Detection of potential overbreak zones in tunnel blasting from MWD data

J. Navarroª'*, J.A. Sanchidriánª, P. Segarraª, R. Castedoª, E. Costamagnab, L.M. Lópezª

ªUniversidad Politécnica de Madrid - E. T.S.l. Minas y Energía, Spain
b DIATI, Politecnico of Turin, Italy

ABSTRACT

The damage from blasting to the remaining rock mass is analyzed with the purpose of developing a drilling index
from measure while drilling (MWD) parameters, able to predict high risk of poten tia! over- and under-excavated
zones produced by blasting in the con tour of a tunnel. A new methodology based on the comparison of scanner
profiles of the excavated sections with the position of the contour blastholes, has been developed to obtain the
excavated mean distance (EMD) between the blasthole and the excavated profiles at each MWD record position,
which may be considered as a damage measure. MWD parameters, that describe the in-si tu rock mass properties
befare the blast, are thoroughly normalized to remove externa! influences that may hide the actual response of
the rig to rock mass properties and lead to wrong interpretations. 54 blasts, which comprise around 1700
contour blastholes, have been compared with more than 4000 excavated sections. A non-linear multiple-variable
power-form model has been developed to predict the excavated mean distance as function of the normalized
penetration rate, hammer pressure, rotation speed, rotation pressure and water flow parameters, and the lookout
distance. These parameters combine the rotational, hydraulic and percussive mechanisms of the drill, and the
confinement of the explosive charge with depth. Sources of uncertainty, unavoidable in the harsh condition in
which the data were measured, such as drilling deviations, the scaling and primary support prior to scanning the
excavated section, possible variations (unrecorded) in the explosive linear density, etc., have been assumed to be
of random nature.

l. Introduction Blast Damage Index (Yu and Vongpaisal, 1996) that correlates the
mechanics and the effects of wave propagation in to the rock mass; and
Rock excavation in mining and tunneling frequently use cautious (iv) failure approach index (Xu et al., 2017) that quantifies the rock
blasting techniques. The primary objective of blasting is to fragment mass damage through numerical simulations for tunnel support design.
rock to allow loading and haulage, without creating extensive damage Overbreak on the contour perimeter, which is defined as the void
to the remaining rock mass. As Andersson (1994) defined: "Cautious created during the excavation in excess of an established perimeter or
blasting is a blasting that does not cause damage to the rock outside of the pay line (Mahtab et al., 1997), is usually correlated with the damage
intended damage distance". extension zone which measures the quality of the blast. Overbreak and
For a cautious blast design, the damage on the perimeter of the underbreak are mainly influenced by the geotechnical condition of the
excavation is mainly induced by contour and buffer blastholes and it is rock mass (rock disturbances and rock strength) and blast design
created by a drop of strength, caused by the opening or shearing of parameters such as the explosive type, the charge concentration, the
newly generated or existing fractures or cracks (Scoble et al., 1997; blast timing, the drill pattern and the drilling deviations (Ibarra et al.,
Ouchterlony et al., 2002; Costamagna et al., 2018). In tunneling, da- 1996; Oggeri and Ova, 2004; Singh et al., 2003; Singh and Xavier,
mage can be categorized as major, minor or no damage, when there is 2005; Hustrulid, 2010; Johnson, 2010). Blasting affects the rock mass
rock falling, chips detachment or no visual damage, respectively structure because of shock wave propagation (vibrations), gas pressure
(Costamagna et al., 2018). Damage assessment is analyzed through four and stress redistribution (Singh et al., 2003; Hu et al., 2014).
main indexes: (i) Rock Tunneling Quality Index (Q-value, Barton et al., Guidelines on construction have established an overbreak magni-
1974) as classification of the ground condition for underground ex- tude of 0.15-0. 2 m and 0.1-0.15 m in crown and sidewalls, respectively
cavations; (ii) Blast Damage Factor (Hoek et al., 2002; Hoek, 2012) that (Manda! and Singh, 2009; Cunningham and Goetzsche, 1990). The
estimates the global rock mass strength and the rock mass modulus; (iii) maximum overbreak distance allowed depends on each national

