Latilla Soft Floor MSC

I N F L U E N C E O F W E A K F L O O R O N
T H E S T A B I L I T Y O F P I L L A R S I N
S O U T H E R N A F R I C A N
C O L L I E R I E S

Jonathan William Latilla

A research report submitted to the Faculty of Engineering, University of the Witwatersrand,
Johannesburg, in partial fulfilment of the requirements for the degree of Master of Science
in Engineering.

Witbank, March, 2003

ii
DECLARATION

I declare that this project report is my own unaided work with the exception of assistance
received from D. Neal of Ingwe Rock Engineering in building and running the MAP3D
Numerical modelling (6.2). Determination of input parameters and interpretation of results
were my own unaided work.

It is being submitted for partial fulfilment of the degree of Master of Science in the
University of the Witwatersrand, Johannesburg. It has not been submitted before for any
degree or examination in any other University.

Information used in this project report was obtained while the author was employed by the
Rock Engineering department of Ingwe Collieries (the South African division of BHP
Billiton Energy Coal).

-----------------------------------
J W Latilla

------------------day of ------------------------------- ----------------

iii
ABSTRACT

A number of pillar collapse cases have been recorded in southern African collieries where
weak floor rock was considered to be influential. This study attempts to analyse a few of
these cases to determine if there are any relationships between them.

Initially a literature survey was carried out which concentrated on local as well as
Australian and United States reports and papers. This was followed by collating data from
four failed and four stable, southern African case studies. Three and two-dimensional
numerical modelling was done for each of the case studies as well as analysing each from a
civil engineering foundation stability point of view.

It became apparent from the literature that pillar settlement and surface subsidence are
more prevalent overseas while locally pillar failure with its corresponding safety
implications is more likely to occur.

A few relationships were established during analysis of the case studies. However, it
appears that only a small number of the currently worked seams in southern Africa have
floors which are prone to failure, and that in these cases, it would be prudent to design
pillars to higher factors of safety as well as to limit the width of panels. A provisional set
of design guidelines have been proposed for the identified problem seams.

Two-dimensional numerical modelling was found to be a useful tool in studying this
problem but considerably more geotechnical investigations are needed to evaluate the civil
engineering design methods.

The problem is a rather complex one and more work, considered to be outside the scope of
this project, will need to be carried out. The additional work required is likely to be costly,
in that a large volume of field data will have to be collected.

iv
ACKNOWLEDGEMENTS

Prof T.R. Stacey of the Mining Engineering Department at the University of the
Witwatersrand for his invaluable input as Mentor for this project.

Dr P.S. Buddery of Strata Engineering Australia and J.J. van Wijk of Ingwe Rock
Engineering for reviewing this project report.

D. Neal of Ingwe Rock Engineering for assistance with MAP3D modelling.

Ingwe Collieries (Messrs M Oppenheimer and E Scholtz) for permission to publish the
Emaswati, Welgedacht and Zululand Anthracite Colliery (ZAC) case studies .

Eyesizwe Coal (Mr J Nel) for permission to publish the Matla case studies.

Ingwe Rock Engineering Department for assistance with access to computer facilities and
programmes.

I. Canbulat of CSIR Miningtek for allowing access to the PHASE
2
program.

Prof. J.N. van der Merwe of Pretoria University for information on the Sigma case study
and papers on foundation failure from the Illinois coalfield in the USA.

B. Jack of CSIR Miningtek for locating two reports and Dr G. Gurtunca for permission to
quote from them.

Dr B. Madden for copies of reports on two Australian cases.

Dr C. Mark of NIOSH for numerous American papers on foundation failure.

Dr E. Sellers of CSIR Miningtek for various rock properties data.

v
TABLE OF CONTENTS

DECLARATION ..................................................................................................... ii
ABSTRACT............................................................................................................ iii
ACKNOWLEDGEMENTS ..................................................................................... iv
TABLE OF CONTENTS ......................................................................................... v
LIST OF FIGURES ................................................................................................ ix
LIST OF TABLES ................................................................................................. xii
LIST OF SYMBOLS ............................................................................................ xiii
LIST OF ABBREVIATIONS ................................................................................ xiv
1 INTRODUCTION ................................................................................................... 1
2 THE GOAL OF THIS PROJECT ............................................................................ 1
3 INVESTIGATION INTO PREVIOUS WORK ON FOUNDATION STABILITY .... 1
3.1 Geological factors ...................................................................................... 2
3.1.1 Seams prone to floor problems ..................................................................... 2
3.1.2 Clay minerals .............................................................................................. 3
3.1.3 Facies characterisation ................................................................................ 3
3.1.4 Conclusions ................................................................................................ 3
3.2 Heave Mechanisms ..................................................................................... 4
3.2.1 Buckling of floor beam and plastic flow modes ............................................ 4
3.2.2 Squeezing of weak floor rock ....................................................................... 7
3.2.3 Floor failure due to horizontal stress ............................................................ 8
3.2.4 Coal pillar shearing through weak floor beam ............................................... 8
3.2.5 Swelling floor strata .................................................................................. 11
3.2.6 Bearing capacity failure around pillar edges ............................................... 11
3.2.7 Hard rock mines ........................................................................................ 12
3.2.8 Summary of floor failure mechanisms ........................................................ 13
3.2.9 Conclusions .............................................................................................. 14
3.3 Geotechnical Testing ............................................................................... 14
3.3.1 Ultimate bearing capacity .......................................................................... 14
3.3.2 Swelling potential ..................................................................................... 16
3.3.3 CMRR based floor rating ........................................................................... 16
3.3.4 Slake durability and Duncan swell tests...................................................... 17
3.3.5 Summary of geotechnical aspects ............................................................... 17
3.3.6 Conclusions .............................................................................................. 18
3.4 Floor Behaviour Monitoring .................................................................... 18
3.4.1 Durban Navigation Colliery ....................................................................... 18
3.4.2 Natal Anthracite Colliery ........................................................................... 19
3.4.3 Central Illinois .......................................................................................... 20
3.4.4 Conclusions .............................................................................................. 21
3.5 Pillar Failures .......................................................................................... 21
3.5.1 Swaziland and KwaZulu-Natal ................................................................... 22
3.5.2 Klipriver Coalfield .................................................................................... 22
3.5.3 Salamon and Munro collapsed cases ........................................................... 23
3.5.4 Sigma Colliery .......................................................................................... 24
vi
3.5.5 Colorado ................................................................................................... 25
3.5.6 Cooranbong Colliery ................................................................................. 26
3.5.7 Conclusions .............................................................................................. 27
3.5.8 General ..................................................................................................... 27
3.6 Design Procedures ................................................................................... 28
3.6.1 Numerical simulations ............................................................................... 28
3.6.2 Civil engineering approach ........................................................................ 30
3.6.3 General ..................................................................................................... 30
3.6.4 Conclusions .............................................................................................. 31
3.7 Main Findings of previous work on foundation stability ......................... 31
4 FAILED CASES .................................................................................................... 32
4.1 Emaswati Colliery, main haulage south ................................................... 33
4.1.1 Introduction .............................................................................................. 33
4.1.2 Geology .................................................................................................... 35
4.1.3 Geotechnical tests ..................................................................................... 35
4.1.4 Background ............................................................................................... 36
4.1.5 Post collapse observations ......................................................................... 40
4.1.6 Discussion ................................................................................................ 41
4.1.7 Conclusions .............................................................................................. 41
4.2 Matla 1, 5 seam, panel R14 South ............................................................ 42
4.2.1 Introduction .............................................................................................. 42
4.2.2 Geology .................................................................................................... 42
4.2.3 Geotechnical testing .................................................................................. 45
4.2.4 Post collapse observations ......................................................................... 46
4.2.5 SIMRAC investigation ............................................................................... 46
4.2.6 4 seam pillar collapse ................................................................................ 48
4.2.7 Conclusions .............................................................................................. 48
4.3 Welgedacht, Alfred seam stooping ........................................................... 49
4.3.1 Introduction .............................................................................................. 49
4.3.2 Geology and geotechnical data ................................................................... 49
4.3.3 Sequence of events .................................................................................... 51
4.3.4 Discussion ................................................................................................ 53
4.3.5 Conclusions .............................................................................................. 54
4.4 ZAC, Mngeni shaft, panel MEN1 ............................................................. 55
4.4.1 Introduction .............................................................................................. 55
4.4.2 Geology .................................................................................................... 55
4.4.3 Geotechnical testing .................................................................................. 57
4.4.4 Background ............................................................................................... 58
4.4.5 Observations ............................................................................................. 59
4.4.6 Remedial work .......................................................................................... 63
4.4.7 Conclusions .............................................................................................. 64
4.5 Other ZAC cases ...................................................................................... 65
4.5.1 North shaft, panel NW2N .......................................................................... 65
4.5.2 South shaft, section 3 ................................................................................ 66
4.5.3 South shaft, chequerboarding ..................................................................... 67
4.6 Conclusions from the study of the failed cases ........................................ 67
5 STABLE CASES ................................................................................................... 67
vii
5.1 Matla 1, 5 seam, Panel R14 West ............................................................. 68
5.2 Matla 2, 5 seam, Panel M12 North ........................................................... 68
5.3 ZAC, Mngeni, panel MN .......................................................................... 70
5.4 ZAC, Maye, poor roof case ...................................................................... 71
5.5 Conclusions of study of stable cases ........................................................ 72
6 NUMERICAL MODELLING ................................................................................ 72
6.1 Rock mass properties ............................................................................... 73
6.2 Three dimensional stress analysis ............................................................ 73
6.2.1 Emaswati, main haulage south ................................................................... 76
6.2.2 Matla 1, 5 seam, panel R14South ............................................................... 78
6.2.3 Welgedacht, Alfred seam stooping ............................................................. 80
6.2.4 ZAC, Mngeni, panel MEN1 ....................................................................... 81
6.2.5 Matla 1, 5 seam, R14West ......................................................................... 84
6.2.6 Matla 2, 5 seam, panel M12North .............................................................. 84
6.2.7 ZAC, Mngeni, panel MN ........................................................................... 86
6.2.8 Summary of MAP3D results ...................................................................... 88
6.2.9 Three-dimensional modelling conclusions .................................................. 88
6.3 Two dimensional stress analysis .............................................................. 89
6.3.1 Emaswati Colliery, Main haulage south ...................................................... 91
6.3.2 Matla 1, 5 seam, panel R14South ............................................................... 95
6.3.3 Welgedacht Colliery, Alfred seam stooping ................................................ 96
6.3.4 ZAC, Mngeni, panel MEN1 ..................................................................... 100
6.3.5 Matla 1, 5 seam, panel R14West .............................................................. 106
6.3.6 Matla 2, 5 seam, panel M12North ............................................................ 106
6.3.7 ZAC, Mngeni, panel MN ......................................................................... 108
6.3.8 ZAC, Maye, soft roof problem ................................................................. 111
6.3.9 Summary of two-dimensional modelling results ........................................ 112
6.4 Numerical modelling conclusions........................................................... 112
7 FLOOR BEARING CAPACITY .......................................................................... 114
7.1 Terzaghis Method ................................................................................. 114
7.2 Vesics Equation .................................................................................... 117
7.3 Conclusions for Terzaghis and Vesics methods ................................... 118
8 CASE STUDY COMPARISONS.......................................................................... 119
8.1 Identification of similarities .................................................................. 119
8.2 Case studies analysed by heave classification ........................................ 123
8.3 Case study findings ................................................................................ 125
9 SUMMARY OF FINDINGS OF THIS RESEARCH PROJECT .......................... 126
9.1 Previous work on foundation stability ................................................... 126
9.2 Case studies ........................................................................................... 126
viii
9.3 Numerical modelling .............................................................................. 127
9.4 Civil engineering based stability analyses ............................................. 127
10 STABILTY RATING ....................................................................................... 128
10.1 Discussion of findings ............................................................................ 128
10.2 Stability rating ...................................................................................... 132
10.3 Design procedure ................................................................................... 134
11 CONCLUSIONS .............................................................................................. 135
12 RECOMMENDED FURTHER WORK ............................................................ 136
13 REFERENCES ................................................................................................. 136
14 BIBLIOGRAPHY ............................................................................................ 142
APPENDIX A. GEOTECHNICAL TESTS .............................................................. a
Impact Splitting ...................................................................................................... a
Floor rock testing .................................................................................................... b
APPENDIX B. FAILED CASES .............................................................................. e
APPENDIX C. STABLE CASES .............................................................................. i
ix
LIST OF FIGURES

Figure 3.1. Type I floor heave. ................................................................................................ 5
Figure 3.2. Type II floor heave. .............................................................................................. 6
Figure 3.3. Position of horizontal stress induced roof and floor damage ................................... 9
Figure 3.4. Newcastle coalfield floor heave model .................................................................. 10
Figure 3.5. Shear failure of a foundation. ............................................................................... 12
Figure 3.6. Stress distribution in a pillar with a weak band at the top. ..................................... 13
Figure 3.7. Stress change in the floor of a longwall tailgate. ................................................... 19
Figure 3.8. Zones of horizontal and vertical floor movement. .................................................. 21
Figure 4.1. Emaswati, main haulage south, showing pillar condition ratings. ........................... 34
Figure 4.2. Section (E-W) through collapsed Area at Emaswati (after Petzer, 1991). ................ 36
Figure 4.3. Emaswati, convergence monitoring. ...................................................................... 40
Figure 4.4. Open topped and closed ridge types of floor heave. ............................................... 41
Figure 4.5. Matla 1, 5 seam pillar collapse area. ..................................................................... 43
Figure 4.6. Matla 1 - showing position of dolerite sill. ............................................................ 44
Figure 4.7. Two pillar sidewall failure mechanisms at Matla. .................................................. 47
Figure 4.8. Welgedacht, Alfred seam, section entrapment area. ............................................... 50
Figure 4.9. Welgedacht. Effect of weathering on bord width. .................................................. 51
Figure 4.10. Near vertical floor buckling mechanism. ............................................................. 53
Figure 4.11. Load distribution through a failing pillar (after Wagner, 1974). ........................... 54
Figure 4.12. Mngeni, pillar deterioration area. ........................................................................ 56
Figure 4.13. ZAC, typical chequerboard layouts. .................................................................... 58
Figure 4.14. Mngeni, deflection and ri de of survey pegs and position of convergence
monitoring stations. ....................................................................................................... 61
Figure 4.15. ZAC, Mngeni, convergence monitoring. .............................................................. 62
Figure 4.16. Mngeni floor failure mechanism (after Oldroyd, 2000). ....................................... 63
Figure 4.17. Mngeni pillar wrapping. ..................................................................................... 64
Figure 5.1. Portion of Maye poor roof area. ............................................................................ 72
Figure 6.1. MAP3D. Model geometry. Mngeni failed case. Perspective view. .......................... 75
Figure 6.2. MAP3D. Emaswati, Main haulage south, horizontal window
zz
. ........................... 77
Figure 6.3. MAP3D. Emaswati, Main haulage south, vertical window,
zz
. .............................. 77
Figure 6.4. MAP3D. Emaswati, Main haulage south, vertical window, U
t
. ............................... 78
Figure 6.5. MAP3D. Matla 1, 5 seam, panel R14South, horizontal window,
zz
........................ 79
Figure 6.6. MAP3D. Matla 1, 5 seam, panel R14South, vertical window,
zz.
........................... 79
Figure 6.7. MAP3D. Welgedacht, Alfred seam, horizontal window,
zz
. .................................. 80
x
Figure 6.8. MAP3D. Welgedacht, Alfred seam, vertical window,
zz
. ...................................... 81
Figure 6.9. MAP3D. ZAC, Mngeni, panel MEN1, horizontal window, step 1,
zz
. .................... 82
Figure 6.10. MAP3D. ZAC, Mngeni, panel MEN1, horizontal window, step2,
zz
. ................... 83
Figure 6.11. MAP3D. ZAC, Mngeni, panel MEN1, vertical window, step 2,
zz
. ...................... 83
Figure 6.12. MAP3D. ZAC, Mngeni, panel MEN1, vertical window, step 2, U
t
. ....................... 84
Figure 6.13. MAP3D. Matla 1, 5 seam, panel R14West, horizontal wi ndow,
zz
. ...................... 85
Figure 6.14. MAP3D. Matla 2, 5 seam, panel M12North, horizontal window,
zz
. .................... 86
Figure 6.15. MAP3D. ZAC, Mngeni, panel MN, before chequerboarding,
zz
. ......................... 87
Figure 6.16. MAP3D. ZAC, Mngeni, panel MN, after chequerboarding,
zz
. ............................ 87
Figure 6.17. PHASE
2
. Emaswati, Main haulage south,
1
(MPa), sensitivity analysis,
adjusted for percentage extraction. ................................................................................. 92
Figure 6.18. PHASE
2
1
(MPa),

