Professional Documents
Culture Documents
Deep Shaft Sinking Limestone, Mudstone, Halite
Deep Shaft Sinking Limestone, Mudstone, Halite
S. Pollak
Arup, New York
ABSTRACT: The Jansen Project is a greenfield potash mine development being undertaken by
BHP in Saskatchewan, Canada. The project, which remains subject to sanction by the BHP Board
for full construction and execution, includes a pair of shafts which have been sunk to a depth of
approximately 1 km. Unlike previous shafts within the province, the Jansen shafts were sunk
using new technology in the form of Herrenknect’s Shaft Boring Roadheader (SBR) machine.
The shaft liner is being constructed via a two-pass method; with the final liner to be constructed
from the bottom up. Shaft stability in the sinking condition is ensured through a combination
of artificial ground freezing and primary ground support. The focus of this paper is the area of
the shaft at a depth of 700 m to 900m composed of limestone, mudstone, and halite which had a
primary support of rock bolts and mesh in a shaft diameter of ~8 to 12 m. This paper covers the
characterization, analysis, primary ground support selection and construction observations made
within these zones.
1 INTRODUCTION
242
Table 1. UCS and Rock Mass Behaviour.
Rock Mass UCS Range Domain Behaviour
Limestone 20-120 MPa HR 4,8 Elastic, massive, local blocks
Dolomitic Mudstone 10-50 MPa HR 6,9 Plastic deformation, shearing
Halite 20-40 MPa HR 5,7 Time dependent creep
Where DISL = damage initiation spalling limit, CCM = convergence confinement method
The in situ stress was defined by micro-fracture testing, observation of borehole break-outs, and
other wireline methods conducted in project boreholes. The testing validated a regional assumption
of a strike-slip fault regime (σH>σv>σh), where σH trends NE-SW. Based on this data, the follow-
ing total stress ratios were adopted in the zone of interest (note that hydrostatic stress states equal
to the overburden stress (σv) were assumed within the Davidson and Hubbard Evaporite layers):
σH=1.18σv; σh=0.81σv.
The output from the characterization process as shown in Table 2 was to define plausible ground
behaviors and the relevant calculation tools to be applied for engineering analysis.
2 CALCULATION METHODOLOGIES
2.1.2 FLAC3D
In some instances, the CCM was validated using the finite difference software program FLAC3D.
The numerical approach overcomes the limits of the CCM, mainly by allowing for an anisotropic
243
in situ stress field and realistic longitudinal displacement at the face. A simple constitutive model
(Mohr-Coulomb) was employed. Results were found to largely agree with the CCM.
2.4 Creep
Two distinct zones of halite exist within the Hard Rock (HR) intervals. The behaviour of halite
is largely time dependent. As such, appropriate constitutive models must be applied. A simple,
single component Power Law model in FLAC3D based on creep parameters (A=3.75E-10 MPan
hr-1, n=3.6) derived from uniaxial tests was developed. Models were initially solved under Mohr-
Coulomb conditions for elastic displacements, after which the creep model was switched on and
run for the allotted time. The output of such models is radial convergence versus time. In some
instances, compressible material has been trialed in conjunction with a structural (rigid) liner in
order to delay load build-up. A key learning with lab based creep measurement during the project
was to ensure that sufficient test time occurred to ensure parameters could be developed effectively
for each stage of creep.
3 GROUND SUPPORT
Selection has been made based on required system durability (lifetime) and maintenance, stress-
strain characteristics with regards to ground behavior, and overall safety provided. Use of fully
resin grouted rock dowels is common in hard rock conditions. The system is compatible with a
jointed, stiff rock mass.
244
For a uniform, homogenous salt (halite) free of fractures and surface slabbing, the situation calls
for a different approach. Early success in evaporite mines was found using mechanical end anchor-
age bolts, which guarantees the full bolt length is available for stretching, but lacks protection from
corrosion. This problem could feasibly be solved by allowing slip to occur between grout-bar over
a discrete portion of the bolt by either using smooth bar, a plastic sleeve, or “zero” strength resin
grout. Ideally, the resin/rock bond strength should commence debonding prior to reaching the UTS
of the bar to avoid a sudden snapping failure. Recent field trials (Neely, 2014) in evaporites demon-
strated that a significant residual capacity through mechanical friction remains post bond failure as
the bar/resin is dragged through the borehole. The bar rib geometry and shear strength/stiffness of
the resin grout play a role in this. Basic criteria were defined as: bar yield strength < resin/rock bond
(slip) strength < ultimate bar strength. This assumed an elongation of 12% (min.). Additionally,
unbonded elongation capacity > 120mm (based on differential displacement along the dowel from
numerical modeling). For durability, all dowels and accessories were specified as galvanized. The
service life of the ground support was limited by creep movements in the halites, with replacement
anticipated after approximately 10 years.