* Corresponding author.
E-mail address: juan.navarro.miguel@upm.es (J. Navarro).
legislation and special terms can be arranged between the two parts in running the machine itself. From the correlation of MWD data with rock
the contract (Olsson, 2010; Costamagna et al., 2018). For example, mass geo-mechanical measures, Schunnesson et al. (2011) suggested a
Scandinavian countries present similar regulations for tunneling ex- model for the hydraulic properties of the rock mass, based on the
cavation requirements (Anlaggnings-AMA in Sweden, InfraRYL in Fin- monitored water flow and water pressure during rotary-percussive
land and the Norwegian Public Roads Administration, NPRA, in drilling. Schunnesson et al. (2012) explained a methodology to assess
Norway; Olsson, 2010; SN, 2004; NPRA, 2004). Consequences of abad rock strength ranges based on a MWD hardness index provided by Atlas
drilling can be short pulls of the rounds, increase of rock reinforcement Copeo software. For that, they used Schmidt hammer to correlate MWD
due to extra overbreak in the rock mass, longer scaling and mucking values with empirical rock strength measurements. Naeimipour et al.
time and bad control of grouting. (2014) developed a void detection algorithm based on MWD para-
To analyze the extension of the overbreak, studies have been carried meters. The algorithm was calibrated on full scale experimental tests in
out by comparing the laser profile of the excavated perimeter with the concrete blocks with various strengths. Kahraman et al. (2016) found a
designed tunnel profile. Kwon et al. (2009) investigated the char- strong correlation of the penetration rate with the uniaxial compressive
acteristics of the excavation damage zone (EDZ) in a tunnel construc- strength (UCS), the Brazilian tensile strength, the point load strength
tion and carried out a sensitivity analysis of the predicted EDZ with and the Schmidt hammer. Van Eldert et al. (2018) assessed the extent of
severa! rock mass parameters obtained from laboratory and in situ tests. the damage zone from MWD parameters and ground penetration radar
They determined that in situ stress ratio, Young's modulus and EDZ measurements recorded along the tunnel wall.
were the three main parameters in rock mass behavior after blasting. Finally, Atlas Copeo AB, Sandvik and Bever Control have developed
Manda! and Singh (2009) measured the overbreak, dividing the contour their own software (Tunnel Manager MWD, iSURE and Bever Control,
profile in three sections: left wall, right wall and crown. They found that respectively), as a too! for planning, administration and evaluation of
the crown is more affected by overbreak, due to the stress conditions in drilling. From the MWD files collected, the blastholes can be re-
this zone. Kim and Bruland (2009, 2015) estimated a tunnel contour presented in 3D, and hardness and fracturing maps are provided. The
quality index (TCI) based on overbreak distances of cross-section theoretical background for this information is however confidential.
scanners, contour roughness and longitudinal overbreak variation in Although there are many studies focused on the overbreak control
each blasted round. Costamagna et al. (2018), used scanned tunnel by blasting effect on the one hand and on geological and geo-me-
profiles to assess the overbreak of the excavated void in relation to the chanical interpretation of the rock mass by using MWD on the other, no
intended theoretical section, and correlated this with the rock mass relation between MWD parameters and overbreak (i.e., under or over
conditions of each round. excavation with respect to the theoretical tunnel contour line) from
Rock excavation techniques are highly influenced by the geo- blasting exists. This paper aims at developing an engineering too!,
mechanical properties of the rock mass (Oggeri and Ova, 2004; Singh based on MWD parameters, able to predict high-risk of overbreak po-
et al., 2003; Singh and Xavier, 2005; Mahdevari et al., 2013). A site tential zones generated on the perimeter of a tunnel face excavated by
investigation for a tunneling project generally includes a description of blasting. For that, scanner profiles of excavated sections have been
the rock condition and a rough estimation of the rock mass structural compared with the contour blastholes position, to obtain the excavated
properties; however, during the operation, unexpected anomalies that mean distance (EMD) created by each hole. Given that blasting vari-
may influence the results of the operation often occur. Such anomalies ables (mainly explosive charge and timing) are constant in contour
can be detected by Measure While Drilling (MWD) system on modern blastholes (as from blast reports), the overbreak and underbreak are
jumbos. This has been described by Schunnesson (1997) as a drill considered mainly influenced by the geotechnical condition of the rock
monitoring system which logs drilling data at predetermined length mass. Since the MWD system can characterize the rock mass condition
intervals providing information of the operational parameters involved of the blastholes, these measures can be correlated with the excavated
in drilling. For rotary drilling, Teale (1965) and Liu and Yin (2001) mean distance (EMD) calculated along the depth of each blasthole. The
introduced the concept of specific energy (SE) as the energy required to analysis will match high over-excavated zones with highly fractured or
excavate a unit volume of rock. Scoble et al. (1989) studied the varia- softer rock masses before the blast, and low excavated measures with
tions of monitored parameters to define different geological formations. competent rock.
Hatherly et al. (2015) compared the MWD data with the geological rock
conditions, obtained by geophysical logs, and demonstrated that if ro- 2. Data overview
tary speed and weight on bit are kept constant, MWD measurements
can determine rock properties. Leung and Scheding (2015) have de- The study has been developed in the underground extension work of
veloped a coal-seam detection model called SEM (modulated specific the municipal wastewater treatment plant in Bekkelaget (Oslo),
energy). Norway. The facility comprises five caverns, a main access gallery of
In relation to studies developed for percussive and rotary-percussive about 850 m length and other sections. The construction was developed
drilling, Schunnesson (1996) used the logged parameters of percussive in competent rock mass, composed by gneiss with small tonalite and
drilling to develop a method for estimating Rock Quality Designation quartzite intrusions. The Rock Tunneling Quality Index (Q-value;
(RQD, Deere and Miller, 1966) based not only on the penetration rate Barton et al., 1974), obtained from visual inspection of the tunnel face,
and torque parameters, but also on their variation, which shows a close was used as classification of the ground condition. Fig. 1 plots the Q-
correlation with the presence of large discontinuities, fractures or major value versus the chainage for the rounds analyzed in the study. They
faults. Schunnesson (1998) and Hjelme (2010) introduced a new comprise 54 rounds located between chainages 294-518 and 560-772
methodology to normalize percussive and rotary-percussive drilling of the main gallery (making up more than 400 m). Fig. 1 also shows the
parameters, to remove externa! influences generated by the blasthole classification of the rock mass based on the Q-value index; rock mass
length and the drill rig performance. Peng et al. (2005) and Tang (2006) condition generally qualified as good and fair.
proposed a method for void/fracture detection and for prediction of the Data from the rig have been used to locate the contour blastholes
rock mass properties based on drilling for roof bolting. They found that and to compare them with the excavated profiles. A three-boom jumbo
the feed pressure is a good detector of anomalies or discontinuities in XE3C, manufactured by Atlas Copeo, equipped with percussive-rotary
the rock and is a good estimator of the rock mass strength. In addition, top hammer drill, using semi-automatic ABC (Advanced Boom Control)
Peng et al. (2005) designed a new methodology for normalizing MWD was used to drill the analyzed blasts. Data comprises production face
parameters based on determining the performance of the machine when drilling holes of short length (4-5.5 m), using single rod (5.5 m length
drilling the air (i.e. for no-load condition, outside the rock mass). This and 38 mm diameter) and 46 mm bit. Eight MWD parameters were
assesses how much feed pressure or rotation pressure is required for logged during the drilling operation with a sampling interval of 0.1 m.
 System (DRS) defined by two vertical planes XctYct, YctZct and a hor-
35 izontal XctZct plane. The angles of the TRS axes with the DRS ones are e,
30 w and y (see Fig. 2).
e: 25 Good (52%)
o
3.1.1. NTM coordinates system
~ 20
ce The position of the jumbo inside the tunnel is first obtained, see
o 15
Fig. 2 (i. NTM coordinate System). The jumbo has a laser scanner in-
10 stalled. In addition, target plates, with known coordinates, are located
5 along the tunnel wall at every 5 m distance. The absolute coordinates of
the jumbo are calculated by trilateration (i.e. distance measurement
300 400 500 600 700 800 from the laser scanner to the target points). In this case study, XNrM,
Chainage (m)
ZNTM coordinates are given in the NTM projection and YNTM is the

Fig. l. Rock Tunneling Quality Index; Rock Tunneling Quality Index versus height above sea leve!.
chainage and classification of the ground condition: poor (1 < Q ,,; 4), fair
(4 < Q,,; 10), good (10 < Q,,; 40). 3.1.2. Tunnel Reference System (TRS)
The drill rig is aligned with the tunnel line (perpendicular line to the
These are described next as an extension of Peng et al. (2005), Beattie face of a new round that defines the orientation of the drilling) in order
(2009), Hjelme (2010), Schunnesson et al. (2011), Schunnesson and to follow the design of the construction. For that, two targets are
Kristoffersson (2011) and Navarro et al. (2017) interpretations; the mounted on one of the booms. The laser beam points to the free face in
acronym and the units for each parameter are given in brackets. the direction of the tunnel axis and the boom is rotated until the laser
beam passes through both targets (Fig. 2, ii. TRS). The boom is so
- Feed Pressure (FP, bar): measure ofthe hydraulic pressure inside the aligned with the tunnel axis and the orientation and inclination of the
cylinders. Feed pressure is required not only to keep the bit in boom are registered in three orthogonal vectors cX;, Y,, Z,) to create a
contact with the bottom of the hole throughout the transmission of coordinate system parallel to the tunnel axis and the free face.
energy, but also to maintain a minimum force between bit and rock The laser scanner also measures the distance from the jumbo to the
to maximize energy transfer to the rock. face of the new round and records the chainage at which it is located
- Hammer Pressure or Percussive Pressure (HP, bar): this is a measure inside the tunnel. This chainage is taken as reference plane of the col-
of the impact pressure acting on the pistan in the rock dril!. laring depth position of the blastholes. Negative depth values are as-
- Damp Pressure (DP, bar): it measures the pressure absorbed by the signed to measurements behind this plane, and positive values, to
drill rig to prevent vibrations or undesired motion in the boom or measurements ahead of this plane.
drill rod.
- Rotation speed (RS, rpm): it is defined as the number of turns of the 3.1.3. Drilling Reference System (DRS)
bit per minute. The blasthole position measured by each boom is calculated in the
- Penetration Rate (PR, dm/min): rate of penetration of the drill bit Drilling Reference System defined by means of three spherical co-
through the rock mass. ordinates: blasthole length (lb), azimuth or lookout direction (Lv) and
- Rotation Pressure (RP, bar): it is the torque pressure required to inclination or lookout angle (Lr) (see Fig. 2):
rotate the bit at a defined speed.
(1)
- Water Pressure (WP, bar): it is the pressure ofthe water used to flush
the drill cuttings from the blasthole. (2)
- Water Flow (WF, l/min): it is the rate of water inflow into the drill
rod (3)
- Hole Length (HL, m): depth at which each sample of the above
The inclination angle varies between O and 90º both for holes drilled
parameters is logged.
upwards or downwards so that the azimuth is between O and 180º for
holes drilled upwards and between O and -180º for holes drilled
The charging of the rounds was carried out with emulsion of dif-
downwards.
ferent linear charge for different types of blasthole: cut, lifter, easer,
Blasthole positioning data logged by the ABC system uses sensors
buffer, con tour. String loading method was used for the charging of the
installed along the boom (outside the blasthole) to measure the azimuth
contour blastholes, with a design linear charge of 0.5 kg/m. This is
and the inclination angles. The semi-automatic ABC total system in-
assumed to be constant in the analysis. The actual charge of the holes
stalled in the drill rig authorizes the operator to move the boom and
may vary from the design value. However, such variations are assumed
feeder manually to follow a predesigned drill plan (Navarro et al.,
to be random so that, although they are a source of indetermination in
2018). Once the boom is placed in the required position and befare the
the analysis, they will not bias the influence of the other parameters in
drilling starts, the measurements of the azimuth and inclination angles
the overbreak from blasting.
are logged in the DRS. These are considered constant as the boom re-
mains still while drilling the blasthole. The end coordinates of the
3. Analysis of the excavated area blasthole are calculated by adding, to their collaring coordinates, the
result from Eqs. (1), (2) and (3). Since deviations beyond the collaring
3.1. Jumbo navigation point cannot be measured by the MWD technology there is a possible
error between the actual end position of the blastholes and the end
Navigation is necessary to locate thejumbo inside the tunnel befare position given by the MWD system. Drill deviations have not been
drilling a new round. For that, the jumbo rig uses three reference sys- monitored so they are a source of indetermination in our analysis that,
tems, sketched in Fig. 2: (i) an absolute coordinate system that refer- as those mentioned in Section 2, is assumed to be of random nature.
ences the position of the jumbo, in this case the EUREF 89 Norwegian
Transverse Mercator (NTM) projection, (ii) a Tunnel Reference System 3.1.4. Transformation system IRS-DRS
(TRS) with one axis parallel to the tunnel axis and the other two in the For Atlas Copeo jumbos, the directional coordinate vectors of the
plane of the tunnel face of the new round, and (iii) a Drilling Reference TRS cX;, Y,, Z,) and the NTM coordinates of the jumbo (XNTM• y NTM•
Fig. 2. Representation of the three reference systems involved in the jumbo navigation: i. NTM Coordinates System, ii. Tunnel Reference System, iii. Drilling
Reference System.