using actual pillar width. ....... 93
Figure 6.19. PHASE
2
, Emaswati, Main haulage south,
3
(MPa). .............................................. 93
Figure 6.20. PHASE
2
, Emaswati, Main haulage south, U
t
(m). ................................................. 94
Figure 6.21. PHASE
2
. Emaswati, Main haulage south, U
t
(m), sensitivity analysis. ................... 94
Figure 6.22. PHASE
2
, Matla 1, 5 seam, panel R14South, (MPa). ............................................. 95
Figure 6.23. PHASE
2
, Matla 1, 5 seam, panel R14South,
3
(MPa). ......................................... 96
Figure 6.24. PHASE
2
, Welgedacht, Alfred seam,
1
(MPa), overview. ...................................... 97
Figure 6.25. PHASE
2
1
(MPa), road nearest goaf. .......................... 98
Figure 6.26. PHASE
2
, Welgedacht, Alfred seam, 1(MPa), first row of superimposed bords. ... 98
Figure 6.27. PHASE
2
, Welgedacht, Alfred seam, 3(MPa), road nearest goaf. ......................... 99
Figure 6.28. PHASE
2
, Welgedacht, Alfred seam, 3(MPa), first row of superimposed bords. ... 99
Figure 6.29. PHASE
2
. ZAC, Mngeni, panel MEN1,
1
(MPa), before chequerboarding. ............ 101
Figure 6.30. PHASE
2
1
(MPa), after chequerboarding. .............. 101
Figure 6.31. PHASE
2
. ZAC, Mngeni, panel MEN1, close-up of
1
(MPa), before
chequerboarding. .......................................................................................................... 102
Figure 6.32. PHASE
2
1
(MPa), after
Figure 6.33. PHASE
2
1
(MPa), 3 pillars away from
Figure 6.34. PHASE
2
3
(MPa), before chequerboarding. ............ 103
Figure 6.35. PHASE
2
3
(MPa), after chequerboarding. .............. 104
Figure 6.36. PHASE
2
3
Figure 6.37. PHASE
2
. ZAC, Mngeni, panel MEN1, U
t
(m), after chequerboarding. .................. 105
xi
Figure 6.38. PHASE
2
t
(m), 3 pillars away from
Figure 6.39. PHASE
2
. Matla 1, 5 seam, panel R14West,
1
(MPa). ......................................... 106
Figure 6.40. PHASE
2
3
(MPa). ......................................... 107
Figure 6.41. PHASE
2
. Matla 2, 5 seam, panel M12North,
1
(MPa). ....................................... 107
Figure 6.42. PHASE
2
3
(MPa). ....................................... 108
Figure 6.43. PHASE
2
. ZAC, Mngeni, panel MN,
1
(MPa), before chequerboarding. ................ 109
Figure 6.44. PHASE
2
1
(MPa), after chequerboarding. .................. 109
Figure 6.45. PHASE
2
3
(MPa), before chequerboarding. ............... 110
Figure 6.46. PHASE
2
3
(MPa), after chequerboarding. .................. 110
Figure 6.47. PHASE
2
. ZAC, Maye, soft roof,
1
(MPa). .......................................................... 111
Figure 6.48. PHASE
2
3
(MPa). .......................................................... 112
Figure 7.1. Foundation bearing capacity factors. .................................................................... 115
Figure 7.2. Determination of cohesive strength. ..................................................................... 116
Figure 8.1. Relationship between slake durability and class of floor heave. ............................ 125
Figure 10.1. Relationships plotted for stable and failed cases. ................................................ 129
Figure 10.2. Safety factor vs. mining height. ......................................................................... 129
Figure 10.3. safety factor vs. panel width. ............................................................................. 130
Figure 10.4. Safety factor vs. seam dip. ................................................................................. 130
Figure 10.5. Safety factor vs. roof impact splitting rating. ..................................................... 131
Figure 10.6. Safety factor vs. average weighted slake durability rating. .................................. 131
Figure 10.7. Safety factor and panel width plotted against FSR. ............................................. 133
Figure 10.8. Floor Stability Rating design chart. .................................................................... 134

xii
LIST OF TABLES

Table 3.1. Sedimentary facies, strength properties and underground behaviour. ........................ 4
Table 3.2. Newcastle coal field; mining geometries and floor properties. ................................ 10
Table 3.3. Comparison between Klipriver and other failed cases. ............................................ 22
Table 3.4. Dimensions of some pillar collapse cases used by Salamon and Munro (1967). ........ 24
Table 3.5. Sigma Colliery failure compared to case studies. .................................................... 25
Table 3.6. Colorado failure compared to case studies .............................................................. 25
Table 3.7. Cooranbong panel North B compared to case studies. ............................................. 27
Table 4.1. Emaswati Colliery, Roof and floor ratings. ............................................................. 37
Table 4.2. Pillar condition rating. ........................................................................................... 39
Table 4.3. Matla 1, 5 seam, roof and floor ratings. .................................................................. 45
Table 4.4. Pillar damage rating. ............................................................................................. 47
Table 4.5. Welgedacht, Alfred floor, geotechnical results. ...................................................... 51
Table 4.6. Mngeni, geotechnical testing. ................................................................................ 57
Table 5.1. Matla 1, panel R14West, floor lithology. ................................................................ 68
Table 5.2. Matla 2, 5 seam, geotechnical results. .................................................................... 69
Table 6.1. Summary of rock mass properties .......................................................................... 74
Table 6.2. MAP3D Material properties. .................................................................................. 74
Table 6.3. MAP3D horizontal to vertical stress ratios. ............................................................ 74
Table 6.4. Summary of MAP3D results. ................................................................................. 89
Table 6.5. UCS from point load (MPa). .................................................................................. 90
Table 6.6. PHASE
2
transversely isotropic material. ................................................................. 90
Table 6.7. PHASE
2
isotropic materials. ................................................................................ 91
Table 6.8. Summary of PHASE
2
results. ................................................................................ 113
Table 7.1. Design rock mass strength for failed and stable cases. ........................................... 115
Table 7.2. Foundation strength by Terzaghis method. ........................................................... 116
Table 7.3. Heave factors for failed and stable cases from Vesic's modified formula ................ 118
Table 8.1. Summary of failed cases. ...................................................................................... 120
Table 8.2. Summary of stable cases. ...................................................................................... 121
Table 8.3. Failed vs. stable cases comparison. ....................................................................... 122
Table 8.4. Relationship between slake durability and floor heave type. ................................... 124
Table 8.5. Case studies compared by heave classification. ..................................................... 124
Table 10.1. Floor Stability Rating matrix. ............................................................................. 132
Table 10.2. Floor Stability Rating values for case studies. ..................................................... 133

xiii

LIST OF SYMBOLS

A Dry mass before slaking (g)
b Bord width (m)
C Dry mass after two slaking cycles (g)
c Pillar centre distance (m)
C
o
Unconfined

compressive strength (MPa)
d
30
Swelling displacement after 30 minutes (mm)
fs Fracture spacing (impact splitting) (m)
hr Mean unit height into roof (impact splitting) (m)
I
d2
Slake durability index
L Initial length of sample (mm)
N
r
& N
q
Bearing capacity factors
Angle of internal friction ()
P
u
Ultimate load applied over width B (MN/m)
r
u
Unit rating (impact splitting)
r
w
Weighted rating (impact splitting)
S
30
Swelling strain after 30 minutes
S
o
Unconfined shear strength (MPa)
T
o
Indirect tensile strength (MPa)
tu Thickness of unit (impact splitting) (m)
U
t
Total displacement
w Pillar width (m)
W Weight density (units)
w/h Pillar width to height ratio
1
Major Principal stress
3
Minor Principal stress
zz
Vertical stress
x
/

Principal stress vector in the horizontal or near horizontal plane

xiv
LIST OF ABBREVIATIONS

ACIRL Australian Coal Industry Research Laboratory
APL Axial point load index
BDS Brazilian disk strength
CM Continuous miner
CMRR Coal Mine Roof Rating
COMRO Chamber of Mines Research Organisation
CSIR Council for Scientific and Industrial Research
DNC Durban Navigation Colliery
DPL Diametral point load index
DRMS Design Rock Mass strength
DS Duncan free swell
DTS Direct tensile strength
FEM Finite element modelling
FR Floor rating
FSR Floor stability rating
GENROC Gencor Rock Engineering Department
HF Heave factor
IRED Ingwe Rock Engineering Department
IS Impact splitting
MBC Mine-floor bearing capacity
MC Natural moisture content
MSHA United States Mine Safety and Health Administration
NIOSH National Institute of Safety and Health in the USA
PLI Point Load Index
PS Pillar stress
RQD Rock Quality Designation
SAIMM South African Institute of Mining and Metallurgy
SANGORM South African National Group on Rock Mechanics
SD Slake durability
SIMRAC Safety in Mines Research Advisory Committee
SSCR Steady state creep rate
UBC Ultimate bearing capacity
UCS Uniaxial Compressive Strength
USBM United States Bureau of Mines
UWF Underclay water factor
ZAC Zululand Anthracite Colliery

1
1 INTRODUCTION

The widely applied coal pillar design formula derived by Salamon and Munro (1967) has
been used by and large successfully over the past 30 years in the design of bord and pillar
workings in southern African collieries. However, over the years there have been several
cases in which workings with relatively high factor of safety pillars have failed and the
presence of soft floor strata has been judged to be a contributory or controlling factor. The
most newsworthy of these was the failure at Emaswati Colliery, Swaziland, in 1991 which
resulted in the temporary entrapment of 26 workers.

In order to avoid the problem in future there is a need to enable rock engineers to identify
potential floor stability problems and adjust their coal pillar design accordingly. To assist
in this, cases in which soft floor strata were associated with pillar failure are analysed in
this project report together with cases in which the pillars remained stable even though soft
floor was present. One case in which it was considered that the pillars were punching into
weak roof strata is also studied.

2 THE GOAL OF THIS PROJECT

It is intended that this project will improve understanding of pillar failures associated with
soft floor rocks in southern African collieries and enable the practical experience gained
from the case studies to be used to improve the design of pillars standing on weak
foundations.

The approach taken in this project report, to achieve its goal of producing design guidelines
that will reduce the likelihood of pillar foundation failures, is outlined below:
o Review of both local and overseas papers and reports.
o Review of at least two accepted civil engineering foundation design
methods to assess their relevance to the solution of the problem.
o Investigation of mining geometries, geology, convergence data and any
anecdotal evidence for selected case studies of both failed and stable cases.
o Geotechnical testing of floor samples obtained close to failed areas (where
possible).
o Numerical modelling using both the two-dimensional PHASE
2
and three-
dimensional MAP3D programmes.
o Identification, if possible, of geotechnical areas in Southern Africa in
which floor failures are most common.
o Production of pillar foundation design guidelines and definition of
limitations to their use.
o Identification of future work to be carried out on the subject.

3 INVESTIGATION INTO PREVIOUS WORK ON FOUNDATION
STABILITY

Several local and overseas reports and papers containing references to cases of pillar
failures in which soft floor strata were present have been studied. In addition, individuals
with experience in coal mine floor stability problems, both local and overseas, were
2
approached for any anecdotal evidence or copies of any relevant publications in their
possession. The information gathered during this review is from South Africa, Australia
and the United States. No comparative data were found from other major coal mining
countries such as the United Kingdom, Germany, India, China or Russia.

An Internet search was also conducted but produced few results.

The following aspects were considered:
o Geological factors
o Floor heave mechanisms
o Geotechnical testing
o Site investigations
o Pillar failure
o Design procedures

These are dealt with in turn in the sections to follow.

3.1 Geological factors

In the following section certain geological factors are considered. These include seams
which have been found to be prone to floor stability problems, the influence of clay
minerals on floor stability and the characterisation of seam floor strata.

3.1.1 Seams prone to floor problems

Problems caused by weak floor during secondary extraction in South African collieries,
mentioned by Fauconnier and Kersten (1982), are:
o The 5 seam of the Springs / Witbank coalfield has, locally, a laminated micaceous
sandstone floor which presents heaving problems. In other places a soft, highly
micaceous siltstone floor causes mining problems because of its slippery and
friable nature.
o In the KwaZulu-Natal coalfields, weak floor and heaving of sediments could
destroy access to mined areas making it impossible to stoop.

Hill (1989) briefly mentions that floor creep was experienced on the Gus seam at Vryheid
Coronation Colliery where the roof was massive sandstone and the parting was 1 to 4m
thick. This used to result in poor roof conditions in the underlying Uppe r Dundas and to a
lesser extent in the Lower Dundas.

Madden and Wagner (1985) introduced a pillar rating system for South African coal mines
which includes punching of the roof and floor as one of three parameters considered most
important. However, when it came to assessing the strength of various seams, pillar
punching was one of the aspects specifically ignored as its contribution was nearly equal
for all seven seams studied. This is significant in that it indicates that of the seven seams,
none were considered to have a clear propensity to foundation failure. The individual seams
are not clearly identified in this paper (they are simply called seams I through to VII) and
so it is impossible to determine whether any of the seams covered in the case studi es (see
sections 4 and 5) were part of this study.
3
3.1.2 Clay minerals

The definition of clay minerals by Whitten and Brooks (1972) is worthy of mention. It
states, inter alia, that: An important characteristic is their ability to lose or take up water
according to the temperature and amount of water present in a system. And notes further
that: The montmorillonite group is especially notable for the way in which it takes up and
loses water. Montmorillonites are formed by the alteration of basic rocks, or other silicates
low in K, under alkaline conditions.

These clay minerals are considered to be influential in the swelling floor of the 5 seam at
Matla 1 (Section 4.2).

3.1.3 Facies characterisation

Jermy and Ward (1988) consider that the UCS, while being easy to determine, provides
very little indication of potential roof conditions. It can however be used to understand
floor behaviour and in designing support pillars. Sedimentary facies likely to be
problematical, when encountered in the floor, are shown in Table 3.1, which is an extract
from Jermy and Wards table of 24 different facies.

There is a wide scatter of strengths for all the facies with the shaly rock being as strong and
in some cases stronger than the sandy facies. The medium and coarse -grained sandstones
have some of the lower UCS results. Jermy and Ward (1988) found that UCS and tensile
strengths are significantly reduced by increasing moisture content. No relationship was
found to be immediately apparent between the UCS of rock and its underground behaviour.

Facies 1 has reportedly the worst geodurability, a similar facies type in northern KwaZulu-
Natal (coal field and seam not specified) decomposes rapidly on exposure, resulting in roof
and floor instability. Jermy and Ward (1988) conclude that: The UCS of the facies will be
of importance when considering heave associated with punching of pillars into the floor,
calculating the size of mining pillars and in the design of underground development in
general. Low durability facies appear to be more prone to floor heave than the more durable
types. Facies that exhibit poor durability and low t ensile strengths are considered very
unstable under mining conditions.

3.1.4 Conclusions

The literature indicates that certain seams have floors more likely to fail than others, the 5
seam of the Witbank, and Highveld coalfields as well as unspecified seams i n KwaZulu-
Natal being especially prone. This is borne out in the findings of the case studies contained
in Sections 4 and 5.

Sandier facies generally form weaker pillar foundations than those consisting of shaly
material, due to the sandier strata being able to absorb more moisture. The relationship
between moisture content and floor stability is explored in more detail in Section 7.2.

4
Table 3.1. Sedimentary facies, strength properties and underground behaviour.
Facies Description Strength
(MPa)*
Properties of rock strata
underground
1 Massive dark grey to black
carbonaceous silt stone.
BDS 5 8
DTS 0.2 0.4
DPL 0.3
APL 2 2.4
UCS 45 - 75

Very poor roof and floor strata due to
low tensile strength and poor
durability. Deteriorates rapidly on
exposure. Roof falls common and
floor heave occurs when depth of
mining exceeds 150m
2 Lenticular-bedded siltstone with
discontinuous ripple cross
lamination.
BDS 7 - 10
DTS 0.3 0.4
DPL 0.5
APL 2 3.5
UCS 60 - 75
3 Alternation of 1cm thick layers of
flat laminated siltstone and fine-
grained sandstone.
BDS 6 - 9
DTS 0.2 0.4
DPL 0 - 1
APL 1 4.5
UCS 80 - 105
11 Massive coarse grained white
feldspathic sandstone.
BDS 3 - 5
DTS 0.8 1.1
DPL 1.5 2.5
APL 1.8 2.5
UCS 45 - 55
Good roof and floor strata.
Decomposes under prolonged
saturation giving rise to stability
problems.
12 Coarse grained and white
feldspathic sandstone with planar /
trough crossbeds.

13 Intensely bioturbated carbonaceous
siltstone or fine-grained sandstone.
BDS 7 - 9
DTS 0.2 2.3
DPL 3.5
UCS 55 - 70
Deteriorates rapidly upon exposure
and saturation to give roof and floor
stability problems.
* BDS = Brazilian disk strength / DTS = Direct tensile strength / DPL = Diametral point load
APL = Axial point load / UCS = Uniaxial compressive strength. Strength quoted is the central 50% of
the data values.

3.2 Heave Mechanisms

To assist with understanding the mechanics of floor failures a number of different floor
heave mechanisms have been considered.

3.2.1 Buckling of floor beam and plastic flow modes

Wuest (1992) identified two major types of floor heave, namely, buckling of the floor beam
and plastic flow of floor material.

A floor made up of stronger individual beams would be more likely to break off near the
pillar edge under load while a relatively thicker beam of weaker rock would tend to flow
out from beneath the pillar. Identification of these mechanisms is made more complicated
in that they can often exist together.

5
Applying civil engineering concepts to coal mine floors (Wuest, 1992) is possible but the
more complex loading system and far greater area of floor exposed need to be borne in
mind.

Two types of floor heave, generally similar to those identified by Wuest (1992), were
analysed by Peng, Wang and Tsang (1995) at two mines in West Virginia. The mechanisms
of the two types are as follows:

Type I

The floor consists of several thin layers of relatively strong rock. The sequence of floor
failure is:
o On development, where vertical stress is not high enough to break the floor, the
horizontal stress will not be able to cause heave.
o During retreat pillaring (stooping) the vertical stress increases. This means a build
up of shear stresses on the bedding planes. Shear failure occurs first on bedding
planes as they have the lowest shear strength.
o The shear failure of the bedding planes changes the floor from a composite beam
into several individual beams (A in Figure 3.1). Tensile failure may then occur as a
result of redistributed tensile stresses in the individual beams (B).
o Vertical stress is relieved by the tensile and shear failures mentioned above. If high
horizontal stress exists this will cause buckling of the broken and separated
individual beams (C).

Figure 3.1. Type I floor heave.
C
A B
Soft floor
More
competent
lower floor
6
o Most floor heaves of this type occur during secondary extraction (retreat pillaring
or stooping) when the abutment stresses on the out -bye pillars cause the floor to
fail. More importantly, the horizontal stress causes the broken floor to heave up.

