Está en la página 1de 9

A Reliability-Based Approach to Open Stope Span Design in Underground Mining

Peter M. Cepuritis
Western Australian School of Mines, Australia
Ernesto Villaescusa
Western Australian School of Mines, Australia

Abstract
The main aim of underground mine planning and design is to develop a mine design that will
accommodate company business objectives and, when implemented, operates within specified
performance criteria such that financial and operational risks are minimised. The inherent uncertainty
associated with rock engineering implies that there is always a risk that designs may not perform as
intended, along with attendant economic and safety implications. The objective of the design engineer is
to ensure that the level of design reliability is commensurate with the stage of project development and
acceptable business risks. This paper presents an optimised rock engineering approach to open stope span
design in underground hard rock mines. The principal philosophy behind the approach is to ensure that
appropriate methodologies are utilised in data collection, data analysis, rock mass model formulation and
stope design at relevant project stages in order to maximise design reliability.

Introduction

In the planning and design of underground mines, corporate strategies and business plans are the main
drivers in advancing a particular project. Management are responsible for formulating policies relating to
acceptable business risks, which then need to be addressed in the design process. The design engineer is
tasked to develop a design that attempts to accommodate these objectives and constraints. Design
reliability (i.e. the probability of achieving the specified performance criteria) is influenced by orebody
geometry, knowledge of local rock mass conditions, and practical and economic constraints of the day.
The initial objective is to generate a design to the required level of reliability and if, during initial
construction, the design does not meet performance objectives then it will be necessary to optimise and
improve design reliability or implement mitigation strategies to manage the risk of potential poor
performance.
1.1

Design Optimisation By Improving Reliability

The key philosophy behind design optimisation is the continual reduction in uncertainty in all facets of the
design process with project development. This is achieved by ensuring that the most appropriate
methodologies are utilised at each stage of the project in data collection, data analysis, rock mass model
formulation and stope design. The key steps of the proposed reliability-based approach are to;
1. Determine the design and performance objectives (and required reliabilities), at each stage of the
project.
2. Assess the suitability of data for analysis. If data is unsuitable, assess the most appropriate data
collection strategy.
3. Assess the most appropriate approach to rock mass modelling.
4. Assess the most appropriate design methodologies.
5. Assess/confirm design reliability, and

6. Quantify the potential economic impact of the design reliability on the project, and if necessary
devise and evaluate risk minimisation strategies.
Figure 1 shows how these steps relate to the open stope span design process. The key steps are generic, as
design objectives, data collection, data analysis, rock mass model formulation and design methods evolve
with project development, as shown in Figure 2. The figure highlights that ongoing rock mass data
collection and back analyses are fundamental to establishing the reliability of data and rock mass models,
and to confirm design reliability.

Figure 1: Generalised rock engineering design process for open stope span design

Data Collection

The principal aims of data collection programmes in open stope mining should ensure that;

Sufficient data density exists around proposed spans. Data are collated and assigned to
preliminary or established domains. Where data are clustered or poorly represented within a
domain, additional sampling locations may need to be considered. The variability within each
domain is ascertained, and modelled if possible.

All required input parameters are available in each design sector for the selected analysis or
design methodology used at each stage of development.

Data is of sufficient quality to reliably undertake span design. Data are validated and erroneous
data removed, and reliability improved by ensuring bias is removed prior to analysis. Statistical
moments are undertaken on each parameter in each domain to evaluate data homogeneity, isotropy
and reliability. The coefficient of variation of each parameter can be used to indicate whether
input parameter variability is within certain tolerable limits in order to provide required design
reliabilities (see section 5).

Figure 2: Evolution and interaction of key activities, together with example ranges of data, rock
mass model and design reliabilities with project development

If existing data are not suitable then an appropriate data collection programme should be developed;

Decide on the most reliable data collection methods. The limitations of each rock mass sampling
scheme (i.e. data types, subjective versus objective, quantitative versus qualitative, etc.) are
assessed relative to the required objectives and level of reliability.

The location and orientation of sampling schemes are optimised to minimise sampling biases and
reduce the occurrence of significant 'unknown' geological features in key areas (i.e. minimise
Type 1 uncertainties).

Maximise volumetric coverage for each parameter by utilising a number of test methods and
developing reliable correlations between indices.

The cost of additional data must be weighed against the likely increase in reliability. In this case,
the amount, and hence precision, of data should be related to the required level of reliability (e.g.
Priest and Hudson, 1981).

Where no opportunity exists to collect further data there is a risk that the design may not meet its required
level of reliability and subsequently pose a significant risk to project viability and therefore needs to be
documented and communicated to management.