Rotation of X1Z1 plane around Yt Axis Rotation of YtZt plane around Xt Axis Rotation of X1Y1 plane around Zt Axis
Rotation towards right and left Uphill - downhill inclination Lateral inclination

Fig. 3. Transformation from drilling reference system cX'd. Xi.Zd) to tunnel reference system cX,. Y,. Z,). Left: Rotation towards right and left, or bearing angle, over
Y,;center: uphill-downhill inclination, or elevation angle, rotation over X,; right: lateral inclination, roll or bank angle, rotation over Z,.

ZNTM) are presented at the end of each MWD file. The three rotation identified by its respective chainage. An interface AutoCAD-Matlab
angles to transform the DRS coordinates of a point in a borehole to the (AutoCad, 2017; Matlab, 2017) has been created to automatically
TRS can be seen in Fig. 2 and are further explained in Fig. 3. compare the excavated profiles with the contour blastholes for each
This transformation is required below in arder to know the exact round. The profile formed by the contour hales (hereinafter named
position of the blasthole collars and their orientation. The location of contour profile) is compared with the scanner profiles of the excavated
the oriented blastholes in absolute coordinates (Xt,NTM• Yt,NTM• Zt,NTM) is sections in arder to obtain the excavated mean distance between the
obtained by adding the NTM coordinates of the jumbo (XNrM, YNTM, blasthole and the scanner section at each depth for which MWD data
ZNTM) to the oriented coordina tes of the blastholes (X,, Y,, Z,) in the TRS are logged. This distance is considered as an indicator of the resulting
system. damage (i.e. over-excavation).
For safety reasons, scanning of the excavated section is done after
scaling and installation of primary supports. During scaling, non-stable
3.2. Superposition of excavated profiles and contour blastholes rocks are removed from wall and roof to avoid rock falls and to ensure
safe work conditions. Next, rock bolts and shotcrete are applied to re-
A laser scanner system has been used to monitor the final profiles of inforce the tunnel walls and to prevent stability problems. These op-
the excavated void from each blast. The laser was located in the center erations obviously modify somewhat the perimeter excavated.
of the excavation to be scanned and target spheres were installed along Shotcrete was automatically applied to the tunnel surface and the
the wall of the main gallery in places with known coordinates. The thickness of the shotcrete !ayer was modified per round according to the
software of the scanner identifies the position of the spheres in a post- geotechnical recognition of the tunnel wall and crown. The thickness of
analysis of the 3D cloud of points and trilaterates the location of the the shotcrete !ayer was known for every round and it was homogeneous
scanner by measuring distances from the later to the spheres. Profiles of over walls and crown, according to the operation reports. This thickness
the excavated void in a direction perpendicular to the tunnel line are has been added to the scanner profiles in the AutoCad files to obtain the
collected at steps of 0.2 m from the 3D cloud of points; each profile is
 - Blasthole Contour profile - Excavated section profile
• Blasthole extended at the chainage of the excavated section

16 16
15 15
14
14
13
13
I12
I12
>- >- 11
11
10
10 9 406
9 8 404
8 7
6 6
2 400
-2 -4 398 o
X(m) -6 Z(m)
398
a) b) -6

Fig. 4. Chainage 399. (a) Contour hales profiles and position of the blastholes; (b) Overlapping between the contour profiles and the excavated section profiles.