Type II

The immediate floor consists of massive soft rock such as mudstone. The vertical stress
exceeds the floor bearing capacity. The sequence of failure is:
o The floor fails once the vertical stress exceeds the bearing capacity.
o After failure, the bearing capacity is reduced to the residual value which remains
nearly constant while the pillar keeps on punching into the floor.
o Due to pillar punching, the roof above the pillars deflects. The magnitude of heave
is dependant on the roof deflection above the pillar.

Figure 3.2. Type II floor heave.
B

Broken zone
A

Residual
bearing
capacity

7
For control of floor heave Peng et al, 1995, recommend that:
o Type I floor heave is best handled by designing the pillar such that tensile stresses
in the floor and shear stresses on the bedding planes are less than the tensile and
shear strengths of the floor and bedding planes respectively.
o The decision must be made as to whether any degree of type II heave is acceptable.
If none is allowed then the floor load must be kept below the floor bearing
capacity, or alternatively, some controlled heave is allowed for in the design stage.

Two control techniques were put forward:
o Cut a vertical slot in the floor to relieve the stored horizontal stress, the main cause
of type I heave.
o Floor bolting to prevent type I heave by forming a composite floor beam and to
prevent failed floor material from intruding into the entry (type II heave).
Mechanical bolting was tried and found to be not ef fective. Floor trusses, with or
without a centre floor cut, are effective for floor heave control.

3.2.2 Squeezing of weak floor rock

In a case study of floor heave in the Zeigler mines near Murdock, Illinois, Speck (1981)
attempted to develop procedures for the detection of potential floor heave problems during
the exploration phase of mine design. These mines lie between 45.7 and 76.2m (150 to
250ft) deep and the seam is 1.8 to 2.1m (6 to 7ft) thick.
Speck (1981) found that underclay being squeezed from under the coal pillar formed bulges
of displaced material along the pillar edges (type II heave). Another result of this squeezing
was that the outward moving underclay often caused the pillar edges to be dragged outward
and break in tension, or rash.

While a large variation in strengths was measured for floor lithologies, the UCS best
correlated with natural moisture content.

The closer the stronger floor units are to the coal pillar the more resistant the floor will be
to floor heave. The presence of swelling clays was not found to play a significant role in
floor heave on these mines.

Due to the in situ underclay having many randomly orientated fissures causing core loss
and breakage during drilling, laboratory tests were biased towards material with hi gher
strengths. However the following test results were obtained for the soft underclay at the
Ziegler mines:
o The majority of underclay samples had triaxial strengths of between 2.7 and
6.9MPa (400 to 1000psi).
o In situ underclay compressive strength was found by means of plate bearing tests to
be 57 to 85% lower than laboratory results.

Speck (1981) then went on to quote a modification to Vesics foundation strength equation
(Vesic, 1975) and concluded that: It therefore appears that the heave factors derived
from the classic bearing capacity equation are capable of predicting mine floor stability
when only underclay water content and thickness and mine depth are known.

8
This hypothesis and the modified equation have been tested for the South African case
study sites and found to predict floor failure or stability in seven of the eight case studies
(this is dealt with in more detail in Section 7.2).

Covering much of the same ground as Speck (1981), Stephenson and Rockaway (1981),
also studied floor heave at the Zeigler mines. They state that as the floor material beneath
the pillar fails the pillar moves down, transferring its load to the adjacent pillars. If the
adjacent pillar foundations are only marginall y stable they may then also be overloaded.
This could lead to the progressive spread of floor heave (not pillar failure) through a large
area. Stephenson and Rockaway (1981) conclude that:
o Underclay thickness up to 1.8m (6ft) causes significant changes in bearing
capacity. For underclay thicker than 1.8m this effect is reduced.
o Floor heave occurred at sites where the natural moisture content of the underclay
was highest.
o The ratio of claystone (the stronger unit underlying the underclay) to underclay
strength, is also an important factor contributing to floor bearing capacity.
o No single set of data at any site will always predict floor behaviour.

3.2.3 Floor failure due to horizontal stress

Gale (1991) maintains that high horizontal stresses cause the floor to fail in tension
manifested by a vertical crack starting at the floor surface. The cracks may start at either
the centre of the roadway (symmetrical loading) or closer to the pillar edge (possible
differential loading). The magnitude of the horizontal stres s may be much higher in one
particular direction.

The effect of mining direction directly influences the effect of high horizontal stress, with
roads parallel to it being least affected while those perpendicular to it may experience roof
gutter and floor failure. The locations of such failures are largely influenced by their
position in relation to the direction of the principal horizontal stress. Figure 3.3 (Strata
Control Technology, 1993) illustrates the posit ion of roof and floor damage with respect to
the principal horizontal stress direction.

3.2.4 Coal pillar shearing through weak floor beam

Floor heave problems in the Newcastle coalfield, Australia, were studied by Seedsman and
Gordon (1991) at three mines, namely, Chain Valley Colliery, Cooranbong Colliery and
Wyee State Mine. Mining geometries and floor conditions are summarised in Table 3.2.
Some of their findings are that:
o The depth of clay extrusion from beneath pi llars is limited to 1 2m.
o The claystone had an undrained elastic modulus of 350 to 750MPa, which is about
50% of laboratory values. The drained modulus is estimated to be in the order of
200 to 300MPa.
o Heave is considered to be due to local bearing capacity failure in the rib spall zone
(pillar edge area) with the edge of the pillar core shearing through the weak beam.
The floor beam is then eccentrically loaded over a distance of 1 to 2m. This
mechanism is sketched in Figure 3.4.
9
o Foundation engineering principles can be applied to the study of weak floors in
coal mines and pillar settlement can be analysed using simple elastic theory.
Modulus values for claystone are an order of magnitude lower than for typical co al
floor rocks.
o In agreement with findings in the Illinois coal fields, referred to in section 3.2.2,
moisture content may be used to predict the UCS and elastic modulus.
o Due to the high safety factors of the pil lars at Cooranbong and Chain Valley,
Seedsman and Gordon (1991) considers that it was unlikely that the spalling was
related to the onset of pillar failure. In the case of Wyee it would appear that the
pillars were under designed when the overburden of the overlying Great Northern
seam was active. Longwalling in the Fassifern seam caused the remnant pillars to
fail in the Great Northern seam.
o For the generally low aspect ratios of the thin claystones seen in the Newcastle coal
field it is unlikely that lateral flow of claystone would tear the pillars apart.

Figure 3.3. Position of horizontal stress induced roof and floor damage

GOOD CONDITIONS
BAD CONDITIONS (Sag)
GUTTERING ON LEFT
HAND SIDE
GUTTERING ON RIGHT
HAND SIDE
Major horizontal stress (MHS)
MHS
MHS MHS
MHS
MHS
Direction
of
driveage
Direction
of
driveage
Direction
of
driveage
Direction of driveage
Stress concentration



10

Table 3.2. Newcastle coal field; mining geometries and floor properties.
Chain Valley Cooranbong Wyee
Cover depth (m) 140 110 - 120 180
Mining height (m) 2.3 2.6 2.9 3.1
Pillar centres (m) 24 x 30 27 x 95 30 x 95
Bord width (m) 5.5 5.5 5.5
Pillar width- of narrowest
side (m)
18.5 21.5 24.5
Extraction (%) 37 25 23
Pillar safety factor
(Salamon)
3.03 4.89 3.13
Pillar width to height ratio
(narrowest side)
8.04 8.27 7.90 8.44
Soft floor thickness (m) 1.08m with 0.3m
coal capping
2.8m with 0.3m coal
capping
0.5m with 1.2m
coal capping
Soft floor strength (MPa) 1.3 23.4 8 - 53 32.7
Maximum floor heave (mm) 280 >300 550
Horizontal floor movement
(mm)
85 34 Not measured

Figure 3.4. Newcastle coalfield floor heave model

Effective
footing
stresses
Roadway
Compressive failure
Rib compression
Abutment zone
Peak stress
Pillar shearing through floor beam.
Zone of rib
spall
Floor coal beam
Claystone
11
3.2.5 Swelling floor strata

A floor heave database was compiled by Hill, Gordon and Madden (1994) in which,
contrary to findings detailed elsewhere (Section 2.3), the claystone mineralogy, in
particular the presence of smectite (swelling clay), is found to be of great significance in
determining floor behaviour in that it determines the propensity of the floor rock to swell.

Hill et al (1994) postulate a new model for floor heave, at least in the low stress
environment of the first workings at Cooranbong. In this model floor heave is primarily due
to unloading and subsequent expansion of the claystone. Pillar punching is thought not to
be a significant mechanism.

The model offers explanations for the following phenomena:
o Floor heave greatly exceeds the vertical displacement of subsidence to the extent
that heave and subsidence cannot be rationally explained in terms of lateral
extrusion of claystone from beneath pillars.
o Bearing capacity concepts borrowed from civil engineering fail to accurately
predict the magnitude of heave.
o Floor heave tends to develop very slowly, whereas stress changes due to mining are
usually quick.
o Dinting of roadways is often not effective in controlling heave.

This model is not applicable within the context of this study as the pillars are not
considered to have any great influence on the floor heave.

3.2.6 Bearing capacity failure around pillar edges

Madden et al (1995) state that foundation failure can take a number of forms depending o n
the strength, thickness and location of the weak stratum within the roof or floor horizons. A
case from Galvin (1995) is included in which, if the floor material is only marginally
weaker than the pillar, high loads may be generated in the pillar prior t o the onset of
failure. Bearing capacity failure then develops around the pillar edges due to:
o Peak pillar loads occurring close to the pillar edge.
o Floor strata being weakened by the removal of vertical confinement.
o The floor strata close to the pillar being free to fail in shear after removal of the
vertical confinement.
o As the shear strength of coal measures rocks is typically only half their
compressive strength, the intact floor strata may fail, resulting in floor heave close
to the pillar edges. Blocks of floor strata then rotate out from under the pillar edges
into the roadway (Figure 3.5).

Loss of bearing capability around pillar edges transfers the load previously supported by
the failed foundation further into the pillar and pillar strength is reduced due to the
reduction of surface area of the end constraints.
There are three factors, according to Madden et al (1995), which interact to determine
whether stability can be re-established, these are:
o Confinement building up under the pillar as failure progresses being sufficient to
arrest the foundation failure.
12
o Increase in pillar load and reduction in strength, being sufficient to induce pillar
failure.
o Whether the effect of increasing bord width, leading to reduced roof stability and in
turn taller, less stable pillars is sufficient to cause pillar failure.
The last two factors mentioned above indicate accelerated deterioration and are not thought
to lead to the re-establishment of stability.

In strata of moderate or higher shear strength, bearing capacity failure tends to progress
gradually, with resistance to the process building up as it progresses. Unless the situation is
one of deadweight loading, the load required to drive the process is not always avai lable. In
the mining situation the load input is generally governed by displacement of the roof strata,
with the roof stiffness controlling the rate of loading of the pillar system. With increased
pillar width, greater confinement is provided to the foundation with an increased
probability of arresting failure.

3.2.7 Hard rock mines

Brady and Brown (1985) state that: Where hangingwall and footwall rocks are
weak relative to the orebody rock, a pillar support system may fail by punching of pillars
into the orebody peripheral rock. The mode of failure is analogous to bearing capacity of a
foundation, and may be analysed in a similar way. This type of local response will be
accompanied by heave of floor rock adjacent to the pillar lines, or extensive fretting and
collapse of roof rock around a pillar.

Figure 3.5. Shear failure of a foundation.

Load
distribution
in a pillar
Shear failure
of foundation
13
They quote a set of equations to calculate the bearing capacity of a cohesive, frictional
material such as soft rock where the angle of friction and unit weight of the floor rock is
known.

Coates (1970) shows the stress distribution in a pillar with a weak band at the top ( Figure
3.6). This geology results in transverse tension at the top of the pillar and in addition, the
compressive stresses increase towards the pillar edge near the base of the pillar. At mid -
height the axial stress is slightly higher towards the pillar centre while near the floor the
highest axial stress is towards the pillar edge. If this model is inverted a weak floor
scenario is represented.

Figure 3.6. Stress distribution in a pillar with a weak band at the top.

3.2.8 Summary of floor failure mechanisms

Floor beam buckling (type I floor heave) is usually confined to stronger floor bands and is
often associated with secondary extraction or high horizontal stresses. Plastic flow (type II
floor heave), on the other hand, i s usually associated with weak rock such as claystone. The
failing floor has been documented as dragging the pillar sides outwards and breaking them
in tension.

Another model, developed for the Newcastle coalfield in Australia, has the floor beam
being loaded eccentrically by 1 to 2m of the pillar edge while the core of the pillar shears

Axial stress
at pillar
height
Axial stress
at pillar
height
Transverse stress
near centreline
Weak layer
14
through the weak floor beam. This failure mode appears to contain certain aspects of the
two types mentioned above.

It is often difficult to differentiate between the var ious mechanisms as they often work in
conjunction.

High horizontal stresses have also been reported as causing the floor to fail in tension.
While basic analysis using civil engineering concepts is possible, it must be recognised that
the underground loading system is more complicated.

A new floor heave model put forward for low stress environments in Australia suggests that
it is primarily due to unloading and subsequent expansion of claystone and not pillar
punching.
Where the natural water content of t he immediate floor (underclay) is highest, heave is
most likely to occur.

3.2.9 Conclusions

Australian and United States experience indicates that while pillar settlement is common in
places, pillar failure is more likely to occur in southern Africa. This is probably due to the
generally more conservative pillar designs practiced in these two overseas countries.

Pillar settlement and floor heave as experienced elsewhere are more of economic
consideration than a safety hazard as they are responsible for surface subsidence and
closure of underground roadways. In southern Africa, pillar failures as a result of
foundation instability can have significant economic consequences but they also potentially
pose a greater threat to the safety of underground employees.

3.3 Geotechnical Testing

From the literature studied, four distinct approaches to analysing floor performance have
been identified. The first of these is to determine ultimate bearing capacity by correlating
laboratory and engineering index properties with plate loading tests. The second approach
is to determine the swelling potential of the floor rock from a battery of soil engineering
tests. In the third approach, the coalmine roof rating system (CMRR) is modified to allow
coal mine floors to be classified. The fourth approach is to carry out modified versions of
the slake durability and Duncan free swell tests, these being modified to allow testing of
weaker strata than would be possible otherwise (Annexure A).

3.3.1 Ultimate bearing capacity

Numerous tests were carried out at six mines in central and southern Illinois by Chugh, Atri
and Dougherty (1988) to determine which parameters could best be used to estimate the
ultimate bearing capacity (UBC) and deformation properties of immediate floor strata.
These included:
15

o Borehole shear tests to obtain unconfined shear strength ( S
o
) and angle of internal
friction ()
o Plate loading tests of immediate floor strata, under as -mined and soaked-wet
conditions, to obtain UBC and deformation modulus at 50% and 90% of UBC
(DM
50
and DM
90
). These tests were done in freshly mined areas with 152, 203 and
305mm (6, 8 and 12in) square plates.
o Engineering index properties, which included: natural moisture content ( MC),
density, particle size analysis, Atterberg limits, clay mineralogy, axial swelling
strain, and slake durability.
o Laboratory strength and deformation properties which included unconfined
compressive strength (C
o
) and modulus of elasticity at 50% of C
o
(E
50
), Point-Load
Index across bedding planes (PLI) and indirect tensile strength (T
o
).

Chugh et al come to the following conclusions and recommendations:
o UBC under test plates cannot be estimated with significant confidence from
unconfined compressive strength data. This approach generally overestimates the
UBC.
o UBC underneath test plates can be estimated with significant confidence from T
o

and values obtained from borehole shear tests. UBC is slightly underestimated
with this approach. In both this and the above conclusion, no mention was made of
adjusting the result either up or down to better reflect the UBC under test plates.
o Angle of internal friction, , is a very important parameter to enable accurate
estimation of UBC under plates or full sized pillars.
o Axial Point Load Index (PLI) does not correlate as well as T
o
with engineering
index properties or strength-deformation properties of floor strata.
o Rock Quality Designation (RQD) does not correlate well with in situ strength-
deformation properties.
o The determination of compressive strength of underclay as proposed by Speck
(1981) may not be valid when the moisture content is less than 6%.
o The following simple engineering tests may be used to characterise floor strata:
natural moisture content, indirect tensile strength, Atterberg Limits and axial
swelling strain. The first two should be conducted at 150mm intervals while the last
two should be at 300mm intervals.
o It is not necessary to determine UCS, clay mineralogy or clay size fractions of the
immediate floor strata.
o An angle of internal friction of 20 may be used with confidence for calculation
of UBC instead of =0 as the probability of being less than 20 was found to be
only about 16%. This statement contradicts the earlier one where the angle of
internal friction is identified as an important paramet er which must be determined
to allow accurate estimates of UBC to be made.
o Cohesive strength of the floor should be estimated from T
o
and values rather than
from C
o
and values when UBC is estimated by Vesics equation.

It will be necessary for plate l oading apparatus to become locally available before the
above method can be verified for southern African conditions.

16
3.3.2 Swelling potential

Shakoor and Sarman (1992) carried out engineering tests on mudrock samples collected in
many different states of t he United States. The properties determined are as follows: grain
size distribution, clay content, clay type, texture, percent absorption, Atterberg limits,
specific gravity of solids, dry density, void ratio, second cycle slake durability index, UCS,
volumetric increase and swelling pressure. The percentage of 0.005mm clay was used to
classify mudrocks into: claystone (>66% clay), mudstone (33-66% clay), mudshales (33-
66% clay, laminated) and siltstones (<33% clay). They conclude that:
o Excessive swelling in mudrocks is not confined to those containing highly
expansive clay minerals. Mudshales, for example, exhibited the lowest adsorption
rates even though some contain expandable clay minerals.
o No single property can be used to predict swelling potential. A correlation (r =
0.84) was found to exist between swelling pressure and Volumetric increase.
o Mudrocks must be categorised on the basis of their geological characteristics (e.g.
percentage of 0.005mm clay) for adequate evaluation of swelling potential.
o Absorption, slake durability, void ratio, UCS and plasticity index, in varying
combinations, can be used as predictive tools for swelling potential.

