LICENTIATE T H E S I S

Sohail Manzoor R

ock Evaluation U

sing Digital Im

ages and Drill M

onitoring Data

Department of Civil, Environmental and Natural Resources EngineeringDivision of Mining and Geotechnical Engineering

ISSN 1402-1544ISBN 978-91-7790-655-1 (print)ISBN 978-91-7790-656-8 (pdf)

Luleå University of Technology 2020

Rock Evaluation Using Digital Images

and Drill Monitoring DataBefore and after rock blasting

Sohail Manzoor

Mining and Rock Engineering

130545-LTU-Sohail.indd All Pages130545-LTU-Sohail.indd All Pages 2020-10-19 16:312020-10-19 16:31

Licentiate Thesis

Rock Evaluation Using Digital Images and

Drill Monitoring Data Before and after rock blasting

Sohail Manzoor

Division of Mining and Geotechnical Engineering

Department of Civil, Environmental and Natural Resources Engineering

Luleå University of Technology

Luleå, Sweden

Published by Luleå University of Technology, 2020

ISSN 1402-1757 ISBN 978-91-7790-655-1 (print) ISBN 978-91-7790-656-8 (electronic)

Luleå 2020

www.ltu.se

i

FOREWORD

This thesis summarizes the research work carried out for the degree of Licentiate in Mining Engineering at the Division of Mining and Geotechnical Engineering, Luleå University of Technology, Sweden. The research was partially funded by Vinnova, the Swedish Energy Agency and Formas through the SIP-STRIM program. Their support is gratefully acknowledged. I would also like to acknowledge the support from the SLIM project funded by the European Union's Horizon 2020 research and innovation programme under grant agreement no. 730294 and the support from Centre for Advanced Mining and Metallurgy (CAMM2), Sweden.

Sincere gratitude is expressed to my supervisor Associate Professor Anna Gustafson for her valuable time, support and guidance throughout the research. My co-supervisors Professor Håkan Schunnesson and Professor Daniel Johansson are also gratefully acknowledged for their enlightening discussions. Hanna Falksund (LKAB, Malmberget), David Vojtech (LKAB, Svappavaara), Dr. Peter Schimek (VA Erzberg, Austria) and Thomas Seidl (Montanuniversität, Leoben) are recognized for their support in data acquisition. I also acknowledge my colleagues for their support in various ways, especially Samaneh Liaghat, Ulf Stenman, Gurmeet Shekhar and Balazs Varannai.

I am most thankful to my Allah, who enabled me to accomplish this milestone, and my parents for their lifetime unconditional love, support and encouragement. I owe my love and gratitude to my lovely wife, who believed in me and stood by my side to pursue my research work. I thank her for the patience and continuous encouragement and inspiration during this research.

Sohail Manzoor

October 2020 Luleå, Sweden

ii

iii

ABSTRACT

Current practices of assessing rock mass usually involve techniques that focus on the surface or outcrop of the rock mass, such as scanline surveys, window surveys, photogrammetry and laser scanning etc. These techniques generally lack the ability to provide sufficient information about the rock mass or bear various inherent constraints, such as safety issues, time requirements, user biasness, equipment requirements and reproducibility of results. The rock fragmentation is predicted using mathematical equations known as fragmentation models. However, these models ignore some key factors that significantly affect the nature of fragmentation in sublevel caving, such as chargeability of blastholes, fan shaped drilling pattern, and they require numerous rock parameters which are not well known in most cases. These models are often site-specific and are mostly developed for surface mines. Therefore, their application in underground mining is not so common.

The aim of this research is to investigate the possibility of improving rock mass assessment and rock fragmentation prediction. Drill monitoring technique was selected as a potential tool to analyse the rock mass and forecast the rock fragmentation.

To test the selected technique, measurement while drilling (MWD) data were collected from three different mines. The variations in the MWD data were analysed to identify zones and structures present in the rock mass. The results were compared to 3D images obtained by close-range terrestrial digital photogrammetry for validation; they showed close agreement. MWD data were also used to classify the rock mass into five classes in a sublevel caving operation: solid, slightly fractured, highly fractured, having cavities and major cavities. The loading operation of the blasted rock was filmed, and digital images of Load-Haul-Dump (LHD) buckets containing blasted rock were extracted from the video recordings. The blasted rock inside the buckets was categorized as fine, medium, coarse and oversize fragmentation based on its median fragment size (X50). A statistical analysis determined the correlation between MWD based rock mass classes and fragmentation classes.

The results showed that the MWD data can be used to differentiate between weak and strong rock mass or between fractured and competent rock mass. It can be used to differentiate joints, cavities, foliations etc. Fine and medium sized fragmentation has better correlation and can be predicted with higher accuracy using MWD data than coarse and oversize fragmentation. The results suggest the drill monitoring technique has the potential to assess rock mass and predict rock fragmentation to some extent. Keywords: Measurement while drilling, Photogrammetry, Rock fragmentation, Mining, Sublevel caving

iv

v

LIST OF APPENDED PAPERS

Paper A

Manzoor, S., Liaghat, S., Gustafson, A., Johansson, D., and Schunnesson, H., 2019. Rock mass characterization using MWD data and photogrammetry. In: Mining goes Digital: Proceedings of the 39th International Symposium 'Application of Computers and Operations Research in the Mineral Industry' (APCOM 2019), June 4-6, 2019, Wroclaw, Poland, p. 217-225. DOI: 10.1201/9780429320774-25

Paper B

Manzoor, S., Liaghat, S., Gustafson, A., Johansson, D., and Schunnesson, H., 2020. Establishing relationships between structural data from close-range terrestrial digital photogrammetry and measurement while drilling data. Engineering Geology, 267 (2020), 105480. DOI: 10.1016/j.enggeo.2020.105480

Paper C

Manzoor, S., Danielsson, M., Schunnesson, H., Gustafson, A., Liaghat, S., and Johansson, D., 2020. Predicting rock fragmentation based on measurement while drilling data: A case study from Malmberget mine, Sweden. Submitted to Scientific Journal.

vi

vii

TABLE OF CONTENTS

1 INTRODUCTION .............................................................. 1

1.1 Background .......................................................................................... 1 1.2 Problem statement ................................................................................ 2 1.3 Aim and objectives ............................................................................... 2 1.4 Research questions ............................................................................... 2 1.5 Scope and limitation ............................................................................. 3 1.6 Research flow ...................................................................................... 3 1.7 Thesis structure .................................................................................... 4 1.8 Authors’ contributions to appended papers ............................................ 4

2 LITERATURE REVIEW ..................................................... 7

2.1 Rock mass assessment ........................................................................... 7 2.1.1 Traditional methods .............................................................................. 7 2.1.2 Remote sensing techniques ...................................................................... 8 2.1.3 Measurement while drilling: A potential rock assessment technique ...................... 9

2.2 Rock fragmentation ............................................................................ 10 2.2.1 Effect of rock fragmentation ................................................................... 11 2.2.2 Fragmentation in sublevel caving (SLC) ................................................... 11 2.2.3 Factors affecting fragmentation in SLC ..................................................... 12 2.2.4 Fragmentation assessment ..................................................................... 13 2.2.5 Fragmentation prediction models ............................................................. 13 2.2.6 Measurement while drilling: A potential fragmentation prediction tool ................ 14

3 TEST SITES .................................................................... 15

3.1 Erzberg mine ...................................................................................... 15 3.2 Leveäniemi mine ................................................................................ 16 3.3 Malmberget mine ............................................................................... 17

4 MATERIALS AND METHODS ......................................... 19

4.1 Measurement while drilling data ......................................................... 19 4.2 Photogrammetric data ......................................................................... 20 4.3 Video recordings ................................................................................ 22 4.4 Data processing ................................................................................... 22

4.4.1 Measurement while drilling data ............................................................. 22 4.4.2 Photogrammetric data .......................................................................... 24

4.5 Statistical analysis ................................................................................ 28

5 RESULTS AND DISCUSSION .......................................... 29

5.1 Pre-blast rock mass assessment based on MWD ................................... 29 5.2 Fragmentation prediction based on MWD data ................................... 34

viii

5.2.1 Correlation test results ......................................................................... 34 5.2.2 Partial least square regression (PLS-R) results ............................................ 36

6 CONCLUSIONS, CONTRIBUTION AND FUTURE RESEARCH ..................................................................... 39

6.1 Conclusions ........................................................................................ 39 6.2 Research contribution ........................................................................ 39 6.3 Future research ................................................................................... 40

7 REFERENCES ................................................................. 41

ix

LIST OF FIGURES

Figure 1.1: Overview of research flow ........................................................................ 3

Figure 2.1: Traditional methods of discontinuity measurements (Noroozi et al., 2015)7

Figure 2.2: Contact free data acquisition for rock face profiling (Laser Technology Inc., 2020) ................................................................................................ 8

Figure 2.3: An oversize fragment bigger than the LHD bucket ................................. 12

Figure 3.1: Areial view of Erzberg mine with the lower right corner showing the location of mine in Austria (Boch et al., 2019) ....................................... 16

Figure 3.2: Leveäniemi open pit mine with marked walls used for this study (modified from hitta.se, 2018) ................................................................................ 17

Figure 3.3: Layout of SLC operation at LKAB mines (LKAB, 2020) ........................ 18

Figure 4.1: Schematic layout of fan shaped drilling pattern in SLC ........................... 19

Figure 4.2: Basic geometric arrangement for data acquisition of a bench face (modified from 3GSM, 2010) ................................................................................. 20

Figure 4.3: Stereoscopic image pair. Two corresponding image points P(u,v) relate to one three-dimensional object point P(X,Y,Z) (3GSM, 2010) ................. 21

Figure 4.4: Bench face with range poles at Erzberg's open pit mine in Austria .......... 22

Figure 4.5: Noisy MWD data parameters plotted against depth of the hole .............. 23

Figure 4.6: MWD data parameters vs. depth of the hole with specified location coordinates and status ............................................................................. 24

Figure 4.7: 3D model of a bench face of Leveäniemi mine with joints, exposed surfaces and scanlines (white dotted lines) marked on the surface ............ 25

Figure 4.8: Scanline data showing coordinates of intersecting structures ................... 25

Figure 4.9: Frames with irrelevant (A, B, C, D, E, F, and G) and relevant data (H, I) .............................................................................................................. 27

Figure 4.10: Frames showing four different classes of fragmentation ......................... 28

Figure 5.1: MWD data variations related to different rock types ............................... 29

Figure 5.2: Bench face showing different types of rock mass ..................................... 30

Figure 5.3: Rise in penetration rate (PR) and rotation pressure (PR) at fracture/joint location .................................................................................................. 31

Figure 5.4: Variations in penetration rate (PR) and rotation pressure (RP) along with corresponding structures ......................................................................... 32

Figure 5.5: MWD data from two holes showing fluctuations in all the drilling parameters .............................................................................................. 33

Figure 5.6: MWD data plot of a borehole and rock face image for poor rock conditions .............................................................................................. 34

Figure 5.7: Scatter plots and linear regression lines corresponding to different fragmentation and rock mass classes ........................................................ 35

Figure 5.8: PLS-R model quality indicators .............................................................. 36

x

Figure 5.9: Comparison of actual and predicted fragmentation (%) based on eq. 1 and eq. 2. ...................................................................................................... 38

1

1 INTRODUCTION

1.1 Background Rock evaluation, whether before or after the rock mass is blasted, has a key role in designing and implementing the pre-blast and post-blast activities in rock engineering operations. In this research, rock evaluation includes rock mass assessment before blasting and rock fragmentation analysis after blasting. Blasting is a dynamic and complex process which requires a detailed description of the rock mass to be blasted. Rock mass properties are among the most important influencing parameters in blasting outcomes (Bhandari, 1997; Latham and Lu, 1999; Azimi et al., 2010; Akbari et al., 2015). These properties include intact rock properties, i.e., physical and mechanical properties of the rock, and discontinuity structures of the rock mass.

The intact rock properties are usually determined from in-situ or laboratory tests, while discontinuity structures are mostly measured in the field (Firpo et al., 2011). Traditional methods for determining the geometrical characteristics of the rock mass in the field require a geological compass and physical access to the measurement locations (Gaich et al., 2006). The traditional methods may suffer from human error or measurement bias (Ferrero et al., 2009; Gigli and Casagli, 2011) and have other constraints, including safety issues, access and time restrictions, difficulty reproducing the results, limited coverage in large target areas, lack of user experience etc.. These constraints make traditional methods unpopular with users (Kemeny and Post, 2003; Ferrero et al., 2009), and they are often replaced by remote sensing techniques. Photogrammetry and laser scanning are currently the most widely used techniques for rock face characterization (Ferrero et al., 2009). Each has advantages and disadvantages (Kemeny and Post, 2003; Jaboyedoff et al., 2009; Sturzenegger and Stead, 2009a; Lato et al., 2010; Fisher et al., 2014; Li et al., 2016), but both can overcome the limitations of traditional methods. They have other advantages as well, including the availability of data, more objective results, the ability to work in magnetic environments etc. The mining industry has specific problems with geomechanical data collection, for example, limited exposure and unsafe rock face conditions. Therefore, a safe, fast and contact-free data acquisition technique is always preferred over a traditional geological mapping technique (Lemy and Hadjigeorgiou, 2003). However, these remote sensing techniques require expensive instrumentation, fieldwork that can hinder the production cycle in mining operation and a software to extract the information from the collected data. In a weathered rock face or if blasting practices are poor, these techniques may also over-estimate rock fracturing.