Rock Mass Characterisation Models

A multitude of approaches exist to modelling the rock mass and its components (see Figure 2), however,
the following questions should be addressed to assess the most appropriate approach;

Is the adopted approach valid, given the amount and distribution of data?

Does the model adequately capture and represent all parameters necessary to assess rock mass
interactions in each domain, given the proposed design method?

In the development of domain boundaries, the following should be considered;

How do we construct domains that best represent spatially homogeneous regions?

Do they effectively minimise parameter variability?

Can domain boundaries be optimised?

Is it possible to sub-divide and categorise the domains based on data density?

It may be that a single approach to rock mass modelling may not be appropriate and that an integrated or
hybrid approach may be required, using different models at a number of scales and for various purposes.
The accuracy of any developed models should be established by validation in the field (through further
mapping) or using alternative data sources. Reliability of spatial models also needs to be assessed using,
for example; cross-validation, Q-Q plots, uncertainty and isoprobability maps (i.e. for models developed
using kriging and conditional simulation techniques).

Design Methods

The most appropriate design methodology is largely a function of;

the level of detail of the study, usually related to project development stage,

specific objectives of the engineering design,

level of reliability or tolerance required,

quality and quantity of input data,

spatial representation and complexity of conceptual rock mass models,

level of understanding of the engineering problem.

Open stope design is fundamentally an evolving iterative process. The rock mass model is a crucial
component in guiding the design engineer in identifying anticipated rock mass responses and selection of
appropriate design methods. Empirical methods do not rely on a detailed understanding of failure
mechanisms and, as such, are generally only appropriate for preliminary designs (i.e. Conceptual to Prefeasibility), however, a number of strategies may be adopted to improve their reliability;

Incorporate spatial variability of empirical parameters to optimise and improve design reliability
for the various regions of the mine. Geostatistical models can be used to spatially model both
input and design parameter variability (Cepuritis, 2004; Villaescusa et al., 2003).

It may be found that parameter weightings of existing techniques do not provide optimal results
under all conditions. Mechanisms for making site-specific adjustments should be explored.

Development of new empirical systems, via formulation of alternate parameters and/or


weightings, may provide improved reliability in predicting rock mass response. Hybrid methods those that incorporate the results of numerical modelling (Villaescusa et al., 1997) or analytical
methods, should also be considered to improve efficacy of empirical based approaches.

Figure 2 highlights that a staged approach to the use of numerical modelling techniques is recommended,
to improve reliability, yet cognisant of the additional effort required (modified from Wiles, 2006);

Refine Parameters; refine input or geometric parameters in linear elastic models, conduct
sensitivity analyses on input parameters, such as magnitudes and directions of the in situ stress
regime, refine pillar and excavation geometries.

Lithology-based criteria; developing simple stress-based criteria for each lithology. This involves
using a simple single material property linear elastic model and defining stress-based criteria for
each lithology by collating back analysis data on a unit-by-unit basis.

Directly model material variability; Geological zones with different material properties are
directly incorporated into the linear elastic model in forward analysis. Ability to incorporate
spatial variability of material properties is also recommended. Examples of this approach include
2-dimensional FEM analyses with input parameters represented as spatial random fields
(Griffiths and Fenton, 1993) and geostatistical models (e.g. block models) used as direct input
into numerical modelling (Stavropoulou et al., 2007).

Directly model material complexity; where significant deformation is anticipated linear elastic
modelling approaches will not be able to capture all mechanisms and, used alone, are unlikely to
provide realistic or reliable instability criteria for design. Complex material models (e.g. nonlinear modelling) that include inelastic fault slip or that incorporate additional mechanisms should
be considered, however, these require substantial effort in model construction and calibration.
Figure 2 also indicates that it may be advantageous to conduct a number of design methodologies
in parallel to account for uncertainty in potential rock response mechanisms.

Design Reliability

Design reliability relates to the probability of the excavation succeeding in fulfilling its intended function
during its duty life. Its intended function are generally based on a set of performance criteria, with a
mechanism to predict conditions of instability (i.e. instability criteria) or undesirable rock mass behaviour

(i.e. performance function), which in open stope span design, is usually controlled by aspects of the design
variable, such as stope geometry and sequence (see Table 1).
Table 1: Example open stope span design criteria
Performance Assessment Criteria
Qualitative;
 Stable, Unstable, Failed
Quantitative;
 ELOS, % dilution
 shape factors
 direct volume comparisons
 measurements of rock mass strain and
displacements

5.1

Instability Criteria or Performance


Function
Empirical Parameters;
 Q, RMR, N, N'
Critical State;
 stress threshold
 strain threshold
 displacement
 displacement rate

