Yi X. (yi_xiaoping@hotmail.com and yi_xiaoping@hotmail.com)

and

Bawden W.F., Dept. of Mineral Engineering, University of Toronto, Canada. (bawdenw@ecf.utoronto.ca)

Rockburst is a dynamic phenomenon that occurs in hard rocks under high stress concentrations. In terms of mechanism it may vary between a strain burst, i.e., instant disintegration of a volume of rockmass, and a micro-earthquake, i.e., instant shear displacement on pre-existing geological features. A rockburst may exert impact, shock wave or seismic wave loading on structures depending on the type and distance of the structure as well as the magnitude of the rockburst. Underground excavations, rock support, surface buildings and tailings dams are common structures at a mine site.

Mining is a mixture of art, engineering and science. It is one of the oldest business on the planet and it is the basis of all other industries. It has had a profound impact on our civilization and the environment. An underground mine may be one of the largest man-made structures on earth. For a long time, mining has been considered as being backward and less amenable to scientific methods. With the help of computers, we can now review and predict the positions of the planets in the solar system for hundreds of years in the past and to the future (see the Planet Finder on www.lightandmatter.com). Similarly, plain and simple scientific methods have been developed for better mine design.

The Hemlo Gold Mines consist of the three mines of Williams Mine, Golden Giant Mine and David Bell Mine. Micro-seismic activities that generate rock sound were common in the mines. Falls of ground occurred frequently in the stopes. Three dimensional stopes and drifts were modelled geometrically in the AutoCAD computer software. These geometry models were then used with the three-dimensional stress analysis software MAP3D (1) to compute the stresses around excavations and caved openings. The results were used for improvements on excavation design, support design and mining plan.

At Williams Mine, underground observations, drill cores exploration and computer stress modelling were used to propose a new shape for the footwall drifts. A blast design software was used to improve the quality of stope blasting, which also served to prevent rockbursts caused and triggered by blasting. At David Bell Mine, stress modelling was done to study the stress distributions around stopes, and the stress pattern change at the C-zone and A-zone junction. This served to aid support design in hanging wall drifts and stopes as well as long term mine planning. At Golden Giant Mine, an innovative stress modelling approach, a stress damage graph and the Mathew’s Stability Graph were used for short and long term planning such as design for stope geometry, cablebolt support and mining sequence.

1. Introduction

The Hemlo mines are located at Hemlo, Ontario, 40 km east of the town of Marathon, on the north shore of Lake Superior. It is located right off the Trans-Canada Highway 17 and is about 15 km north of the Pukaskwa National Park. The gold deposit was discovered in 1981. As an example, the Golden Giant produces 3,000 tonne of ore per day and it employed a total of 345 people. Annual production was 1,092,000 tonnes of ore yielding 340,000 ounces (9.639 tonnes) of gold. Mineralization is associated with the Moose Lake Porphory intrusive, a highly strained and metamorphosed multiphase felsic intrusive body. The ore zone may consist of feldspathic, sericitic and baritic ore units.

Most of the stopes in Williams Mine are 10 - 50 m thick and dip 70 degrees, whereas most of the stopes in David Bell Mine are 2 - 17 m and dip 53 - 65 degrees. Stopes in the Central Golden Giant Mine are somewhere in between. Typically, they trend at 110 degrees and dip 65 degrees to the north. Long hole open stoping and delayed backfill mining methods were used. Large blocks ore zones were first established to provide multiple production faces. Details of the mining methods for each block vary somewhat between the three mines. For example, primary stopes were filled with cemented rockfill at Williams Mine and primary stopes were filled with hydraulic sand fill in David Bell Mine, but, continuous retreating with pastefill was employed in Golden Giant Mine. For the Hemlo deposit, the ore grade decreases with ore thickness from east to west. The immediate hanging wall of a few meters thickness is usually a sericitic metasediment and beyond it the hanging wall rockmass is a competent metasediment. The footwall rockmass is usually a Quartze Eye schist or a pink feldspar porphyry.