Without sufficient understanding of the rock mass, any blasting activity is prone to a high degree of uncertainty. Evaluation of blast outcome is critical to improve and optimize blast design as it dirctly affects the downstreem processes, such as loading, hauling and crushing (Badroddin et al., 2013; Liu et al., 2020). The criterion of a good blast result depends on many factors, such as good heave, uniform fragmentation, loose muck, muck profile angle etc. (Jha et al., 2015). Rock fragmentation is defined as the size distribution of blasted rock and is often used as a key parameter to measure blast performance because of its association with drilling, blasting, loading, hauling and

2

crushing activities (Hunter et al., 1990). In sublevel caving (SLC), fragmentation is a very important factor; it affects the gravity flow of the blasted material, the loading at the draw points, the efficiency of Load-Haul-Dump (LHD) operations, the ore-pass efficiency and the energy requirements of crushers (Wimmer, 2010; Danielsson et al., 2017). Rock fragmentation is generally assessed by either a direct method involving sieving analysis or an indirect method involving observational, empirical and image-based methods. Empirical methods have many benefits, as they help to target required rock size distribution after blasting (Morin and Ficarazzo, 2006), but the uncertainty and even unavailability of the required data to estimate rock properties cause errors. Another limitation of fragmentation models is their origin; i.e., they are mostly developed for surface mines (Ouchterlony and Sanchidrian, 2019) and lack application in underground environments. For example, they do not incorporate several essential factors relevant to fragmentation in SLC mining, such as the shape of the blast fan, confined nature of blasting, drilling accuracy, chargeability etc.

1.2 Problem statement The current techniques of pre-blast rock mass assessment are generally time consuming, costly and limited to the exposed surface. Therefore, these techniques do not provide accurate insight into the hidden volume of the rock mass. This uncertainty in rock mass description can lead to poor blast results, as well as slope stability issues. Similarly, the prediction models used for rock fragmentation have various limitations, especially in an underground environment. The models’ parameters which are related to rock mass properties are usually not well known and are estimated using empirical formulae developed under certain circumstances (Ouchterlony and Sanchidrian, 2019). However, the uncertainty and even, in some cases, unavailability of complete data for the estimation of these parameters cause errors in these fragmentation models. As these fragmentation models have been developed mainly for surface blasting, they do not consider several important factors relevant to fragmentation in SLC mining, such as the design and success of drilling and blasting operations. It is important to find a solution to these limitations with pre-blast rock mass assessment and post-blast fragmentation prediction.

1.3 Aim and objectives The aim of this thesis is to investigate the potential of MWD data for improving pre-blast rock mass assessment and post-blast rock fragmentation prediction and, thus, help to overcome the limitations of current techniques. The specific objectives of the thesis include the following:

• To assess rock mass quality using MWD data and validate the results. • To find the relationship between rock fragmentation and MWD-based rock

mass types in sublevel caving.

1.4 Research questions This thesis addresses the following research questions:

1. What are the constraints/limitations of the current rock evaluation techniques?

3

2. How can rock mass structures be correlated with MWD data using established methods of rock mass characterization?

3. How does fragmentation correlate with MWD data?

1.5 Scope and limitation This thesis is focussed on pre-blast rock assessment and post-blast rock fragmentation prediction. It investigates the potential of the measurement while drilling technique to assess the hidden volume of the rock mass and to forecast the expected fragmentation after a blast in SLC operation. The pre-blast assessment research in this thesis is limited to surface mines. In addition, the analysis looks at the MWD data of each hole. It does not provide a large-scale analysis of MWD data for rock mass assessment. The research on rock fragmentation prediction is limited to underground operation in sublevel caving mines, but it can be expanded to other mining methods as a fragmentation prediction tool.

1.6 Research flow The objective of the research was to find an alternate tool for rock mass assessment and fragmentation prediction. Figure 1.1 shows the path adopted to achieve the research objective. It consists of a comprehensive literature review, data acquisition, data processing and the documentation of research results.

Problem statement

Literature review

Data acquisition for pre-blast rock mass assessment

Data analysis and documentation of

results

Formulation of research questions

Post-blast fragmentation evaluation

Pre-blast rock mass assessment

Correlating MWD and structural data

Correlating MWD and fragmentation data

Paper A and B

Paper C

Selection of required data and potential test sites

Processing measurement while drilling, image and filming data

Data acquisition for post-blast fragmentation evaluation

Measurement while drilling data

Digital images of bench faces

Measurement while drilling data

Filming of material handling

Figure 1.1: Overview of research flow

4

1.7 Thesis structure This thesis is divided into six (06) chapters briefly discussed below.

Chapter 1 includes background of the research area, problem statement, aim and objectives of the research, research questions answered by the thesis, scope and limitation of the research, structure and flow of the research and different authors’ contributions to the appended papers.

Chapter 2 covers the literature review of the available information, previous studies related to the topic and potential improvements.

Chapter 3 contains the information on the case-study mine sites. It describes their location, the principal mining methods applied, production details etc.

Chapter 4 describes the type of data acquired for the study, the acquisition methods and how different data types were processed to get relevant information.

Chapter 5 documents the results of the study. It includes the results for pre-blast rock mass assessment using MWD data and post-blast rock fragmentation evaluation using MWD data.

Chapter 6 summarizes the conclusions drawn from the study and explains how the results can be used to improve mining practices. It also includes the contributions of the thesis to the research community and gives insight into possible research areas/activities that can be explored/performed in the future to benefit both the Swedish and the global mining community.

A complete list of references studied/consulted is given at the end of the thesis.

1.8 Authors’ contributions to appended papers Table 1-1 includes the details of the activities carried out by each author in the papers/manuscripts. Major activities include:

1. Literature review 2. Problem definition 3. Data acquisition: digital images/filming of material handling operation 4. Data acquisition: measurement while drilling data 5. Data processing: image/filming data 6. Data processing: measurement while drilling data 7. Data analysis and results 8. Drafting and editing manuscripts 9. Supervision, reviewing and editing manuscripts 10. Manuscript submission, revision and final acceptance

5

Table 1-1: Authors' contributions to appended papers

Authors Paper A Paper B Paper C

Sohail Manzoor 1,2,3,5,7,8,10 1,2,3,5,7,8,10 1,2,3,5,7,8,10

Samaneh Liaghat 6 3,6,8 6

Anna Gustafson 2,3,9 2,3,9 2,3,9

Daniel Johansson 9 9 9

Håkan Schunnesson 2,3,4,9 2,3,4,9 2,3,4,6,9

Markus Danielsson --- --- 3,5

6

7

2 LITERATURE REVIEW

An extensive literature review identified the current and potential techniques to improve pre-blast rock mass assessment and post-blast fragmentation prediction. The findings are summarized in the following sections.

2.1 Rock mass assessment Pre-blast rock mass assessment is very important in rock engineering projects, as it directly affects the blast and slope designs. The existence of weak zones or discontinuities within a rock mass should be investigated carefully in any rock engineering or geotechnical project. The knowledge of the physio-mechanical properties, as well as the geometry of the rock mass, is of prime importance in rock engineering, as these traits control rock failure mechanisms (Firpo et al., 2011; Gigli and Casagli, 2011; Chen et al., 2016; Buyer and Schubert, 2017; Riquelme et al., 2017). With insufficient understanding of the geometrical setup of the rock mass, any sort of rock failure analysis will suffer a high degree of uncertainty. Physio-mechanical properties are usually determined from in-situ or laboratory tests, while geometrical parameters are measured mostly in the field using traditional methods or remote sensing techniques.

2.1.1 Traditional methods The traditional methods for determining the geometrical characteristics of the rock mass in the field require a geological compass and physical access to the measurement locations (Gaich et al., 2006). These traditional methods for discontinuity characterization include one dimensional scanline survey and two dimensional window surveys (Sturzenegger and Stead, 2009a). Figure 2.1 shows an example of these traditional methods.

Figure 2.1: Traditional methods of discontinuity measurements (Noroozi et al., 2015)

These surveys are typically performed by a geologist or field engineer and thus may suffer from human error or measurement bias (Ferrero et al., 2009; Gigli and Casagli, 2011). There are several other constraints, such as safety issues, access and time restrictions, difficulties in reproducing the results, limited coverage in large target areas, user experience etc., that make these methods less preferred by users (Kemeny and Post,

8

2003; Ferrero et al., 2009). The mining industry has specific problems with geomechanical data collection, for example, limited exposure and unsafe rock face conditions. Therefore, a safe, fast, and contact-free data acquisition technique is always preferred over a traditional geological mapping technique (Lemy and Hadjigeorgiou, 2003). Accordingly, remote sensing techniques are gaining widespread acceptance in these applications.

2.1.2 Remote sensing techniques With the recent technological advancements, digital elevation modelling and geomorphological terrain analysis have undergone an evolutionary transformation in the last decade (Westboy et al., 2012). Several techniques are now available to determine the geometrical properties of the rock mass without physical access, as shown in Figure 2.2, thus increasing user safety and overcoming access and time restrictions.

Figure 2.2: Contact free data acquisition for rock face profiling (Laser Technology

Inc., 2020)

These techniques involve the development of 3D geometrical models of the rock surface and analysing the models using specially developed computer software to determine the geometry of the rock mass (Feng et al., 2001; Haneberg, 2008; Sturzenegger and Stead, 2009a, b; Bonilla-Sierra et al., 2015; Riquelme et al., 2015). These contact-free techniques date back to the 1960s when early researchers used terrestrial photographs for geological mapping (Linkwitz, 1963; Rengers, 1967). However, image-based measurements have been advancing rapidly since 1990s (Roberts and Poropat, 2000). Other possibilities include the use of laser scanners or a combination of laser scanners and digital images to construct rock faces and extract the required information (Kemeny et al., 2003; Feng and Roeshoff, 2004; Lemy and Hadjigeorgiou, 2004).

9

Photogrammetry and laser scanning have emerged as the most widely used tools for rock face characterization (Ferrero et al., 2009; Jaboyedoff et al., 2012; Westboy et al., 2012; Salvini et al., 2013; Colomina and Molina, 2014; Chen et al., 2016; Menegoni et al., 2018). Both techniques can overcome the limitations of traditional methods, and they provide other advantages as well (see Kemeny and Post, 2003; Jaboyedoff et al., 2009; Sturzenegger and Stead, 2009a; Lato et al., 2010; Fisher et al., 2014; Li et al., 2016), such as the following:

• They provide more representative data than other conventional methods, as they are not restricted to the base of the outcrop.

• They offer easier access to steep and high rock faces. • They are safer in poor rock conditions, as the surveyor does not need to

physically access the rock face. • They are applicable in magnetic environments where a conventional compass-

clinometer method fails to work.

2.1.2.1 Photogrammetry Many years ago, ISRM (1978) suggested photogrammetry can be a viable alternative to traditional methods in potentially dangerous and inaccessible areas. In the last two decades, aerial photogrammetry and close-range terrestrial digital photogrammetry (CRTDP) have evolved as powerful tools and are now widely used for 3D topographic modelling (Remondino and El-Hakim, 2006; Matthews, 2008; Fraser and Cronk, 2009; Salvini et al., 2013; Assali et al., 2014). “Close-range” denotes up to 300 m from the camera/scanner location to the object under consideration (Wolf and Dewitt, 2000). Photogrammetry is very useful in determining 3D structural parameters from stereoscopic images (Vasuki et al., 2014). It has been applied for many years in both surface and underground mining operations (Fekete et al., 2010). It is becoming even more powerful, conquering limitations faced in the past through the recent advancements in digital camera technology and the use of unmanned aerial vehicles (UAVs) for digital photography in remote areas (Bemis et al., 2014). The use of UAVs can considerably improve data acquisition in inaccessible areas where other techniques face limitations (Salvini et al., 2017; Francioni et al., 2018). However, photogrammetry requires an expensive camera, fieldwork that can hinder the production cycle in mining operation, and a software to extract the information from the images. In addition, it can over-estimate the rock fracturing because of weathering of the rock face or poor blasting practices.

2.1.3 Measurement while drilling: A potential rock assessment technique The measurement while drilling (MWD) technique has the potential to provide valuable insight into the rock mass as it monitors the drilling process. It does not require additional equipment like laser scanners or cameras to collect data. It is considered a low-cost approach to get high resolution data in less time than other subsurface exploration techniques (Khorzoughi, 2013). Since data are recorded during the drilling operation, the technique does not slow down the production process. It records various parameters, for example, drilling depth, penetration rate, percussive pressure, feed pressure, rotation pressure, rotation speed, damper pressure, flushing pressure etc., at

10

specified intervals along the length of the borehole as the drill bit penetrates the rock mass (Schunnesson, 1996; van Eldert et al., 2019).