actual excavated contour from the blast. However, an unknown un- 3.3. Excavated mean distance from the blasthole
certainty still remains.
The comparison between excavated and contour profiles must take The overbreak created around the blasthole by blasting is caused,
into account the lookout angle and lookout direction values for each among other factors, by the combination of the explosive and the rock
blasthole, which make them to be outward oriented. Since no deviation mass condition around the blasthole (Hustrulid, 2010; Johnson, 2010).
data of the blasthole is available, it is assumed that the lookout in- Considering the contour blastholes position per round, an Excavated
creases linearly with depth. Eqs. (1) and (2) are used to calculate the Mean Distance (EMD) has been defined (Fig. 5). It corresponds to the
theoretical position of the blasthole at each excavated profile depth in area between the midpoint of the spacing on both sides of the hole and
the local Drilling Reference System (DRS). Variables Lr and Lp are ob- the excavated profile, normalized by the distance between the mid-
tained from the MWD files and lb is the length of the blasthole from the points of the spacing on both sides of the blasthole. When two adjacent
collaring position to the respective excavated profile depth (Zp, see holes are on the same side of the excavated profile, the EMD is calcu-
Fig. 2), obtained with Eq. (3). lated by (Fig. 5, EMD 1):
Irregularities on the free face of a new round cause the collars of the
blastholes not to be in the same plane, so they have different collar EMD=~
depths. In addition, sorne excavated profiles (mainly profiles at the 21. + ~ (4)
2 2
beginning of the round) are not influenced by ali contour holes of the
current blast but for from the previous one. Each blasthole is extended where 5 1 and 52 are the spacing between the current blasthole and the
from the foremost collaring hole to the depth of the deepest blasthole of
each round to calculate the excavated area for ali profiles included in a • Mid-spacing - Excavated Profile
round. Fig. 4a represents the contour profiles (cyan lines), the position
0 Blasthole • lntersection from the mid-spacing
of the blastholes (black lines) and their extensions (tiny blue dots). with the excavated profile
To carry out the analysis, both contour and excavated profiles must
be overlaid. Excavated profiles from AutoCAD files are drawn in a
vertical .xy plane, where the Y coordina te is referred to the Y NTM ab-
EMD 1 (Eq. 4) EMD 2 (Eq. 5)
solute coordinate and the hypothetical Z coordinate, i.e. depth of the .xy
plane, is indicated by the chainage at which it is located. The contour
blastholes coordinates in the DRS must be rotated to the TRS. Since the
o
rounds studied belong to the main gallery and they are excavated in a 3 2
straight line, the .xy and xz planes for both TRS and DRS coincide
(angles y and w are O); only plane y17, is rotated in case the tunnel axis is
uphill (positive e angle) or downhill (negative e angle), see Fig. 3. The
. 1S2/2
rotation of the DRS contour blastholes coordinates (Xd, Yd, Zd) to the
TRS (X,, Y,, Z,) is obtained by introducing the three angles (B, y = O, Js112
w = O) in a 3D rotation matrix. The translation of the Y, and Z, co-
ordinates is carried out by adding the YNTM and the chainage values of 1
the round studied, respectively. Fig. 4b sketches the overlapping ofboth
excavated (red lines) and contour holes (blue lines) profiles for a round.
o
Void
Fig. 5. Calculation of the Excavated Mean Distance.
adjacent ones (they are generally around 0.7 m); An is the area ex- monitoring sensors, the hole length and/or the drill rig performance
cavated by each blasthole defined by points 1, 2, 3 and 4 in Fig. 5, EMD (Schunnesson, 1998). The control system of the jumbo, during the ad-
1; it is positive when there is over-excavation and negative for under- justment of the parameters while drilling, induces systematic variations
excavation. in the MWD data. Ali together, they add uncertainty to the data that
In case two consecutive blastholes are located one inside and the must be previously normalized in order to highlight changes in the
other outside of the excavated profile, corresponding to over and under- parameters by the rock effect. This process comprises: (i) filtering out of
excavation (Fig. 5, EMD 2), the total excavated area is obtained by unrealistic data, (ii) removing of the ramp-up section of the logs, (iii)
adding both areas with their respective sign: correction of systematic variations in the MWD parameters such as the
effect ofhole length and feed pressure influence and (iv) normalization
EMD = AT2 + An with the standard deviation to account for fluctuations in the signals.
La Lb (5)
Fig. 7 shows a flow chart of the filtering and corrections applied to the
where Ar2 is the area defined by 5, 6 and Pint (Fig. 5, EMD 2); AT3 is the raw MWD data, with the acronym used after each step.
area defined by 7, 8 and Pint (Fig. 5, EMD 2); Pint is the intersection point
between the excavated profile and the line joining two adjacent blast- 4.1. Filtering of unrealistic values
holes; La and Lb are the distances between Pint and the mid-spacing point
between the current blasthole and the adjacent one. Production data often includes unrealistic high and low perfor-
The scanner profiles and the EMD values per blasthole are evaluated mance values of the jumbo, which may lead to a wrong interpretation of
at 0.2 m intervals. The MWD sample interval is 0.1 m and the collaring the MWD data (Ghosh et al., 2015). A new fil ter of the data is developed
chainage of each contour blasthole differs due to irregularities of the before any further analysis. For that, the empirical probability dis-
free face. Thus, the actual chainage of the MWD logs of each hole varies tribution function of each MWD parameter is built from the complete
so that the position of the measurements recorded by the MWD system data set values (54 blasts, comprising more than 6500 blastholes). The
does not coincide with the depth of the calculated EMD values. A pie- 95% confidence interval of MWD parameters defines the interval to
cewise cubic Hermite interpolating polynomial (Fritsch and Carlson, retain data for the analysis. Fig. 8 shows the cumulative distribution
1980) is used to interpola te the EMD values at the specific depths of the functions (CDF) of the eight parameters, where the black dashed hor-
MWD logs. izontal lines mark the 95% coverage and the vertical lines mark the
As an example, Fig. 6a shows the area of influence of each blasthole corresponding limits of the parameters, given in Table l.
(ATl or AT2 + AT3 in Fig. 5), calculated every 0.2 m for a round be-
tween the chainages 398 and 406 m; over- and under-excavated areas 4.2. Removing of ramp-up operation mode
are plotted in different colors and the contour blastholes are marked in
black. Fig. 6b represents the EMD values for one of the blastholes in The control system is normally based on three main operational
graph a (that includes the backwards and forwards extensions, as ex- modes (Schunnesson, 2017; Navarro et al., 2018): (i) collaring, (ii)
plained above) versus chainage. It shows the EMD calculated (black ramp-up, both of which control the increase of the drilling pressure to
dots), the cubic interpolation of the EMD calculated (red line), and the minimize hole deviations, and (iii) normal drilling, which controls the
estimated EMD values corresponding to the MWD depths logged (blue performance of the parameters to optimize the operation and minimize
dots). damages to the drill system. As can be seen in Fig. 9, the feed pressure
(FP) shows, initially, a sharp rise (ramp-up mode) until it reaches a pre-
4. MWD data processing set threshold at which it stabilizes (normal drilling mode). Only values
included in the normal drilling mode are considered for the analysis, as
The response of MWD parameters is often affected by externa! in- data in the ramp-up mode are not representative of changes in the rock
fluences different than the rock mass, such as the calibration of the mass conditions. Signals of the 8 parameters have been divided

Over-excavated Area - Under-excavated Area Estimated EMD values at MWD logs depth
- Contour blasthole • Measured EMD values - lnterpolation line

0.3

0.2
16
0.1
14 Io o
~
w
-0.1

10 -0.2

-0.3
8
6
-0.4 ~-~--~--~--~--~-~--~--~
398 399 400 401 402 403 404 405 406
Chainage (m)
X(m)
a) b)
Fig. 6. (a) Area excavated along each blasthole; (b) EMD cubic interpolation and EMD value according to the MWD depth.
 Raw data
1 Mwol----------IMWDc1I
i.Filtering of unrealistic values &
ii.Removing of the ramp-up section

iii.Feed pressure
J iii.Hole length
correction
iv.Signal fluctuation correction
IMWOcJI---- IMWDc2I
Fig. 7. MWD data processing flow chart.

3,:u-: :ZI
1
Ramp-up , Normal drilling
..--.-----.
1
Ramp-up , Normal drilling
..--.-----.

f ~I~ 1i~~I) 1
0_1
o 20 40 60 o 50 100 150 200 250
PR(dm/min) HP (bar)

s":r- --) 1

o 50 100 150 o 50 100 150

FP (bar) DP (bar)

8
O:
o
[,,,,,,==="''~7:..__~L¡_
100 200
RS(rpm)
300
_j

400
[ .Zi
o 50 100
RP (bar)
150
1:1~1 ~:~1~ ~ 35 ,J: 1
150 1 30 1

10 20 30
[]
o 50 100 150 200
WF(l/min)
250
r]~I
1

i :~1\:J\;JI. u. 58
¡::
1

':"
1
~-
11.

