back analysis techniques for assessing open stope performance

Back Analysis Techniques for Assessing Open StopePerformance

P Cepuritis1 and E Villaescusa1

ABSTRACT

A methodology and a number of software tools have been developed to assess theoperational and geotechnical factors affecting the performance of open stopingoperations. Open stope performance is generally measured by the ability to achievemaximum extraction with minimal dilution. The paper describes the tools and techniquesfor collecting and analysing common factors affecting performance, such as drill andblast, development undercutting, stress induced damage, rock mass quality and large-scale geological features. Common back analysis techniques, such as empirical stabilitygraph methods, are limited in their ability to identify and quantify the relativecontributions of the various factors that influence excavation performance. The paperproposes a methodology that enables the evaluation of a variety of contributing factorssimultaneously. The proposed methodology also enables the evaluation of spatialvariability in various parameters under consideration. Example data has been collectedand analysed with some results presented.

INTRODUCTION

The Western Australian School of Mines (WASM) is currently conducting research into optimising thedesign and extraction sequence of open stoping excavations in highly stresses rock masses. A largecomponent of this study will involve back analysis of stoping activities from a number of participatingmines. In this regard, the project initially aims to identify and assess the contributing factors on largeexcavation performance. As part of this research, a number of techniques and software tools have beenused to assist in the back analysis of stoping performance at a number of open stoping operations inAustralia. Open stope performance is generally measured by the ability to achieve maximum extractionwith minimal dilution. Hence, the success of the open stoping method relies on the stability of large(mainly un-reinforced) stope walls and crowns as well as the stability of any exposed fill masses(Villaescusa, 2004). The success or performance of an open stope can therefore be judged on the actualoutcome versus the planned outcome, in terms of the final volume, tonnage and grade of materialextracted, and the timeliness of extraction, compared to the planned design and schedule.

OVER-BREAK AND UNDER-BREAK

The two main physical criteria for assessing stope performance are:

• the over-break volume, and

• the under-break volume.

Over-break refers to the volume of material excavated in excess of the planned stoping volume (or‘reference’ volume), and under-break refers to the volume of intact rock material left unexcavatedrelative to the planned stoping volume.

Australian Mining Technology Conference 26 - 27 September 2006 261

1. Western Australian School of Mines, PMB 22, Kalgoorlie WA 6430.

Reference volume

The ‘reference’ volume usually consists of the stope design and in-place development; however, thismay need to be modified if any changes were implemented during excavation. For example, additionalholes may have been drilled and fired that were not on the production plan or, conversely, drilled holeswere not fired due to bridging/blockages or due to a decision to leave a pillar. The ‘reference’ volume istypically represented by a primitive triangulated irregular network (TIN) wireframe model in mineplanning software.

Cavity monitoring surveys

To obtain a quantitative measure stope performance, final excavated volumes need to be obtained andcompared to the stope design boundaries. Cavity monitoring system (CMS) surveys have been a wellestablished method for surveying inaccessible open cavities in underground mining operations for wellover ten years. The CMS was developed jointly by Noranda Technology Centre (NTC) and OPTECHSystems, Canada (Miller et al, 1992). This system is widely used in Australia’s undergroundoperations, especially in Western Australia (Jarosz and Shepherd, 2000).

Final stope void volume

A typical CMS stope survey can generate in the order of 30 000 - 60 000 data points. This is anextremely large amount of data to process, and in most cases contains a significant amount of redundantdata. This is especially apparent in close proximity to the CMS set-up, where excavations are typically‘over-sampled’. In addition, some regular surface features, such as relatively uniform flat planes,which may only required to be represented by a few points, can also be over-sampled. In order toconduct meaningful post-processing volume calculations, this data needs to be filtered down to apractical level, without unduly misrepresenting the final void geometry. The general process fordeveloping a final stope void volume generally consists of the following:

• data acquisition utilising CMS survey techniques;

• data filtering of redundant points, including CMS outlier/spike identification and removal;

• CMS amalgamation (where required), involving merging/combining a number of CMS surveystaken from a variety of sites; and

• final void modelling of CMS data.