The propensity for a particular floor rock unit to swell and by how much, is considered
more important than being able to quantify precisely why swelling takes place. Duncan free
swell tests are occasionally carried out on core with a perceived tendency to swell.
However this has only been done for one of the eight case study sites (Section 4.1) and
consequently the link between floor swelling and foundation strength cannot be determined
at this stage.

Hill (1996) reports work carried out on coal mine shales by Eeckhout in the United States
who concludes that the strength of the shales is severely reduced by humidity fluctuations.
In an environment with humidity lower than 60% the shale sample strength remains
relatively high while expansion and contraction are kept to a minimum. This implies that
where the humidity exceeds 60% shale would be susceptible to weakening due to expansion
and contraction. Hill quotes measurements taken in South African Collieries in 1995 which
show humidity readings varying from 80% at shaft bottom to 100% in the working section.

Hill further states that: Although foundation failure is more common at depth due to
higher pillar stress, it has occurred at depths of less than 40m with resultant pillar failure.
Foundation stability can be evaluated by examining factors such as thickness, strength and
durability of the immediate floor, however this is a complex mechanism which requires
further study.

3.3.3 CMRR based floor rating

The well-known coal mine roof rating (CMRR) developed in the United States by Molinda
and Mark (1994) was modified by Riefenberg (1995) to allow classification of coal mine
floors.

The CMRR is based largely on the premise that geological discontinuities largely determine
the structural integrity of a coal mine roof. Other geotechnical parameters used in
determining the CMRR are UCS and moisture sensitivity. Adjustments are also made to
take into account the presence of a strong beam within the bolted horizon, the condition of
unit contacts and ground water. A strength of the method is that the CMRR can be
17
determined quickly and simply from any roof exposure higher than the bolted interval, such
as in roof falls or air-crossings. No time consuming and expensive laboratory tests are
required nor is it necessary to drill boreholes. UCS for example, is estimated from the type
of damage sustained by the rock after a blow with a ball -peen hammer.

The modifications to the CMRR to produce the floor rating (FR) include:
o The CMRR has an adjustment factor to take into consideration the strengthening
effect of a strong band within the bolted horizon. This has been eliminated for the
floor strata rating.
o Where the CMRR rates discontinuities and unit strength as 70% and 30% of total
rating respectively, the FR assumes that the rock strength is probably as important
a factor in floor stability analysis as discontinuities. For this reason the two aspects
have equal weighting for the FR.

Both CMRR and FR have been determined (where possible) as part of the analysis of the
case studies for this research (Sections 4 and 5). While the method is quick and simple to
use, the lack of access to most of the failed and stable sites resulted in less than half being
rated. From the limited data available no correlat ion is apparent and, therefore, no further
work on either the CMRR or FR was undertaken.

3.3.4 Slake durability and Duncan swell tests

Buddery and Oldroyd (1992) published a methodology for rating coal mine floors as
determined from slake durability and Duncan free swell tests. This work was re-evaluated
by Latilla et al (2002). The system is relatively quick and simple to use and allows testing
of the typically weak floor rock encountered beneath some seams. These tests have been
used on borehole cores drilled at or near some of the case study sites and are more fully
described in Appendix A.

3.3.5 Summary of geotechnical aspects

Indirect tensile strength (T
o
), natural moisture content, Atterberg limits and axial swelling
strain may be used to characterise floor strata.
When estimating UBS by means of Vesics equation, (Vesic, 1975) indirect tensile strength
(T
o
) should be used to estimate the cohesive strength of floor rock while an internal angle
of friction () of 20 can be used with confidence.
No single property can be used to predict swelling potential. Swelling of mudrocks is not
confined to those containing highly expansive clay minerals.
From the literature it became obvious that many of the tests carried out failed to predict
known floor conditions.
Indications are that it is preferable to avoid carrying out numerous costly and time
consuming laboratory based geotechnical tests and rather concentrate on evaluating other
simpler, more cost effective tests, such as Duncan swell, slake durability and impact
splitting, Buddery and Oldroyd (1992) or the floor rating of Riefenberg (1995).

18
3.3.6 Conclusions

Cost constraints preclude the drilling of boreholes close to existing failed and stable areas
to carry out further geotechnical tests. As a result, the most useful tests are those which can
be carried out simply and quickly on relatively small irregular samples taken underground
as close to the areas of interest as possible. Of these, the most useful is the determination
of natural moisture content. As noted previously, this value, utilised in Specks
modification of Vesics formula, has been used successfully to determine foundation
stability in seven out of eight of the case study sites (section 7.2).

Natural moisture content, slake durability and Duncan swell should be routinely collected
for core samples of the floor, especially in seams with a history of floor failure, namely,
the Highveld coalfield 5 seam, the 3 seam of the Free State coalfield, the Alfred seam in
KwaZulu-Natal and the Main seam of the Zululand and Swaziland coalfields.

3.4 Floor Behaviour Monitoring

Three cases where the behaviour of the floor was monitored were identified during the
literature survey, two in South Africa and one in the United States.

3.4.1 Durban Navigation Colliery

Ozan (1991b) studied floor behaviour in the tailgate of a longwall at Durban Navigation
Colliery (DNC) as the face approached.
Panel parameters were as follow:
Cover depth: 218m
Seam thickness: 0.9 to 1.0m
Mining height in gateroads: 2.0m
Longwall face width: 160m

The roof consisted of shale and sandstone while the floor was sandstone.
During gateroad development (single entry) 100 to 150mm of floor heave had been
experienced and it was hoped that by monitoring both displacement and stress, and tying
these in with the results of numerical modelling, a better understanding of floor failure
mechanisms could be obtained. However, due to technical difficulties full data were not
obtained. The conclusions that could be drawn are as follows:
o Of the total closure (subsequent to initial development) of 22mm, only 2mm were
measured in the floor.
o Although the exact stress change could not be obtained from some of the cells, it
was found that the horizontal stress changed by 4.76MPa in a direction
perpendicular to the tailgate, while the vertical stress changed by 2.9MPa. This is
illustrated in Figure 3.7.
o The increase in horizontal stress indicates a type II (closed ridge) floor heave
mechanism.
o The 3MPa abutment stress increase was not enough to cause floor instability. A
contributing factor was thought to be that the floor had been broken by heavy
machinery during development.

19

Figure 3.7. Stress change in the floor of a longwall tailgate.

3.4.2 Natal Anthracite Colliery

A study was carried out into the behaviour of the surrounding strata during pillar extraction
by Ozan and Prohaska (1992).
Panel parameters were as follow:
Cover depth: 210m
Seam thickness: 1.8m
Pillar centres: 23x23m and 23x46m
Bord width: 6.0m
Safety factor (Squat): 2.48
Pillar width: 17m
Pillar width/height ratio: 9.44

They chose two test sites in the centre of a panel where the highest stresses were expected
due to the arrowhead shape of the pillar extraction. They found that, as the stooping line
approached the instrumented site, the roof deflection was fairly uniform at all stations
(maximum 20mm) while the floor heave was greatest in the centre of the roadways (average
120mm). It was concluded that, as the roof applied increasing load on the pillars, the pillar
foundations settled and floor heave (buckling) occurred. No differential horizontal floor
movement was recorded, the whole thi ckness of floor simply moved upwards. This floor
damage is consistent with type I (open crack) floor heave.

Tailgate
Chain pillar
Longwall face
Horizontal stress
increase in floor:
4.76MPa Vertical stress
increase in floor:
2.9MPa
20
3.4.3 Central Illinois

Chugh, Chandrashekhar and Caudle (1988) studied laboratory and field results and applied
finite element modelling (FEM) to try to understand the interaction of floor, pillar and roof
strata. This study was carried out in two central Illinois mines in order to:
o Analyse stress changes in coal pillars and floor strata.
o Determine the character of horizontal and vertical floor movements.
o Correlate floor heave, roof sag, pillar settlement and opening convergence.
o Study the variation of pillar settlement and floor heave rates as a function of time,
mining geometry and location within the panel.
o Establish the functional relationships between pillar settlement and floor heave.

Available panel parameters were as follow:
Seam thickness: 1.8m (assumed) (6ft)
Pillar centres: 27.7 and 32.3m
Bord width: 5m (assumed)
Pillar width: 22.9 and 27.4m (75 to 90ft)
Pillar width/height rat io: 12.5 and 15.0m

It was found (inter alia) that:
o Pillar settlement of 25 to 50mm is relatively uniform across the panel except for the
pillars adjacent to the barrier.
o A stress relief zone extended 0.9 to 1.2m (3 to 4ft) into the pillars. Sloughing of
coal from the lower half of the pillars was common at all sites. The zone of
horizontal floor movement extended about 3 to 4.3m (10 to 14ft) into the floor
(Figure 3.8).
o Horizontal floor movements were generally directed across entries towards the
centre with a magnitude of between 7 and 22mm (0.3 to 0.9in). The lateral
movements were associated with significant shearing stresses.
o Horizontal floor movement is consistent with type II (closed ridge) floor heave.
o Negligible horizontal movement was observed parallel to the entries.
o Most sub-floor vertical movement was limited to the upper 2.4m (8ft) of underclay
(Figure 3.8).
o The roof appeared to be undergoing slow, massive movement as no significant bed
separation in the roof was detected.
o The ratio between total closure and pillar settlement was found to be between 2.0
and 3.5.
o Closure rates were found to be 20 to 88mm (0.8 to 3.5in) per month initially,
slowing to about 5mm (0.2in) per month after 60 to 90 days. These rates were
found to be a function of pillar and bord widths.
o A strong correlation (r=0.79) was found to exist between the steady state creep rate
(SSCR) and pillar stress (PS) where:
o SSCR(in/month) = -1.42+0.0036PS(psi)
o Potential for cutter roof behaviour is increased by the presence of weak floor
strata. Cutter roof is defined by Mark and Mucho (1991) as; Compressive type
roof failure (commonly called cutter roof, guttering, shear, snap top, pressure
cutting or kink roof). In thinly bedded (laminated) rock, classic cutter roof
develops as the progressive, layer-by-layer crushing and buckling of individual
beds. Surface subsidence was considered to be primarily due to pillar settlement
and appears to be generally 80 to 90% of the pillar settlement value. This
hypothesis was generally borne out by measured trends at each site.

21
As was the case in the work by Stephenson and Rockaway (1981), there was no mention
made of pillars actually failing due to foundation failure.

3.4.4 Conclusions

In the three documented floor monitoring cases, while there was floor heave and even some
pillar damage, no mention of subsequent pillar failure is made. It is assumed that the
longwall chain pillars did not fail as no mention is made of such an occurrence.

In the DNC and Central Illinois cases the floor heave was type II (closed ridge) as
movement or increase of stress in the horizontal plane was detected. The other case (Natal
Anthracite Colliery) was type I heave (open crack).

The magnitude of floor heave was l arger for the type I heave at 120mm compared to 2mm
and between 7 and 22mm in the two type II cases.

Figure 3.8. Zones of horizontal and vertical floor movement.

3.5 Pillar Failures

A number of cases have been studied where pillar collapse occurred with soft floor being
present. These are more applicable to the current study than the cases of floor heave only.
Pillar stress
relief zone
Coal sloughing from
lower half of pillar.
Zone of
horizontal
floor
movement
Zone of
vertical
floor
movement
22
Similarities between these cases and the failed case studies (Section 4) are also highlighted
where they are considered to be significant.

3.5.1 Swaziland and KwaZulu-Natal

Oldroyd and Buddery (1988) report on two cases of non-violent pillar failure at collieries
in Swaziland and KwaZulu-Natal. One of these was considered in the case studies (Section
4.5.1). They conclude that the load is largely controlled by the panel geometry and the
elastic moduli of the roof strata and coal seam (barriers). The pillar response is determined
by the load/deformation characteristics of the pillars themselves and the physical properties
of the roof and floor. Depending on the strength of the roof, pillars and floor, failure may
be accompanied by roof fracturing or fl oor heave.

3.5.2 Klipriver Coalfield

A number of cases of falls of ground in the Klipriver coalfield were investigated by
Buddery and McGregor (1990), one of which is deemed to have been caused by weak floor.

The mining geometry is compared to those of the failed cases (Section 8) in Table 3.3:

Table 3.3. Comparison between Klipriver and other failed cases.
Klipriver Failed cases (Section 8)
Cover depth (m) 26.1 to 26.2 32.5 to 165
Mining height (m) 2.1 to 2.2 2.4 to 3.8
Pillar centre distance (m) 12 11 to 24.4
Bord width (m) 5.9 to 6.06 5.9 to 8
Pillar width (m) 6.1 to 5.94 5.1 to 18.1
Safety factor (Salamon) 3.63 1.51 to 2.3
Pillar width to height ratio 2.7 to 2.9 2.1 to 5.6

As can be seen from the above, the pillar centre distance, pillar width, pillar width to
height ratio and bord width fall within the range determined for the other failed cases. Of
these, bord widths are not considered influential as they are generally similar for most
southern African coal mines.

Their observations include the following:
o The carbonaceous and micaceous shale/mudstone floor (sometimes rather sandy),
was very wet and highly weathered near the fall. The floor was reported to be
making water prior to the fall and tramming was a problem.
o Consolidation tests performed on samples gathered near the fall area revealed that
the floor had little tendency for consolidation when loaded under natural moisture
content condition and there is little swelling upon saturation at an applied stress of
125kPa. However, the floor material undergoes a greater rate of consolidation when
saturated. Upon unloading and when saturated, the rock shows a greater tendency
to swell.
o This type of floor material with its high silt content has a very low strength when
wet.
23
o There was considerable evidence of floor heave and failure near the fall and the
pillars appeared to have punched into the floor to a depth of 0.5 to 1m.
o Joints had opened up, generally sympathetic to the pillar sides, resulting in
considerable spalling of the coal pillars.
o The fall area was the lowest portion in that area of the mine. This result ed in water
collecting there which saturated the floor rock. From borehole information, the
saturated material would have been 3.5m thick.
o The average pillar stress after mining was 2.5MPa. Based on consolidation tests on
saturated floor rock, the resulting settlement beneath the pillars would be 80mm
with a corresponding 200mm of swell in the (unloaded) bords.
o Release of load at the pillar edges allowed the floor to swell, which assisted the
spalling process. The increased load on the remaining pillar core resulted in greater
settlement. About 900mm of spalling would effectively double the pillar load
resulting in 110mm of settlement. The load/settlement/spalling sequence would
have continued until the floor and pillars failed.
o The tendency for parts of the floor to settle (beneath pillars) and others to swell
(beneath bords) would have created internal shear stresses, which would contribute
to floor failure.
o The presence of major joints would have contributed to the fall as they would result
in localised stress changes and provide planes of weakness along which pillars and
roof could fail.

It is clear from the above that the floor failure was of the class II type and that the presence
of swelling materials combined with water was influential in causing f oundation failure.

3.5.3 Salamon and Munro collapsed cases

Madden and Hardman (1992) studied the long-term stability of coal pillars. They quote
Salamon as giving natural causes as one of the reasons for the scatter of pillar strength and
load values, that is, variations in coal strength, seam structure and roof and floor quality.

Table 3.4 is an extract from the dimensions of collapsed pillar cases used by Salamon and
Munro (1967)

From Table 3.4 it can be seen that all three cases fall within the (admittedly broad) range of
32 to 165m for cover depth while they all have lower safety factors than those of the case
studies (Section 8). The pillar width for case 55 (3.27m) is less than that of the collapsed
cases in the case studies (5.1 to 18.1m), This case also has a lower mining height (1.5m -
case studies 2.4 to 3.8m). The pillar width to height ratio for case 58 is considerably lower
(1.42) than that of the case studies (2.1 to 5.6).

Due to the low safety factors of these pillars it is considered most likely that the
documented floor heave was only a manifestation of overloaded pillars and that the weak
foundations were not the primary cause of the pillar failure, i.e. the pillars would have
probably failed irrespective of the nature of the floor.

24
3.5.4 Sigma Colliery

Two cases of pillar failure were reported by van der Merwe (1998), both on the No 3 seam
at Sigma Colliery.

Case 1 occurred in bord and pillar workings at a depth of 120m with a 3m mining height.
The pillar centres were 24m and the bord width 6.6m. The floor was reportedly a micaceous
sandstone similar to that of the Witbank and Highveld No 5 seam. The section was very we t
and the sandstone turned to a clay-like substance, more liquid than plastic in
performance. The floor sandstone flowed out from beneath the pillars and opened cracks
through the centre of pillars dividing them in half.

The sandstone is generally 0.2 to 0.5m thick in this area, but in this case it was 1m thick.

Case 2 was in a longwall panel very close to completion, mining into the equipment take -
off road. The face stopped for one shift due to belt problems. At this stage the pillar
between the face and take-off road was 4m wide and 100m long. This long slender pillar
punched into the soft floor, which heaved, trapping the shearer. The face supports went
iron bound resulting in a six month production delay. The mining geometry of the Sigma
Colliery case 1 is compared to those of the failed sites from Section 8 in Table 3.5.

Table 3.5 shows that Sigma Colliery case 1 fal ls within the mining geometry ranges of the
failed cases for all but pillar width to height ratio.