The foregoing discussion also highlights the fact that a shorter bolt is less likely to fail than
a longer bolt, which is subjected to higher relative displacements. The shortest bolt possible
that is able to fulfill the mission of halite retainment and mesh support should be selected. The
key in defining an appropriate combination of bonded/unbonded length and bar size will be
in measuring the actual bond strength to the evaporites, and how the load develops in the bar
post-slip. The Factor of Safety can then be defined as: FoS = mobilized mechanical friction
load/UTS
The overall rock reinforcement system (bolt, plate, nut, washer, resin) was considered together
to identify any weak links and select the appropriate accessories for the bar. The results of project
specific testing, as defined by ASTM F432, which had the plates pulled over a 100mm hole is shown
in Figure 1. A 200mmx200mm domed plate gave similar performance when compared against the
bar only.
Figure 1. Stress-strain behavior of various rock reinforcement systems as tested for the Jansen Project.
245
4 ROCK MASS ANALYTICAL BEHAVIOR AND OBSERVATIONS
Rock mass behavior is discussed by rock type and subdivided by the local domain (HR) within the
following section. A summary of predicted analytical behavior and observed behaviour is shown
in Table 3:
Photos of the range of ground conditions described are shown in the Appendix.
Surface support in the form of steel welded wire fabric (WWF) or synthetic “polymesh” is
required to restrain small fragments of rock in between dowel faceplates. The typical maximum
block size which can fall between a bolting pattern of 1.0-1.5m is approximately 20 kN. The load
transfer mechanics involve the detached rock exerting tension on the mesh (acting as a “basket”),
which is then transmitted to the rock dowels via the face plate. Measures should be taken to prevent
cutting of the mesh wires by the plate, which could represent the weak link in the ground support
system. Testing suggests that the mesh stiffness is controlled by slippage at the bearing plates, and
by weld and wire failure. The size of the bearing plates had a significant effect on the performance
of the welded mesh. For a 200x200 size plate at 1.2m bolt spacing, an average peak load of 20 kN is
confirmed at a stiffness of approximately 50 kN/m. Therefore, some “bagging” of the mesh may be
expected as peak strength is mobilized.
Polymesh is another surface support option. It is more durable than steel mesh for permanent
applications, but also susceptible to tear and large displacements. For hard rock application, using a
layer of each type against the shaft represented the best option to achieve the aims of high stiffness
and reliability (WWF) with decreased aperture and better corrosion potential (poly). In order for the
system to retain higher stiffness, the WWF is installed over the polymesh layer.
246
Figure 2: 3DEC BBM model for HR8 spall prediction.
members. The limestone is anticipated to be relatively stable and massive, with localized rock falls
as dictated by intersection of random sub vertical joints and bedding. Due to massive (high GSI)
composition and increased depth/stress, the spall potential is classified as low-medium depending
on the use of lower or upper bound compressive strength.
A 3DEC bonded block model was created for HR8, incorporating test results to derive a tensile
strength distribution. The model was carried out under in situ stresses consistent with -890m depth
and as described in Section 1.2. The results are shown in Figure 2.
The black lines represent broken contacts between elements which correspond to cracks with an
opening width >0.1mm. From the plan view, the cracks are largely surficial along the shaft wall only
and do not propagate into the rock mass. Therefore, surface retainment ground support was deemed
sufficient. Confinement quickly increases away from the shaft wall to limit crack propagation to
<0.3m deep.
As a check on the bonded block model, an analytical DISL calculation was considered. It is
important to recognize that the two methods are independent and derived differently from each
other. The envelope for spalling is delineated by three regions as follows: (a) CI or crack initiation
which occurs under very low confinement and is commonly between 35-60% UCS. For HR8, a
value of 37% was determined. (b) CD or limit of stable crack propagation (onset of strain localiza-
tion), found at point of reversal in volumetric strain curve. Commonly between 70-90% UCS. For
HR8, a value of 70% was determined. (c) a transition limit links the two preceding curves and has
a slope between 7-9. Once the intact rock strength and spalling envelope is constructed, the stress
path history of a point of interest can be added from histories extracted from the 3DEC model. As
the model was excavated incrementally in 2m advances, the stress history of both maximum and
minimum principal stresses can be tracked through the entire process of shaft sinking. The intact
rock strength envelope and various GSI rock mass strength envelopes were plotted along with the
spalling limit as shown in Figure 3. The shaded gray area indicates conditions under which spalling
is likely to occur. Three points of interest were tracked for the analysis: at shaft wall, 1m away from
wall, 1.5m away from wall.
The spalling assessment yielded the following conclusions:
• spalling, and not shear, is the failure mechanism for GSI>70.
• corebox photos and televiewer data suggest a massive rock mass with GSI>80; thus spalling
controls
• spalling at shaft wall is predicted, which corroborates the 3DEC model
• confinement increases to such an extent after 1m that spalling is inhibited. This also validates the
empirical spall prediction depth of <1m.
The overall conclusion is that the analytical approach aligns well with the numerical approach,
despite the difference in background and development of the two methods. Arriving at the same
conclusion by two different analyses provides added confidence in the result. The selected ground
247
Figure 3. Dawson Bay Limestone analytical spall prediction with lower and upper bound transition limits (in
red) and associated stress paths.
support was a stiff resin grouted 1.5m long, #8 bolt installed on a 1.0mx1.5m pattern with WWF.