The final void volume generally consists of generating a TIN wireframe ‘model’ in mine planningsoftware, based on the CMS survey data. Common practice is to generate slices through the raw CMSdata, filter the slices to reduce redundant points and re-generate the volume. In some circumstances,this process can severely affect the precision and accuracy of final void model. In addition, thetreatment of ‘blind-spots’ or shadow areas, as well as presence of broken material or rill in the stopewill significantly impact on the accuracy of the final void shape.

Calculation and reporting of over-break and under-break

Over-break and under-break volumes are generally calculated by intersecting the ‘reference’ volumewith the final stope void volume utilising mine planning software, typically using triangulationintersections of the relevant TIN wireframes. Depending on the relative configuration and aspect ratiosof individual triangles within the wireframe models, errors may occur during triangulation intersectionprocess (principally due to floating point precision), resulting in an unsolvable volume intersection.

262 26 - 27 September 2006 Australian Mining Technology Conference

P CEPURITIS and E VILLAESCUSA

Tonnage and grade values of dilution and ore loss can also be calculated utilising the ‘reference’and final stope void volumes in conjunction with a block model of reserve grades and densities,including backfill materials. These data can then be stored in a database for further reconciliationreporting and presentation (Morin, 2006).

FACTORS AFFECTING STOPE PERFORMANCE

A brief summary of major factors controlling open stope performance are provided below (after Clarkand Pakalnis, 1997):

• stope geometry (size, shape, orientation),

• location of existing development (ie under-cutting),

• rock mass characteristics,

• stress conditions,

• large-scale geological structures,

• rock reinforcement,

• drill and blast processes, and

• time dependency.

In order to assess the influence of each of these factors on stope performance, it is necessary to beable to adequately characterise and/or quantify the factors both prior to and during excavation process.This requires all personnel involved in the design and production stages to record relevant information,such that it can be reviewed on completion of the excavation (Villaescusa, 1998).

Stope performance reviews

Stope performance reviews act as a technical audit to the stope design process (Villaescusa, 2004).Aside from being an historical record of the achieved physicals (ie volume, tones and grade), animportant purpose of these reviews is to understand and assess the factors that affect final stopeperformance. These findings can then be used to develop improved design and engineering tools foroptimising stope and pillar design, layout and sequencing, as well as implementation practices. In thisregard, it would be beneficial to develop a database of these major factors, in conjunction with thephysicals obtained from CMS analyses.

BACK ANALYSIS TECHNIQUES

Interest has been shown in trying to understand the relative influence of each of the factors mentionedabove for use in optimising the planning and design of open stopes. A variety of back analysistechniques have been explored. For example, Wiles (2006) uses back analysis results from linearelastic numerical modelling to indicate the reliability of stress-related damage and failure criteria,Suorineni et al (1999) investigated the influence of the location and orientation of fault structures onopen stope over-break using 2D hybrid boundary-finite element studies, statistical and empiricalstability graph techniques have been used by Stewart (2005) who attempts to relate blast hole design,stress damage and relaxation to the amount of over-break in narrow vein stopes, and Wang (2004)looks at the effect of undercutting on over-break. These approaches may be appropriate if the dominatemechanism can be identified; however, they may not be considered comprehensive as each factor isinvestigated separately, with the influence of other contributing factors not readily accounted for. Amethodology, therefore, is required to account for and assess the relative influence of each individualfactor on stope performance.


BACK ANALYSIS TECHNIQUES FOR ASSESSING OPEN STOPE PERFORMANCE

PROPOSED METHODOLOGIES

A methodology is proposed to assist in the confirming, or otherwise, of hypotheses intimating therelative contribution of various factors and their influence on open stope performance. Thismethodology can utilise, for example, the results of:

• candidate criteria developed from numerical modelling,

• over-break and under-break analysis,

• modelling of large-scale geological structures, and

• rock mass quality modelling.