Fig. 1 shows the longitudinal of the mined-out stopes as of the end of year 1999. Mining occurred upto 1200 m below surface. The Golden Giant shaft as shown in the middle of the figure was 1344 m deep. The mines were in their final stages of production since the main ore zones were more than 80 % mined-out. Due to stress redistribution on the remaining solid stopes, falls of ground and unplanned caving in the stopes were common occurrences. . There were a couple cases where scoops of a quarter million dollars each were buried in the rocks at Williams Mine. These cases were recorded as unusual occurrences.

Figure 1. Longitudinal of the mined out stopes at Hemlo Mines as of September, 1999.

Detailed experimental and analytical studies of mechanical rockbolts under impact loading were performed by Yi and Kaiser (2, 3). It was found that rockbolts may undergo elastic and plastic deformations. Within the elastic range, a wave is generated in the rockbolt by the impact loading of a drop weight, which consisted of a sinusoidal carrier wave and a higher frequency triangular rider wave. The rider wave was due to elastic wave oscillations within the two ends of the steel rod. The carrier wave is of course due to the vibration movement of the drop weight on the steel rod spring. However, In the plastic range, the waves disappeared all together. The plastic deformation of low carbon steel absorbs a lot of energy, but, it may lead to its final breaking. An ideal rock support is one that can yield under impact load but can not break. A practical alternative is to use tiers of different support elements and systems to achieve the same goal, that is, to yield but not break. For rockburst hazard prevention, there are other measures in addition to ground support. There had been successful applications of Destress Blasting in the world, where the rockmass was gently blasted in controlled conditions in anticipation of rockburst hazard

A rockburst of about 2 magnitude on the Nuttli scale occurred in the 9415 level footwall access drift between 26 and 35 stopes on August 31-September 1, 1999. The author went to see the rock pile at 26 intersection on September 1. A large back failure occurred in the pink feldspar porphyry rock to the east side of the cablebolted 26 intersection. More piles of failed rocks were seen in the drift towards the west. It was estimated that the segment of the footwall drift from 26 intersection to 35 intersection came down instantly. There was a footwall orepass at the 33 stope. The rockpile on the floor at the 26 intersection were made of small pieces of rocks. The standard support was 2.1 m resin rebars and weld mesh. Some intersections were shotcreted and cablebolted. Based on prior knowledge of rockbursts and two years of experience at the Hemlo Mines, this author’s opinion is that the rockmass in the roof of the drift that consisted of partly broken rock and partly highly stressed rock finally disintegrated. In other words, it was a large strain energy type of rockburst that is similar to a pillar burst.

The gold price went down to historically low in the year 1998-1999 prior to the above rockburst event. Block #4 was a key production block and it was where the above rockburst occurred. There was a lot of discussion between engineers, consultant, workers and management on how to control the ground with minimum or affordable cost. Cost reduction measures were the main topic of the management. To the author’s recollection, the elastic stresses for the above intersections of the footwall drift were computed and they showed high values in large areas compared to case histories. The stress plots were shown to the designated engineers and consultant. After discussion with engineering and production personnel, the mine management came to a decision to shotcrete and cablebolt the intersections. Unfortunately, the above rockburst occurred after the additional support was installed. Fortunately, no one was in the segment of the drift when the rockburst occurred and no secondary disaster was caused as a result of the rockburst. It is comforting to know that the Canadian mining industry considers people invaluable. The author was informed that since mine production in 1983, there had been no fatality in routine operations in the three underground mines with a total daily production of 11000 tonnes. There was an extraordinary electrification of a student on a high voltage transformer installed in a protected station in Williams Mine. It should be pointed out that the mining industry is essentially closed to the public and outside experts. A mine usually hires only one consultant for a certain purpose during a certain period of time. The author believes that effective and affordable measures could have been found to prevent the above rockburst, had more outside experts been involved in the discussions. Stress modelling or simulation for underground mine design are increasingly to be accepted as necessary procedures by mine engineers.

Rock fragments were often observed under the screen in the footwall drifts particularly in the sharp south corner where all service tubing was located. Mine engineers observed that pieces of loose gradually appeared in the back at a few metres behind the excavation face and drift rehab was required with the elapse of time. Although blast damage and stress increases during and after excavation were known to be the causes, the contribution of each was not clear. This study was conducted to understand the causes and to develop solutions to the rehab problem.

