneling projects. PhD Thesis. Massachusetts Institute
of Technology, Cambridge, p. 589.
Sousa, L.R., 2012a. Report for the State Administration of Foreign Experts Affairs.
SKLGDUE, Beijing, p. 54.
Sousa, R.L., 2012b. Risk assessment in tunnels using Bayesian Networks. In: Sousa, Vargas,
Fernandes, Azevedo (Eds.), Innovative Numerical Modelling in Geomechanics. CRS
Press, London, pp. 211244.
Sousa, R.L., Einstein, H., 2012. Risk analysis during tunnel construction using Bayesian
Networks: Porto Metro case study. Tunn. Undergr. Space Technol. 27, 86100.
Sousa, L.R., Miranda, T., Roggenthen, W., Sousa, R.L., 2012. Models for geomechanical characterization of the rock mass formations at DUSEL using data mining techniques. US
Rock Mechanics Symposium, Chicago, ARMA 12-120, p. 14 (in CD-Rom).
Tang, S., Tong, M., Hu, J., He, X., 2010. Characteristics of acoustic emission signals in damp
cracking coal rocks. Min. Sci. Technol. 20, 143147.
Vlasov, S.N., Makovsky, L.V., Merkin, V.E., 2001. Accidents in transportation and subway
tunnels. Construction to operation. Russian Tunneling Association, Moscow, p. 198.
Wang, J., Zeng, X., Zhou, J., 2012. Practices on rockburst prevention and control in
headrace tunnels of Jinping II hydropower station. J. Rock Mech. Geotech. Eng. 4
(3), 258268.
Wu, S., Feng, X., Sousa, L.R., 2010. Jinping II mega hydropower project, China. Int.
Conference on Hydroelectric Schemes in Portugal. A New Cycle, Porto, pp. 223231.
Yagiz, S., 2009. Assessment of brittleness using rock strength and density with punch penetration test. Tunn. Undergr. Space Technol. 24 (1), 6674.
Yagiz, S., Candan, G., 2010. Application of fuzzy inference system and nonlinear regression
models for predicting rock brittleness. Expert Syst. Appl. 37 (2010), 22652272.
Yan, P., Lu, W., Chen, M., Shan, Z., Chen, X., Zhou, Y., 2012. Energy release process of surrounding rocks of deep tunnels with two excavation methods. J. Rock Mech. Geotech.
Eng. 4 (2), 160167.

Вам также может понравиться