During construction, only minor flaking was noted at hitch transitions where the excavation profile
was slightly enlarged. Otherwise, the limestone was massive and intact.
248
Figure 4. Range of calculated rock mass convergence in Red Beds versus observed values in the analogous
Watrous Formation from -500m.
displacements between 0.8-1.5% were estimated using the empirical methods previously discussed.
This falls within the range of minor squeeze which is generally dealt with by rock dowels and shot-
crete or wire mesh. Applying the CCM, a plastic zone thickness of 2m and convergence of 24mm
prior to support placement 9m back from the face was predicted. At time of installation, the equilib-
rium support pressure was calculated as 74 kPa. Split sets were envisioned as adequate short term
support given their stress-strain behavior. The bolts slip until the pressure drops to match their 30
kPa slip load limit, inducing a further 3mm of displacement.
During construction, localized sloughing (toaster sized fragments) was observed to be loading
the mesh in the NE-SW quadrants at -859m. At this point, the bench was at -880.2m. There was no
indication of plate bending, or significant WWF deformation. The failed depth was estimated to be
on the order of 0.3-0.5m, over an area of 2m x 3m circumferentially.
The mechanism is unclear, but suggests some component of time degradation as the bench was
~2D beyond the area of interest when the ground began to slough. A soft/stiff effect could also play
a role given the presence of the contact between the two different rock types. It was noted on later
inspection that the bagging appears to have “bulked up” over the course of 1 month, but the failure
extent has remained unchanged. The “bulk up” or bulging of mesh may indicate yielding of the
wires, although ample capacity remains until ultimate strength is reached. Away from the contact,
the mudstone was competent and only minor surface flaking noted.
249
Based on the CCM, a 2.1m thick plastic zone may develop and an additional 3mm of radial dis-
placement is imposed on the 2.6m long, #8 resin grouted dowels following installation 9m back
from shaft bottom. The equilibrium pressure of 216 kPa is significantly higher than HR6 due to
stiffer support, but still within capacity for both yield strength and elongation. Therefore, the pro-
posed support type is acceptable from a technical standpoint. Three SEPT tests (0.4m bond length)
were conducted in HR9 to assess bond capacity. All three tests allowed for full mobilization of the
dowel tensile strength without pulling out.
Observations made from the shotcrete deck showed raveling of up to 0.5m in the NW-SE quad-
rants which is consistent with the assumed NE-SW maximum horizontal stress trend. Minor peel-
ing and flaking was present throughout as was anticipated from Figure 4. Overbreak/raveling was
predominately observed to occur towards the top of the layer and up to the contact with the overly-
ing limestones. The bottom 1m of limestone was also spalled, a manifestation of the stiffness dif-
ference between the two materials and the resulting additional stress concentration in the limestone.
Similar behavior has been observed around stiff dykes intruding into softer host rock mass.
Minor difficulty installing ground support was noted due to caving drillholes and break-up
around the collar. However, once in place there was no sign of bolt or mesh bagging – only minor
loose scat caught in place.
4.3 Halites
Figure 5. Radial displacements at various depths into halite after 75 years creep time, taken at evaporite layer
midpoint.
250
The value of differential displacement after 75 years of creep time amounts to:
Δε = 642mm - 533mm = 109mm < 120mm stretch capacity of partially bonded dowel
The assumption of a 75 year design life is extreme as the support only needs to maintain integ-
rity of the surface until placement of the final liner, 1-2 years after initial excavation. However, the
calculation serves to highlight the conservatism built into the design to address uncertainty in use
of lab derived creep test data versus in situ primary and steady-state rates. An important learning
from this analysis is the downward creep that can also occur on the final liner due to the differential
lateral movement between halites and surrounding rock.
During construction, short encapsulation pull tests (SEPTs) were carried out on the #8 partially
debonded rock dowels in order to confirm bond strength. A minimum bond of 4.8 MPa was deter-
mined, with several dowels being loaded up to the 30t jack capacity without slip. Therefore, the
0.9m bond length and 3.2 MPa bond stress assumed during design was confirmed, with potential for
further optimization which could increase service life by allowing more elongation.
5 SUMMARY
A detailed rock mass characterisation program using direct and indirect methods was the foun-
dation for determining analytical rock mass behaviour. Multiple analytical methods were used to
provide robust designs for ground support selection. The selected support was tested in the lab and
in the field to validate design input assumptions. Frequent observations were made during sinking
which generally supported the predicted behaviour from rock mechanics anlayses. The shafts were
successfully bottomed out in 2018 at a depth of ~1km and represent a turning point in how mod-
ern-day shaft sinking can be safely and effectively undertaken.
Appendix: Photo Templates
251
Photo 2. Typical Dolomitic Mudstone.
REFERENCES
ASTM F432-13. Standard Specification for Roof and Rock Bolts and Accessories. ASTM International. West
Conshohocken, PA. 2013. www.astm.org.
Neely, D. (2014). Failure mechanism of resin anchored rebar in potash. (unpublished Master’s thesis). University
of Saskatchewan.
252