This case was used to demonstrate the existence of stress-induced rock damage in the absence of blast damage. The study of this case helped understand failures in the drifts where blast damage is inseparable from stress-induced damage.

Stope caving in the form of fine muck occurred in 30 stope from 9370 up to 9415 levels between July and September 1997 (Fig. 2). At the beginning of caving, the stope stresses were relatively low. Back and wall failure initiated from the soft transverse Lamp. dykes on both east and west walls where there was no support, and propagated upwards along the hanging wall as mucking continued.

Figure 2. Longitudinal view of the Williams Mine 3D stress model for the 30 stope area in 1997.

32 mm-diameter diamond drill cores were retrieved from the rockmass near the cave in 30 stope 9415 level to study stress-induced rock damage (Fig. 3). The rockmass was not subject to any blast damage. The west to east hole was drilled from 32 to 30 crosscuts on the 9415 level. The hole direction was perpendicular to principal stress direction (north to south). Stresses were computed around the cave to relate stresses to rock damage.

Figure 3. Photograph of the west to east drill cores from 32 crosscut towards 30 stope on 9415 level in Williams Mine showing five sixth of the core panel length. Cores at the bottom were from the footwall of the caved opening.

In Fig. 3, the cores on top of the photograph were from the east side of the 32 crosscut and the cores on bottom were from the vicinity of the caved surfaces. The cores in the last 15 m length near the caved area were broken, and intermittent core discing was observed in the last 10 m length in the footwall of the cave. Stress modelling showed that severe core damage did not occur in the currently stressed ground but in the partially relaxed ground that experienced high stresses beforehand. Previous studies at the Golden Giant Mine revealed the same phenomenon (4). The cores clearly showed that rock fracturing did occur in the absence of blast damage. Stress induced damage was visible on the drill cores only after the ground was relaxed/destressed, although it could still be said that rock damage was occurring in the highly stressed ground.

Fig. 4a shows one surveyed shape for the footwall drift as well as the existing design shape. The design shape employs "shanty" upper walls but vertical lower walls with the intention to conform to the foliation planes while maintaining sufficient usable space for vehicles. Underground observations were made for the footwall drifts on the four different levels of 9660, 9450 (mine starting level), 9390 (block 4 sill pillar), and 9190 (block 5) at increasing depths of 660, 870, 930 and 1130 m, respectively. Surveyed back and floor lines were compared to design, which revealed drift overbreak during excavation but not additional failures afterwards. The excavated drift shapes, rock units, and the occurrences of geological features were also included in the observations. A total length of 2450 m of footwall drifts was observed and the study aimed at finding out the effects of stresses as well as geology on drift overbreak and rock failures.

The common deviations from design occurred in the forms of drift back overbreak, failures in the sharp south corner, and wall failures that conformed to the foliation planes. Table 1 presents a quantitative measure of the deviations from design. The table shows that the deviation from design generally increased with increasing depth when blast damage could be assumed to be constant. Since the 9450 level was the starting level of the mine. It was excavated when the mining induced stresses were minimal. It was understandable that back overbreak was less than the 9660 level above. This study has shown once again that overbreak and failures in the drifts were strongly influenced by depth or stress levels.

Table 1. Observations of drift back overbreak and wall failures.

Figure 4. Profiles of footwall drifts in Williams Mine, (a) between 30 and 31 XC on 9370 level, (b) between 48 and 49 XC one 9555 level, (c) improved shape design.

In large stopes, the stope back often caved near the hangingwall to form a "shanty" back (Fig. 5). Caving occurred at the hanging wall because of (i) the hanging wall contact being a Muscovite shear zone, (ii) gravity effects, (iii) stress induced fractures in the back being normal to the foliation planes, and (iv) possible existence of C-set joints dipping south at about 20 degrees.

Fig. 6 shows the major principal stress contours for a typical drift shape with walls failing to the foliation planes. There is stress concentrations in the back and floor, and the stresses are oriented parallel to the free surfaces. On the other hand, the stresses in the immediate walls are less than the in situ stresses, i.e., the wall rockmass is relaxed.