The methodology relies on the ability to query the volume of rock around the excavation underanalysis to select regions fitting various candidate criteria thought to contribute to over-break. Theadvantage of this system to other empirical approaches lies in the ability to test various candidatecriteria simultaneously and also account for spatial variability of component parameters.

Implicit surfaces

The proposed methodology involves fitting implicit surfaces, defined by radial basis functions (Carr etal, 2001), to candidate criteria and other features of the rock mass. The use of mathematical radial basisfunctions allow for relatively complex mathematical intersections and/or unions of these implicitsurfaces or volumes. Implicit surfaces can be used in a variety of back analysis processes, for example,over-break and under-break analysis, where they are superior compared to the known issues associatedwith triangulation intersections. For example, Figure 1 shows the highly detailed implicit surfacemodel, which honours all CMS data points, compared with a traditional filtered triangulation model.Due to the highly detailed nature of the output, it can also assist in identifying localisation of over-breakor under-break. This information can provide information on the morphology, extent and nature ofover-break. For example, the morphology can indicate whether the over-break was localised andtetrahedral in shape, possibly indicating a small structurally controlled wedge-type failure, or arciformand protracted, possibly indicating stress-related failure and/or subsequent ‘arching’ response.



a) b)a) b)

FIG 1 - Design stope volume (dark grey) and CMS (light grey) for (a) triangulated final void volume and(b) final void volume based on implicit surfaces.

Implicit surfaces can be used to model, for example, large-scale geological structures and their controlon rock mass quality and subsequent influence on the levels of over-break. Figure 2 (a) shows two faultstructures modelled from geological data, highlighting the occurrence of major over-break on the stopewall with its intersection with ‘Fault B’. Figure 2 (b) shows the distribution of rock mass quality, in thiscase fracture frequency interpolated from borehole data, on the over-break surface. Figure 2 (c) showshow the intersection of the two fault surfaces locally controls an increase in fracture frequency (isosurfacevalue of 12 fractures per metre) in the area highlighted by the circle shown in Figure 2 (b).

Candidate criteria

The results from numerical modelling can also be incorporated into the back analysis methodology.Candidate surfaces based on say, a maximum shear stress criteria and/or confinement-based criteria,can be displayed alongside other candidate criteria and over-break data. Figure 3 (a) shows the resultsof linear elastic numerical modelling represented as implicit isosurfaces, together with over-break.Figure 3 (b) shows contours of low confinement on the over-break surface, whilst Figure 3 (c) showsmaximum shear stress, plotted as both contours on the over-break surface and as an isosurface.



a) b) c)

isosurface

faults

Fault

A

Fault

B

a) b) c)

isosurface

faults

Fault

A

Fault

B

FIG 2 - Results of geological modelling of (a) two faults (b) contours of fracture frequency on over-break surface,and (c) interaction of faults, rock mass quality and over-break.

Isosurface(?

3=0.75MPa)

b)a) c)

Isosurface

(?max=20MPa)

Isosurface( 3=0.75 MPa)

b)a) c)

Isosurface

(max=20 MPa)

FIG 3 - Results of numerical modelling showing (a) minor principal stress isosurfaces (b) contours of minor principalstress on over-break surface, and (c) isosurface of maximum shear stress and contoured on over-break surface.

Volumetric queries

The implicit surfaces can then be used to generate queries within the rock mass based on intersectionsand unions of volumes (similar to Boolean ‘and’ and ‘or’ operations, respectively). For example, basedon the data presented above, it is possible to select a volume of the rock mass around the excavationbased on stress-based criteria and/or rock mass quality and/or distance to a prospective geologicalstructure. In this example, a query was constructed using the following criteria;

• maximum shear stresses greater than 15 MPa,

• distance less than 10 m from ‘Fault B’, and

• fracture frequency greater than seven.

The resulting volume is shown in Figure 4 and provides a very good correlation between the queryvolume and the location of over-break experienced during mining. Although ‘Fault B’ transects theentire stope, Figure 4 (b) also highlights that its presence alone is not an indication that over-break willoccur. Figure 4 shows that intersection and union functions of implicit surfaces can be used as avaluable tool in determining the relative influence of various candidate criteria on stope performance.