MWD is a useful technique for gathering data to characterize the structural and mechanical properties of the penetrated rock mass (Monteiro et al., 2009; Khorzoughi and Hall, 2016; Rodgers et al., 2018; Vezhapparambu et al., 2018). It permits fast and accurate insight into the geo-mechanical parameters of the rock mass, far in advance of the downstream processes (Rai et al., 2016).

The MWD technique, combined with other exploration methods, provides an improved description of the subsurface rock properties (Segui and Higgins, 2002). Schunnesson (1996) interpreted the MWD parameters to identify the fracture zones along the borehole in a railway tunnel and calculate rock quality designation (RQD). Different studies have used MWD to characterize the rock mass (Khorzoughi, 2013; van Oosterhout, 2016; Ghosh et al., 2017a; Khorzoughi et al., 2018; Van Eldert et al., 2019), showing that MWD parameters respond differently to different rock mass structures. For example, a fractured rock mass exhibits higher penetration rate and increased rotation pressure (Vezhapparambu et al., 2018). Penetration rate is the most effective drilling parameter to characterize the rock mass using MWD (Khorzoughi, 2013). Penetration rate increases when the drilling encounters a fracture and demonstrates high variation in an extensively fractured rock mass (Schunnesson, 1996). The torque/rotation pressure also shows an increase when a broken rock mass is being drilled, as it has the tendency to stall the drill bit (Schunnesson, 1996). The rock strength, fractures, and voids can affect the penetration rate (Segui and Higgins, 2002). The peaks on the logs of penetration rate to depth show the presence of fractures, and the width of these peaks indicate the fractures’ aperture size (Khorzoughi, 2013). As the drill bit encounters an open fracture, the penetration rate increases, and the rotary torque decreases, as there is no resistance based on the aperture size (Khorzoughi, 2013). The feed pressure and pull-down pressure show a slight drop as well because there is no engagement of the drill bit with the rock mass (Khorzoughi, 2013).

2.2 Rock fragmentation Breaking the in-situ rock mass into smaller fragments that can be easily loaded and transported to downstream facilities is an important part of the mining process (Bamford et al., 2017). It is usually achieved by drilling and blasting the rock mass using commercially available explosives and is considered the most efficient way of extracting hard rocks (Elahi and Hosseini, 2017). Rock blasting is a complex operation, as it is influenced by numerous variables, including explosive parameters, drilling and charging parameters, and rock mass parameters (Hunter et al., 1990; Bhandari, 1997; Latham and Lu, 1999). It is a crucial stage in the mining process, as it affects the operational costs related to downstream processes, including loading, hauling, and crushing (Badroddin et al., 2013). To analyse the efficiency of the blasting process, it is vital to develop a reliable technique to assess the blast results. The efficiency of a mining process is generally analysed as total production cost per ton of excavated material (Hunter et al., 1990). This requires a detailed evaluation of unit operations, i.e. drilling, blasting, loading, haulage, and crushing, based on a common parameter (Hunter et al., 1990). Fragmentation is a key parameter to measure the blast performance, as it is related to all

11

these unit operations’ costs. It is a function of blast design, rock strength, and the natural discontinuity framework (Hunter et al., 1990; Azimi et al., 2010).

2.2.1 Effect of rock fragmentation Rock fragmentation is a measure of fragment size distribution of the post-blast broken rock material (Cho and Kaneko, 2004). Its assessment has always been of keen interest to mining engineers seeking to optimize the blasting operations. It has a key influence on the production rates and the machines’ performance (Thurley, 2013). Its impact on downstream processes has been widely studied for half a century by various researchers (Seccatore, 2019). The purpose of these studies has been to optimize rock fragmentation so that the overall mineral extraction can be optimized (Onederra et al., 2015). The development of a reliable fragmentation prediction and evaluation method would facilitate blast optimization (Roy et al., 2016; Babaeian et al., 2019; Liu et al., 2020). It could help in the selection of drilling tools based on the rock characteristics or in the choice of the best blasting strategy. It could also assist in estimating the grindability of the rock mass to optimize mill operations (Casali, 2001).

Fragmentation plays an important role in controlling the costs of any mining operation (Siddiqui et al., 2009; Monjezi et al., 2009). These costs can be divided into pre-blast costs, i.e., the costs related to drilling, charging, and blasting the rock mass, and post-blasts costs, i.e., the costs related to loading, hauling, and crushing the blasted rock. In general, reduced post-blast costs and increased pre-blast costs are the result of finer fragmentation and vice versa (Mackenzie, 1967, in Morin and Ficarazzo, 2006).

2.2.2 Fragmentation in sublevel caving (SLC) Fragmentation is a very important aspect of sublevel caving (SLC); in fact, it can be considered one of the primary parameters in the success of this mining method (Wimmer, 2010). In SLC, fragmentation affects the gravity flow of the blasted material, the loading at the draw points, the efficiency of load-haul-dump (LHD) operations, the ore-pass efficiency and the energy requirements of crushers (Wimmer, 2010; Danielsson et al., 2017). Wimmer (2010) found finer fragmentation increased the mobility of the material and resulted in more uniform gravity flow from the blasted ring. Meanwhile, coarser fragmentation, especially oversize fragments, hinders the material flow and may cause hang-ups in the ring, resulting in reduced ore recovery if the hang-ups cannot be released; unsafe working conditions are another possible result (Danielsson et al., 2017). The term oversize fragment is defined by Singh and Narendrual (2010) as “any fragment produced from primary blasting, which cannot be adequately handled by the standard loading, hauling and crushing equipment used in an operation”. Figure 2.3 shows an example of an oversize fragment in SLC operation.

12

2.2.3 Factors affecting fragmentation in SLC Fragmentation variation is an inherent characteristic of an SLC operation (Shekhar et al., 2016), depending on the design of the blast ring (Dustan and Power, 2011), the confined nature of blasting (Johansson, 2011), and the specific charge (Hustrulid and Kvapil, 2008). The design of the blast ring is important, as holes are concentrated at the bottom and more widely spread in the upper part of the ring (Dunstan and Power, 2011). The upward drilling of long holes will also increase the probability of borehole deviation in the upper part of the ring (Ghosh et al., 2017b). This design of the blast ring and the probable borehole deviation lead to an uneven burden and specific charge along the boreholes, resulting in an uneven energy distribution to break the rock mass. Consequently, variations in fragmentation are observed at different stages of material loading from the blasted ring (Power, 2004; Brunton et al., 2010; Wimmer et al., 2012). The chargeability of SLC rings is another key factor in fragmentation and boulder generation. Danielsson et al. (2017) found a decrease in chargeability would cause a lower local specific charge and, thus, reduced fragmentation. Ghosh et al. (2017b) showed borehole stability is another important consideration when analysing fragmentation, as it is directly linked to the amount of explosives in the borehole. If the borehole is blocked at some point because of the caving of the borehole wall, it may not be possible to restore or re-drill the borehole, which causes insufficient charging of the explosive to break the rock mass effectively (Lundin, 2020). Borehole instability is

Figure 2.3: An oversize fragment bigger than the LHD bucket

13

also likely to be worsened by high stress states, production blasting, rock bursts or mine seismicity (Zhang, 2016).

2.2.4 Fragmentation assessment The importance and uncertainty of fragmentation make it a critical consideration in attempts to optimize any mining operation. However, the assessment of rock fragmentation has some practical limitations. There are several direct and indirect techniques for fragmentation assessment, including sieving analysis, observational methods, and image-based analysis (Roy et al., 2016; Elahi and Hosseini, 2017; Babaeian et al., 2019), but they have limitations when it comes to continuous monitoring of fragmentation, especially in an underground environment (Campbell and Thurley, 2017). Sieving analysis involves a tedious process of taking rock samples from muck piles, passing them through a stack of progressively smaller mesh screens with square openings, and weighing and classifying the rock fragments into different classes based on which mesh size they do not pass through (Thurley, 2013). Sieving analysis is considered the most accurate method to determine the rock fragmentation. However, it is an impractical tool for routine fragmentation assessment because of slow feedback, high time consumption, high cost, and the interuptive nature of the process; it is also not possible to sieve a full scale blasted muck pile in the mine (Thurley, 2013; Roy et al., 2016; Bamford et al., 2017; Campbell and Thurley, 2017). The observational methods include visually based counting and classification of the rock fragmentation; they are greatly influenced by the observer’s experience and bias and thus may be less accurate (Babaeian et al., 2019). Image-based methods are the most common ones for rock fragmentation analysis (Chatterjee, 2010). Photographing the muckpile and using image analysis methods to determine fragmentation have proven to be competitive tools in size distribution studies (Singh et al., 1991). However, image-based methods have some limitations, especially in underground environments. These limitations include particle delineation error, sub-resolution particle error, segregation and group error, overlapped particle error, capturing error, profile error, weight estimation error, and sampling error (Thurley, 2013). Although manual edge detection for rock fragments using Split-Desktop is a relatively precise technique to remove particle delineation error, it is highly time-consuming (Elahi and Hosseini, 2017). Fragmentation measurements using image-based methods have been criticized for sampling bias (Devgan, 1992). Sampling errors are affected by three parameters: type, scale, and number of images (Hunter et al., 1990).

2.2.5 Fragmentation prediction models Fragmentation is mostly predicted before blasting using empirical formulae generally known as fragmentation models. Fragmentation prediction has many benefits, as it helps to target required rock size distribution after blasting (Morin and Ficarazzo, 2006) and to optimize the mining operation. Ouchterlony and Sanchidrian (2019) presented a comprehensive review of prediction models for rock fragmentation. They defined a fragmentation model as a combination of equations used to describe the shape and position of the particle size distribution after blasting. The fragmentation model also describes the influence of various factors, such as the amount and strength of explosive geometry of the blast and properties of the rock mass, on particle size distribution.

14

According to Ouchterlony and Sanchidrian (2019), the dependence between the various factors and rock fragmentation generally takes a “multiplicative form”, which means the factors contribute significantly to particle size distribution. The parameters specifying the factors related to explosive and blast geometry are mostly well known and controllable in any blast, but those related to rock mass properties are usually not well known and are estimated using empirical formulae developed under certain preconditions. The uncertainty and even, in some cases, unavailability of all the required data cause errors in these fragmentation models. Another limitation of the models is their origin; i.e., they are mostly developed for surface mines (Ouchterlony and Sanchidrian, 2019) and lack application in underground environments. For example, they do not incorporate several essential factors relevant to fragmentation in SLC mining, such as the shape of the blast fan, confined nature of blasting, drilling accuracy, chargeability etc.

2.2.6 Measurement while drilling: A potential fragmentation prediction tool The MWD technique is currently being applied in several sublevel caving mines to monitor the drilling process. Ghosh et al. (2015) showed MWD data can be used to estimate the borehole stability after a hole has been drilled. Ghosh et al. (2018) and Navarro et al., (2019) developed geological models of the rock mass and assessed the chargeability of production boreholes in SLC mines by analysing the MWD parameters. Ghosh et al. (2018) used MWD data to classify the rock mass into different categories, i.e., solid rock mass, fractured rock mass, highly fractured rock mass with potential cave ins, rock mass with minor cavities, and rock mass with major cavities. The research suggests the possibility of exploring the relationship between these rock mass types based on MWD data and fragmentation.

15

3 TEST SITES

Based on the literature review, three types of data were selected to be collected from mines for further research: measurement while drilling data, photogrammetric data containing digital images of the bench faces, and video recordings of the loading operation to capture images of the blasted rock fragments. The required data were collected from three different mines based on the nature of the required data. Details of the mines are described in the following sections.

3.1 Erzberg mine The Erzberg mine is located in Eisenerz, an old mining town of Styria in the central-western part of Austria. It is a large open-pit iron ore mine located about 260 km south-west of the capital, Vienna. Erzberg is believed to have the largest reserves of iron ore in the country, with approximately 260 million tons of iron ore. It is the world’s largest deposit of siderite (FeCO3) mineral, mixed with dolomite (CaMg(CO3)2) and ankerite (CaFe(CO3)2). The major minerals in the deposit are siderite, ankerite, dolomite, and calcite. Due to the presence of different mineral compositions in the mine, the iron concentration ranges from 22-40%, with an average value of 33%.

The Erzberg mine is one of the major mines in central Europe which is considered as the icon of Austrian industrialization. The deposit has been mined for more than 1300 years. The deposit was mined out using surface and underground mining methods until 1986, when the underground operations ended. As per current production and reserve estimation, the deposit will be depleted in the next 30-35 years. The Austrian mining company VA Erzberg is presently conducting the mining activity with a workforce of about 250 persons. The mine has a total of 31 bench levels, of which 21 are currently active. An overview of the mine is shown in Figure 3.1.