WP (bar) 56 1
o 2 4 o 2 4
Fig. 8. Cumulative distribution function of the MWD parameters. Units and HL(m) HL(m)
acronyms are given in Table l.
Fig. 9. Drilling operation modes in MWD signa! from blasthole 71 at chainage
349. Units and acronyms meaning are given in Table 1; HL is hole length.
Table 1
Range of reasonable values of the MWD parameters as from their 95% coverage.
analyzed.
Parameter Acronym Units Range According to incidence reports, the sensors that monitor the posi-
tion of boom 3 were out of calibration in most of the rounds, and the
Penetration rate PR dm/min 1-38
Hammer pressure HP bar 130-235
blastholes drilled by this boom had a lower lookout and lookout di-
Feed pressure FP bar 20-80 rection angles than recorded by the MWD files. This resulted in a sys-
Damp pressure DP bar 45-100 tematic under-excavation in the contour profiles drilled by boom 3. The
Rotation speed RS rpm 170-310 reports indicated a more intensive scaling on this side to meet the
Rotation pressure RP bar 35-80
12-22
contour requirement. This results in a significant distortion of the ex-
Water pressure WP bar
Water flow WF l/min 54-175 cavated profile, hence in the over-excavated values calculated. Since
the precise blasts for which boom 3 was out of calibration were not
available, ali data from boom 3 has been discarded for the analysis.
considering the point at which the feed pressure stabilizes. Values be- The correction of the hole length influence (to obtain a signa!
fare this point comprise the ramp-up data set and forward values the MWDc2 , see Fig. 7) is done by:
normal drilling data set. The latter will be referred as MWDci·
MWD~2 = [MWD~1-MWD~,,HL] + MWDfit,HL; with i = 1, 2, ···,N
(6)
4.3. Hale length and feed pressure corrections
where i indicates each measurement in a hole log, N being the number
Systematic variations generated by the drilling system and other of these. MWDfe,HL is a polynomial regression with hole length, of the
parameters can be estimated and removed by averaging, for a large average value at every 0.1 m hole length for the entire data of each
amount of data, the response of the parameters (Schunnesson, 1998; MWDc1 parameter. MWD 1fit;HL is the intercept of the fit, i.e. the value at
Hjelme, 2010). Schunnesson (1998), Hjelme (2010) and Ghosh (2017) depth zero.
described a significant hole length dependence in sorne parameters for Fig. 10 shows the average MWDc 1 signa! (blue lines), the polynomial
percussive, rotary-percussive and Wassara water-hydraulic DTH drilling regression (MWDfit;HL, green lines) and the hole length normalized
modes, respectively. This phenomenon is related to the increase of the average signa! (MWDc 2 , red lines), at every 0.1 m hole length for boom
frictional resistance between the drill rods and the blasthole walls, the l. The hole length normalization is not applied for the hammer pressure
reduction of the available pressure over the hammer, the reduction of and for the water flow because there is no noticeable effect of depth in
the flushing efficiency with depth and the bit wear (Schunnesson, the average signa! (see Av.HPci;c 2 and Av.WPci;c2 in Fig. 10).
1998). The average value of the eight MWDc 1 parameters at every 0.1 m Previous work by the authors (Navarro et al., 2018) analyzed the
hole length has been calculated for the entire data from the 54 blasts relationship between MWD parameters to limit the number of the MWD
 ~217.51 M
~ 200
2501
tl.-...
<(
c..
::e
217

>216.5 '-----~---~--~ ~150-~

~ 100L__ _ _ _ _ _ _ __

~
<(
~
N
~ 71.8

u.. 71.6
> 71.4
721

~
(,)
D..
1001

50~
-
~.~--
,,....r.J
,,.._

~1-~_-
~-f: . -v-'<_,.~_\.\
N 230
2601

~'"~ ::~
M
~ 228
701
"
ti)
ci:: 226 __
~ 220 ----~--~~--~
~ 224---~---~--~
40----~--~---~

~ ~
~:1~ ~<(::~ ~::~
181 801

20 40 60 80 20 40 60 80
o 2 4
HL(m) FPc2 FPc2
Fig. 10. Correction of hole length influence in MWD parameters (pass from Fig. 11. Correction of the feed pressure influence in MWD parameters (MWDc2
MWDc 1 to MWDcz, see Fig. 7); blue lines: average signals, MWDci (obtained conversion to MWDc 3 , see Fig. 7); blue lines: average signals, MWDc2 (corrected
after filtering unrealistic values and removing ramp-up section, see Fig. 7); for hole length influence, see Fig. 7); green lines: polynomial regression; red
green lines: polynomial regression; red lines: average normalized signals, lines: average normalized signals, MWDc 3 . Units and acronyms of the para-
MWDcz. Units and acronyms of the parameters in the ordinates are given in meters are given in Table 1. (For interpretation of the references to color in this
Table l. (For interpretation of the references to color in this figure legend, the figure legend, the reader is referred to the web version of this article.)
reader is referred to the web version of this article.)
4.4. Analysis of fl.ucmations in the MWD signals
variables, and to select the more significant ones that are required for a
sound rock mass characterization. They did a cross-correlation analysis Schunnesson (1996, 1997) claimed that when discontinuities in
between the signals of the eight parameters in order to find leadership rock are drilled, penetration rate and rotation pressure show significant
behaviors. The results pointed to the feed pressure as the lead para- fluctuation, resulting in a noisy signa!. Following this reasoning, para-
meter that drives the adjustment of the other variables to optimize the meters involved in the rotational mechanism of the jumbo (rotation
drilling. They showed that penetration rate, hammer pressure, damp pressure and rotation speed) and the penetration rate have been pro-
pressure and rotation pressure are influenced by the feed pressure. cessed. The procedure is carried out for the signa! of each hole in-
According to this, the feed pressure generates systematic variations in dividually, where the MWDc 3 values in a hole are divided by the
these parameters that may hide the rock dependence on them. On the standard deviation of the entire signa! of that hole. The resulting signals
contrary, rotation speed, water flow and water pressure are little in- for these three parameters are MWDc4 •
fluenced by the feed pressure, thus being considered independent. MWD;
MWD~ 4 =
3
The same methodology followed for the hole length influence is now d( c ) ; with i = 1, 2, · ··,N
used to correct the feed pressure influence. The average value of the st MWDc 3 (8)
seven MWDc 2 parameters (feed pressure is not included) is calculated at where i indicates each measurement in a hole log, N being the number
steps of 1 bar feed pressure value for the 54 blasts for boom l. Similar to of samples per signa!.
Eq. (6), the correction of the feed pressure influence (to obtain a signa!
MWDc3 , see Fig. 7) is done by:
5. Detection of potential overbreak zones
MWD~3 = [MWD~2-MWD},,,FP] + MWDfit,FP; with i = 1, 2, ···,N (7)
The processed penetration rate (PRc 4 ), hammer pressure (HP c3 ) and
where i indicates each measurement in a hole log, N being the number rotation pressure (RP c4 ) result in rock dependent parameters. The
of these. MWDfit;FP is a polynomial regression with the feed pressure, of normalized rotation speed (RSc 4 ) and water flow (WFc3 ) are in-
the average value at every 1 bar feed pressure for the entire data of each dependent parameters sensitive to rock variations (Navarro et al.,
MWDc2 parameter. MWD 1fit;FP is the intercept of the fit, i.e. the value at 2018). The response of these normalized parameters can detect varia-
the minimum feed pressure. tions in the rock and for equal blasting conditions they may explain
Fig. 11 shows the average MWDc 2 signa! (blue lines), the polynomial variations in EMD data. The feed pressure is not considered as it has
regression (MWDfit;FP, green lines) and the feed pressure corrected been used during the normalization.
average signa! (MWDc3 , red lines), as function of the feed pressure. In The lookout distance (i.e. distance from the collaring to the position
line with Navarro et al. (2018), it can be seen that penetration rate of the hole in the XY plane at each depth, see Fig. 2) is also considered
(PR), hammer pressure (HP), damp pressure (DP) and rotation pressure for the analysis as it may reflect the confinement effect by depth.
(RP) parameters have a strong dependence from the feed pressure (FP). Fig. 12a shows, as an example, variations of EMD with the hole length
For the case of rotation speed (RS), water pressure (WP) and water flow and lookout distance for the holes of the blast located at the chainage
(WF), the influence of feed pressure is considerably less, and these data 500. As can be seen, there is a negative influence of the lookout in the
are not normalized for the subsequent analysis; Fig. 11 shows no large EMD which means that the excavated area in relation with the blasthole
differences between MWDc2 and the resulting MWDc 3 signals for these position decreases with the lookout distance; this EMD relation with
parameters. lookout distance is the same as with hole length since the lookout
distance increases with depth. An increase of confinement with lookout
 - Hole6 Hole 8 - Hole 11
0.5 ,----~---~--~---~--~-~

.., 200
2201
~ 180
:z:: 160
I o
140
e 120'---~-~-~-~-~-

:e
w

.., 1201
100
~ 80
-0.5 '----~---~---~---'----~-_J e 60
o 2 3 4 5 40
'----~-~-~-~-~
Hole length (m)

201 ""'"40
501
~ 15
~ 30
o:: 10 o:: 20
5 --~-~-~-~-~- 10L_~-~-~-~-~

Ie o
:e

~
w
801
~ 60
40L--~-~-~-~-~-

2 3 4 5 0 2 3 4 5
Lookout distance (m) HL(m) HL(m)
a) b)
Fig. 12. Preselection ofthe data for blastholes ofbooml at the chainage 591; (a) EMD vs. hole length and lookout distance; (b) MWD signals for blastholes 6, 7 and
11. See Table 1 for the acronyms of MWD parameters; subscripts c3 and c4 refer to the corrections shown in Fig. 7.