Candidate query reliability

Using an approach described by Wiles (2006), understanding uncertainty in observational backanalyses can assist in determining the reliability of any forward analyses. Wiles (2006) describes theprocedure for determining the reliability of candidate stress-based strength criteria and its use in failureprediction. This approach can also be incorporated into this methodology by generating implicitisosurfaces of various prediction criteria, at a number of values, which could be used to representprobabilities of failure.



a) b)a) b)

Beyond limit

of rock massmodel

FIG 4 - Results of intersection of multiple candidate criteria for stope AP02 looking (a) north west and (b) south east.

Stope performance database

Data generated from the tools above can be linked to a purpose designed stope performance database,along with CMS derived physicals and qualitative stope performance information. The databasecontains relevant fields for quantitative and qualitative information in the following areas:

• stope details,

• design details,

• rock mass and boundary condition data,

• drilling and blast,

• CMS and performance,

• rock reinforcement, and

• extraction and filling data.

The qualitative information, mainly derived from the results of stope performance reviews,provides an important role in confirmation or verification of the impact of the various factors indicatedusing the volumetric querying techniques described in preceding sections. A brief overview of themethodology for incorporating the data in a back analysis is provided in Figure 5.



Stope Performance Database

Development, stopereinforcement data

Development offorward analysis

criteria

testing

ConfirmCriteria?

linear/non-linearcontinuum/discontinuum

Observation andmonitoring

Design details

Summary rock mass

parameters

Yes

No

Statistical

Analysis/Qualitative

Evidence

isosurfaces

e.g. Fault surfaces

Rock MassModelling

Stope Design

StopeDevelopment

StopeProduction

NumericalModelling

Rock MassParameterVolume(s)

StopeDesignVolume

DevelopmentVolume

Large ScaleStructures

StopePerformance

Review

DevelopCandidate

Criteria

Hypothesistesting

Final VoidVolume

ReferenceVolume

Over-break/

VolumesUnder-break

CriteriaIsosurfaces

Stress/strain

e.g. UCS, ff/m

CMS data, physicals

FIG 5 - Framework for back analysis of stope performance utilising numerical modelling, volumetric queries andstope performance database.

Additional software tools

The software tools include existing proprietary tools as well as a number of tools developed by WASM.The tools developed by WASM consist of a number of database development applications andscripting tools to existing mine planning software, such as Surpac Vision. Examples include:

• tool for quickly ascertaining geometric properties of stope surfaces,

• tools to generate the production of the following over-break and under-break data:

• depth (ie perpendicular distance to reference volume) of over-break and under-break,

• volume of over-break and under-break, and

• area of over-break and under-break, and

• tools for querying and graphing various stope performance data.

STOPE PERFORMANCE CASE STUDY

The systems developed at WASM have been used to back analyse stope performance data from BarrickAustralia’s Kanowna Belle Gold Mine (KBGM). Open stoping operations at the KBGM commencedaround January 1998. The mine is divided into four main mining areas, each with distinct open stopingmining methods and sequencing;

• Block A:

• large multi-lift stopes, primary/secondary centre-out sequence, and

• shallow to medium depth, faulted footwall.

• Block B:

• bottom-up, downhole benching, central retreat, and

• narrow ore widths.

• Block C:

• bottom-up, pyramidal sequence,

• initially double lift primary/secondary 1-3-5 pyramidal sequence with medium sized stopes,

• continuous pyramidal sequence (after July 2003), medium single lift stopes, and

• faulted hanging wall.

• Block D:

• bottom-up, continuous pyramidal sequence, small single lift stopes, also mining panels fromhanging wall to footwall, and

• depths in excess of 1000 m, faulted hanging wall.