A total of about 12 million tons of rock is mined out annually using conventional drill and blast method; this includes 4 million tons of ore and 8 million tons of waste. After blasting, the waste is dumped in a nearby dump area, and ore is hauled to the crushers. Two gyratory crushers installed at the mine site break the boulders and bigger pieces of rock to a maximum size of 15 centimetres; the crushers have a capacity of 1200 tons per hour. The crushed material is processed further to remove the waste from the ore. This processing phase includes sensor-based sorting for the high-quality ore (iron content > 30%) and dense media separation for the low-quality ore. The ore is crushed again to achieve the final product size of 2-10 mm in diameter using 4 crushers and 7 sieves. The final product of 3 million tons of fine ore is transported to Linz and Donawitz annually for processing into high quality steel.

16

3.2 Leveäniemi mine The Leveäniemi iron ore mine is located in Svappavaara, an old mining village in Kiruna Municipality in the northern part of Sweden. It is an open-pit mine, owned and operated by Luossavaara-Kiirunavaara AB (LKAB). Mining started in the village around 1950, and full-scale open-pit mining of the ore was in operation during 1964-1983. After that, the mine was closed and flooded. In 2015, the mine was reopened and all the water was pumped out. Today, the planned annual production is 12 million tonnes of magnetite iron ore with an average grade of 44%. There are approximately 15 years of remaining mine life as per reserve estimates.

Due to the varying nature of ore qualities, different short-term and long-term plans have been made using deposit models to maintain a good balance of iron concentration for the processing purposes. After blasting, the waste is dumped in a nearby dump area, and ore is hauled to the crushers. The crushed material is processed further to remove the waste from the ore. The processing phase has three stages: sorting, concentrating, and pelletizing. Magnetic separation is used in sorting and concentrating phases to separate iron from the waste. The final product in the form of pellets is transported to customers for processing into high quality steel.

One wall of the mine with poor rock conditions and one with good rock conditions and pre-split blast holes (see Figure 3.1) were selected for the study. The wall with poor conditions was selected to check the response of MWD data to highly fractured rock; the wall with pre-split blast holes was selected for the following reasons:

Figure 3.1: Areial view of Erzberg mine with the lower right corner showing the location of mine in Austria (Boch et al., 2019)

17

• It was intact with minimum blast-induced fractures because of pre-split blasting. • Half casts of the pre-split blast holes were identified on the wall, permitting

improved correlation with MWD data.

3.3 Malmberget mine LKAB’s Malmberget iron ore mine is located in Gällivare municipality in northern Sweden. The mine consists of 20 ore bodies which spread over a large underground area of 2.5 by 5 km (Lund, 2013). Of these, 13 are currently being mined (Shekhar, 2020). Large-scale mining operation started in 1888 after the completion of railroad. The annual production in 2019 was 16 million tons of ore. The ore bodies are comprised of Precambrian volcanic rock. Magnetite is the main iron ore extracted from the mine, but there are different amounts of hematite and apatite in the various ore bodies (Lund, 2013). Mining operation is carried out at several levels because there are multiple ore bodies. The main haulage levels are at 600 m, 1000 m, and 1250 m. To achieve high productivity, large-scale sub-level caving (SLC) is the principal method of ore extraction. Blasted rock from the drawpoints is hauled by LHD machines to the ore passes where the ore is transferred from the production to the haulage level. The ore is transported to underground crusher stations by trucks, and after primary crushing, the ore is hoisted to the processing plant at the surface. A general layout of the complete

Figure 3.2: Leveäniemi open pit mine with marked walls used for this study (modified from hitta.se, 2018)

18

operation is shown in Figure 3.3. The only modification at Malmberget mine is at stage four (4) (see Figure 3.3) where trucks are used for chute loading and transport instead of locomotives.

Figure 3.3: Layout of SLC operation at LKAB mines (LKAB, 2020)

Emptying & Crushing

Hoisting

19

4 MATERIALS AND METHODS

The data for pre-blast rock mass assessment including MWD data and photogrammetric data were acquired from the Erzberg and Leveäniemi mines. The data for post-blast fragmentation assessment including MWD data and video recordings of the loading operation were acquired from the Malmberget mine. The details of the acquired data are discussed in the following sections.

4.1 Measurement while drilling data In the Erzberg and Leveäniemi mines, drilling is done with fully mechanized Epiroc (Atlas Copco) smartROC D65 drill rigs equipped with pneumatic Down-The-Hole (DTH) hammers. The Epiroc drill monitoring system retrieves and stores MWD data. The recorded data on the D65 drill rigs include time (YYYY-MM-DD hh:mm:ss), depth (m), penetration rate (m/min), feed pressure (bar), rotation pressure (bar), and percussive pressure (air pressure measured by bar). The measured and recorded data are stored in two .xml files using IREDES format. The drill quality “DQ” files (one for each drill plan) contain hole and drill plan identification etc. The MWD files (one for each hole) contain all MWD parameters for each hole. The sampling interval along the borehole can vary from 2 cm and up and was in this case set to 5 cm. MWD data are stored on each drill rig on a USB memory stick. At the Erzberg mine, MWD data were collected from two different bench faces for a total of 14 boreholes. The length of the boreholes was approximately 30 metres. At the Leveäniemi mine, MWD data were collected from nine bench faces for a total of 473 blast holes. The length of the boreholes was approximately 30 m for the pre-split blast holes and 15 m for the production blast holes.

The fragmentation study was carried out in two orebodies in the Malmberget mine. In Malmberget mine, drilling is done using Epiroc’s Simba WL6C rigs equipped with Wassara hydraulic ITH hammers. The drilling is done in a fan shaped pattern. A schematic layout of fan shaped drilling pattern is shown in Figure 4.1. 16 rings were selected from both orebodies due to the limitation of video recordings. for this study. The shortest holes are less than 20 m long and the longest are close to 50 m. Most of the rings consist of eight boreholes, but a few close to the foot wall side, have nine boreholes. This resulted in a total number of 133 boreholes. The diameter of a borehole was 115 mm with a burden (the distance between two subsequent rings) of 3.5 m. The MWD data collected from the Simba drill rigs include time (YYYY-MM-DD hh:mm:ss), depth (m), penetration rate (m/min), feed pressure (bar), rotation pressure (bar), and percussive pressure

Figure 4.1: Schematic layout of fan shaped drilling pattern in SLC

Boreholes

Drifts

20

(water pressure measured by bar). For this study, the sampling interval was set to 3 cm.

4.2 Photogrammetric data Photogrammetric data include stereoscopic images of the two bench faces from Erzberg mine and nine bench faces from Leveäniemi mine. The method adopted to acquire these images was the same for both mines. The images were taken using a DSLR camera following the standard procedures prescribed by 3GSM (2010). The camera used for this study was the Nikon D70s which could cover an imaging area of approximately 40 m height and 50 m width. Figure 4.2 shows the principal geometric arrangement for taking pictures of a bench face.

The distance from the imaging position to the bench face, D, should be long enough to cover the area to be surveyed. The camera has two lenses: a Sigma lens and a Nikon lens. The Nikon lens is standard; it comes with the camera and has a focal length ranging from 18-70 mm. The Sigma lens is used to capture images in more confined or restricted spaces and has a focal length ranging from 10-20 mm. The Sigma lens has a shorter focal length than the Nikon lens and a wider angle of view but lower magnification. If the camera position is close to the bench face and cannot go farther away because of restrictions like bench width and thus cannot cover the whole area of interest with the standard Nikon lens, the Sigma lens should be used. However, the Nikon lens should be used if the distance, D, does not restrict the choice of lens, as this lens has less distortion than the Sigma lens and allows more light to enter the lens, making it more useful in low light conditions. The mining environment often restricts the distances and has more confined spaces, as in this study. Therefore, the Sigma lens was used to capture the whole area with a shorter imaging distance, D.

The principle of taking a stereoscopic image pair is shown in Figure 4.3. The stereoscopic images collected from the mine sites were used to generate 3D models of the bench faces. A 3D model is developed by combining the spatial information with the information in 2D images. To generate a 3D model using ShapeMetrix3D®, the standard procedure is to take two pictures instead of multiple pictures of the area under

Figure 4.2: Basic geometric arrangement for data acquisition of a bench face (modified from 3GSM, 2010)

21

observation as data acquisition is easier and faster. However, several images were taken parallel to the bench face so that the best image pair giving maximum accuracy in the 3D model generation could be used for analysis purposes.

Collecting spatial information and incorporating it in the model is another crucial aspect of 3D model generation. This is done by marking and surveying the control points in the model. It should be done carefully to ensure the metric accuracy of the 3D model (3GSM, 2010). At least three control points are required to reference the model in a global coordinate system (geo-referencing). However, up to six points increase the reliability of the measurements (3GSM, 2010). Therefore, six control points, in the form of target disks (red coloured), were used in each model for referencing purposes. Figure 4.4 shows one of the benches of the Erzberg mine with range poles used for referencing purposes. The coordinates of these control points were determined using a GPS system. The distance between two control points was fixed and measured and then used to scale the model.

Figure 4.3: Stereoscopic image pair. Two corresponding image points P(u,v) relate to one three-dimensional object point P(X,Y,Z) (3GSM, 2010)

22

4.3 Video recordings Video recordings of the loading operation were collected from two orebodies of the Malmberget mine. To monitor fragmentation, cameras were installed in the selected drifts to record each LHD bucket filled with blasted material. Cameras were configured to record a short video whenever there was movement in the recording area. To ensure better quality videos, cameras were assisted with good lighting. However, the dust in the underground environment sometimes reduced the quality of recording, and the headlights of the LHD machines sometimes distorted the frame pixels. The collected data cover approximately two years.

4.4 Data processing

4.4.1 Measurement while drilling data In the first step of the MWD data processing, the recorded drilling parameters were extracted in a tabular form. After the most significant features of the samples were imported into the table, the input data were filtered to remove incorrect or faulty values. Noise and outliers can appear at different steps of the data collection process. In this study, the recorded samples causing noise in the data were divided into three categories:

• Samples recorded in the beginning of each hole before achieving a steady drilling process

• Samples recorded during the addition of a new rod to the drill string • Samples with negative or abnormally high values

All these noisy data samples were removed from the input data.

Figure 4.4: Bench face with range poles at Erzberg's open pit mine in Austria

23

Figure 4.5 shows the measured parameters for one hole before filtering. The figure indicates that both feed pressure and percussive pressure are reduced during collaring, which occurs during the drilling of the first rod (6 m). The reduction of feed pressure results in a reduction of the rotation pressure during the same interval. The figure also shows that the penetration rate reaches a value of 150 m/min at the beginning of the hole, but this is impossible in real drilling. Because of this high value, the rest of the curve disappears, as the mean value is less than 1 m/min. Finally, all pressures drop after a regular interval of 6 m, when a new rod is added to the drill string. Information on the location of the holes is required to map the MWD datasets to the structural data obtained from the 3D models.

Epiroc drill monitoring machines record drill quality (DQ) files in addition to the MWD files. These files contain the specifications for each borehole, i.e. borehole identification, plan identification, borehole name, type of borehole, start point coordinates of borehole, end point coordinates of borehole etc. The start and end point coordinates of the borehole represent the x, y, and z for the first and last recorded samples in each borehole. Borehole identification and plan identification are two common features of the DQ files and MWD files. The files can be combined based on their common features for each sample in the MWD file, and the start and end point coordinates of the hole in which the sample is placed can be precisely determined. The approximate coordinates of each sample can be obtained by adding the depth value to the z coordinate of the related borehole start point or by subtracting the depth value from the z coordinate of the related borehole end point. Another feature recorded for each borehole in the DQ files is status, representing the status of the specified borehole with respect to the following values: undrilled, fail, and success. Since the recorded data from undrilled or fail holes are not reliable, the samples related to these holes were removed from the main dataset. Figure 4.6 shows the data filtering, the start and end point coordinates, and the status of the borehole.

Figure 4.5: Noisy MWD data parameters plotted against depth of the hole

24

Two different methods were applied to filter out the noisy MWD data for the fragmentation study of the Malmberget data. ‘Method A’ was adopted from Ghosh et al. (2018) and ‘Method B’ from Liaghat et al. (2021). The rock mass surrounding the boreholes in Malmberget mine was classified into five classes: solid, fractured, highly fractured with potential cave-ins, cavities, and major cavities using an average MWD value for the whole blast ring as described by Ghosh et al. (2018).

4.4.2 Photogrammetric data The photogrammetric data were processed and analysed using ShapeMetriX 3D, a commercial software package developed by 3GSM and used to extract structural data from stereoscopic images (Gaich et al., 2006; Haneberg, 2008; Vasuki et al., 2014; Beyglou, 2016; Buyer and Schubert, 2017). 3D models were constructed from stereoscopic images using one of the software modules. Once a 3D model is constructed, the software displays the result along with a quality indicator. In this study, only the models with best quality were used for further analysis. Another quality check for the 3D model is the base length ratio indicated in the results. The base length ratio is the ratio of the base length, B, and the mean distance, D, as shown in Figure 4.2. It should range between 12.5 and 20 for accurate measurements. The models complying with this condition were chosen for further analysis.