Design Variables
Size;
 Critical span
 Hydraulic radius
 Radius factor
Orientation
Shape
Sequencing

Estimates Of Initial Open Stope Design Reliability In Forward Analysis

It is generally assumed that the selected design methodology adequately captures the anticipated rock
mass interactions that will control behaviour. In this case, the reliability of the derived design variables
(and design performance) will be controlled by the uncertainty in the input variables. The basic approach
to establishing design reliability in forward analysis is provided below (Baecher and Christian, 2003);
1. Establish an analytical model.
2. Estimate statistical descriptions of input parameters.
3. Calculate statistical moments of the performance function.
4. Calculate the reliability index and probability of failure.
Different analysis methods should produce different means and variances in the performance function, and
thus different probabilities of failure and design reliability. A reliability-based design approach provides a
quantitative method to establish and compare the relative reliability of each design method. Currently, the
most effective way of quantifying reliability of empirically derived design parameters is to assess the
predictive capability of the empirical database using probabilistic and likelihood methods (Mawdesley et
al., 2001; Suorineni et al., 2001). Mean input parameters can be used to assess the average likelihood of
stable, unstable or failed stopes, given the chosen span and empirical performance function (e.g.
RMR, Q, N', etc.). Where an analytical solution exists, classical reliability-based methods can be
implemented, using the mean and variance of capacity and demand inputs, such as First Order Second
Moment (FOSM) or First Order Reliability (FORM) methods. Other methods to evaluate reliability, for
relatively simple analytical and numerical techniques, using input parameter variability include, point
estimation method (PEM) (Rosenblueth, 1975), alternative point estimation method (APEM) (Harr,
1989) and stochastic simulation methods (e.g. Monte Carlo and Latin Hypercube).
5.2

Confirming Design Reliability Based On Back Analysis

The main objective of back analysis is to provide instability criteria or performance functions that can be
correlated to past performance and used as a reliable predictor to future stope performance. Confirmation
of design reliability can be made by back analysing quantitative performance assessment criteria, such as
depth of failure or ELOS against design variables such as span width, hydraulic radius or radius factor. By
analysing stope performance data the likelihood of obtaining a certain amount of dilution for a given span
dimension can be ascertained. Unfortunately, classical statistical approaches may not be viable, due to
small sample sizes, significant variability or no apparent fit to standard models. For example, Figure 3a

shows stope performance from the data from the Mt Marion Gold Mine (Villaescusa, 2010). Although
there is a general trend, it is difficult to ascertain meaningful correlations through linear regression. In this
case, it is proposed that non-parametric (distribution-free) statistical approaches may be warranted to
estimate probability density functions;

Data is first separated into a number of bands based on the dependent design variable or
performance function.

The percentage of the data within each band over a certain threshold of the performance
assessment criteria is then calculated.

The results are then plotted as a relative frequency histogram or line, or the probability density is
estimated using kernel density estimators.

Figure 3b shows the example data represented as probabilities of depth of failure versus hydraulic radius.
In this way, the probability of depth of failure exceeding 1m, 2m and 3m for, say, a hydraulic radius of 8m
can be estimated as 80%, 53% and 26%, respectively.

Figure 3: Plot of a) depth of failure versus hydraulic radius for Mt Marion data (Villaescusa, 2010)
and b) probability of depth of failure versus hydraulic radius for 1m, 2m and 3m
The reliability of instability criteria based on linear elastic modelling be quantified based on the method
proposed by Wiles (2006). The method can be applied to alternate candidate instability criteria with their
corresponding Cv compared. This is an effective tool to compare the reliability of various instability
criteria options. Numerical methods are advantageous in that the geometrical aspects of the stope design
are intrinsically captured in the method. The design engineer assesses results from various models or
design alternatives, by comparing one or more stope geometries, layouts or sequence options. Contouring
(or isosurfacing) of modelling outputs at critical values of the instability criteria, such as critical strain or
stress, can indicate the predicted performance geometry (e.g. over-break). The difficulty lies in
quantitatively comparing all geometric aspects, however, Cepuritis (2011) provides a method to
quantitatively define and compare size and shape of predicted and actual over-break.

Quantifying the Potential Economic Impact of the Design on the Project

Once the reliability of controlling instability criteria or design has been established and accepted, future

design geometries and sequence options can then be evaluated and compare, which considers all
operational and financial aspects of the project. However, the major factor influencing the choice of
design geometry/sequence is the anticipated excavation performance and it's subsequent economic impact;

The range of predicted over-break volumes and their likelihood (i.e. expected volume of overbreak) needs to be evaluated. For example, for simple critical stress based criteria, the shape of the
critical stress isosurface (hence volume) is determined, as well as their probability of occurrence
(Wiles, 2006).