The location of the main mining areas and sequence is summarised in Figure 6. Analyses to date haveconcentrated on Block A and Block C. Even simple queries from the stope performance database can beof benefit to understanding the controls on performance. For example, simple analysis of volumetric datafor both mining blocks, is shown in Figure 7. Figure 7 (a), shows the exposed area of over-break on thestope wall surface with respect to the entire area of the stope wall surface. Figure 7 (a) indicates thatmajority of over-break occurring on stope wall surfaces does not affect the entire surface. Although stopesurfaces are larger in Block A, it shows that there is significant variability in the surface area of over-breakin both mining areas. This figure indicates that, although rock mass and stress conditions are differentbetween the mining blocks, the observed response of over-break within each mining block, in terms ofsurface expression, can be quite varied, ranging from localised over-break (ie small area of over-break) toover-break occurring almost across the entire stope surface.



Figure 7 (b) displays a plot of average depth of over-break (volume of over-break divided by area ofover-break exposed) versus stope wall surface area. Intuitively, given the same rock mass, boundaryand implementation conditions, one would expect the depth of over-break to increase as the stope wallsurface increases, however, the data highlights the potential chaotic or unpredictable ‘yielded’response of over-break, or ‘fall-off’. This behaviour, together with the data shown in Figure 7 (a), mayexplain why empirical techniques which relate the normalised amount of dilution (to stope surfacearea) with rock mass classifications, such as the ELOS stability graph technique (Clark and Pakalnis,1997), often show poor correlations.

Figure 7 (b) does, however, highlight the different response, in terms of average depth of over-breakwith stope wall surface area, between the two mining blocks. In general, Block C stope surfaces exhibiteddeeper zones of over-break to Block A stopes. Block C stopes show high variability in average depths ofover-break, with significant amounts of over-break even with relatively small stope wall surface areas.Conversely, Block A stopes generally show no increase of average depth of over-break with increasingstope surface area, with depths of over-break generally below 2 m. This would indicate increased stopessizes may have been achievable without significantly increasing the amount of dilution.



Jan ’98 - Dec ‘99

Jan ’00 - Mar ’01

Apr ’01 - May ‘02

May ’02 - Jul ‘03

May ’02 – May ‘03

Jun ’03 - Jun ‘03

Jul ’03 - Nov ‘05

Nov ’05 - Present

BLOCK D

BLOCK C

BLOCK A

BLOCK B

9600mRL

9800mRL

10000mRL

9400mRL

Open PitSurface at 10360mRL

Blocks B, C and D

Block APrimaries

Secondaries

Jan ’98 - Dec ‘99

Jan ’00 - Mar ’01

Apr ’01 - May ‘02

May ’02 - Jul ‘03

May ’02 – May ‘03

Jun ’03 - Jun ‘03

Jul ’03 - Nov ‘05

Nov ’05 - Present

BLOCK D

BLOCK C

BLOCK A

BLOCK B

9600mRL

9800mRL

10000mRL

9400mRL

Open PitSurface at 10360mRL

Blocks B, C and D

Block APrimaries

Secondaries

FIG 6 - Long section of Kanowna Belle underground operations showing main mining areas and stoping sequence.

Block C data was investigated further and separated into primary and secondary stopes. To provideconsistency end wall data was removed from the primary stope data set, as end walls in the secondarystopes were not evaluated (being fill). The resulting data is plotted in Figure 8. This figure shows thatover-break in secondary stopes, generally occurs over larger areas of the stope surface, possiblyindicating the effects of stress and/or blast damage.



0

1

2

3

4

5

6

7

8

9

10

0 1000 2000 3000 4000 5000 6000

Stope wall surface area (m2)

Av

era

ge

de

pth

of

ove

r-b

reak

(m)

Block A

Block C

a) b)

0

1000

2000

3000

4000

5000

6000

0 1000 2000 3000 4000 5000 6000


Exp

os

ed

are

ao

fo

ver-

bre

ak

(m2)

Block A

Block C

0

1

2

3

4

5

6

7

8

9

10

0 1000 2000 3000 4000 5000 6000


Av

era

ge

de

pth

of

ove

r-b

reak

(m)

Block A

Block C

a) b)

0

1000

2000

3000

4000

5000

6000

0 1000 2000 3000 4000 5000 6000


Exp

os

ed

are

ao

fo

ver-

bre

ak

(m2)

Block A

Block C

FIG 7 - (a) Plot of exposed area of over-break versus stope wall surface area and (b) average depth of over-breakversus stope wall surface area.