distance and depth is associated with an increase of the difficulty in Table 2

breaking the rock, resulting then in a decrease of the over-excavated Statistic of the MWDc 3 and MWDc4 parameters for boom l.
area, hence the EMD, which may be even negative (under-excavation). Parameter Units Mean Std." Min. Max.
Therefore, the larger the lookout distance and the deeper the drilling,
the more under-excavation is created in relation with the position of the Penetration ratec 4 m/min 13.01 9.40 0.04 57.10
contour blastholes. In sorne cases (data for holes 6, 8 and 11, Fig. 12a), Hammer pressurec3 bar 211.81 15.04 63.00 230.0
Feed pressurec3 bar 73.94 5.71 21.20 79.89
the general trend is not followed. Fig. 12b shows the respective MWDc 3 Damp pressurec3 bar 65.31 12.76 48.23 86.63
and MWDc4 logging for holes 6, 8 and 11. One may think that local rock Rotation speedc4 rpm 34.73 30.91 0.18 98.58
conditions may cause such results and that MWD parameters should Rotation pressurec4 bar 26.55 12.99 0.09 128.05
show a peak or sorne kind of variation in the signa!. This is effectively Water flowc 3 l/min 81.99 28.20 53.70 145.56

observed for hole 6, where MWDc 3 -c4 parameters show a significant Water pressurec3 bar 16.43 4.87 13.57 21.84
Lookout m 0.27 0.15 0.02 0.94
peak or signa! fluctuation at depth 3--4 m, in line with the over-ex- EMD m o.25b 0.30 -0.63 1.09
cavation peak represented in graph a for this hole. However, no dis-
tinctive variation appears in the MWDc 3 -c4 records of holes 8 and 11. ª Std. is standard deviation.
Hole 8 shows an over-excavated section at 2.5-3.5 m, whereas hole 11 b Mean absolute Value.
indicates under-excavation at depth 1-2 m, though no significant
changes can be observed in the MWD signa! of these two holes at any estimated for each parameter are significantly different from zero, i.e.
depth. Other causes like drill deviations, malfunctioning of blasting, their p-value is < 0.0001 for ali cases, which means that the analysis is
scaling, etc. may be behind such EMD outlier profiles, uncorrelated statistically significant. According to the value and sign of the coeffi-
with the MWD logs; these outliers, that happen at most in one or two cients, the MWD parameters affect EMD differently:
holes per blast (many blasts do not show any) are removed from the
analysis. - The normalized penetration rate (PRc4 ) and rotation pressure (RPc 4 )
Table 2 shows the statistics for the MWDc 3 and MWDc 4 parameters, show negative correlation (negative sign in A1 and A4 ) with EMD.
the lookout distance, and the EMD values for boom l. According to Schunnesson (1996, 1997), high fluctuations in the PR
A power function of the MWD parameters to predict the EMD is and RP signals indicate soft/fractured rock. High tluctuations mean
considered: high standard deviation so that, according to Eq. (8), PRc4 and RPc 4
show low values. This means that low PRc4 and RP c4 values indica te
(9) soft/fractured rock, prone to suffer over-excavation. The contrary is
where Lctist is the lookout distance. also true.
Since EMD has positive and negative values, an additive constant A0 - The normalized hammer pressure (HP c3 ) also shows a negative
has been included. The normalized damp pressure and water pressure correlation with EMD (negative sign in A2 ), which indicates that,
have been removed since their contributions were found minimal. The when discontinuities or soft rock is drilled, the hammer of the boom
non-linear regression model has been programed with Matlab 2016b by needs a lower pressure to hit the rock. As explained by Peng et al.
a Levenberg-Marquardt non-linear ordinary least squares method (2005) and Navarro et al. (2018), when open fissures or softer rock
(Matlab, 2016). The model coefficients are given in Table 3. are found, the feed pressure shows a sharp drop to the leve! of
The determination coefficient ofthe fit R2 is 0.74. Coefficient values drilling the air. This behavior is also noticed in the hammer
Table 3
Non-linear regression model coefficients.
PRc4 HPc3 WFc3
A, A2 As

Coefficient -2.8432 -0.0341 -0.1188 0.0352 -0.0108 0.0834 -0.0600

pressure; in open fissures, there is no resistance far the hammer to and lower prediction band at a 95% confidence leve!. The residuals are
hit the rock which results in a reduction of its pressure to avoid normally distributed (see Fig. 13b).
damage in the system. In this way, low hammer pressures values are Fig. 13c shows the distribution of the root mean square error
related with over-excavated areas and vice versa. (RMSE) far each of the 54 blasts. The median and the 25th and 75th
- The normalized rotation speed (RSc 4 ) shows positive correlation percentiles of RMSE are 0.142m, 0.102m and 0.186m, respectively.
with EMD (positive sign in A3 ). Schunnesson (1998) claimed that at An illustration of the application of the model is shown in Fig. 14 far
high feed pressure, the rotation pressure required far the rotation of three blasts, representing the 75% (large error, Fig. 14a), 50% (ex-
the bit increases and sometimes a reduction in the rotation speed pected error, Fig. 14b) and 25% (small error, Fig. 14c) of the RMSE
can be appreciated. The higher resistance to rotate may also trans- values; they correspond to blasts located at the chainages 388, 500 and
late into an increase in the signa! fluctuation. According to Eq. (8), 591, respectively. In the three cases, the EMD predicted is compared
the higher the standard deviation of the RS parameter, the lower the with the EMD measured far blastholes of boom l. Five different over-
RSc4 and the lower the excavated area is created, hence the positive excavation ranges have been defined. For high RMSE values (Fig. 14a),
sign of its exponent in Eq. (9). For the case of discontinuities, visual differences between the predicted EMD and the EMD measured
Hustrulid (1968) showed that at low feed pressure, the bit will not are apparent; these are considerably reduced when representing
be in constant contact with the bottom of the hole, resulting in a free medium and low RMSE values (Fig. 14b and c). Fig. 14c shows a pre-
rotation of the bit that may show a lower fluctuation in the rotation dicted EMD generally in line with the EMD measured though still sorne
speed. This way, the lower the standard deviation of the RS para- light differences exist. Considering the noise of the MWD data - un-
meter, the higher the RSc 4 and the greater the excavated area. avoidable in the harsh conditions where such data are measured - and
- The normalized water flow (WFc 3 ) has a positive correlation with the additional uncertainty brought by drilling deviations, the scaling
EMD (positive sign in A5 ). This can be explained by the fact that, and primary support done befare scanning the excavated section, sen-
when discontinuities or soft rock is drilled, the control system re- sors potentially out of calibration, possible variations in the explosive
quires a higher water flow due to the higher amount of drill cuttings linear density, etc., the quality of the fit is outstanding.
or the water leakage to the discontinuities.
- The lookout distance presents a negative correlation with EMD,
which means, as discussed befare, that the excavated area in rela- 6. Conclusions
tion with the blasthole position decreases with an increase of depth
and lookout. The overbreak of the remaining rock mass in tunnel blasting has
been analyzed in the light of MWD records, with the purpose of de-
The results of the EMD predicted with the suggested model versus veloping a prediction model of over- and under-excavation depths from
the EMD data are plotted in Fig. 13a. The linear regression obtained has blasting. Such predictive model may also be seen as a drill or rock index
a slope of one with a zero-constant term. Fig. 13a also shows the upper that could be used to identify zones of potentially high geotechnical risk
(those far which the over-excavation prediction is high). By comparison