The software provides an option to check the quality and accuracy of the model referencing as well. Statistics include the standard deviation of easting, northing, and elevation, as well as total standard deviation of coordinates. Range poles with a fixed length were used during data acquisition for scaling purposes. A special software tool, the SMX Normalizer, was used to scale the 3D model using the range pole length; this is very important for the accuracy of the model. Figure 4.7 shows a 3D model of one of the benches of Leveäniemi mine with pre-split blast holes.

Figure 4.6: MWD data parameters vs. depth of the hole with specified location coordinates and status

25

Figure 4.7: 3D model of a bench face of Leveäniemi mine with joints, exposed surfaces

and scanlines (white dotted lines) marked on the surface

The 3D models were analysed using the Analyst module of the software package. Scanlines were drawn on the bench faces to determine the depth of structures marked on the bench faces. The scanlines were drawn exactly on the location of the boreholes to achieve higher accuracy when comparing photogrammetric data with MWD data. For example, if a borehole had the coordinates with easting, 182617.9, northing, 7504762.9, and elevation, 281.1 m, the scanline was added using the same easting, northing, and elevation values in the 3D model. An example of scanline data is shown in Figure 4.8.

Figure 4.8: Scanline data showing coordinates of intersecting structures

26

4.4.3 Video recordings In the first step of processing recorded videos, maximum possible frames were extracted from each video. The “Frame Rate” is a parameter representing the number of extracted frames from a video per second; here, it was a self-adapting parameter, set automatically according to the processing ability of the system running the code. In this work, the frame rate was about 30 frames per second for each input video. Further experiments showed the RGB photos did not provide more information about rock fragmentation, so the frames in RGB format were converted to grayscale. This helped to reduce data size and increase processing speed without any considerable effect on the results.

The cameras used for the study were sensitive to motion and thus recorded any sort of movement in the filming area. Movements were caused by LHDs, personnel cars, drifters, water sprinklers, scrappers etc. or by the oscillating ventilation duct during blasting activities. All videos recording movements other than LHDs were discarded and not considered for further analysis. Similarly, the videos of moving LHDs had two types of recordings: recordings when the LHD machine was going into the drift with an empty bucket to load the material and recordings when it was coming out of the drift with a bucket filled with blasted material. The recordings with empty LHD buckets were discarded. This left only those videos containing LHD machines with buckets filled with material to be further processed.

A MATLAB code was used to extract the relevant images (frames) from the videos. Depending on the length of the videos, 500-1200 frames were extracted from each video, yielding too many images to be handled. Most of the frames did not contain images of buckets with the blasted material. Instead, they showed different machine parts, or they were empty; i.e., the machine was entering or leaving the filming area. Few of the extracted frames from each video (~5-10) contained the LHD bucket filled with material. Some frames showed partial buckets, as the videos were recorded during LHD motion. Only frames with full LHD buckets were forwarded to fragmentation analysis; all the rest were disregarded. Figure 4.9 shows an example of frames with relevant and irrelevant data. Frames labelled A, B, C, D, E, F, and G are irrelevant to the analysis; frames labelled H and I are relevant.

27

4.4.3.1 Quick rating system The raw video data were processed, and 5908 images of LHD buckets filled with blasted rock were selected for fragmentation assessment. These types of images can be analysed using commercially available software, e.g., Split-Desktop® (Kemeny, 1994), WipFrag™ (Maerz et al., 1996), GoldSize (Kleine and Cameron, 1996), FragScan (Schleifer and Tessier, 1996), TUCIPS (Havermann and Vogt, 1996), CIAS® (Downs and Kettunen, 1996), PowerSieve (Chung and Noy, 1996), IPACS (Dahlhielm, 1996), Fragalyst (Raina et al., 2002) etc. Image analysis using commercial software is normally time consuming; for example, analysing a single image using Split-Desktop® can take two to three hours depending on the quality of the image (Petropoulos, 2015). Therefore, it was considered impractical, or even impossible, to analyse this volume of images using commercial software. Instead, the study used a quick rating system (QRS) similar to that used by Petropoulos (2015), Wimmer et al. (2015), and Danielsson et al. (2017) to assess median fragment size (X50) of the material inside the LHD buckets. In the QRS method, the actual images of fragmented rock are compared to reference images to assess fragmentation (Wimmer et al., 2015). It is similar to the Compaphoto method (Cunningham, 1996), but is customised to evaluate fragmentation in LHD buckets. QRS allows a faster estimation of fragmentation based on reference images (Wimmer et al., 2015). Reference images were prepared using Split-Desktop®. They were categorised into four classes based on X50. Class 1 referred to fine fragmentation,

Figure 4.9: Frames with irrelevant (A, B, C, D, E, F, and G) and relevant data (H, I)

28

X50 < 50mm; class 2 was medium fragmentation, X50 = 50-400 mm; class 3 was coarse fragmentation, X50 = 400-1000 mm; class 4 was oversize fragmentation, X50 > 1000 mm. Malmberget mine defines oversize fragments as rock blocks bigger than 1x1x1 m (Gustafson et al., 2016; Danielsson et al., 2017). All images were analysed manually using the QRS approach in which they were compared to reference images and categorised as one of the four classes. An example of the classes is shown in Figure 4.10.

4.5 Statistical analysis The study used correlation analysis to analyse the relationship between the type of rock mass determined from MWD data and the type of fragmentation calculated from QRS. Correlation analysis is a statistical approach used to assess the association between two sets of variables, e.g., independent and response variables, where the correlation coefficient “R” quantifies the strength of the relationship (Franzese and Luliano, 2019). Values range from −1 to +1; in this rating system, -1 shows a perfect negative correlation, +1 shows a perfect positive correlation between the pair of variables, and 0 indicates a complete independence and absence of any correlation (Franzese and Luliano, 2019). The intermediate values from 0 to ±1 represent a partial correlation of variables; the correlation can be significant or weak (Franzese and Luliano, 2019). However, significant correlations can have values as small as ±0.4 in practical applications (Kleinbaum et al., 1998).

The study used multivariate correlation analysis, as it can handle multiple variables and consider all the associations between variables simultaneously (GeiB and Einax, 1996). Some of the most commonly used multivariate methods are Principal Component Analysis (PCA), Discriminant Analysis (DA), Multiple Regression Analysis, Factor Analysis, Logistic Regression, Partial Least Squares (PLS) Regression, Cluster Analysis, Log-Linear models, Multivariate Analysis of Variance etc. The study used Partial Least Squares (PLS) because of its minimal demands on measurement scales and sample size, as well as its ability to suggest where relationships might or might not exist. Partial Least Squares (PLS) helps to build models predicting more than one dependent variable (Lorber et al., 1987). PLS regression was introduced by Wold in 1966; it generalises and combines features from Principal Component Analysis and Multiple Regression (Colombani et al., 2012). PLS does not assume the predictors are fixed, unlike multiple regression; this means the predictors can be measured with error, making PLS more robust to measurement uncertainty (Cramer, 1993). A detailed description of the strengths and weaknesses of the PLS method is given by Cramer (1993).

Figure 4.10: Frames showing four different classes of fragmentation

29

5 RESULTS AND DISCUSSION

5.1 Pre-blast rock mass assessment based on MWD To define the MWD response to various geomechanical features, the coordinates of the scanlines and boreholes were matched to the images of the walls. Structural data and MWD data were then plotted against the depth of the borehole to identify the connections between the data and the geological and geomechanical features. Finally, areas showing similarities were identified and marked manually. Some examples of the results from the test walls are presented below.

As the literature suggests (see Section 2.1.3), two of the drilling parameters, percussive pressure, and feed pressure, normally don’t respond to the rock behaviour unless there is a void or open joint where the drill bit loses its contact with the rock mass. The other two parameters, penetration rate and rotation pressure, are more sensitive to the changes in the rock mass. Higher penetration rate indicates a weak or soft rock mass which is easier to penetrate and vice versa for hard rock mass. An increased rotation pressure indicates a fractured or broken rock mass which tries to resist the rotation of the drill bit. The rotation pressure is then increased to maintain the pre-set rotation speed. These assumed responses are used to mark the variations in MWD parameters as shown in Figure 5.1.

Figure 5.1: MWD data variations related to different rock types

30

The MWD parameters in Figure 5.1 show four distinct responses.

1. A zone of higher penetration rate than the mean value indicates the presence of softer or weaker rock mass (rock mass A) at the start of the borehole.

2. A sharp increase in penetration rate and drop in all the other parameters indicates an open joint around 9 m depth.

3. A zone of lower penetration rate than the mean value (after the joint) indicates a harder rock mass (rock mass B).

4. A zone of increased penetration rate, as well as an increased rotation pressure, indicates a weaker and fractured or broken rock mass (rock mass C) in the lower half of the borehole

Information extracted from the 3D image of the bench face (Figure 5.2) validates the findings of the MWD data in Figure 5.1. In the upper right part of the bench face, the rock is foliated/fractured (marked in yellow in Figure 5.1). This rock mass is separated from the underlying rock by a bedding plane (marked in green). This plane causes the sudden increase in the penetration rate and a drop in the other three drilling parameters shown in the figure. Under the bedding plane, the rock is intact and competent, without any fractures or joints. The decreased penetration rate observed in this region (Figure 5.1) indicates this rock mass is harder than the rock mass above the bedding plane. The lower part of the bench face has a highly fractured and broken rock mass (marked in red). This section has a higher penetration rate and higher rotation pressure (Figure 5.1). The last part of the MWD data response cannot be identified in the image since sub-drilling and loose material at the toe of the bench covers the bottom part of the hole.

Figure 5.2: Bench face showing different types of rock mass

31

Figures 5.3 and 5.4 show the response of MWD data when a hole is drilled through a rock with fractures or joints with a very small aperture. In these conditions, it is primarily penetration rate (PR) and rotation pressure (RP) that respond to fractures or joints.

These parameters show a slight increase in their values when the drilling process encounters discontinuities. As the rock is of competent nature and the aperture of these fractures/joints is not very prominent, the increase in drilling parameters is also not very prominent. Statistically, the mean value of penetration rate is 0.85 m/min; it increases to approximately 2 m/min, when it intersects the discontinuities, as shown in the figures. Similarly, the mean value of rotation pressure is 60.78 bar, and it increases up to 70 bar when drilling through the discontinuities.

Figure 5.3: Rise in penetration rate (PR) and rotation pressure (PR) at fracture/joint location

32

Figure 5.5 shows another type of MWD data response, indicated by sections or regions of noisy data with fluctuations for all drilling parameters. The figure indicates that this type of behaviour occurs when there are cavities with removable material, or a wedge is no longer available on the rock face after the blast. The fluctuations in the MWD data in this case last longer than the fluctuations for fractures or joints. Sticky, highly fractured, or loose material likely causes this extensive fluctuation of the MWD data.

Figure 5.4: Variations in penetration rate (PR) and rotation pressure (RP) along with corresponding structures

33

The third type of behaviour shown by the MWD data occurs when drilling is carried out in foliated rock or in a rock mass with closely spaced bedding planes. An increase in penetration rate can be seen in this type of rock mass. Figure 5.6 presents MWD data for a highly fractured rock face, with the corresponding image. As the figure shows, the foliations do not cause any abrupt fluctuations in the MWD data. There are four spikes in the penetration rate because of four fractures/joints (marked as black lines) in the rock face image, but the regions between them show almost constant behaviour for the foliations or bedding planes. Statistically, the mean value of the penetration rate has increased, i.e. 0.95 m/min, showing the low strength or fractured nature of the rock. The end part of the hole is sub-drilling, so it cannot be seen in the images. The rock was blasted with a normal production blasting method, so the actual borehole location was removed due to back break.

Figure 5.5: MWD data from two holes showing fluctuations in all the drilling parameters

34

5.2 Fragmentation prediction based on MWD data This section describes the results of the correlation tests and partial least squares regression for fragmentation classes and rock mass types based on the MWD data collected from a sublevel caving mine.

5.2.1 Correlation test results Table 5-1 compares the correlation coefficients for MWD filtering methods A and B (see Section 4.4.1) to determine any quality differences between the methods. As the table shows, the correlation coefficients calculated for filtering method B are relatively higher (bold typeface) than for method A. Therefore, the MWD dataset generated by filtering method B was used for the rest of the calculations.