Table 3.4. Dimensions of some pillar collapse cases used by Salamon and Munro
(1967).
Case No. 54 55 58
Colliery Welgedacht Blesbok South Witbank
Seam Springs Witbank No. 5 Witbank No. 5
Depth (m) 62.5 65.0 39.6
Pillar width (m) 6.10 3.27 5.20
Bord width (m) 7.60 5.83 6.40
Mining height (m) 2.44 1.50 3.66
Pillar width to height
ratio
2.50 2.18 1.42
Safety factor 1.17 0.76 1.33
Date of working ? 1954 1957
Date of subsidence 1917 1955 1959
Time interval ? 4 months 1.5 yrs
Surface area affected
(ha)
0.61 4.05 3.64
Number of pillars 74 714 348
Remarks Scaling of pillars and
floor uplift noticed prior
to collapse. Sudden rise
in seam (elevation?)
coincided with one end
of the collapsed area.
Sudden increase in super
incumbent strata due to
seam dipping and surface
rising in the same area.
Further aggravated by large
dykes in close proximity.
Pillars talking and floor
heave two weeks before
collapse.
Irregular shaped pillars and
small portion of area mined
at 6.1m bords with 6.1m
square pillars. Floor
punching and pillar scaling
observed several weeks
before the collapse.

25
Table 3.5. Sigma Colliery failure compared to case studies.
Sigma Colliery case 1 Failed cases (Section 8)
Cover depth (m) 120 32.5 to 165
Mining height (m) 3 2.4 to 3.8
Bord width (m) 6.6 5.9 to 8
Pillar width (m) 17.4 5.1 to 18.1
Pillar width to height ratio 5.8 2.1 to 5.6

van der Merwe (1995) states that foundation failure may occur where the pillar loa d is more
than twice the UCS of the floor rock. Lateral displacement of the floor material in these
cases resulted in the pillars being torn apart with wide vertical cracks being visible.

The lack of UCS results for the case study sites and the fact that obtaining samples from
most of these areas is unlikely in the foreseeable future makes it difficult to verify this
statement for the case study sites.

3.5.5 Colorado

In a mine fatal accident report published in 1999 by the Mine Safety and Health
Administration (MSHA), it is stated that severe floor heave was present during pillar
extraction at a mine in Colorado. In this area the floor heaved 0.9 to 1.2m (3 to 4ft) during
the final stages of pillar extraction. At the time a bounce was experienced which is
described as, when the pillar actually yields, you can feel the floor give. Substantial floor
heave was reported in the section prior to the accident. The D seam overlies the previously
worked B seam by 82.3m (270ft). The roof consisted of interbedded shal es, sandstones and
coal while the floor was described as primarily shale. The mining geometry at this site is
compared to that of the failed case study sites in Table 3.6.

Table 3.6. Colorado failure compared to case studies

Colorado case
Failed cases
(Section 8)
Cover depth (m) 228.6 (750ft) 32.5 to 165
Mining height (m) 3.66 (12ft) 2.4 to 3.8
Pillar centre distance (m) 25.9 x 27.4 (85 x 90ft) 11 to 24.4
Bord width (m) 5.5 (18ft) 5.9 to 8
Pillar width narrowest side (m) 20.4 5.1 to 18.1
Pillar width to height ratio 5.57 2.1 to 5.6

Three other conditions are worthy of not e:
o A burn line or area of severe devolatilisation (burning) some 213m (700ft)
away.
o Severe pillar sloughage or spalling was observed close to the extraction line
while further away it was described as variable.
o Cutter roof or guttering was present in many places in the section, some of
which had occurred some time before pillar extraction commenced.
26

Only mining height and pillar width to height ratio of the Colorado case fall within the
range for the failed case study sites. The presence of dolerite intrusions is common to the
failed cases and the burn line was probably caused by a similar feature. It was not clear
from the report if the pillar sloughage was associated with floor movement or not.

3.5.6 Cooranbong Colliery

In a study at Cooranbong Colliery, Hill (1994) concludes that:
o Heave and subsidence were only loosely connected. This is due to swelling of
smectite clay and buckling of the floor as a result of horizontal stress. Neither can
be explained in terms of extrusion of claystone beneath the pillars.
o Subsidence was largely attributed to consolidation of the claystone as water was
forced out resulting in consolidation and an increase in the shear strength of the
claystone.
o Subsidence tends to occur fairly rapidly after mining and then slows considerably
with very small movements still taking place several years after mining.
o A conservative pillar design approach was recommended, namely:
i. Pillar safety factors should be in excess of 3.
ii. Percentage extraction not to exceed 50%.
iii. Minimum pillar width to height ratio of 5.
iv. Regular pillar geometries to be adopted.
v. Substantial barriers to be left between panels, with w/h >10.
vi. Leave floor coal to strengthen the base of pillars.
vii. Restrict total extraction in proximity to areas where the surface subsidence
must be prevented.

Considering that the range for pillar safety factors of the failed case studies (Section 8) is
1.51 to 2.3, that percentage extraction is between 48.8 and 78.5% and that the pillar width
to height ratio ranges between 2.1 and 5.6, items i, ii and iii in the above list could have
been compiled with the southern African failed cases in mind.
It is important to point out that a large number of perfectly stable panels in South African
mines would not comply with the above guidelines and if implemented would substantially
reduce their viability and reserve utilisation. The guidelines do however form a point of
departure when considering the design of bord and pillar workings in seams with a
perceived propensity to foundation failure.
Of interest in this study is panel North B. This area was mined in 1983 with 12m pillars,
7m bord widths and a cover depth of 75m. After bottom coaling the final mining height was
4m. The pillar sizes were irregular wit h many pillars less than 10m wide and some even
less than 6m. In addition, the panel was surrounded by goaf on three sides. It was
concluded that the weak floor contributed to the subsequent pillar collapse as the low level
of confinement at the base of the pillar would have created tensile zones in the ribsides
leading to spalling. Heave of up to 1.5m had occurred and surface subsidence was 1.3m.

The mining geometry at this site is compared to that of the failed case study sites in Table
3.7.

The Cooranbong case falls within the range determined for the case studies for all but the
mining height. In common with two of the four failed cases from Section 4, secondary
extraction had been carried out adjacent to the failed area. This would have reduced the
safety factor for at least two rows of pillars along the goaf edge.
27
Table 3.7. Cooranbong panel North B compared to case studies.
Cooranbong case Failed cases (Section 8)
Cover depth (m) 75 32.5 to 165
Mining height (m) 4 2.4 to 3.8
Bord width (m) 7 5.9 to 8
Pillar width (m) 12 5.1 to 18.1
Pillar width to height ratio 3 2.1 to 5.6

3.5.7 Conclusions

o Pillar safety factors vary greatly in the South African collapsed cases, from 0.76 to
3.36 while in the failed case studies (Section 4) the safety factors range from 1.51
to 2.3. In the two overseas cases the safety factor (Salamon) is between 1.37 and
1.92.
o Average pillar width to height ratios are 2.96 and 4.29 for the South African cases
and US / Australian cases, respectively. If the Sigma Colliery case is excluded the
South African average drops to 2.25. For the failed case study sites the width to
height ratio is 2.1 to 5.6.
o The presence of water (saturated floor rock) or dolerite intrusions (burning) is
recorded in a significant number of cases both here and overseas.
o Most of the cases experienced type II floor heave with associated accelerated pillar
spalling.
o Other problems recorded as occurring in two or more cases are horizontal stress
driven failures and irregularly sized pillars.
o Considering that the range for pillar safety factors of the failed case studies is
1.51 to 2.3, for percentage extraction is 48.8 to 78.5% and for pillar width to height
ratio is 2.1 to 5.6, the recommendations made by Hill ( 1994) could have been
compiled with the southern African failed cases in mind. These are: that the safety
factor should exceed 3, the percentage extraction should not be more than 50% and
the minimum pillar width to height ratio should be 5. These guidelines should only
be considered as a point of departure when considering the design of South African
bord and pillar panels with a perceived propensity to foundation failure.

3.5.8 General

Of great significance is the fact that, of all the United Stated papers s tudied, none reported
pillar failure associated with weak floor rock. Mark (2001), when asked about this, stated
that:

o Generally weak floor has never really been considered a safety issue. It is however
a big problem when it comes to subsidence.
o Pillar collapses are relatively rare in the United States due to their generally high
width to height ratios.
o Lately there have been about 12 collapses, mostly in high coal (2.5 to 3m).
o Generally in the United States a pillar failure (collapse) means a squeeze , a
generic term for roof and floor closure.
28
o One case of a massive pillar collapse was experienced in a trona mine that was
clearly attributable to floor failure. It occurred at a depth of 490m and 1.6km
2

collapsed at once causing one fatality and generati ng a Richter 5.3 earthquake.

3.6 Design Procedures

A number of different design approaches have been identified in the literature and are
summarised in the following section.

3.6.1 Numerical simulations

Ozan (1991a) looks in detail at various civil engineering approaches to foundation design
and found that they have a significant shortcoming in that the influence of adjacent pillars
is not taken into account. He concludes that the methods have promise if this shortcoming
could be overcome.

Ozan (1991a) reports a programme of numerical modelling using the FLAC code in which
applied load, material properties and mining height were kept constant while pillar width
and centre distance were changed. The ratio of centre distance and pillar width ( c/w) varied
between 3 and 1.25.

w
b w
w c
+
= / Equation 1

Where:
c = pillar centre distance (m)
b = bord width (m)
w = pillar width (width of footing) (m)

A bord and pillar system at a depth of 280m was modelled with an elas tic modulus of the
floor kept as 2000MPa. Internal friction angle was varied between 0 and 20. It is
interesting to note the findings of Chugh, Atri and Dougherty (1988) where it is stated that
in only 16% of cases, the angle of internal friction is less than 20 (section 3.3.1).

As would be anticipated, as c/w decreases the expected settlement of the floor sub-strata
also decreases. The case of c/w =3 has a pillar width of 15m, bord width of 30m and a
resulting settlement of 4mm while at the other end of the scale, for c/w =1.25, w is 24m, b
is 6m and the resulting settlement 0.3mm. The latter bord width value is far more
representative of the underground situation. In fact for all the cases where b is 6m the
modelled settlement ranges from about 1mm down to 0.3mm. The modelled settlement of
0.3mm to 1mm is considered to be on the low side when compared with the 25 to 50mm
quoted by Chugh, Chandrashekhar and Caudle (1988) in section 3.4.3.

In addition, the influence of the angle of internal friction () was modelled (also with
FLAC) and it was found that there is less floor settlement where the angle of friction is
highest. The modelled settlement is 460mm for = 0, dropping rapidly to 4mm at 10
29
then reducing further to 2.7mm at 20. As stated earlier, Chugh, Atri and Dougherty (1988)
found that only 16% of cases had an angle of internal friction lower than 20 .

Ozan (1991a) states that, in general, the three main r easons for floor heave are:
o Swelling rock in the floor strata.
o High horizontal stresses.
o Pillar punching of smaller pillars in high vertical stress environments.

In another numerical simulation Ozan (1991a) used the Flac code to model a soft floor
(E=20Mpa) beneath a single 10 x 40m footing, loaded with net intensity of 50kPa. After
approximately 1500 time steps equilibrium was reached with displacement of 23.5mm. The
average foundation settlement was found to be about 20mm. The loads and strengths used
in this simulation are far removed from the reality of underground coal mining and the
results should be viewed in this light.

Looking at the above items in more detail, he goes on to state that problems can be
expected in roadways where the floor consists of clay material, particularly
montmorillonite, but notes that no design methods applicable to this aspect of floor heave
could be found. The following remedies are proposed:
o Prevent water from entering floor layers.
o Provided the layer is not too thick, remove the swelling rock. This could prove
difficult from an operational point of view as the swelling rock would need to be
either handled by the washing plant or stowed in the workings.
o Leave a floor coal cover, practical in thicker seams only.
o Remove swelling floor layer and replace with non-swell material. This may only be
practical in the case of a main travelling road as the cost of this operation would be
relatively high.

The correct value of the k ratio (horizontal/vertical stress) for South African coal mines
was little known at the time that Ozans report was written, but he quotes it as varying from
0.5 to over 2 dependent on geological structure (Fauconnier and Kirsten, 1992) while
Santos and Bieniawski (1989) are reported to have stated that in shallower mines k is well
above 1.0.

An interesting solution is put forward by Peng and Tsang (1989) after finite element
modelling of coal pillars standing on weak floor. This is that stable workings could be
achieved by designing pillars to yield totally. However during pillar yield, tensile stresses
are increased in the roof and floor, as a result of which, additional attention to roof and
possibly floor stability would be essential. As the yield pillars would have higher vertical
stresses on them than the non-yield pillars they would also be more likely to punch into the
floor and accelerate floor heave.

Finite element modelling (FEM) was used by Su and Hasenfus (1999) to illustrate the site -
specific nature of coal pillar design and the use of modell ing techniques to accommodate
these factors. Of interest for this project is that at high width to height ratios (w/h = 16)
weak floor rock may reduce the ultimate pillar strength by as much as 50%. Their FEM soft
floor curve virtually flattens off at a peak pillar strength of 29.6MPa (4300psi),
corresponding to w/h = 8 and (as would be expected) correlates reasonably well with the
given soft floor field data. At width to height ratios more typical to South African collieries
there is no discernable reduction in pillar strength at w/h between 2 and 4 and a 21%
reduction at w/h = 6. The reduction in strength due to soft floor for larger width to height
ratios predicted by the FEM method is borne out by the (admittedly limited) field data.

30
While the use of numerical modelling has been evaluated in a number of studies no
definitive set of guidelines has been established to assist with the design of pillars standing
on soft floors. This aspect has been studied in more detail in Section 6.3 where it is found
that the two-dimensional package PHASE
2
gives encouraging results.

3.6.2 Civil engineering approach

Terzaghis (1964) method for calculating the bearing capacity of foundations (which is
widely used in civil engineering) is evaluated by Madden et al (1995). This will not be
discussed in detail here, as the civil engineering approach to the problem is investigated in
Section 7.1.

Rockaway and Stephenson (1982) state that various studies have at least attempted to
qualitatively describe the floor failure process and suggest techniques to overcome the
problem. The varying success of the methods at different sites points to the problem being
multi-faceted. Using Vesics equation (Vesic, 1975) they determine factors of safety, which
correlate well with observed mine floor conditions. It should be noted that in both papers
by Rockaway and Stephenson (Rockaway and Stephenson, 1982 and Stephenson and
Rockaway 1981), no mention is made of the foundati on failures leading to pillar failure,
although surface subsidence is a problem above areas of punching pillars.

Speck (1981) quoted a modification to Vesics foundation strength equation that has been
applied as part of this research (Section 7.2).

3.6.3 General

York, Canbulat and Jack (2000) note that even though pillars are designed according to
current accepted practice there are still cases of failure due to weakness in the surrounding
strata. They identified roof and floor stability as one of nine factors affecting coal pillar
strength. However, a way to incorporate these weaknesses into a design methodology is not
established and further research into foundation stability of coal pillars is deemed to be
necessary. The discussion of the modes of foundation failure, detailed by Madden et al
(1995), is repeated in their report.

Following five years of extensive research in Illinois Basin coal mines, Chugh and Pytel
(1992) came up with a ten-step method to calculate a factor of safety against foundation
failure in which:
o It considers a case where the floor consists of a weak layer underlain by a stiff
layer.
o The influence of adjacent pillars and non-zero values of internal friction are
accommodated.
o A modified bearing capacity factor is calculated.
o The bearing capacity can either be measured using plate load tests or estimated
using natural moisture content data.
o The bearing capacity of the underlying stronger layer is determined from bearing
capacity equations.

31
Coal pillar foundation failure is covered by Peng (1978) although he also makes no mention
of pillar failures associated with weak floor. In the majority of United States references,
floor heave is considered more of a production nuisance than a safety is sue. He quotes the
following equation to determine the ultimate bearing capacity of a pillar foundation using
the Rankine wedge concept.

q r
u
u WhN N
WB
B
P
P + = = A
2

Equation 2.

Where:
P
u
= the ultimate load applied along a width B
W = Weight density of the foundation soil
h = Pillar height
N
r
and N
q
are bearing capacity factors dependent upon the frictional characteristics of
the soil. For = 20, N
r
= 5 and N
q
= 7.

In this book Peng also describes the bearing capacity test and associated equipment. The
complexity and size of the apparatus is such that these tests will only be carried out in
South African collieries if floor failure becomes a far more prevalent problem.

Galvin (1995) states that operators should be aware of the following:
o On the basis of soil and foundation engineering principles, the ratio of pillar width
to weak floor thickness has a major controlling influence on the development of
bearing capacity failure. The bearing capacity of a weak foundation decreases
markedly with an increase in weak floor thickness. Increasing pillar width increases
the pillar strength, reduces load and hence reduces the probability of bearing
capacity failure.
o With time, soft strata may undergo significant consolidation under the effects of
pillar load. Differential floor displacement and subsidence due to consolidation
should not be taken as indicators of pillar or bearing capacity failure.
o The long-term effect of water on the strength of weaker foundation material needs
to be determined.

3.6.4 Conclusions

The extensive use of numerical modelling has failed to produce a definitive set of
guidelines for the design of pillars standing on soft floors.

The bearing capacity of a weak foundation decreases markedly with an increase i n weak
floor thickness. Increasing pillar width increases the pillar strength, reduces load and
lessens the likelihood of bearing capacity failure.

3.7 Main Findings of previous work on foundation stability

A number of publications and reports dealing with pillar foundation failure were studied to
identify any learning points which may assist with understanding and analysing the case
32
studies in sections 4 and 5 of this project report. Most of the reports and papers studied
were from South Africa, Australia and the United States of America.

The literature indicates that certain seams have floors more likely to fail than others, the 5
seam of the Witbank, and Highveld coalfields as well as the Alfred and Main seams in
KwaZulu-Natal being especially prone. Sandier facies generally form weaker pillar
foundations than those consisting of shaly material because they are able to absorb more
moisture. The relationship between moisture content and floor stabilit y is explored in more
detail in Section 7.2.