·-b:,,-
3.5 ~---~---~----~---~

f_:~f -1 ~-----------------------~
>.
u
;
::::¡
3

2.5

2
1.5 ~---~---~---~---~---~
e
L1.
1.5

I
"C
....Q)
,,,::::¡ 0.5 -0.5 o 0.5
m Residuals (m)
Q)
b)
E
e o
:!!:
w

-0.5

-1 ~---~----~---~----~---~
1 lm[l]m
0.1 0.15 0.2
1 •

0.25
•

0.3
•* •·
0.35
I

-1 -0.5 o 0.5 1.5

EMD predicted (m) RMSE (m)
a) e)
Fig. 13. Result of the EMD predicted model: (a) predicted versus measured EMD values (upper graph: residuals); (b) Box-plot of the RMSE values per blast.
 EMD measured RMSE = 0.19 m EMD predicted

13 13
12 12
11 11

.s
-10
.s
-10

------""'"€>
:;; 9 :;; 9

- ----E)._,
1- 1-
z ·. z
>- 8 >- 8

7 7

6 <> 6
---a..~ o o
5 5
394 394 392 390
386
a) Chainage (m)

EMD measured RMSE = 0.14 m EMD predicted

~'
13 13

12 12

11
~~~....
'
_11
.s
:;;
10 I10
:;; ~,,.
1- 1- 9 ~~ ~
z 9 z
>-
8
>- 8
~
~
...
~
~
~

.. '
7

r~
7

6 6

5 5
508 508
-2 506 ~
506 -2
504
C¡,a¡ 502 2
11a9 e 500
(1¡¡) 498 6 1' \fO)
b)
EMD measured RMSE = 0.10 m EMD predicted
9
···':~
9

' ' r;~ (~

.. -~
8 8
<~ ~
7
'
..... """'v ~ I
7
·~ ~
~~~
6
I:;; 6
5
(~~
-..;; :;; 5

~~
1-
1-
z ~ 'El z
e...v >- 4
>-
-:::
4
""'€>
~
3
-:::: 3
-- ~

·-
2 2 ~

1
596 (~
~ -2
1
596
~(
~
~ -2
594 o 594 o
592 2 592 2

~ ' """ ~ ' """

Cha¡11a 590 4 Cha¡11a 590 4
e) 9e (tn) 588 6 9e (tn) 588 6

EMD < -0.3 e-0.3 < EMD < -0.15 -0.15 < EMD < 0.15 e 0.15 < EMD < 0.3 e 0.3 < EMD
O Collaring - Blasthole
Fig. 14. Representation of the predicted EMD model (right graph) and the EMD measured (left graph) for blastholes of the boom 1: (a) blast in chainage 388, 75th
percentile of the RMSE; (b) blast in chainage 500, median value of the RMSE; (e) blast in chainage 591, 25th percentile of the RMSE.