Table 5-1: Comparison of correlation coefficients for two filtering methods, A and B

Variables Fine Medium Coarse Oversize

Method A

Method B

Method A

Method B

Method A

Method B

Method A

Method B

Solid 0.546 0.634 -0.625 -0.710 -0.291 -0.360 0.279 0.264 Slightly fractured -0.352 -0.546 0.494 0.657 0.062 0.250 -0.470 -0.366

Highly fractured -0.631 -0.658 0.689 0.728 0.387 0.406 -0.225 -0.294

Cavities -0.612 -0.641 0.632 0.690 0.417 0.388 -0.094 -0.151 Major cavities -0.525 -0.585 0.488 0.590 0.432 0.419 0.087 -0.043

Figure 5.7 shows the scatter plots and linear regression plots with the percentages of different fragmentation categories from the image analysis (fine, medium, coarse,

Figure 5.6: MWD data plot of a borehole and rock face image for poor rock conditions

35

oversized) on the Y-axis and percentages of different rock types from the MWD analysis (solid, slightly fractured, highly fractured, cavities, major cavities) on the X-axis. In general, the figure shows less scatter for fine and medium fragmentation than for coarse and oversize fragmentation. That explains why the correlation coefficients listed in Table 5-1 have higher values for the fine and medium fragmentation categories. The amount of fine material increases when the percentage of solid rock mass increases, and it decreases for all other rock types. Similarly, medium sized rock fragments decrease when the amount of solid rock mass increases and increase for all other rock types. The relationship for coarser material is similar, but the scatter of data does not support a reasonable relationship. Nor does the scatter of data support a reasonable relationship for the oversize fragmentation category.

Figure 5.7: Scatter plots and linear regression lines corresponding to different fragmentation and rock mass classes

36

Table 5-2 lists the Pearson correlation coefficients (R), p-value, and coefficients of determination (R2) for the various relationships. The null hypothesis in this study states there is no relationship between the variables under observation, while the alternative hypothesis states there is a relationship between the variables under observation. A significance level of 0.05 is used to reject the null hypothesis; i.e., a p-value less than 0.05 (typically ≤ 0.05) is statistically significant. There is strong evidence against the null hypothesis, as there is a probability lower than 5% that the null hypothesis is correct (and the results are random). R2 shows the amount of variability captured by the model and indicates goodness of model fit. Values of R, p-value, and R2 for fine and medium categories (bold typeface in Table 2) indicate a better correlation than for coarse and oversize categories. The Spearman correlation method suggests a monotonic relationship; i.e., the relationship between the variables under observation is not strictly linear. However, the coefficients calculated by either the Pearson method or the Spearman method suggest a similar relationship; i.e., fine, and medium categories have better correlation than coarse and oversize fragmentation categories.

Table 5-2: Summary of correlation test using Pearson correlation method

Variables Fine Medium Coarse Oversize

R p-value R2 R p-

value R2 R p-value R2 R p-

value R2

Solid 0.634 0.008 0.402 -0.71 0.002 0.504 -0.36 0.171 0.130 0.264 0.323 0.070 Slightly fractured -0.55 0.029 0.298 0.657 0.006 0.431 0.250 0.351 0.062 -0.37 0.163 0.134

Highly fractured -0.66 0.006 0.432 0.728 0.001 0.530 0.406 0.119 0.165 -0.29 0.269 0.086

Cavities -0.64 0.007 0.411 0.690 0.003 0.477 0.388 0.137 0.151 -0.15 0.576 0.023 Major cavities -0.59 0.017 0.342 0.590 0.016 0.348 0.419 0.106 0.176 -0.04 0.876 0.002

5.2.2 Partial least square regression (PLS-R) results PLS regression is applied to model the relationship between fragmentation categories and rock mass types. Figure 5.8 illustrates the quality indices of the PLS regression model for the first two model components. The PLS components are linear combinations of the independent variables that maximize their covariance with dependent variables. The figure shows that the Q2 cumulative index which represents the global goodness of fit and predictive quality of the model has the highest value for the first component and further components do not improve the quality of the model. This indicates that the first component captures maximum information and provides the best model fit. Q2 cum has a relatively low value, suggesting the quality of the fit varies quite a bit, depending on the dependent variables. The R²Y cum and R²X cum measures the cumulative fraction of the variation of the Y (fragmentation) and X (rock type) variables,

Figure 5.8: PLS-R model quality indicators

37

respectively, explained after the selected component. In this case, they show increasing values for increasing components.

Table 5-3 lists the cumulative Q2 quality index for all four dependent variables, i.e., fine, medium, coarse, and oversize fragmentation. It also shows the components after the first component deplete the quality of the model i.e. the PLS model with the first component is a better fit than higher components. The predictive models for fine and medium size rock fragmentation have a reasonable goodness of fit for the first component (bold typeface in Table 5-3), but the models for coarse and oversize fragmentation are very poor fitting. Similar behaviour was observed in correlation tests.

Table 5-3: Cumulative Q2 quality indices for 1st two components

Component Comp 1 Comp 2 Fine 0.346 0.263 Medium 0.447 0.417 Coarse 0.078 -0.033 Oversize -0.059 -0.085 Total 0.203 0.140

Table 5-4 lists the model parameters for all dependent variables and R2 values for the respective models. The R2 values for the models of fine and medium fragmentation (bold typeface) suggest a better fit than the R2 values for the models of coarse and oversize fragmentation, as the latter are very low.

Table 5-4: Model parameters and R2 values for fragmentation prediction models

Variables Fine Medium Coarse Oversize Intercept 23.956 61.226 10.284 4.380 C1 - Solid 0.292 -0.255 -0.047 0.012 C2 – Slightly fractured -0.531 0.464 0.086 -0.023 C3 – Highly fractured -2.190 1.913 0.353 -0.093 C4 - Cavities -1.335 1.166 0.215 -0.057 C5 - Major Cavities -1.095 0.956 0.176 -0.046 R2 0.420 0.510 0.148 0.057

The predictive models for fine and medium fragmentation are given in equation 1 and equation 2.

Fine(%) = 23.956 + 0.292(C1) − 0.531(C2) − 2.190(C3) − 1.335(C4) − 1.095(C5) Eq 1.

Medium(%) = 61.226 − 0.255(C1) + 0.464(C2) + 1.913(C3) + 1.166(C4) + 0.956(C5) Eq 2.

The PLS model parameters support the correlation test results, as seen in Table 5-4. Fine fragmentation has a positive relationship with solid rock mass and a negative relationship with other rock types. Similarly, medium fragmentation has a negative relationship with solid rock mass and a positive relationship with the rest of the rock types. The indices in Table 5-3 and Table 5-4, i.e., cumulative Q2 and R2 values, suggest fine and medium rock fragmentation can be predicted better than coarse and oversize

38

fragmentation. Figure 5.9 shows the comparison model of actual versus predicted fragmentation values for fine and medium categories based on statistical analysis of eq. 1 and eq. 2.

A possible explanation of the poor correlation of coarse and oversize fragmentation categories is their origin. Earlier studies have shown a relative coarsening of the fragmentation with increasing material extraction, up to 80%, from the blasted ring, after which it becomes unpredictable (Power, 2004; Brunton et al., 2010; Wimmer et al., 2012). This indicates the probability of most of the coarse or oversize fragmentation coming from the upper part of the ring where there are higher chances of drilling problems, chargeability issues, low specific charge, and increased ore dilution. Ore dilution is the mixing of waste caved rock into the ore; dilution increases with increased material extraction from the blasted ring, and it has been argued that diluted material from the caved hanging wall is coarser than the blasted ore. MWD data do not provide any information about the waste caved rock mass outside the ring geometry. Thus, it becomes difficult to see the exact relationship with these two fragmentation classes. Charging of boreholes is an important consideration as occasionally lesser boreholes are charged than drilled. The information about boreholes charging was unavailable in this study, which could have possibly improved the prediction results otherwise.

Figure 5.9: Comparison of actual and predicted fragmentation (%) based on eq. 1 and eq. 2.

39

6 CONCLUSIONS, CONTRIBUTION AND FUTURE RESEARCH

6.1 Conclusions The research questions stated in Chapter 1 are answered and concluded in this section.

RQ 1: What are the constraints of the current rock evaluation techniques?

The current pre-blast rock discontinuity evaluation techniques hinder the production cycle, over-estimate rock fracturing in weathered or poorly blasted rock, get affected by user experience and are limited to exposed surface of the rock mass. They provide an uncertain information about the geology where holes are drilled and blasted. The uncertainty or unavailability of complete data for accurate rock mass description cause undesired blast outcomes and errors in fragmentation prediction models. The existing fragmentation models are mostly developed for surface mines under certain preconditions and are not suitable for predicting fragmentation in sublevel caving mines.

RQ 2: How can rock mass structures be correlated with MWD data using established methods of rock mass characterization?

To correlate MWD data with rock mass structures on the bench face, scanlines should be drawn exactly along the drillholes location. Traditional scanline survey can be used for this purpose, but it was impractical because of time, access and safety constraints in this study. Therefore, terrestrial digital photogrammetry was used to locate drillholes on the bench face and to identify rock structures. The variations in drilling parameters were also marked in MWD data to identify different rock structures. The results obtained from MWD data showed a close agreement with the results of photogrammetry and suggested the possibility of using MWD for enhancing the information about the geological and geomechanical setup of the rock mass.

RQ 3: How does fragmentation correlate with MWD data?

Correlation tests showed a positive correlation between fine fragmentation and solid rock mass i.e. an increase in percentage of solid rock mass caused an increase in percentage of fine fragmentation. In contrast, fine fragmentation had a negative correlation with other rock types i.e. an increase in percentage of other rock mass categories (fractured, highly fractured, cavities and major cavities) caused a decrease in fine fragmentation. Medium fragmentation had a negative correlation with solid rock mass and a positive correlation with other rock types. However, coarse, and oversize fragmentation did not show any particular relationship with different rock types as observed for fine and medium fragmentation. Cumulative Q2 quality indices and R2 values from PLS regression modelling also showed better models for fine and medium fragmentation as compared to coarse and oversize fragmentation categories.

6.2 Research contribution The research identifies different constraints of current techniques for geometrical assessment of rock mass and fragmentation prediction in sublevel caving. The research contributes towards

40

• Improving rock mass description before blasting using drill monitoring technique • Improving fragmentation prediction in sublevel caving by exploring the

relationship between drill monitoring and fragmentation

6.3 Future research • Improving MWD system:

Data processing showed different samples in MWD data that represented false information like samples at the start of the drilling, samples with unrealistic values for different drilling parameters etc. Research can be carried out to improve existing drill monitoring systems and remove noises in MWD data.

• Improving blast design formulae: The current blast design formulae or available software do not incorporate the information from MWD data. Also, these formulae generally provide a single solution to the whole blast. With a more general and thorough understanding of MWD data, the blast design formulae can be improved by improving the rock mass description. Also, the blast design can be made more flexible and adaptable to each blasthole depending upon the MWD data of each blasthole.

• Improving fragmentation prediction in SLC: This research covers 16 blast rings from the sublevel caving mine. The amount of data can be increased in future to increase datapoints for statistical analysis to have more dependable results. The research in this thesis only considered the nature of rock mass to predict fragmentation. However, other parameters like blasthole stability, chargeability and explosives information can be incorporated to have better prediction in future.

• Automating rock fragmentation analysis: A primary limitation in extending the study to a greater number of blast rings was the time required for fragmentation analysis. The current available software for automatic image analysis are too much expensive. A cost as well as time effective method for automatic image analysis method will help to analyse larger datasets in future.

• Predicting oversize fragments using MWD data: This research analysed the relationship of oversize fragmentation category with median fragment size (X50) greater than 1000 mm. This category can have other fragmentation types as well like fine, medium, or coarse fragments. However, an analysis can be performed to see the predicting ability of MWD data considering only the oversize fragments which are the least desired outcome of the production blast.

41

7 REFERENCES

3GSM, 2010. ShapeMetriX 3D: User Manual. 3G Software and Measurement GmbH., Graz.

Akbari , M., Lashkaripour, G., Yarahamdi Bafghi, A. and Ghafoori, M., 2015. Blastability evaluation for rock mass fragmentation in Iran central iron ore mines. International Journal of Mining Science and Technology, 25(1), pp. 59-66.

Assali, P., Grussenmeyer, P., Villemin, T., Pollet, N. and Viguier, F., 2014. Surveying and modeling of rock discontinuities by terrestrial laser scanning and photogrammetry: Semi-automatic approaches for linear outcrop inspection. Journal of Structural Geology, 66, pp. 102-114.

Azimi, Y., Osanloo, M., Aakbarpour-Shirazi, M. and Aghajani, B. A., 2010. Prediction of the blastability designation of rock masses using fuzzy sets. International Journal of Rock Mechanics and Mining Sciences, 47(7), pp. 1126-1140.

Babaeian, M., Ataei, M., Sereshki, F., Sotoudeh, F. and Mohammadi, S., 2019. A new framework for evaluation of rock fragmentation in open pit mines. Journal of Rock Mechanics and Geotechnical Engineering, 11, pp. 325-336.

Badroddin, M., Khoshrou, H. and Siamaki, A., 2013. Prediction of fragment size distribution from blasting: artificial neural networks approch. In: Proceedings of the 36th International Symposium on Application of Computers and Operations Research in the Mineral Industry (APCOM). 4-8 November 2013, Porto Alegre, Brazil.