Relations between rockmass foliation planes and stress magnitudes and directions have important implications in rockmass failures. It is common

Figure 5. Caved stope back for the 28 stope 9370-9415 levels in March and May of 1997 in Williams Mine.

knowledge that a foliated rockmass has the highest compressive but the least tensile (or dilation) strength in the direction normal to the foliation planes. In the drift back, and particularly in the corners, the compressive stresses generally cause spalling and shear type of failures. Failure often initiates at thin foliation dykes whose surfaces are planar, smooth and coated with calcite, and stress-induced fractures in these thin foliation dykes are often normal to the foliation planes (Fig. 4a). Fig. 4b shows that the back failed to a "shanty" shape due to the existence of a 100 mm thick foliation shear zone, i.e., the failure surface tended to be normal to the weak foliation plane.

Despite the low stresses in the drift walls, more failures in the form of separation of foliation planes and buckling were observed in the north (hanging) wall than the back in the highly foliated rock units, such as Muscovite schist. This is because foliation planes became open due to stress relaxation in the north wall. The ground condition in the north wall becames worse if there were foliation shears or dykes (Fig. 4b). On the other hand, more failures in the drift back are usually observed in the semi-massive, hard felsic porphyry because stress relaxation in the north wall usually does not cause failure in the absence of closely spaced foliation planes. Support design for the foliated rockmass in the north wall should help the rockmass form a thick beam.

The observation that pieces of loose gradually form at a few metres behind the excavation face can be explained by the analytical results that stresses in the drift back increase with distance behind the face to reach the maximum at 3 times the drift diameter. However, that rehab was often required with the elapse of time was due to increasing induced stresses with the advance of mining.

Once it was realised that stresses and geological features affect rock failures in the footwall drifts, a new drift shape design as shown in Fig. 4c was proposed, which has the following advantages:

1. the "shanty back" was designed not only to direct stresses to flow normal to the foliation planes but also eliminate potential wedges due to C-set joints and C-set like stress induced fractures,
2. The "shanty footwall" was used to eliminate foliation undercutting so as to prevent gravity induced failures by maintaining beam action and increasing support effectiveness.

Figure 6. Contour plots of major principal stresses for a typical footwall drift in Williams Mine (stresses computations are accurate outside 1 m of free surfaces).

The "shanty back" for the new drift design as shown in Fig. 3c is only partially shanty because it took into considerations of the facts that the 2.1 m rebars in the standard support can support a small wedge in the back and that the development cost increases with increasing drift size. This was one of the examples where science, cost and work force acceptance worked together to reach an engineering solution. A full "shanty back" was used in cut-and-fill stopes in Australia (5).

Figs. 7 shows part of an improved drift. Rebars in the south wall are required only to reinforce the foliated rockmass to form a thick beam. No south wall and corner failures have occurred for the improved drift design during its excavation.

The drift geometry study presented above was completed in 1998 (6). Since then, some new drifts were excavated in the new shape as shown in Fig. 7. For certain severely damaged drifts, such as the west end of the 9215 level where it was only at 15 m distance from the ore zone, the back was rehabbed to be shanty (see Figs. 4b and 5). Such segments were stable as of the time when the author left the mines in October of 1999. In the summer of year 1999, a rockburst of about 2 magnitude on the Nuttli scale occurred in the 9415 level footwall drift between 16 and 30 stopes. This segment of the drift was in the pink feldspar porphyry rock. The back of this entire segment came down instantly, and the final shape was like the top of the caved stope as shown in Fig. 5. Therefore, if the drifts were excavated or rehabbed to the "fully shanty shape" (entire back being perpendicular to the foliation plane), the rockburst in the important access drift might have been prevented. It should be noted that the author has not had the opportunity to study the details of this rockburst.

Figure 7. Photograph of the south half of an improved drift design in Williams Mine (looking West).