0

500

1000

1500

2000

2500

3000

0 500 1000 1500 2000 2500 3000


Ex

po

se

dA

rea

of

Ov

er-

bre

ak

(m2)

Primary Stopes

Secondary Stopes

FIG 8 - Plot of exposed area of over-break versus stope wall surface area for Block C stopes.

CONCLUSIONS

An improved methodology to understand the relative influence of the various factors that influenceopen stope performance has been proposed. The method allows for the simultaneous investigation ofvarious factors, such as rock mass parameters and the results of numerical modelling, the integration ofthe spatial variability of parameters, and ability to query the rock mass volume for a variety ofcandidate criteria. The methodology is not a replacement for traditional analytical or numericaltechniques, yet provides a framework for integrating and interrogating results from these techniques.WASM are currently in the process of further refining the interrogative methodology using thetechniques described in this paper. Currently, meta-data derived from the volumetric analysis and stopeperformance reviews is kept separate from the geometric entities (ie stopes, faults, over-break, etc) inthe stope performance database. Potential exists to integrate both the geometrical entities and meta-datain object-oriented databases and possibly explore 3D-GIS capabilities to generate complex queries tofurther enhance back analysis techniques.

ACKNOWLEDGEMENTS

We wish to thank Kanowna Belle Gold Mine, Barrick Australia, for kindly providing us with theirCMS data and allowing us to publish this paper. We wish to acknowledge financial assistance of ourother industry partner, BHP Billiton – Cannington and the ARC for a Linkage Grant. We would alsolike to acknowledge Zaparo Pty Ltd for allowing the use of their software to assist in this work.

REFERENCES

Carr, J C, Beatson, R K, Cherrie, J B, Mitchell, T J, Fright, W R, McCallum, B C and Evans, T R, 2001.Reconstruction and representation of 3D objects with radial basis functions, in Proceedings ACMSIGGRAPH 2001, pp 67-76, Los Angeles, USA, 12 - 17 August.

Clark, L M and Pakalnis, R C, 1997. An empirical design approach for estimating unplanned dilution from openstope hangingwalls and footwalls, in Proceedings 99th AGM – CIM, pp 25 (Canadian Institute of Mining,Metallurgy and Petroleum).

Jarosz, A P and Shepherd, L, 2000. Open stope cavity monitoring for the control of dilution and ore loss, inProceedings MPES2000, pp 63-66, Athens, Greece, 6 - 9 November.

Miller, F, Potvin, Y and Jacob, D, 1992. Laser measurement of open stope dilution, CIM Bulletin, 85:96-102.

Morin, B, 2006. Cutting edge long hole open stope analysis at Perilya’s Broken Hill operations, in ProceedingsSecond International Seminar on Strategic Versus Tactical Approaches in Mining 2006, Sect 41, pp 1-17(Australian Centre for Geomechanics: Perth).

Stewart, P C, 2005. Minimising dilution in narrow-vein mines, PhD thesis, JKMRC, University of Queensland, 261 p.

Suorineni, F T, Tannant, D D and Kaiser, P K, 1999. Determination of fault-related sloughage in open stopes, IntJ Rock Mech Min Sci, 36(7):891-906.

Villaescusa, E, 1998. Geotechnical design for dilution control in underground mining, in Proceedings MinePlanning and Equipment Selection, (ed: Singhal) pp 141-149 (Balkema: Rotterdam).

Villaescusa, E, 2004. Quantifying open stope performance, in Proceedings MassMin2004, pp 96-104, Santiago,Chile, 22 - 25 August.

Wang, J, 2004. Influence of stress, undercutting, blasting and time on open stope stability and dilution, PhD thesis,Department of Civil and Geological Engineering, University of Saskatchewan, 279 p.

Wiles, T D, 2006. Reliability of numerical modelling predictions, Int J Rock Mech Min Sci, 43(3):454-472.



back analysis techniques for assessing open stope performance

Documents