Economic value of the expected volume over-break is then determined. Predicted over-break
volumes can be imported into grade models and economic value of the additional unplanned
volume determined (i.e. grade and tonnes) and subsequent direct additional mining (load, haul and
fill) and milling costs established and compared.

The indirect economic impact of over-break and damage are much harder to define; blocked drawpoints, secondary breakage and damaged brows, re-drilling of blast holes, additional rehabilitation
costs, ore loss (pillars), additional rises, lost opportunities caused by short term diversion of
resources, damage to development access and infrastructure. Empirical databases of past indirect
consequences of over-break and damage should be used to assess likely impacts to operations and
indirect costs.

After the reliability of design parameters and economic impacts have been assessed, a quantitative-risk
based design approach can be adopted to assess design alternatives (Lilly and Villaescusa, 2001).

Conclusions

A reliability-based design optimisation approach has been presented. The key philosophy behind the
approach is the continual reduction in uncertainty in all facets of the stope design process. The approach
presented attempts to achieve this by ascertaining and improving reliability by;

removing errors, reducing biases and appropriate planning of data collection programmes.

optimise domain boundaries to improve statistical homogeneity, select appropriate approaches to


rock mass modelling, and establish reliability of rock mass models.

adopting staged approaches to the selection of the most appropriate design methods based on the
level of understanding of rock-excavation interactions.

utilise methods to assess design reliability based on; performance criteria, instability criteria,
direct volumetric comparisons.

Assessing the economic impact of design reliability to determine adequacy of the design and/or to
determine best alternative between design/sequencing options.

References
Baecher, G. and Christian, J. (2003). Reliability and Statistics in Geotechnical Engineering. John Wiley &
Sons, Chichester, England.
Cepuritis, P. (2004). Three-dimensional rock mass characterisation for the design of excavations and
estimation of ground support requirements. In Villaescusa, V and Potvin, Y (eds.), Ground Support in
Mining and Underground Construction, A.A. Balkema, Rotterdam, pp. 115-128.
Cepuritis, P. (2011). Technical Note: The use of scale independent shape measures to assess open stope
performance. International Journal of Rock Mechanics and Mining Sciences, 48, pp.1188-1195.
Griffiths, D. and Fenton, G. (1993). Seepage beneath water retaining structures founded on spatially

random soil. Geotechnique, 43, pp.577-587.


Harr, M. (1989). Probabilistic estimates for multivariate analyses. Applied Mathematical Modelling, 13,
pp.313-318.
Lilly, P. and Villaescusa, E. (2001). Optimising Pit Slope Angles and Underground Stope Spans Using a
Risk-Based Approach. In Strategic Mine Planning Conference, Aus.I.M.M, Perth, W.A, pp. 71-74.
Mawdesley, C., Trueman, R and Whiten, W.J. (2001). Extending the Mathews stability graph for openstope design. Transactions of the Institute of Mining and Metallurgy (Section A: Mining industry),
110, pp.A27-A39.
Rosenblueth, E. (1975). Point estimates for probability moments. Proceedings of the National Academy of
Science U.S.A., 72, pp.3812-3814.
Stavropoulou, M., Exadaktylos, G and Saratsis, G. (2007). A Combined Three-Dimensional GeologicalGeostatistical-Numerical Model of Underground Excavations in Rock. Rock Mechanics and Rock
Engineering, 40, pp.213-243.
Suorineni, F., Kaiser, P and Tannant, D. (2001). Likelihood statistic for interpretation of the stability graph
for open stope design. International Journal of Rock Mechanics and Mining Sciences, 38, pp.735744.
Villaescusa, E. (2010). Underground Rock Engineering, Master of Mining Geomechanics Class Notes.
Western Australian School of Mines, 255p.
Villaescusa, E., Cepuritis, P, Li, J, Heilig, J, Wiles, T and Lund, T. (2003). Open Stope Design and
Sequences at Great Depth at Kanowna Belle. Unpublished Research Report for Placer Dome Asia
Pacific. p221.
Villaescusa, E., Tyler, D and Scott, C. (1997). Predicting underground stability using a hangingwall
stability rating. In Lee, HK, Yang, HS and Chung, SK (eds.), Proc. 1st Asian Rock Mechanics
Symposium, Environmental and Safety Concerns in Underground Construction, Seoul, South Korea,
pp. 171-176.
Wiles, T. (2006). Reliability of numerical modelling predictions. International Journal of Rock Mechanics
and Mining Sciences, 43, pp.454-472.

También podría gustarte