Australian and United States experience indicates that while pillar settlement is not
atypical in certain areas, pillar failure is more likely to occur in southern Af rica. This is
probably due to the generally more conservative pillar designs practised in these two
overseas countries.

Pillar settlement and floor heave as experienced elsewhere tend to be more of economic
consideration than a safety hazard as they lead to surface subsidence and underground
roadway closures. In southern Africa, pillar failures resulting from foundation instability
can have significant economic consequences but they also potentially pose a greater threat
to the safety of underground personnel.

It is relatively simple and inexpensive to determine moisture content as well as slake
durability and Duncan free swell indices (Appendix A). As such, this data may be routinely
gathered from core samples of the floor, especially in seams with a history of floor failure,
such as the Highveld coalfield 5 seam, the 3 seam of the Free State coalfield, the Alfred
seam in Kwa-Zulu Natal and the Main seam of the Zululand and Swaziland coalfields.

In the three documented floor monitoring cases, while there was floor heave and even some
pillar damage, no mention of subsequent pillar failure is made.

Average pillar width to height ratios are 2.96 and 4.29 for the South African cases and US /
Australian cases, respectively. If the Sigma Colliery case is excluded the South African
average drops to 2.25. For the failed case study sites the width to height ratio is 2.1 to 5.6.

The presence of water (saturated floor rock) or dolerite intrusions (burning) is recorded in a
significant number of cases both here and overseas.

Most of the cases experienced type II floor heave with associated accelerated pillar
spalling.

While the use of numerical modelling has been evaluated in a number of studies a clear set
of guidelines has not been established to assist with the design of pillars standing on soft
floors. Numerical modelling as a means of analysing the problem is dealt with further in
Section 6.

4 FAILED CASES

The study of a number of failed cases is considered impor tant to enable the identification of
any factors common to the various sites. While only four sites have been chosen it is felt
that these represent a good starting point towards a fuller understanding of this problem.
33

Failed cases are defined as where t here has been total pillar collapse or else severe pillar
damage (indicating that the pillars have passed their peak load carrying capacity) and
where, at the same time, the floor has shown signs of severe instability.

The following aspects, identified during the literature survey, have been investigated or
carried out at the case study sites (where practical):
o The case study floor failures have been classified as type I, II or Newcastle,
according to Peng, Wang and Tsang (1995) or Seedsman and Gordon ( 1991).
o Coal mine roof rating (CMRR) and floor rating (FR) have been determined at most
accessible sites.
o Floor strata have been classified according to the facies descriptions of Jermy and
Ward (1988).

In the following sections the circumstances surroundi ng the failures are described and
available pertinent geotechnical and mining geometry data given. In addition, the factors
considered most likely to have caused the failures are detailed for each of the case study
sites.

4.1 Emaswati Colliery, main haulage south
4.1.1 Introduction

On Saturday 8
th
June 1991, a major collapse of ground occurred at Emaswati Colliery
situated near Mpaka in the Kingdom of Swaziland. Twenty-six underground employees of
the mine were trapped by the collapse and were subsequently all r escued uninjured by the
Chamber of Mines Rescue Brigade using the Rescue Drill based at the Colliery Training
College, Witbank.

An entire drill and blast sections equipment was lost in this incident but work resumed
with equipment obtained from other mines until financial factors resulted in the closure of
the mine in 1992.

Panel parameters were as follows ( Figure 4.1):
Cover depth: 65m
Seam thickness: 2.5m average (1.9 to 3.1m)
Pillar centres: 16m
Bord width: 6.3m
Safety factor (Salamon): 2.3 average (2.0 to 2.6)
Pillar width: 9.7m
Pillar width/height ratio: 4.1 average (3.1 to 5.1)

Another pillar failure in the eastern portion of the mine was documented by Oldroyd and
Buddery (1988). This is not included in this study as there was no evidence of floor failure
associated with it. Roof to floor convergence rates in this case were very low at between
0.34 and 8.83mm/annum

34

Figure 4.1. Emaswati, main haulage south, showing pillar condition ratings.
35
4.1.2 Geology

All coal produced at Emaswati was mined from the main seam which is on average 3.1m
thick (range, 1.5 to 4.3m) and dips towards the east at 1 to 2 (locally up to 6). The
immediate seam roof is a shale band overlain by a thin coal streak. This unit is usually 0.5
to 0.6m thick but can reach 1.1m in places. It forms a very poor roof and is usually mined
with the seam. Above this lies a false roof band consisting of coarse grained, carbonaceous
shaly sandstone. When thin, this band tends to part with the production blast. Overlying
these two weaker zones is massive, coarse-grained sandstone up to 14m thick. The floor is
comprised of interlaminated bands of shale, siltstone and fine -grained sandstone which is
shaly or silty.

The structural geology of the mine is fairly complex, with numerous dolerite dykes from 1
to 30m thick and sills up to 12m thick. Occasional faults with displacement up to 10m
occur and there are numerous joints in the seam associated with the aforementioned major
structural units.

The failed pillars lay within an area completely enclosed by two dolerite sills and a dyke.
As a result of the intrusions, the seam within this area was devolatilised. The thicker of the
two dolerite sills (7m thick) lies between 30m and 70m below the seam ( Figure 4.2).

The sediments within the failed area were tilted towards the south and east due to dolerite
activity. It is postulated that an elevated horizontal st ress environment was created by these
intrusions. Numerous slickensided joints, parallel to the bedding planes, were caused by
these stresses and weakened the floor strata.

4.1.3 Geotechnical tests

As this mine is closed it is not possible to obtain access to the collapsed area. However,
some exploration borehole cores were tested at Emaswati prior to the collapse (Oldroyd and
Latilla,1989) of which three, EM18, 19a and 39) were within 500m of the failed area.

After the failure more boreholes were drilled along the line of a proposed bypass
development (Latilla, 1992). Two of these, (EM53 and 54) were within 500m of the
collapse area.

Impact splitting (IS) was carried out on the roof rock while slake durability (SD) and
Duncan free swell (DS) tests were performed on selected floor cores (Buddery and
Oldroyd, 1992). These tests are described in Appendix A, while the Emaswati results are
summarised in Table 4.1.

Where floor cores had no perceived propensity to slake or swell, they were not tested.
Borehole positions in relation to the failure area as well as roof and floor ratings are shown
in Table 4.1.

A weighted average of the slake durability index and resultant rating for t he immediate
0.5m of floor is also given as a means of comparing the various cases. Where no slake
durability index is given due to the strata being rated as good, a value of 5 has been
assumed for the slake durability index.

36

Figure 4.2. Section (E-W) through collapsed Area at Emaswati (after Petzer, 1991).

4.1.4 Background

Evidence of failure was first reported to, and investigated by, the Regional Rock
Engineering Manager for Genmin, Mr D.C. Oldroyd (1990). He concluded that the area in
question had been recently developed through an area of burnt coal and that the burnt area
was clearly demarcated on the coal volatile matter plan and was apparently caused by an
underlying dolerite sill. He further stated that the pillars had been weakened by burning
even though the safety factor was high at 2.3.

At the time that Mr Oldroyds report was written, the pillars were showing signs of being in
an advanced state of deterioration with large blocks having already fa llen from the pillar
sides. Many exhibited wide fractures, extending some distance into the pillars. The number
of broken mine poles in the panel indicated that considerable convergence had occurred

37
Table 4.1. Emaswati Colliery, Roof and floor ratings.
Borehole
number
Distance from
collapse area
Roof rating
(impact splitting)
Floor rating
EM 18 355m SW Good: IS=98

0.32m carbonaceous
sandstone overlain by
2.31m medium to coarse-
grained sandstone.
0.31m moderate - SD 14-26,
overlying good floor.
Weighted SD for 0.5m of floor 14.9
(Moderate)
0.21m interlaminated and micaceous
sandstone underlain by 0.10m
medium-grained sandstone and 0.50m
sandy shale.
EM 19a 440m SSW Very good: IS=127

0.21m carbonaceous and
micaceous sandstone
overlain by 1.99m coarse
grained sandstone.
0.08m good DS <1, overlying
1.49m poor to very poor - SD 26.1-36
or DS 3.1- 15).
(Poor)
0.08m carbonaceous and micaceous
sandstone underlain by 0.72m highly
micaceous, carbonaceous shaly
sandstone (swelling in the core-box)
and 0.77m micaceous, interlaminated
sandstone and shale
EM 39 500m NE Good: IS=101
0.83m coarse sandstone
overlain by 0.24m fine
gained sandstone (sub
vertically fractured) and
0.15m shaly sandstone.
Good SD <14)
Weighted SD for 0.5m of floor 5
(Good)
0.12m shale / sandy shale underlain
by 0.22m sandstone / sandy shale and
0.11 shale / sandstone.
EM53 460m W Excellent: IS=191

0.57m carbonaceous
sandstone overlain by
1.62m sandstone (remarks:
very hard roof).
Upper 0.6m of the seam
described as very broken.
0.27m poor SD=23.4 and DS=3.8)
underlain by 0.65m moderate
DS=1.1)
(Moderate)
0.27m shale, underlain by 0.38m
sandstone / shale and 0.91m
sandstone.
EM 54 480m W Very good: IS=118

0.23m shaly sandstone
overlain by 2.20m
sandstone, fractured to the
bottom.
Seam has numerous
slickensided slips 45 to 60
in lower 1m.
2.52m very poor SD=50.2 and 89.8
Weighted SD for 0.5m of floor 50. 2
(Very poor)
0.09m shale, broken, underlain by
2.27m sandstone of which the upper
1.33m is weak.
Averages Very good: IS = 127 Weighted SD for 0.5m of floor 23.1
(Poor)

A pillar rating exercise (Figure 4.1) revealed that one pillar could be regarded as
completely failed, 10 close to failure, 25 showing signs of distress, 49 slightly scaled and
19 still intact. A ten-point pillar condition rating, developed for use at Zululand Anthracite
Colliery (ZAC) (van Wijk and Latilla, 2000), has been applied in this case in order to
achieve a uniform approach. This is summarised in Table 4.2.
38

Oldroyd stated that convergence would continue and even accelerate but that a pillar ru n
was not expected due to relatively high width to height ratios and the fact that no spalled
coal had been cleaned up. Roof to floor convergence monitoring stations were established.

Mining was continuing to the east and south of this area, with the sout herly workings
separated from the failure area by long pillars along a dyke. After considering developing a
second access way and wrapping the pillars with mesh and haulage rope it was decided that
the best option would be to fill three roadways and the splits in between with waste
material. Soil or sand supplied through a borehole from surface was considered best, as it
would have a lower void space requiring less compaction. This was to be placed to as close
to the roof as possible. The effect of the wast e filling would be to provide lateral restraint
to the pillars as well as to limit convergence once the waste began to be compacted.

Convergence monitoring indicated that although convergence was continuing, it was at a
reduced rate, (Latilla, 1990). The amount of sand filling required was subsequently reduced
to filling in between clusters of four pillars as shown in Figure 4.1. The outsides of each
cluster of four pillars were to have been wrapped with diamond or weld mesh, attached to
the pillars by roofbolts and lined on the inside with brattice. This would ensure that the
retaining wall of brattice was more flexible than if mine poles were used. This remedial
work was started but not finished before the collapse occurred. Unfortunately, no record
can be found showing which pillar clusters had been completed but indications are that
three out of seven had been filled.

In March 1991 an increase in convergence rate was detected as can be seen in Figure 4.3.
At this stage it was recommended that the wire mesh wrapped around the pillar clusters
should be pulled closer to the pillar sides (Latilla, 1991).

At 10am on Saturday 8 June 1991, a major fall of ground occurred in thi s haulage trapping
26 men. With the assistance of the Chamber of Mines rescue drill they were all brought to
safety by 5pm the following day. A detailed site investigation was carried out (Buddery,
1991), and it was observed that the failed area was completely enclosed by two dolerite
sills and a dolerite dyke as shown in Figure 4.1. The collapsed panel was in contact with
these features for a great proportion of its perimeter.

Devolatilisation of the coal had occurred as a result of this dolerite activity, particularly
close to sill S1 (Figure 4.1) as this sill turns under the seam after intersecting it. The
resultant tectonic stresses caused tilting of the sediments to the south and east resulting in
major east dipping joints in the coal seam. The floor was weakened during this process by
the development of closely spaced slickensided joints perpendicular to the floor bedding
planes (Buddery, 2002).

Prior to this collapse, no concern existed that a violent collapse could take place, for the
following reasons:
o For 12 years a panel elsewhere on the mine had been yielding in a stable
manner even though its safety factor and width to height ratios were relatively
low, 1.56 and 2.58 respectively, (Buddery and Oldroyd, 1988).
o Two cases of pillar failure had been experienced at ZAC previously (similar
coal field), where pillars with lower safety factors and width to height ratios
than those in this case failed in a stable manner ( section 4.5.1)
o In both the above cases, dolerite intrusions and/or major faults were present.
Convergence rates varied between 400 and 8760mm/annum.

39

Table 4.2. Pillar condition rating.
ZAC
rating
number
ZAC pillar condition description
Equivalent
condition rating for
Emaswati.
1 Intact. CM cutter drum marks visible over more than 90% of
pillar.
Intact
2 Light spalling, <0.1m deep, <50% of pillar affected. Corner
damage (<0.3m) on non-spalled pillars.

3 Light spalling, <0.1m deep >50% of pillar affected. Pillar
corner damage (>0.3m) on non-spalled pillars.
Slightly scaled
4 Moderate spalling, 0.1 to 0.3m deep <50% of pillar affected.
Large pillar corner failures 1m deep.

5 Moderate spalling, 0.1 to 0.3m deep >50% of pillar affected.
Blocks slipped out along joints, less than 50% of pillar
affected.

6 Heavy spalling, 0.5 to 1m deep <50% of pillar affected.
Blocks slipped out along joints, more than 50% of pillar
affected. Jagged, vertical fracturing along cleats.
Showing signs of
distress
7 Heavy spalling, 0.5 to 1.0m deep >50% of pillar affected.
Dilation cracks <20mm wide, open gap at top over less than
50% of pillar, hourglass shape.

8 Very heavy spalling, >1m deep <50% of pillar affected.
Dilation cracks 20 to 50mm wide, open gap at top over more
than 50% of pillar. Zones of pillar shattering present.
Close to failure
9 Very heavy spalling, >1m deep <50% of pillar affected. Coal
toes touching, severe damage, dilation cracks >50mm wide,
closure probably 2% of mining height.

10 Failed. Joints mobilised and open by more than 50mm, large
dilation cracks, closure probably over 3% of mining height.
Spalled material up to roof.
Completely failed

Sand filling as recommended earlier, was not prioritised, as the closure rate during 1990
was generally less than 50mm/annum. The major concern then was that the haulage would
not be able to sustain 40% convergence.
In addition, the rehabilitation / prevention work was severely hampered earlier in 1991 by
labour unrest and by subsidence in the haulage shaft after a period of heavy rainfall. By
May, convergence readings showed that the accelerating convergence trend was continuing,
up to 200mm/annum being measured at one station.

Shortly before the collapse it had been decided to mesh and lace the pillars with three
strands of cables and use timber lagging and steel straps to spread the cable tension.
Cluster stick packs were to be placed in the splits along the haulage and travelling roads. In
both cases, the intention was to limit the convergence at which equilibrium would be
reached and restrict localised roof damage. Before any of the above additional support
could be installed, the collapse occurred.

40
-50
-45
-40
-35
-30
-25
-20
-15
-10
-5
0
J
a
n
-
9
0
M
a
r
-
9
0
M
a
y
-
9
0
J
u
l
-
9
0
S
e
p
-
9
0
N
o
v
-
9
0
J
a
n
-
9
1
M
a
r
-
9
1
M
a
y
-
9
1
Date
C
o
n
v
e
r
g
e
n
c
e

(
m
m
)
stn 1
stn 2
stn 2a
stn 3b
stn 4a
stn 5a
stn 6

Figure 4.3. Emaswati, convergence monitoring.

4.1.5 Post collapse observations

The northern edge of the collapse was plotted during the investigation, (Buddery, 1991)
while the position of the southern edge is as reported by the trapped miners ( Figure 4.1).

In places it was observed that the roof had fallen out to a height of 1 to 2m and intact
pillars could be seen in the collapsed area, as could a ventilation wall and mine pole (items
I and X in Figure 4.1). The roof above the fall appeared reasonably intact and seemed to
have subsided as a unit.

The boundary of the fall was characterised by significant amounts of floor failure due to
opening of slickensided joints in the floor. An older area of floor heave was observed to the
west of sill S1 (Figure 4.1), while extensive floor heave was reported near peg 5030 in
1985/6 during re-opening of this area. The Acting Mine Overseer, Mr R Nthlengethwa,
reported that no roof slabbing occurred prior to the fall but that there was extensive floor
heave and opening of floor cracks. Some mine poles had broken while others appe ared
intact. In some cases, they had punched up to 150mm into the floor. Vertical parting
between sill S1 and the seam was noticed, but this appeared to have happened before the
event.

The area was inspected from a geological point of view by Petzer (1991) who, in addition
to corroborating many of the points above, stated that the surface cracks terminate sharply
along the southern dyke while tensile cracks are most pronounced along the eastern and
southern margins which were also bounded by dolerite. It i s thus evident that the failure
tended to terminate along dolerite contacts. He went on to state that floor failure and heave
were the major observed features and that no serious goafing could be seen, as roof
sediments were largely intact.