of scanner profiles of the excavated sections with the blasthole posi- may be considered a damage measure. Sources of uncertainty such as
tions, a methodology has been developed to obtain an Excavated Mean drilling deviations, the scaling and primary support done before scan-
Distance (EMD) between the blasthole and the excavated profile, which ning the excavated section, possible variations (unrecorded) in the
explosive linear density, etc., are assumed to be of random nature, of Computers and Operations Research in the Mineral Industry: 23/05/2015-27/05/
unavoidable in the harsh condition in which such data are measured. 2015.
Ghosh, R., 2017. Assessment of rock mass quality and its effects on charge ability using
Quantitative predictions for different conditions would require a re- drill monitoring technique. Doctoral dissertation. Lulecl Tekniska Universitet.
calibration of the model for the new site, following the methodology Hatherly, P., Leung, R., Scheding, S., Robinson, D., 2015. Dril! monitoring results revea!
described here. geological conditions in blasthole drilling. Int. J. Rock Mech. Min. Sci. 78, 144-154.
Hjelme, J.G., 2010. Dril! parameter analysis in the Loren tunnel. M.Sc. thesis in
Given that blasting factors are constant (as from blast reports) the Geosciences. University of Oslo, Department of Geosciences.
overbreak and underbreak are considered mainly influenced by the Hoek, E., Carranza-Torres, C.T., Corkum, B., 2002. Hoek-Brown failure criterion - 2002
geotechnical condition of the rock mass. Such rock mass properties are edition. In: Proceedings of the Fifth North American Rock Mechanics Symposium,
Toronto, Canada, vol. 1, pp. 267-273.
assessed from MWD parameters that show the response of the jumbo to
Hoek, E., 2012. Blast Damage Factor D. Technical note for RocNews, February.
the rock before blasting. A thorough transformation of the MWD logs Hu, Y., Lu, W., Chen, M., Yan, P., Yang, J., 2014. Comparison of blast-induced damage
has been carried out to filter out systematic variations due to the nature between presplit and smooth blasting of high rock slope. Rock Mech. Rock Eng. 47
(1), 1307-1320.
of the drilling process and to highlight the dependence of the rock in the
Hustrulid, W.A., 1968. Theoretical and experimental study of percussive drilling of rock.
parameters. This transformation includes: (i) filtering out of unrealistic Doctoral Thesis. University of Minnesota.
data, (ii) removal of the ramp-up section of the logs, (iii) normalization Hustrulid, W., 2010. Sorne comments regarding development drifting practices with
of systematic variations in the MWD parameters so that the influence of special emphasis on caving applications. Min. Technol. 119 (3), 113-131.
!barra, J.A., Maerz, N.H., Franklin, J.A., 1996. Overbreak and underbreak in underground
hole length and feed pressure are corrected and (iv) normalization with openings part 2: causes and implications. Geotech. Geol. Eng. 14 (4), 325-340.
the standard deviation to account for fluctuations in the signals. Johnson, J.C., 2010. The Hustrulid bar- a dynamic strength test and its application to the
A non-linear power-form model has been developed that predicts cautious blasting of rock. Doctoral Thesis. Department of Mining Engineering,
University of Utah.
the excavated mean distance as function of the normalized penetration Kahraman, S., Rostami, J., Naeimipour, A., 2016. Review of ground characterization by
rate, hammer pressure, rotation speed, rotation pressure and water flow using instrumented drills for underground mining and construction. Rock Mech. Rock
parameters, and the lookout distance. They combine the rotational, Eng. 49 (2), 585-602.
Kim, Y., Bruland, A., 2009. Effects of drilling and geological parameters on contour
hydraulic and percussive mechanisms of the drill, and the confinement quality in a dril! and blast tunnel. Tunn. Undergr. Space Technol. 24 (5), 584-591.
of the rock mass by depth. The model has a determination coefficient of Kim, Y., Bruland, A., 2015. A study on the establishment ofTunnel Contour Quality Index
0.74, with the coefficients of the model strongly significant. Residuals considering construction cost. Tunn. Undergr. Space Technol. 50, 218-225.
Kwon, S., Lee, C.S., Cho, S.J., Jeon, S.W., Cho, W.J., 2009. An investigation of the ex-
are essentially normally distributed. The signs of the exponents indica te
cavation damaged zone at the KAERI underground research tunnel. Tunn. Undergr.
that normalized penetration rate, hammer pressure, rotation pressure Space Technol. 24 (1), 1-13.
and the lookout distance inversely influence the excavated distance i.e. Leung, R., Scheding, S., 2015. Automated coa! seam detection using a modulated specific
energy measure in a monitor-while-drilling context. Int. J. Rock Mech. Min. Sci. 75,
high values for them reflect hard, unaltered rock. On the other hand,
196-209.
normalized rotation speed and water flow are directly correlated with Liu, H., Yin, K. Karen, 2001. Analysis and interpretation of monitored rotary blasthole
the excavated distance, so that high values for them indicate soft, dril! data. Int. J. Surf. Min. Reclam. Environ. 15, 177-203.
fractured rock. Mahdevari, S., Haghighat, H.S., Torabi, S.R., 2013. A dynamically approach based on
SVM algorithm for prediction of tunnel convergence during excavation. Tunn.
Undergr. Space Technol. 38, 59-68.
Acknowledgements Mahtab, M.A., Rossler, K., Kalamaras, G.S., Grasso, P., 1997. Assessment of geological
overbreak for tunnel design and contractual claims. Int. J. Rock Mech. Min. Sci. 34
(3-4), 185-el.
This work has been conducted under the project "TUÑEL" Manda!, S.K., Singh, M.M., 2009. Evaluating extent and causes of overbreak in tunnels.
(PCD16264900008) funded by the Centre for Industrial Technological Tunn. Undergr. Space Technol. 24 (1), 22-36.
Development (CDTI, Government of Spain). The authors would like to Matlab, 2017. The MathWorks !ne., Natick, MA.
Naeimipour, A., Bahrampour, S., Rostami, J., Dogruoz, C., 2014. Ground characterization
thank OSSA Obras Subterráneas SA, for providing the necessary data. while drilling roofbolters. In: Proceedings of North American Tunneling Conference,
The support of MAXAM Civil Explosives is also acknowledged. Special 22-25 June, Los Angeles, CA, 11, pp. 51-59.
thanks are due to Prof. Hakan Schunnesson of Lulea University of Navarro, J., Sanchidrián, J.A., Segarra, P., Castedo, R., López, L.M., Cienfuegos, R., 2017.
MWD parameters and Drilling Control system. In: Proceedings World Tunnel
Technology, for his guidance in the MWD normalization analysis. We Congress 2017, 9-15 June 2017, Bergen, Norway, pp. 194-203.
would also like to thank the anonymous reviewers for their valuable Navarro, J., Sanchidrián, J.A., Segarra, P., Castedo, R., Paredes, C., López, L.M., 2018. On
suggestions regarding the analyses and the discussion of the results. the mutual relations of drill monitoring variables and the drill control system in
tunneling operations. Tunn. Undergr. Space Technol. 72, 294-304.
NPRA - Norwegian Public Road Administration, 2004. Road Tunnel. Manual 021. ISBN
Appendix A. Supplementary material 82-7207-540-7.
Oggeri, C., Ova, G., 2004. Quality on tunneling: ITA-AITES Working Group 16 final re-
Supplementary data associated with this article can be found, in the port. Tunn. Undergr. Space Technol. 19, 239-272.
Olsson, M., 2010. Tunnel blast design for minimizing of the damage zone. A guidance
online version, at https://doi.org/10.1016/j.tust.2018.08.060. document for SKB and Posiva. Swebrec Report 2010-4, Stockholm, Sweden.
Ouchterlony, F., Olsson, M., Bergqvist, l., 2002. Towards new Swedish recommendations
References for cautious perimeter blasting. Fragblast 6 (2), 235-261.
Peng, S.S., Tang, D., Sasaoka, T., Luo, Y., Finfinger, G., Wilson, G., 2005. A method for
quantitative void/fracture detection and estimation of rock strength for underground
Andersson, P., 1994. The damage zone at tunnel driving, seminar Oct. 1992, SveBeFo mine roof. In: Proceedings 24th International Conference on Ground Control in
report 8. Swedish Rock Engineering Research, Stockholm. Mining, 2-4 August, Lakeview, WV, pp. 187-195.
AutoCAD, 2017. Autodesk, !ne. Schunnesson, H., 1996. RQD predictions based on dril! performance parameters. Tunn.
Barton, N., Lien, R., Lunde, J., 1974. Engineering classification of rock masses for the Undergr. Space Technol. 11, 345-351.
design of tunnel support. Rock Mech. 6 (4), 189-236. Schunnesson, H., 1997. Drill process monitoring in percussive drilling for location of
Beattie, N.C.M., 2009. Monitoring-while-drilling for open-pit mining in a hard rock en- structural features, lithological boundaries and rock properties, and for dril! pro-
vironment: an investigation of pattern recognition techniques applied to rock iden- ductivity evaluation. Doctoral Thesis. Dept. of Environmental Planning and Design,
tification. Doctoral Thesis. Dept. of Mining Engineering, Queen1s University, Division of Applied Geology, Lulea University of Technology, Lulea.
Kingston, Canada. Schunnesson, H., 1998. Rock characterization using percussive drilling. Int. J. Rock Mech.
Costamagna, E., Oggeri, C., Castedo, R., Segarra, P., Navarro, J., 2018. Assessment of Min. Sci. 35, 711-725.
contour profile quality in D&B tunneling. Tunn. Undergr. Space Technol. 75, 67-80. Schunnesson, H., Kristoffersson, T., 2011. Rock mass characterization using drill and
Cunningham, C.V.B., Goetzsche, A.F., 1990. The specifications of blast damage limita- crushability monitoring-A case study. Int. J. COMADEM. 14 (2), 44-52.
tions in tunnelling contracts. Tunn. Undergr. Space Technol 5, 193-198. Schunnesson, H., Elsrud, R., Rai, P., 2011. Drill monitoring for ground characterization in
Deere, D.U., Miller, R.P., 1966. Engineering Classification and Index Properties for Intact tunnelling operations. In: 20th International Symposium on Mine Planning and
Rock. Illinois University at Urbana, Department of Civil Engineering. Equipment Selection, 12-14 October, Almaty, Kazakhstan.
Fritsch, F.N., Carlson, R.E., 1980. Monotone piecewise cubic interpolation. SIAM J. Schunnesson, H., Poulopoulos, V., Bastis, K., Pettersen, N., Shetty A., 2012. Application of
Numer. Anal. 17, 238-246. computerized drill jumbos at the Chenani-Nashri tunnelling site in Jammu-Kashmir,
Ghosh, R., Schunnesson, H., Kumar, U., 2015. The use of specific energy in rotary drilling: India. In: Proceedings 2lst International Symposium on Mine Planning and
The effect of operational parameters. In: International Symposium on the Application Equipment Selection, 28-30 November 2012, New Delhi, India, pp. 729-751.
Schunnesson, H., 2017. Personal communication at Lulecl Techniska Universitet, Sweden. Engineering, University of West Virginia.
Scoble, M.J., Peck, J., Hendricks, C., 1989. Correlation between rotary dril! performance Teale, R., 1965. The concept of specific energy in rock drilling. Int. J. Rock Mech. Min.
parameters and borehole geophysical logging. Min. Sci. Technol. 8 (3), 301-312. Sci. 2, 57-73.
Scoble, M.J., Lizotte, Y.C., Paventi, M., Mohanty, B.B., 1997. Measurement of blasting Van Eldert, J., Schunnesson, H., Johansson, D., Saiang, D., 2018. Measurement While
damage. Mining Engineering, June 1997, pp. 103-108. Drilling (MWD) technology for blasting damage calculation. In: Proceedings 12th
Singh, P.K., Roy, S.K., Sinha, A., 2003. A new blast damage index for the safety of un- International Symposium on Rock Fragmentation by Blasting, 11-13 June 2018,
derground coa! mine openings. Min. Technol. 112 (2), 97-104. Lulea, Sweden, pp. 139-148.
Singh, S.P., Xavier, P., 2005. Causes, impact and control of overbreak in underground Xu, D.P., Feng, X.T., Chen, D.F., Zhang, C.Q., Fan, Q.X., 2017. Constitutive representation
excavations. Tunn. Undergr. Space Technol. 20, 63-71. and damage degree index for the layered rock mass excavation response in under-
SN - Schweizer Norm, 2004. Conditions générales pour constructions souterraines. SIA - ground openings. Tunn. Undergr. Space Technol. 64, 133-145.
Société suisse des ingénieurs et des architectes, Zurich. Yu, T.R., Vongpaisal, S., 1996. New blasting damage criteria for underground blasting.
Tang, X., 2006. Development of real time roof geology detection system using drilling CIM Bull. 89 (998), 139-145.
parameters during roof bolting operation. Doctoral Thesis. Department of Mining