The author created the computer software RocCAD that runs on the AutoCAD platform (7). It was the number one software for underground drill and blast design in the world. Compared to manual work without the software, technicians and engineers were able to perform the design and layout work in one fifth of time. With the saved time, they were able to spend more time underground to observe and survey the results (3-dimensional cavity survey proved very useful), compared more engineering alternatives, and were quick to change the design with up-to-date stope information. Therefore, the RocCAD helped improve overall stope blast quality, which helped prevent stope caving and other rockbursts. In addition, RocCAD were used as a visual database (GIS) to link AutoCAD objects to text data, model building and visualization tool for MAP3D, and microseismic data display in AutoCAD, etc..

The Mathews Stability Graph method (8) had been used at the Golden Giant Mine by the engineers on site since 1990 for the purposes of stope planning and cablebolt design. This resulted in successful stope planning and production. With increasing extraction ratio and depth of mining, the rockmass strength in the stopes was often exceeded at the Hemlo Gold Camp. For certain stopes, the stresses were so high that the stope rockmass strength was exceeded before extraction. The resulting stope caving could possibly not only lead to increased mining cost due to ore dilution but also the inability to meet the production goal.

The stress induced rockmass damage was related to the stress changes that the stope had experienced. This type of damage was expressed as fracturing in hard rock that led to a reduction of the ratio (RQD/Jn) in the expression, Q’ = (RQD/Jn) (Jr/Ja). The resulting expression for the revised Rockmass Quality Index was as follows:

Q’ = (RQD/Jn) D (Jr/Ja) (1)

Where RQD is the rock quality designation, Jn the joint sets number, Jr the joint roughness and Ja joint alteration. The damage factor D was a function of the mining induced stresses. It was found that D is best related to the "Extra Stress Deviator" parameter as defined as follows:

The above Extra Stress Deviator parameter was used independently to compare different mining sequences and mining directions for Block 4 East stopes in the Golden Giant Mine for long term planning. It was integrated into the Mathew’s Stability Graph to aid short term stope planning such as stope size and cablebolt support specifications. Cablebolts that were instrumented to measure displacements at different depths were used together with regular cablebolts.

Figure 8. Stress damage graph.

Fig. 9 shows the stope geometry after the mining of the Q14-15 stope 4566-4600. The Extra Stress Deviator for the stope was 56 MPa from stress modelling, and the stress damage factor D was found to be 0.2. With the stress damage correction, the back point dropped from the "Support Required Zone" to the "Caved Zone" (Fig. 10). Indeed, the back on the 4600 level caved about 5 m before backfilling was complete despite cablebolting using double bulge cables.

When this analysis was done, the planning for the Q14-15 stope 4566-4600 could not be changed to avoid the back failure on the 4600 level. However, a planning decision was made to increase the 2E stope width in order to decrease the hydraulic radius for the Q14-15 stope back on the 4633 level.

Figure 9. Longitudinal showing the stope geometry after the mining of the Q14-15 pillar stope 4566-4600 levels in Golden Giant Mine.

Figure 10. Stress Damage Integrated Mathews Stability Graph for the Q14-15 pillar stope 4566-4600 levels in Golden Giant Mine.

Underground observations were made in May of 1998 before mucking in the J2 stope between 11C and 10A sublevels commenced (Fig. 10). On the 10A sublevel, ground conditions in the H/W drift between J1 and J2 X/C deteriorated considerably after the blasting of the J2 stope 11C-10A. Discontinuous seismic sound was heard. An open foliation fracture was seen in the three crosscuts of I4, J1, and J2, with an increasing opening, from one to tens of millimetres, at 8 m distance from the hanging wall contact. The direction of shear displacement was identified on the east walls of the J1 and J2 crosscuts, indicating that the footwall of the fracture displaced towards the west. This fracture coincided with the hanging wall porphyry sill/dyke in the I4 crosscut. Two additional foliation fractures appeared in the J2 crosscut in the less hard sericitic rock unit situated to the north of the porphyry sill. At the J2 intersection, an open foliation fracture (max. aperture 10 mm) was seen on the floor close to the north wall.