41

4.1.6 Discussion

Buddery (1991) suggests that:
Coal devolatilisation and severe jointing gave rise to pillar yield.
Floor failure may also promote pillar yield as lateral floor movement peels off the
pillar edges.
At some point (beneath the surface depression), the floor underwent massive failure
resulting in an extension of this failure mode around the perimeter. At this stage, the
pillars, then still largely intact, would have punched into the floor.
The failed floor at Emaswati is structurally weak rather than compris ing of very weak
rock.
The remedial measures being implemented were designed to control pillar failure and
their ability to arrest floor deterioration is not known. Cluster stick packs, being very
stiff, may have contributed to floor damage.

Figure 4.4. Open topped and closed ridge types of floor heave.

4.1.7 Conclusions

The surrounding dolerite intrusions are considered to be the primary cause of the pillar
failure in this case, in parti cular, the underlying sill which resulted in the seam becoming
devolatilised. Tectonic forces due to the dolerite intrusions induced slickensided vertical
joints in the laminated floor strata. As the floor heave was predominantly open topped and
not closed (as illustrated in Figure 4.4) it is evident that little or no flow of weak floor from
Floor beams remain the same
length
Floor beams elongated by
material squeezing out
beneath pillar
OPEN TOPPED
CLOSED RIDGE
42
below the pillars occurred in this case. The overall failure mode was more akin to that
experienced in the USA and Australia in that the pillars largely remained intact and the
roof failed around them. However, in those two countries this type of failure is normally
associated with type II (closed ridge) floor heave whereas in this case the floor heave was
of the type I (open topped) variety. This new classification of floor heave types is
considered useful and will be done for all the failed and stable case studies.

The pillars exhibiting the greatest damage were mostly in the central third of the panel,
which would be the position of highest deflection of the massive sandstone roof beam.

Convergence rates were highest where less or no remedial work had been carried out. This
would indicate that the relatively low restraint offered to both the pillars and floor by the
limited sand filling that had been completed, had a significant effect in slowing down
convergence.

4.2 Matla 1, 5 seam, panel R14 South
4.2.1 Introduction

Surface subsidence of 1 to 1.3m was observed at the N
o
1 shaft at Matla Colliery (Matla 1)
on 26 January 1995 during an airborne gas pipeline inspection carried out on behalf of
Gascor. It was found to have been caused by a collapse of pillars in the abandoned 5 seam
bord and pillar workings. There was also an Eskom power line pylon over the affected area.

The area was mined in 1981 by continuous miner. Due to roof control problems, operations
on the 5 seam were curtailed in 1983, from which time, only the underlying 4 seam was
mined. Since the pillar collapse, the 5 seam workings have been sealed off close to shaft
bottom (Latilla, Oldroyd and Wevell, 1997).

The mining geometry at this site was as follows ( Figure 4.5):
Cover depth: 32.5m
Mining height: 2.4m
Pillar centres: 11m
Bord width: 5.9m
Safety factor (Salamon): 2.26
Pillar width: 5.1m

4.2.2 Geology

Three boreholes were drilled around the 5 seam workings at Matla 1 after the pillar
collapse, of which number H85/99 was closest to the failed area.
The seam in this area is 2. 07m thick and has a substantial (0.70m thick) in-seam shale band
(sub-vertically fractured) only 0.08m below the roof. Cover depth at this borehole is
35.65m. A 0.06m grit / shale band forms the immediate roof and this unit is overlain by
0.27m of micaceous sandy shale and 0.19m of glauconitic and micaceous, shaly sandstone.
Above this generally poor to moderate lower roof, there is competent, coarse grained,
43

Figure 4.5. Matla 1, 5 seam pillar collapse area.

44
massive sandstone, which is 0.90m thick. The overlying strata consist chiefly of shaly
sandstone and sandstone. Weathering extends to a depth of nearly 20m.

The seam floor consists of 0.51m of sub-vertically fractured shale / sandy shale underlain
by 1.48m massive, medium grained sandstone. As part of their ongoing quality control
program, the mines Geology department carried out underground channel sampling which
included measurements of roof and floor contamination. The floor contamination in this
panel varied between 0.14 and 0.49m with an average of 0.31m while in the stable adjacent
panel it was significantly less, at 0.06 to 0.11m, with an average of 0.08m.

A dolerite dyke lies to the south of the collapsed area. Poor roof conditions were
experienced when developing through it, as evidenced by the number of irregular and long
pillars which had been left (Figure 4.5). Currently only the 4 seam is being mined at Matla
1 in the south eastern reserve block. The previously mined Matla 1 reserves overlie a
dolerite sill. This sill lies about 50m below the 4 seam ( Figure 4.6) and has possibly caused
slight devolatilisation of the 4 seam in places. Numerous dykes extend up from the sill an d
have caused associated slips and minor faults.

Figure 4.6. Matla 1 - showing position of dolerite sill.
East
mine
A
Sill
5 seam
collapse
area
Sketch not to scale
2 seam
Sill weathered away above
present surface level
A
B
5 seam
4 seam
2 seam
5 seam
4 seam
West
mine
Dolerite
sill
South
eastern
reserves
South
mine
B
N
45
4.2.3 Geotechnical testing

As the 5 seam workings are sealed off, it is not possible to gain access to the area to obtain
samples for test purposes. In addition no exploration drilling is planned anywhere near this
area in future as the only other viable seam (4 seam) has already been mined out for large
distances. The three boreholes drilled soon after the collapse occurred were geotechnically
tested (Buddery, 1995c). Only impact splitting and slake durability tests were carried out
(Appendix A - Geotechnical Testing). Borehole positions relati ve to the failure area, as
well as roof and floor ratings are shown in Table 4.3.

Floor samples were taken at shaft bottom (the only accessible 5 seam workings at Matla 1)
and tested to determine their moisture cont ent as outlined in Appendix A. The floor was
saturated at the sampling site which accounts for the samples remaining roughly the same
weight after immersion for 48hrs. The samples which were dried for 48hrs showed weight
losses of between 9.92 and 10.38%.

Table 4.3. Matla 1, 5 seam, roof and floor ratings.
Borehole
number
Distance
from
collapse
area
Roof rating Floor rating
H85/99 180m ENE Good: IS=112.6 (Rating biased by
presence of very strong massive
sandstone 0.52m into the roof.)
Roof strata SD 4.3 to 9.7 - class A

0.33m micaceous, sandy shale, gritty
to base, overlain by 0.19m micaceous,
glauconitic shaly sandstone and 0.90m
massive, coarse-grained sandstone.
0.51m class D (very poor - SD 70.1)
overlying 1.48m class A (good - SD
5.8)
(Very poor)
0.07m micaceous shale overlain by
0.44m micaceous and sub-vertically
fractured shale, with subordinate
sandstone bands and 1.48m massive,
medium grained sandstone.
H85/98 360m SW Good: IS=108.7. (Rating biased by
presence of very strong massive
sandstone 0.39m into the roof.)
Roof strata SD 4.5 to 9.9 class A

0.24m micaceous, sandy shale, gritty
at base, overlain by 0.15m micaceous,
glauconitic shaly sandstone and 1. 00m
massive coarse-grained sandstone.
0.42m class D (very poor - SD 42.1-
78) overlying 0.59m class A (good -
SD 4.2)
(Very poor)
0.16m sandy shale overlying 0.26m
interlaminated micaceous shale and
sandstone and 0.59m medium to coarse
grained, micaceous sandstone.
H85/100 480m
NNE
Moderate: IS=71.9
Roof strata SD 10.3 (sandy shale and
shaly sandstone) to 45.9 (coaly shale)

0.64m coaly shale, sandy at base,
overlain by 0.39m micaceous sandy
shale, gritty at base and 0.19m
micaceous, glauconitic sandstone.
Above this thicker weak zone lies a
massive, coarse-grained sandstone
0.78m thick. The lower 0.64m of the
roof in this borehole would have been
included in the coal seam in the other
two boreholes.
0.44m class D (very poor - SD 38.7-
84.4) overlying 1.03m class A (SD 6.2)
(Very poor)
0.09m broken, sub-vertically fractured
shale, overlying micaceous, vertically
fractured shale with subordinate
sandstone bands and 1.03m fine-
grained sandstone and shaly
sandstone.
Averages
Good: IS =97.7 Weighted SD for 0.5m of floor 58.7
(very poor)

46
4.2.4 Post collapse observations

The collapsed area was initially investigated by Buddery (1995a and b). He found that the
collapse was due to the failure of the 5 seam floor. Investigations for a feasibility study
into mining of torbanite on the 5 seam to the south west of the Matla reserves in the late
1980s and experience on the 5 seam at Matla, indicate that the floor in this area is weak,
with a pronounced propensity to swell.

It appeared that the floor had extruded from beneath the pillars. This led to the opening up
of joints and cracks in the pillars. The weakened pillars then sank into the floor causing the
failure of the roof sandstone. A clust er of dykes was considered to have contributed by
channelling water to the floor strata. Conditions in the only other panel mined on the 5
seam at Matla 1 (R14 West) were found to be good with only slight floor heave but no roof
or pillar instability (section 5.1 stable cases).

Buddery concluded that only ash filling could assist with providing long-term stability to
the workings. Pillar wrapping or shotcreting of unfailed pillars were not considered viable
options as they would require people to work for considerable periods in a hazardous area.

The small size of the pillars in question was the main problem. The small 5m pillars could
not accommodate the reduction in width caused by the floor extruding into the bords and
thus failed. Widespread floor heave and pillar cracking indicated that the failure process
was continuing and that extensions to the fall could be expected.

4.2.5 SIMRAC investigation

CSIR Miningtek studied the problem in more detail as part of a SIMRAC project (Jack and
Madden, 1995). In addition to the observations of Buddery they found that there were
isolated cases of roof damage and intersection failures soon after mining. Roof support in
this panel consisted of between four and five mechanically anchored bolts per row.

The vast majority of falls occurred after the area was sealed off and consisted of either roof
slabbing or channel falls (Figure 4.5). Slabbing generally terminated 0.3 to 0.4m int o the
roof at the interface of the massive sandstone. Some of these falls are up to 40m long and
while many isolated ones occur in intersections, many run through both bords and
intersections.
Channel falls generally lie between roofbolts or between the pillar sides and closest
bolts. While the height into the roof of these falls is not recorded, they appear from the
photographs accompanying the report to terminate in the weaker lower portion of the roof
(0.2m). The channels are also randomly orientated with the longest being about 30m
long. Many of them were observed to have changed direction through 90. The lack of any
clearly defined directional trends for both types of roof falls indicates that there is no over -
riding horizontal stress direction. The main types of pillar failure indications were vertical
cracking, sidewall spalling, pillar dilation and pillar corner failure. In many instances the
pillars contained vertical cracks close to the edge of the failure area. These cracks were
essentially wedge shaped, being wider at the base of the pillar ( Figure 4.7). This type of
failure is consistent with weak floor flowing from beneath the pillar and tearing the pillar
side apart. Slabs were between 250 and 500mm wide, being narrower at the top than at the
base.
47

Figure 4.7. Two pillar sidewall failure mechanisms at Matla.

In places, where near vertical cracks occurred closer to the pillar skin the pillars appeared
to have spalled. This spalling was due to a similar mechanism to that mentioned above, in
that deterioration was driven by the floor flowing beneath the pillar edges.

Some pillars exhibited what at first looked like stress induced pillar dilation with
horizontal cracking at mid-height (Figure 4.7). This was actually a slab buckling type
failure initiated in the floor but with the roof contact remaining intact.

Pillar corner failure was also common and had a simil ar mechanism to that of sidewall
spalling. In some cases the slabs were wedge shaped (wider at the base) most likely as a
result of the strength of the pillar / roof interface.

In addition, the pillars were rated according to the degree of damage observed (Figure 4.5).
As in the Emaswati case, an attempt has been made to fit these to the ten point rating scale
in Table 4.4. There was no significant connection between pillar conditions and distance
from the major collapse area with patches of moderate to severe pillar damage as much as
200m away. These patches were largely confined to the central region of the panel.

Table 4.4. Pillar damage rating.
Damage
rating
Description of damage
10 pt
rating
Description
Nil No visible damage 1
Intact, CM cutter drum marks
visible over more than 90% of
pillar surface.
Minimal
One corner cracked through
or
One side spalling
2 Pillar corner damage only
Moderate
One corner cracked through
and
One side spalling
4
Spalling, more than 50% of pillar
affected
Severe
More than one corner cracked
through and one or more sides
spalling
6
Failure along joints more than 50%
of pillar affected, hourglass shape.
Very severe
Vertical cracks through the
centre of the pillar
8
Pillar dilation, open gap at top
over more than 50% of pillar.
Sidewall
slabbing
Pillar
dilation
48

It is likely that the amount of roof material on the floor disguised some of the floor heave.
In addition, the floor had been damaged by mechanised mining and appeared loose and
powdery.

Floor heave was detected in some areas where no significant pillar damage had occurred.
Conditions in panel R14 West were near perfect even though minor floor heave was
observed. The only major difference between this panel and the failed one is that the pillar
centres were 12 instead of 11m.

Impact splitting results showed that the seam itself was rated as largely good in the
borehole most representative of the good panel but poor t o moderate in the collapsed one.
The coal is often not geotechnically tested as the geologist generally samples it for coal
quality purposes before any geotechnical work can be done.

MINLAY computer analysis of the failed and stable panels showed that the stable panel had
lower stress levels by 5 and 6% at the edges and centres of the pillars respectively.

4.2.6 4 seam pillar collapse

In March 2002 a collapse covering 74 pillars occurred in 4 seam workings about 1000m to
the east of the 5 seam collapse area. The collapsed pillars had safety factors of about 1.4
and pillar width to height ratios of approximately 2.

Severe pillar spalling of up to 0.7m was observed in nearby panels. The 4 seam at Matla 1
has been recognised for many years as being more prone to spalling than the 4 seam
elsewhere. Even at Matla Colliery 4 seam pillar conditions are far better at number 2 and 3
shafts as well as in the new South Eastern reserves at Matla 1.

It is possible that the underlying dolerite sill had some influence on the behaviour of the 4
seam. This pillar collapse is mentioned because the underlying sill may have influenced the
strength of the sandy shale floor of the 5 seam.

4.2.7 Conclusions

The soft sandy shale floor was further weakened by water and dykes as well a s (possibly)
the underlying dolerite sill. This weak floor material was then squeezed out from beneath
the harder coal pillars. The floor damage has been classified as closed ridge type heave.

As the pillars were only 5.1m wide the effect of 200mm to 500mm wide slabs being torn
off is significant. This was caused by the pillar foundations flowing into the bords. In the
worst case, this would leave pillars with an effective width of only 4.1m. The resultant
safety factor and width to height ratio of the r educed pillars would have been 1.3 and 1.7
respectively.

The initial failure occurred between the south eastern barrier and the long pillars left along
the dyke as shown in Figure 4.5. This would be consistent with the load being applied by an
elastic beam resting on the previously mentioned barrier and long pillars.
49
4.3 Welgedacht, Alfred seam stooping
4.3.1 Introduction

A combined goaf overrun / pillar failure on 14
th
June 1996 resulted in a whole continuous
miner sections equipment being trapped during stooping operations on the Alfred seam.

Cover depth: 165m
Mining height: 3.8m
Pillar centres: 24.4m (80ft)
Bord width: 6.3 8.0m*
Safety factor (Salamon): 1.18 1.51
Pillar width: 16.4 18.1m
Pillar width/height ratio: 4.32 4.76

*Due to weathering induced deterioration at the top of these old pillars (mined pre -
metrication, i.e. before 1971), the actual bord width was as much as 8m in places, as shown
in Figure 4.9.

4.3.2 Geology and geotechnical data

The roof of the 3.8m thick Alfred seam is competent sandstone while the immediate floor
consists of 3.5 to 4m of interlaminated shale and sandstone with thin coal bands. As there
are no boreholes in the immediate vicinity due to the mountainous nature of the terrain, this
geological information is based on observations in an Alfred to Gus seam box-hole 250m
away. Below this parting lies the 2m thick Gus seam, which had been mined over about
half of the area (Marais, 2002).

The Gus and Alfred seam pillars were superimposed due to the parting being thin. The Gus
seam pillars were felt to be considerably stiffer when compared to those on the Alfred seam
(Wevell, 2002). At one site within the failure area the floor heave was cleared away to
reveal a medium to coarse grained sandstone lying about 0.46m into the floor. This
sandstone bed was visibly discoloured and appeared to be weathered (Wevell, 1996a).

During September 1996 floor samples were taken from holes blasted into the floor in a
panel in which stooping was being investigated. Of the sites sampled, site A was
considered the most relevant as it was closest to the pillar failure area. Slake durability
index and moisture content were determined for various floor bands (Wevell, 1996b) and
are shown in Table 4.5.

No other geotechnical data could be traced for this area and as it is no longer accessible it
is unlikely that further data will be obtained at this stage, particularly since the mine was
sold to Kangra Mining by Ingwe Collieries in about 1998.

50

Figure 4.8. Welgedacht, Alfred seam, section entrapment area.

51

Figure 4.9. Welgedacht. Effect of weathering on bord width.

Table 4.5. Welgedacht, Alfred floor, geotechnical results.
Band Thickness
(m)
Lithology Slake
durability
index
Moisture
content
(%)
1 0.08 Laminated siltstone and shale (often missing)
2 0.23 Black carbonaceous shale, highly fractured.
Interfaces of fractures, wet muddy and
discoloured. Water seeping from contact between
bands 2 and 3.
11.4 3.03
3 0.19 Coal with thin shale band at centre. The coal
above the shale band was highly fractured, wet
and discoloured.

4 Siltstone / shale laminated. Well preserved and
competent.
1.14 1.36

4.3.3 Sequence of events

The following is the sequence of events leading up to the collapse and a brief description of
the equipment recovery stage (Oldroyd, Wevell and Kruger, 1996).

Day 1
On the morning of 14
th
June 1996, the miner heard bumps coming fr om the floor during
stooping operations. Later in the shift, more bumps occurred and the CM was trapped by
sidewall movement.