Bamford, T., Esmaeili, K. and Schoellig, A. P., 2017. A real-time analysis of post-blast rock fragmentation using UAV Technology. International Journal of Mining, Reclamation and Environment, 31(6), pp. 439-456.

Bemis, S. P., Micklethwaite, S., Turner, D., James, M. R., Akciz, S., Thiele, S. T. and Bangash, H. A., 2014. Ground-based and UAV-based photogrammetry: A multi-scale, high resolution mapping tool for structural geology and paleoseismology. Journal of Structural Geology, 69, pp. 163-178.

Beyglou, A., 2016. On the operational efficiency in open pit mines. Licentiate Thesis. Luleå University of Technology, Luleå, Sweden.

Bhandari, S., 1997. Engineering rock blasting operations. A. A. Balkema, Rotterdam, Brookfield.

Boch, R., Wang, X., Kluge, T., Leis, A., Lin, K., Pluch, H., Mittemayr, F., Baldermann, A., Böttcher M. E. and Dietzel, M., 2019. Aragonite–calcite veins of the ‘Erzberg’ iron ore deposit (Austria): Environmental implications from young fractures. Sedimentology, 66(2019), pp. 604-635.

Bonilla-Sierra, V., Scholte`s, L., Donze´, F. V. and Elmouttie, M. K., 2015. Rock slope stability analysis using photogrammetric data and DFN–DEM modelling. Acta Geotechnica, 10, pp. 497-511.

Brunton, I. D., Fraser, S. J., Hodgkinson, J. H. and Stewart, P. C., 2010. Parameters influencing full scale sublevel caving material recovery at the Ridgeway gold

42

mine. International Journal of Rock Mechanics and Mining Sciences, 47, pp. 647-656.

Buyer, A. and Schubert, W., 2017. Calculation the spacing of discontinuities from 3D point clouds. Proceedia Engineering, 191, pp. 270-278.

Campbell, A. D. and Thurley, M. J., 2017. Application of laser scanning to measure fragmentation in underground mines. Mining Technology. Transactions of the Institutions of Mining and Metallurgy: Section A, 126(4), pp. 240-247.

Chen, J., Zhu, H. and Li, X., 2016. Automatic extraction of discontinuity orientation from rock mass surface 3D point cloud. Computers and Geosciences, 95, pp. 18-31. http://dx.doi.org/10.1016/j.cageo.2016.06.015.

Cho, S. H. and Kaneko, K., 2004. Rock fragmentation control in blasting. Materials Transactions, 45(5), pp. 1722-1730.

Chung, S. H. and Noy, M. J., 1996. Experience in fragmentation control. In: Proceedings of the Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, Quebec, Canada, 23-24 August 1996. Franklin, J.A. and Katsabanis, P.D. (eds.). Balkema, Rotterdam. pp. 247-252.

Colombani, C., Croiseau, P., Fritz, S., Guillaume, F., Legarra, A., Ducrocq, V. and Robert-Granié, C., 2012. A comparison of partial least squares (PLS) and sparse PLS regressions in genomic selection in French dairy cattle. Journal of Dairy Science, 95(4), pp. 2120-2131.

Colomina, I. and Molina, P., 2014. Unmanned aerial systems for photogrammetry and remote sensing: A review. ISPRS Journal of Photogrammetry and Remote Sensing, 92(2014), pp. 79-97.

Cramer, R. D., 1993. Partial Least Squares (PLS): Its strengths and limitations. Perspectives in Drug Discovery and Design, 1, pp. 269-278.

Cunningham, C., 1996. Lessons from the Compaphoto technique of fragmentation measurement. In: Proceedings of the Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, Quebec, Canada, 23-24 August 1996. Franklin, J.A. and Katsabanis, P.D. (eds.). Balkema, Rotterdam. pp. 53-58.

Dahlhielm, S., 1996. Industrial applications of image analysis - The IPACS System. In: Proceedings of the Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, Quebec, Canada, 23-24 August 1996. Franklin, J.A. and Katsabanis, P.D. (eds.). Balkema, Rotterdam. pp. 59-66.

Danielsson, M., Ghosh, R., Navarro, J., Johansson, D. and Schunnesson, H., 2017. Utilizing production data to predict operational disturbances in sublevel caving. In: Proceedings of the International Symposium on Mine Planning and Equipment Selection (MPES2017), Luleå, Sweden. pp. 139-144.

Devgan, A., 1992. Analysis of rock fragmentation using digital image processing. M.Sc. Thesis. University of Arizona.

Downs, D. C. and Kettunen, B. E., 1996. On-line fragmentation measurement utilizing the CIAS® system. In: Proceedings of the Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, Quebec, Canada, 23-24

43

August 1996. Franklin, J.A. and Katsabanis, P.D. (eds.). Balkema, Rotterdam. pp. 79-82.

Dunstan, G. and Power, G., 2011. Sublevel caving. SME Mining Engineering Handbook, Third Edition. Darlin, P. (ed.). Society for Mining, Metallurgy and Exploration, Inc., pp. 1417-1437.

Elahi, A. T. and Hosseini, M., 2017. Analysis of blasted rocks fragmentation using digital image processing (case study: limestone quarry of Abyek Cement Company). International Journal of Geo-Engineering, 8(16).

Fekete, S., Diederichs, M. and Lato, M., 2010. Geotechnical and operational applications for 3-dimensional laser scanning in drill and blast tunnels. Tunnelling and Underground Space Technology, 25, pp. 614-628.

Feng, Q. and Roeshoff, K., 2004. In-situ mapping and documentation of rock faces using a full coverage 3D laser scanning technique. International Journal of Rock Mechanics and Mining Sciences, 41(3).

Ferrero, A. M., Forlani, G., Roncella, R. and Voyat, H. I., 2009. Advanced geostructural survey methods applied to rock mass characterization. Rock Mechanics and Rock Engineering, 42, pp. 631-665.

Firpo, G., Salvini, R., Francioni, M. and Ranjith, P. G., 2011. Use of digital terrestrial photogrammetry in rocky slope stability analysis by distinct elements numerical methods. International Journal of Rock Mechanics and Mining Sciences, 48, pp. 1045-1054.

Fisher, J. E., Shakoor, A. and Watts, C. F., 2014. Comparing discontinuity orientation data collected by terrestrial LiDAR and transit compass methods. Enginnering Geology, 181, pp. 78-92.

Francioni, M., Salvini, R., Stead, D. and Coggan, J., 2018. Improvements in the integration of remote sensing and rock slope modelling. Natural Hazards, 90(2), pp. 975-1004.

Franzese, M. and Luliano, A., 2019. Correlation analysis. Encyclopedia of Bioinformatics and Computational Biology. Ranganathan, S., Gribskov, M., Nakai, K. and Schönbach, C. (eds.). Elsevier Inc., pp. 706-721.

Fraser, C. S. and Cronk, S., 2009. A hybrid measurement approach for close-range photogrammetry. ISPRS Journal of Photogrammetry and Remote Sensing, 64, pp. 328-333.

Gaich, A., Pötsch, M. and Schubert, W., 2006. Acquisition and assessment of geometric rock mass features by true 3D images. In: Proceedings of the 41st U.S. Symposium on Rock Mechanics (USRMS): "50 Years of Rock Mechanics - Landmarks and Future Challenges". Golden, Colorado, American Rock Mechanics Association.

GeiB, S. and Einax, J., 1996. Multivariate correlation analysis - a method for the analysis of multidimensional time series in environmental studies. Chemometrics and Intelligent Laboratory Systems, 32(1), pp. 57-65.

44

Ghosh, R., Zhang, Z. X. and Nyberg, U., 2015. Borehole instability in Malmberget underground mine.. International Journal of Rock Mechanics and Rock Engineering, Issue 48, pp. 1731-1736.

Ghosh, R., Danielsson, M., Gustafson, A., Falksund, H. and Schunnesson, H., 2017a. Assessment of rock mass quality using drill monitoring technique for hydraulic ITH drills. International Journal of Mining and Mineral Engineering, 8(3), pp. 169-186.

Ghosh, R., Schunnesson, H. and Gustafson, A., 2017b. Monitoring of drill system behavior for water-powered In-The-Hole (ITH) drilling. Minerals, 7(121).

Ghosh, R., Gustafson, A. and Schunnesson, H., 2018. Development of a geological model for chargeability assessment of borehole using drill monitoring technique. International Journal of Rock Mechanics and Mining Sciences, 109, pp. 9-18.

Gigli, G. and Casagli, N., 2011. Semi-automatic extraction of rock masss tructural data from high resolution LIDAR point clouds. International Journal of Rock Mechanics and Mining Sciences, 48(2), pp. 187-198.

Gustafson, A., Jonsson, K., Johansson, D. and Schunnesson, H., 2016. “From Face to surface” - a fragmentation study. In: Proceedings of the Seventh International Conference and Exhibition on Mass Mining – MassMin. Sydney, Australia, 9-11 May 2016. The Australasian Institue of Mining and Metallurgy. pp 555-562,

Haneberg, W. C., 2008. Using close range terrestrial digital photogrammetry for 3D rock slope modeling and discontinuity mapping in the United States. Bulletin of Engineering Geology and Environment , 67(4), pp. 457-469.

Havermann, T. and Vogt, W., 1996. TUCIPS - A System for the estimation of fragmentation after production blasts. In: Proceedings of the Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, Quebec, Canada, 23-24 August 1996. Franklin, J.A. and Katsabanis, P.D. (eds.). Balkema, Rotterdam. pp. 67-72.

hitta.se., 2018. Retrieved October, 2018, from https://www.hitta.se/kartan!~67.63314,21.03247,14z/tr!i=Ti8rUooE/tileLayer!l=1/search!q=vargbacken!b=67.62555:20.99127,67.64072:21.07367!sg=true!t=combined!ai=2000242286!aic=67.64091:21.04226?usergeo=1

Hunter, G. C., McDermott, C., Miles, N. J., Singh, A. and Scoble, M. J., 1990. A review of image analysis techniques for measuring blast fragmentation. Mining Science and Technology, 11, pp. 19-36.

Hustrulid, W. and Kvapil, R., 2008. Sublevel caving – past and future. In Proceedings of the 5th International Conference and Exhibition on Mass Mining – MassMin. Luleå, Sweden, 9-11 June 2008. Luleå University of Technology Press. pp. 107-132.

ISRM, 1978. Suggested methods for the quantitative description of discontinuities in rock masses. International Journal of Rock Mechanics, Mining Sciences and Geomechanics Abstracts, 15, pp. 319-368.

45

Jaboyedoff, M., Couture, R. and Locat, P., 2009. Structural analysis of Turtle mountain (Alberta) using digital elevation model: Toward a progressive failure. Geomorphology, 103, pp. 5-16.

Jaboyedoff, M., Oppikofer, T., Abellan, A., Derron, M.-H., Loye, A., Metzger, R. and Pedrazzini, A., 2012. Use of LIDAR in landslide investigations: A review. Natural Hazards, 61, pp. 5-28, DOI: 10.1007/s11069-010-9634-2.

Jha, A. K., Jha, M. K. and Kumar, S., 2015. Instrument based blast performance evaluation. Journal of Mines, Metals and Fuels, 63(7), pp. 190-192.

Johansson, D., 2011. Effects of confinement and initiation delay on fragmentation and waste rock compaction. Doctoral Thesis. Luleå University of Technology, Luleå, Sweden.

Kemeny, J. M., 1994. Practical technique for determining the size distribution of blasted benches waste dump and heap leach sites. Mining Engineering, 46(11), pp. 1281-1284.

Kemeny, J., Mofya, E. and Handy, J., 2003. The use of digital imaging and laser scanning technologies for field rock fracture characterization. In: Proceedings of the 39th U.S. Rock Mechanics Symposium, Verlag Glückauf, Essen.

Kemeny, J. and Post, R., 2003. Estimating three-dimensional rock discontinuity orientation from digital images of fracture traces. Computers and Geosciences, 29, pp. 65-77.

Khorzoughi, M. B., 2013. Use of measurement while drilling techniques for improved rock mass characterization in open-pit mines. Masters Thesis. University of British Columbia, Vancouver, Canada. DOI: 10.14288/1.0073823.

Khorzoughi, M. B. and Hall, R., 2016. Processing of measurement while drilling data for rock mass characterization. International Journal of Mining Science and Technology, 26(6), pp. 989-994.

Khorzoughi, M. B., Hall, R. and Apel, D., 2018. Rock fracture density characterization using measurement while drilling (MWD) technique. International Journal of Mining Science and Technology, 28, pp. 859-864.

Kleinbaum, D. G., Kupper, L. L., Muller, K. E. and Nizam, A., 1998. Applied regression analysis and other multivariable methods. Third Edition. Pacific Grove. Duxbury Press.

Kleine, T. H. and Cameron, A. R., 1996. Blast fragmentation measurement using Goldsize. In: Proceedings of the Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, Quebec, Canada, 23-24 August 1996. Franklin, J.A. and Katsabanis, P.D. (eds.). Balkema, Rotterdam. pp. 83-90.