A cavity survey that was done when the stope was mucked out revealed 2 - 4 m of hanging wall caving in the centre section of the stope. The hanging wall cables which had a 2.5 m row spacing and a 2 to 2.5 toe spacing ruptured during mucking. In the 10 sublevel rail-haulage, minor spalling was observed in the hanging wall of the J2 stope. The haulage drift is about 20 m from the stope as measured from the haulage south wall to the stope hanging wall contact. Fresh slabbing of about 0.5 cubic metre volume occurred in an unsupported blast raise located at approximately 25 m from the hanging wall contact of the east edge of the stope. Additional caving of the stope hanging wall occurred as the mucking of the J2 stope was completed.

The above observations indicated an extraordinarily high amount of hanging wall caving for this secondary stope. It was suspected that the mined-out A zone was the main cause, but, a complete understanding was essential for the mining plan of upper C zone stopes.

Figure 11. Mining sequence in C-zone from 11C-I4 to 10A-I4 stope in David Bell Mine.

Stress modelling of the mining sequence for C zone was performed. Modelling results in terms of stress deviator contours agree remarkably well with reactions that took place during the mining sequence from 11C-I4 to 10A-I4 stopes (9). It showed that the worst stope hanging wall stress conditions occur one sublevel below and one sublevel above the bottom elevation of the mined-out A zone. Ground stress conditions in the C-Zone will not worsen as stopes are mined above the 10B sublevel (Fig. 12).

Figure 12. (a) Stope geometry when the 10A-10B J2 stope is mined in the future in David Bell Mine.

Figure 12. (b) stress deviator contours when the 10A-10B J2 stope is mined in the future in David Bell Mine.

Art is the use of reasoning and judgment to create something new. Engineering is the applications of rules or laws derived from past practices, of human judgement and of scientific principles to make things. Science is the discovery of rules or laws of things by using measurements, observations, reasoning, analysis and / or computation. These rules or laws may be used to predict future events. Mining is a combination of art, engineering and science. The main conclusions relating to the production of Hemlo mines are listed as follows.

1. The improved drift shape design for Williams Mine was expected to reduce repeated rehab work and potential wedge failures in the footwall access drift due to mining induced stresses. The possibility of wedge failures in the drift back will be substantially reduced. The use of RocCAD for stope blast design helped improve overall blast quality and served to prevent rockbursts.
2. In Golden Giant Mine, stress modelling was integrated into the Mathews Stability Graph as a practical engineering tool for the purposes of stope planning and cablebolt design, where non-elastic rockmass deformation occurred on a large scale.
3. In David Bell Mine, the worst stope hanging wall stress conditions for C zone mining should occur one sublevel below and one sublevel above the bottom elevation of the mined-out A zone. Ground stress conditions in the upper C-Zone will not worsen as stopes are mined above the 10B sublevel.

[1] Wiles T., MAP3D Stress Analysis Software by Mine Modelling PTY Ltd., Australia (1997).

[2] Yi X. and Kaiser P.K., Elastic stress waves in mechanical bolts subject to impact loading. International Journal of Numerical and Analytical Methods in Geomechanics, 18(2), pp. 121-131 (1993).

[3] Yi X. and Kaiser P.K., Impact testing for rockbolt design in rockburst conditions. International Journal of Rock Mechanics and Mining Science & Geomech. Abstr. 31(6), pp. 671-685 (1994).

[4] YI X., Ground Stability Study of Block 4 Stopes at Golden Giant Mine. Report to Battle Mountain Gold, Golden Giant Mine (1997).

[5] Hoek and Brown, Underground Excavations in Rock. Revised and reprinted 1982. (Institution of Mining and Metallurgy, London, England (1980).

[6] YI X., Ground Stability Study of Block 3 and 4 Stopes at Williams Mine. Report to Williams Operating Corporation (1998).

[7] Yi X. RocCAD for 3D mine design, microseismic visualization, database, drill and blast design (runs on AutoCAD 14 platform). Copyright Yi Xiaoping 1997-2004.

[8] Potvin Y., Empirical open stope design in Canada. PhD. Thesis, Dept. Mining and Mineral Processing, University of British Columbia, Canada (1988).

[9] Sprott D., Toppi M., Yi X. and Bawden W., Integration of stress damage factor into Mathew’s Stability Graph. CIM 14^th mine Operator’s Conference, Bathurst, New Brunswick, February 21-25 (1999).