Latham, J. P. and Lu, P., 1999. Development of an assessment system for the blastability of rock masses. International Journal of Rock Mechanics and Mining Sciences, 36, pp. 41-55.

Laser Technology Inc., 2020. Lasertech.com. Retrieved September, 2020, from https://www.lasertech.com/Blast-Design-Rock-Face-Profiling.aspx

46

Lato, M. J., Diederichs, M. S. and Hutchinson, D. J., 2010. Bias correction for view-limited Lidar scanning of rock outcrops for structural characterization. Rock Mechanics and Rock Engineering, 43(5), pp. 615-628.

Lemy, F. and Hadjigeorgiou, J., 2003. Discontinuity trace map construction using photographs of rock exposures. International Journal of Rock Mechanics and Mining Sciences, 40, pp. 903-917.

Lemy, F. and Hadjigeorgiou, J., 2004. A field application of laser scanning technology to quantify rock fracture orientation. In: EUROCK 2004 and 53rd Geomechanics Colloquium. Salzburg, pp. 435-438.

Liaghat, S., Gustafson, A., Johansson, D. and Schunnesson, H., 2021. A novel method of pre-processing and filtering measurement while drilling data. Unpublished article.

Linkwitz, K., 1963. Terrestrisch-photogrammetrische Kluftmessung. Rock Mechanics and Engineering Geology, 1, pp. 152-159.

Liu, M. et al., 2020. A comprehensive evaluation method of bench blast performance in open-pit mine. Applied Sciences, 10, 5398(2020).

Li, X., Chen, J. and Zhu, H., 2016. A new method for automated discontinuity trace mapping on rock mass 3D surface model. Computers and Geosciences, 89, pp. 118-131.

LKAB., 2020. LKAB.com. Retrieved February, 2020, from https://www.lkab.com/en/about-lkab/from-mine-to-port/mining/our-underground-mines/

Lorber, A., Wangen, L. E. and Kowalski, B. R., 1987. A theoretical foundation for the PLS algorithm. Journal of Chemometrics, 1(1), pp. 19-31.

Lund, C., 2013. Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model. Doctoral Thesis. Luleå University of Technology, Luleå, Sweden.

Lundin, M., 2020. Laddbarhet och borrhålsstabilitet i produktionsborrhål - Kirrunavaara gruva. Masters Thesis. Luleå University of Technology, Luleå, Sweden.

Maerz, N. H., Palangio, T. C. and Franklin, J. A., 1996. The Wipfrag image based granulometry system. In: Proceedings of the Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, Quebec, Canada, 23-24 August 1996. Franklin, J.A. and Katsabanis, P.D. (eds.). Balkema, Rotterdam. pp. 91-99.

Matthews, N. A., 2008. Aerial and close-range photogrammetric technology: providing resource documentation, interpretation, and preservation. Denver, Colorado. Bureau of Land Management.

Menegoni, N., Meisina, C., Perotti, C. and Crozi, M., 2018. Analysis by UAV digital photogrammetry of folds and related fractures in the Monte Antola Flysch formation (Ponte Organasco, Italy). Geosciences, 8,299. DOI:10.3390/geosciences8080299.

47

Monjezi, M., Rezaei, M. and Varjani, A. V., 2009. Prediction of rock fragmentation due to blasting in Gol-E-Gohar iron mine using fuzzy logic. International Journal of Rock Mechanics and Mining Sciences, 46, p. 1273–1280.

Monteiro, S. T., Ramos, F. and Hather, . P., 2009. Conditional random fields for rock characterization using drill measurements. In: Proceedings of the Eighth International Conference on Machine Learning and Applications, Miami, Florida, USA, December 13-15, 2009, IEEE.

Morin, M. A. and Ficarazzo, F., 2006. Monte Carlo simulation as a tool to predict blasting fragmentation based on the Kuz–Ram model. Computers and Geosciences, 32, pp. 352-359.

Navarro, J., Schunnesson, H., Ghosh, R., Segarra, P., Johansson, D. and Sanchidrián, J., 2019. Application of drill-monitoring for chargeability assessment in sublevel. International Journal of Rock Mechanics and Mining Sciences, 119, pp. 180-192.

Noroozi, M., Kakaie, R. and Jalali, S. E., 2015. 3D geometrical-stochastical modeling of rock mass joint networks: case study of the right bank of Rudbar Lorestan dam plant. Journal of Geology and Mining Research, 7(1), pp. 1-10.

Onederra, I., Thurley, M. J. and Catalan, A., 2015. Measuring blast fragmentation at Esperanza mine using high-resolution 3D laser scanning. Mining Technology. Transactions of the Institutions of Mining and Metallurgy: Section A, 124(1).

Ouchterlony, F. and Sanchidrián, J. A., 2019. A review of development of better prediction equations for blast fragmentation. Journal of Rock Mechanics and Geotechnical Engineering, 11, pp. 1094-1109.

Petropoulos, N., 2015. Burden dynamics and fragmentation. Licentiate Thesis. Luleå University of technology, Luleå, Sweden.

Power, G. R., 2004. Modelling granular flow in caving mines: large scale physical modelling and full scale experiments. PhD Thesis. University of Queensland, Brisbane, Australia.

Raina, A. K., Choudhury, P. B., Ramulu, M., Chrakraborty, A. K. and Dudhankar, A. S., 2002. Fragalyst - An indigenous digital image analysis system for fragment size measurement in mines. Journal of the Geological Society of India, 59, pp. 561-569.

Rai, P., Schunnesson, H., Lindqvist, P.-A. and Kumar, U., 2016. Measurement-while-drilling technique and its scope in design and prediction of rock blasting. International Journal of Mining Science and Technology, 26(4), pp. 711-719.

Remondino, F. and El-Hakim, S., 2006. Image-based 3-D modelling: a review. The Photogrammetric Record, 21, pp. 269-291.

Rengers, N., 1967. Terrestrial photogrammetry: A valuable tool for engineering geological purposes. Rock Mechanics and Engineering Geology, V, pp. 150-154.

Riquelme, A., Cano, M., Tomás, R. and Abellán, A., 2017. Identification of rock slope discontinuity sets from laser scanner and photogrammetric point clouds: A

48

comparative analysis. In: proceedings of the ISRM European Rock mechanics Symposium-EUROCK 2017, Ostrava, Elsevier Ltd.

Riquelme, A. J., Abellán, A. and Tomás, R., 2015. Discontinuity space analysis in rock masses using 3D point clouds. Engineering Geology, 195, pp. 185-196.

Roberts, G. and Poropat, G., 2000. Highwall joint mapping in 3-D at the Moura Mine using SIROJOINT. In: Proceedings of the Bowen Basin Symposium, 2000, Rockhampton, pp. 371-377.

Rodgers, M., McVay, M., Horhota, D. and Hernando, J., 2018. Assessment of rock strength from measuring while drilling shafts in florida limestone. Canadian Geotechnical Journal, 55(8), pp. 1154-1167.

Roy, M. P., Paswan, R. K., Sarim, M., Kumar, S., Jha, R. R. and Singh, P. K., 2016. Rock Fragmentation by Blasting – A review. Journal of Mines, Metals and Fuels, 64(9), pp. 424-431.

Salvini, R., Francioni, M., Riccucci, S., Bonciani, F. and Callegari, I. (2013). Photogrammetry and laser scanning for analyzing slope stability and rock fall runout along the Domodossola–Iselle railway, the Italian Alps. Geomorphology, 185, 110-122.

Salvini, R., Mastrorocco, G., Seddaiu, M., Rossi, D. and Vanneschi, C., 2017. The use of an unmanned aerial vehicle for fracture mapping within a marble quarry (Carrara, Italy): Photogrammetry and discrete fracture network modelling. Geomatics, Natural Hazards and Risk, 8(1), pp. 34-52.

Schleifer, J. and Tessier, B., 1996. Fragscan, a tool to measure fragmentation of blasted rock. In: Proceedings of the Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, Quebec, Canada, 23-24 August 1996. Franklin, J.A. and Katsabanis, P.D. (eds.). Balkema, Rotterdam. pp. 73-78.

Schunnesson, H., 1996. RQD prediction based on drill performance. Tunneling and Underground Space Technology, 11(3), pp. 345-351.

Seccatore, J., 2019. A review of the benefits for comminution circuits offered by rock blasting. REM - International Engineering Journal, 72(1), pp. 141-146.

Segui, J. B. and Higgins, M., 2002. Blast design using measurement while drilling parameters. Fragblast: International Journal for Blasting and Fragmentation, 6(3-4), pp. 287-299.

Shekhar, G., 2020. Draw control strategy for sublevel caving mines: A holistic approach. Doctoral Thesis. Luleå University of Technology, Luleå, Sweden.

Shekhar, G., Gustafson, A., Boeg-Jensen, P. and Schunnesson, H., 2016. Draw control optimisation along the production drift in sublevel caving. In: Proceedings of the Seventh International Conference and Exhibition on Mass Mining – MassMin. The Australasian Institue of Mining and Metallurgy, Sydney, Australia, 9-11 May 2016. pp. 241-249.

Siddiqui, F. I., Shah, S. M. A. and Behan, M. Y., 2009. Measurement of size distribution of blasted rock using digital image processing. JKAU: Engineering Science, 20(2), pp. 81-93.

49

Singh, A., Scoble, M., Lizotte, Y. and Crowther, G., 1991. Characterization of underground rock fragmentation. Geotechnical and Geological Engineering, 9, pp. 93-107.

Singh, S. P. and Narendrual, R., 2010. Causes, implications and control of oversize during blasting. In: Proceedings of the 9th International Symposium on Rock Fragmentation by Blasting. Granda, Spain.

Sturzenegger, M. and Stead, D., 2009a. Close-range terrestrial digital photogrammetry and terrestrial laser scanning for discontinuity characterization on rock cuts. Engineering Geology, 106, pp. 163-182.

Sturzenegger, M. and Stead, D., 2009b. Quantifying discontinuity orientation and persistence on high mountain rock slopes and large landslides using terrestrial remote sensing techniques. Natural Hazards and Earth System Sciences, 9(2), pp. 267-287.

Thurley, M. J., 2013. Automated image segmentation and analysis of rock piles in an open-pit mine. In: Proceedings of the International Conference on Digital Image Computing: Techniques and Applications (DICTA 2013): Hobart, Australia, 26-28 Nov. 2013, IEEE Communications Society, 2014, Article ID 6691484.

van Eldert, J., Schunnesson, H., Johansson, D. and Saiang, D., 2019. Application of measurement while drilling technology to predict rock mass quality and rock support for tunnelling. Rock Mechanics and Rock Engineering, pp. https://doi.org/10.1007/s00603-019-01979-2.

van Oosterhout, D., 2016. Use of MWD data for detecting discontinuities. Masters Thesis. Delft University of Technology, Delft, Netherland.

Vasuki, Y., Holden, E. J., Kovesi, P. and Micklethwaite, S., 2014. Semi-automatic mapping of geological structures using UAV-based photogrammetric data: An image analysis approach. Computers and Geosciences, 69, pp. 22-32.

Westboy, M. J. et al., 2012. ‘Structure-from-Motion’ photogrammetry: A low-cost, effective tool for geoscience applications. Geomorphology, 179(2012), pp. 300-314.

Vezhapparambu, V. S., Eidsvik, J. and Ellefmo, S. L., 2018. Rock classification using multivariate analysis of measurement while drilling data: Towards a better sampling strategy. Minerals, 8(384).

Wimmer, M., 2010. Gravity flow of broken rock in sublevel caving (SLC) – State-of-the-art. Swebrec - Swedish Blasting Research Centre, Luleå University of Technology, Luleå, Sweden.

Wimmer, M., Nordqvist, A. and Ouchterlony, F., 2012. 3D mapping of sublevel caving (SLC) blast rings and flow disturbances in the LKAB Kiruna mine. Swebrec internal report, Luleå University of Technology, Luleå, Sweden.

Wolf, P. R. and Dewitt, B. A., 2000. Elements of photogrammetry with applications in GIS. Third Edition ed. Boston: McGraw Hill Higher Education. ISBN: 0072924543.

50

Zhang, Z. X., 2016. Rock Fracture and blasting: Theory and applications. First Edition. Butterworth-Heinemann.

LICENTIATE T H E S I S

Sohail Manzoor R

ock Evaluation U

sing Digital Im

ages and Drill M

onitoring Data

Department of Civil, Environmental and Natural Resources EngineeringDivision of Mining and Geotechnical Engineering

ISSN 1402-1544ISBN 978-91-7790-655-1 (print)ISBN 978-91-7790-656-8 (pdf)

Luleå University of Technology 2020

Rock Evaluation Using Digital Images

and Drill Monitoring DataBefore and after rock blasting

Sohail Manzoor

Mining and Rock Engineering

130545-LTU-Sohail.indd All Pages130545-LTU-Sohail.indd All Pages 2020-10-19 16:312020-10-19 16:31