The CM was trammed back but became trapped again resulting in the use of two he -men
to pull the machine out. These are powerful winches with, in this case, pulling capacities of
6.3m
8m
52
250 and 300tonnes. They are housed within what are essentially longwall powered support
units set tight between roof and floor.

Due to the resumption of intense bumping the section was abandoned. The Mine Overseer
investigated the area and due to the noises coming from the panel felt that the inter -seam
raises about 250m away could be overrun.

Day 2
The General Manager and Mine Overseer tried to assess conditions but could not get close
to the equipment. Loud bumps occurred every 5 to 10 minutes.

Day 3
Personnel could get closer to the stooping line where it was observed that the belt and some
other roads had been severely affected by spalling. In one road the roof had collapsed
totally while in others slabbing had occurred. The trapped machinery was not visible.
Bumps were still occurring and it was decided to monitor the situation only and not attempt
retrieval of equipment.

Day 4 and beyond
The section was visited by rock engineering personnel and photographed while the access
road started to be re-opened. At this stage the bump frequency had reduced noticeably.

Floor heave and cracking were visible within 100m of the goaf edge and roof cracks and
falls were observed. Pillar sides were damaged and visible movement was clearly occurring
along prominent joints in pillars. It appeared that pillars had been punched into the floor,
thus initiating the failure process.

A week later, the area was much quieter and the access road cleaning and re -support had
advanced a further 25m. It was possible to see the trapped CM some 50 to 60m further in.
Floor heave was as high as 1.5m in places with some floor slabs standing nearly vertical
(Figure 4.10). A shuttle car had been pushed right up against the roof of an initially 4m
high bord.

The sandstone roof was severely cracked in places and roof falls (slabbing 0.3 to 0.5m
high) had occurred in several roads. Severe pillar spalling had occurred, trapping another
shuttle car.

Retrieval
While most of the equipment was fairly easy to pull out once exposed, the CM required
three he-men pulling simultaneously to remove it. The recovery was more difficult as the
CM was trapped by sidewall movement inside the pillar it was busy e xtracting at the time
of the incident.

A roofbolter was buried in a major roof fall and a shuttle car was lost as a result of
secondary pillar failure caused by the cleaning up operations associated with the recovery.
It was considered too dangerous to continue with the recovery of the shuttle car.

After the equipment was extracted, movement commenced back-bye and the coal started
burning.
53

Figure 4.10. Near vertical floor buckling mechanism.

4.3.4 Discussion

Additional loads imposed by stooping caused the pillars to punch into the floor. After this
the pillars began to fail in a fairly stable fashion, as they had a high width to height ratio
(4.32 to 4.76).

It is possible for pillar failure to take place with very little convergence, often less than 2%
strain. In this case, it would be 76mm of convergence (Oldroyd, Wevell and Kruger,
1996). For complete stability due to recompaction, the pillars would need to be compressed
by 30 to 40% of their original height. This would then mean that the overlying strata would
break up, as coal measures rocks at shallow depth will not deform to this extent in an
elastic manner. The recompaction of the failed pillars close to the goaf edge crea ted,
temporarily, a relatively safe environment in which the equipment could be recovered.

Due to the roof being cracked, severe pillar scaling and substantial floor heave, it was
necessary to clean up spalled coal as part of the recovery operation. This lead to a change
in equilibrium and made it vital that cleaning up was kept to a minimum.

Work done by Wagner (1974) shows that as a pillar fails the load on the core increases
(Figure 4.11). At later stages of pillar failure, the load generated may be high enough to
cause floor punching.

An alternative mechanism is that the edges of stable pillars are pulled apart as they punch
into the floor, eventually causing failure.

It may be that both mechanisms occur and it may be difficult to determine which is the
dominant cause.

Initial
buckling
Further flow of floor
rock from beneath pillar
Near
vertical
buckling
A
B
54

Figure 4.11. Load distribution through a failing pillar (after Wagner, 1974).

The floor heave was seen to form near vertical ridges in places. These ridges possibly
indicate horizontal movement of floor beams as sketched in Figure 4.10 and are a severe
form of the closed ridge type floor heave described in 4.1.7.

From the surface contours on Figure 4.8 it can be seen that the collapse area coincides
with the top of a hill on surface. Goaf behaviour is erratic at Welgedacht and it is not
uncommon for the thick sandstones, which comprise the mountains above the mine, to
bridge across total extraction panels. The fact that pillar conditions remained stable to the
north east of the failure area may be due to the rapid decrease in cover depth on this side.

The underlying Gus seam workings may have influenced the collapse and it is interesting to
note that much of the trapped equipment lay close to the edge of the underlying Gus seam
workings. This may, however, be purely coincidental. Numerical modelling carried out in
Section 6.3.3 was used to try and shed light on this problem.

4.3.5 Conclusions

The load on the pillars in this case was greatly increased by the following:
o The goaf was suspected not to have reached surface.
o The resulting bridging would have increased the load on the pillars awaiting
extraction.
o The collapsed area lay beneath the highest point of the mountain.

Coupled to this was the pillar weathering problem outlined earlier and illustrated in Figure
4.9. This results in the actual safety factor of the pillars being as low as 1.18 even before
the effect of additional loading by the bridging goaf is taken into account.

The immediate floor, consisting of 2m laminated sandstone and shale, was not capable of
supporting the additional pillar load and started to flow out from beneath the pillar
resulting in closed ridge type floor heave. Subsequently severe floor heave and
accelerated pillar spalling occurred.

55
As the spalling was generally not cleaned up it provided considerable lateral constraint.
The confinement, coupled with the pillar width to height ratio of 4.32, resulted in the
pillars failing in a slow and controlled fashion.

4.4 ZAC, Mngeni shaft, panel MEN1
4.4.1 Introduction

Pillar deterioration was first noticed during June 2000 in panel MEN1 at the Mngeni shaft
of ZAC soon after completion of chequerboard pillar extraction of a block of 40 pillars on
the eastern side of the area under discussion. Development was planned in-bye of this area
and a B10 (stone work development) crew was busy some distance in-bye of the pillar
deterioration area. In light of the initially measured rapid closure of between 4 and 15mm
over a two-day period, it was decided to suspend the B10 operations ahead of this area
until further investigations had been carried out.

Cover depth: 102 to 151m (average 126.5)
Seam thickness: 2.74 to 2.80m (average 2.77)
Pillar centres: 22m
Bord width: 6.11 to 6.41m (average 6.26)
Safety factor (Salamon): 1.72 to 2.70 (average 2.20)
Pillar width: 15.59 to 15.89m (average 15.74)

4.4.2 Geology

The roof of the 2.7m thick Main seam in this ar ea comprises of about 10m of massive to
cross-bedded, coarse-grained sandstone. The floor composition is variable, described as
3.65m of stratified sandstone in borehole Mg39. In borehole Mg5 the floor is 0.22m of
shale underlain by 0.71m of sandstone and 0.22m of coal.

As is common at ZAC, the area is structurally quite complex with the three major features
being:
o A 3m upthrow fault along the western side of the panel.
o A 3m wide dolerite dyke running along part of the western side.
o Another fault with a 4m upthrow running through the panel with a SW NE trend
and delineating the north-western limit of the pillar deterioration area.

Numerous other small faults (typically with throws of 0.1 to 0.3m) lie within the area of
interest and generally follow the t rend of the SW to NE 4m fault, mentioned above. Open
roof cracks roughly trending E to W occur close to the chequerboarded area.

The seam dips about 6 towards the northeast. It is this dip, plus a rising surface
topography, that lead to a steady increase in cover depth to a point, further in-bye in the
same panel, where the cover depth is in excess of 300m.

56

Figure 4.12. Mngeni, pillar deterioration area.
57
4.4.3 Geotechnical testing

No boreholes have been drilled in this area since the problem was identified and it is
unlikely that any will be in future. A few boreholes were geotechnically logged and tested
near the Mngeni incline shaft box-cut, (Buddery, 1993 and 1994). Results obtained are
summarised in Table 4.6.

Table 4.6. Mngeni, geotechnical testing.
Borehole
number
Distance from
collapse area
Mg17 620m SSW Excellent: IS=200 / RQD=75.5.
Roof strata SD 5.5 - class A
11.35m massive, feldspathic,
coarse-grained sandstone.
0.55m good - SD=9.5
(Good)
0.55m interlaminated shale / shaly
sandstone and sandstone.
Mg18 590m WSW Excellent: IS=269 / RQD 86.2
Roof strata SD 1.9 class A
0.88m good - SD 2.4 to 3.4
(Good)
0.17m interlaminated shale and sandy
shale underlain by 0.71m
interlaminated sandstone and shaly
sandstone.
Mg19 380m SE Excellent: IS=221 / RQD = 99.3
Roof strata SD 2.2 class A

0.08m moderate - SD 12.6 underlain
by 0.15m moderate - SD 14.8
(Good)
0.08m interlaminated shale and coal
underlain by 0.15m shale / sandy
shale.
Mg23 550m SW Excellent: IS=169 / RQD=85.2
100
Roof strata: SD 2.3 6.3
class A
0.09m shale (false roof band
to be mined down) to expose -
0.16m very hard sandstone with
subordinate shale bands, has
distinct upper parting and will
need to be supported from the
main sandstone roof, consisting
of 11.88m coarse grained,
feldspathic sandstone.
0.74m good to moderate SD 4.9 11.5
/ RQD 50.0 24.1
(Good)
0.20m vertically fractured shale
underlain by 0.54m interlaminated
shaly sandstone and sandstone.
Averages Excellent: IS=215 Weighted SD for 0.5m of floor 7.5
(Good)

The following pertinent comments were made with respect to the likely floor conditions:
o Floor strata were different in each tested borehole and as such, floor conditions
may be variable over the entire reserve area.
o The floor in borehole Mg19 could tend to break due to blast damage and tramming.
o The floor in borehole Mg23 is of low strength with a slight tendency to slake and
could break up under regular tramming of vehicles.

Moisture content tests (Annexure A) indicated that the total moisture content (saturated)
was 2.27% made up of 0.93% existing and 1.34% absorbed.

58
4.4.4 Background

Mngeni shaft has always been problematical due to the high incidence of large faults,
(typically from 3 to 7m throws) and dolerite dykes. As a result, a shortage of pit room is a
common problem, which has often resulted in smallish areas of pillars being extracted near
active developments. The three main forms of pillar extraction at ZAC are as follow
(Figure 4.13):
o Chequerboarding: in which every second pillar is extracted.
o Turbo Chequerboarding: in which, in addition to the pillars extracted by
chequerboarding, the remaining pillars are reduced in size by additional sliping.
o Large pillar chequerboarding: where, in order to reduce roof spans, a reduced size
pillar is left at the corner of each pillar. This method is used for pillars 24m wide
(centres 30m).

It is required in all these methods that the pillar remnants have a width to height ratio of at
least four.

Figure 4.13. ZAC, typical chequerboard layouts.

Chequerboarding Turbo chequerboarding
Large pillar chequerboarding
Pillar
remnant
Mined
59

Turbo chequerboarding had been carried out in an area of 12 pillars long and 4 deep on
the right hand (eastern) side of the affected panel (Figure 4.12).

Soon after the chequerboarding had been completed, signs of pillar deterioration were
noticed. The area was visited during July 2000 by J Latilla and DC Oldroyd and at that
stage, it was considered that most of the deterioration had probably already occurred. As a
precaution, the establishment of convergence monitoring stations was recommended.

No spalled coal was loaded out from the area and coal in the travelling way was pushed
into the splits.

Convergence stations, consisting of post grouted mechanically anchored roofbolts,
anchored 0.5m into both the roof and floor, were established towards the end of October
2000.

Initial readings were taken by both rock engineering and mine personnel on 23 October
while a second set was taken by the mine personnel two days later. It was at this time that
the rapid closure mentioned earlier was measured and a decision taken to remove the B10
crew from in-bye of this area.

4.4.5 Observations

Extensive underground mapping at the end of October 2000, (van Wijk and Latilla, 2000)
revealed that the majority of pillar damage lay between the belt road and the
chequerboarded area (Figure 4.12). On the previously mentioned 10-point scale, the rows of
pillars had the following average ratings, moving away from the chequerboarded area; 6.7,
5.9, 4.9, 4.2, 4.1, 3.3, 2.9, 2.8, and 2.1.

The maximum pillar damage was, as expected, along the chequerboarding goaf edge
while the least damage occurred along the western (left hand) barrier. In chequerboard
extraction the area of extracted pillars is referred to as goaf but in contrast to total pillar
extraction, the roof normally stands up for some time. Roof slabbing in the goaf often
only occurs as much as four rows of pillars in-bye of the row being extracted.

Extensive roof slabbing, up to 3m high, was observed within the southern two thirds of the
goaf. In addition, there were places within the goaf where slabs of sandstone roof, up
to 0.5m thick, had slid up to 5m down dip along the floor.

Floor heave was common to the east of the travelling road and within 50m of the NE to SW
trending fault to the north west of the pillar deterioration area. No heave was observed in
the travelling road, most probably due to tramming of equipment. It may be that the 0.3m
of compacted road material was also capable of preventing heave. Floor heave was found to
be far more prevalent in bords (80%) than in intersections ( Figure 4.12).

Floor cracks were also common, the largest observed being 170mm wide and at least 1m
deep. These were largely confined to the east of the travelling road. In a large number of
places, the cracks were associated with zones of shattering of the pillar sides. The floor
cracks are generally absent in intersections and usually run perpendicularly across the
60
roadways. Floor plates riding over each other were another notable feature and were mostly
confined to the east of the travelling road. As may be expected, the ride direction was
largely down dip, the dip in this area being 6 in a SW to NE direction.

Open roof cracks were observed, many of which appeared to be freshly opened. These were
largely confined to within 50m of the goaf edge and trending towards it. The roof
appeared stable over the whole area and observations up pilot holes (drilled to check roof
composition during development) and borehole Mg58P, confirmed that there were no open
partings or lateral movement visible.

A very large number of mine poles and cluster stick packs had been installed. Many of
these had broken or buckled.

Roof pegs along the travelling road were re-surveyed (by the mines survey department) to
establish how much roof closure had occurred since development ( Figure 4.14). The
maximum deflection measured was 95mm and a significant amount of ride (25 to 86mm)
was also picked up. The directions of the larger ride components were roughly at right
angles to the large (4m) fault and towards the goaf, and were measured between the
major and smaller (0.3m) faults. The ride direction south of the smaller fault is up -dip and
parallel to the roadway. Slippage along the fault with a slight scissors effect (more
movement near the goaf) largely explains this ride, which is also smaller at between 25
and 27mm.

An earth tremor was reported at 7pm on Thursday 26
th
October. According to the article in
the Zululand Observer and a subsequent telephone conversation with the reporter, the
tremor moved inland from an epicentre some 40km off the coast at St Lucia. It registered
3.1 on the Richter scale and was caused by a tectonic slump along a fault parallel to the
KwaZulu-Natal coast. As this tremor occurred about 1 day after the major convergence was
measured it would seem to indicate that it was not connected to the pillar damage.

Contrary to the sudden movement measured soon after the monitoring stations were
established, Figure 4.15 indicates that the convergence became more consistent. Occasional
sudden rapid movement took place in localised areas and this occurred almost as a ripple
effect. Overall convergence measured on a four to six week cycle does not show this jagged
effect but rather a smoother curve.

No dramatic movement was detected along any of the faults in this area, although 1mm
movement and slight damage within 1m of the large 4m fault was evident (Oldroyd,
2000).

Roof and floor tell-tales were installed next to one of the convergence stations. The floor
tell-tale indicated 1mm of movement over about 6 weeks. It was subsequently damaged and
could no longer be used for monitoring. No roof separation was detected.

Sixty metres of joint scan lines were mapped and these revealed that the dip of slips in the
pillars ranged from 45 to 60 and their frequency was between 1.45 and 1.9 per metre.

While the floor appeared dry it must be borne in mind that during development, the faces in
the eastern half of the panel always had water in t hem. It was common for roof cracks to
make water until they were about 30m or so back-bye of the faces. This water would then
lie in the down dip faces and could have saturated and weakened the floor strata.

61

Figure 4.14. Mngeni, deflection and ride of survey pegs and position of convergence
monitoring stations.
62

Mngeni convergence
-60
-50
-40
-30
-20
-10
0
O
c
t
-
0
0
N
o
v
-
0
0
D
e
c
-
0
0
J
a
n
-
0
1
F
e
b
-
0
1
M
a
r
-
0
1
A
p
r
-
0
1
M
a
y
-
0
1
J
u
n
-
0
1
J
u
l
-
0
1
A
u
g
-
0
1
S
e
p
-
0
1
O
c
t
-
0
1
N
o
v
-
0
1
D
e
c
-
0
1
J
a
n
-
0
2
F
e
b
-
0
2
M
a
r
-
0
2
A
p
r
-
0
2
M
a
y
-
0
2
J
u
n
-
0
2
J
u
l
-
0
2
A
u
g
-
0
2
S
e
p
-
0
2
O
c
t
-
0
2
N
o
v
-
0
2
Date
M
o
v
e
m
e
n
t

(
m
m
)
stn 0 stn 1 stn 2 stn 3 stn 4 stn 5 stn 6

Figure 4.15. ZAC, Mngeni, convergence monitoring.

The area was visited separately by Madden and Oldroyd in December 2000 and some of
their observations and conclusions are quoted below:
Madden (2000) stated, inter alia, that:
o Pillar deformation occurred because of increased load after chequerboarding in
the eastern portion of the panel. The thick competent sandstone allowed load
transfer effects to be seen for up to six rows of solid pillars from the goaf
edge.
o Heave and tension cracks in the floor may be due to pillar dilation driving the
weak or layered floor strat a. The effect on the floor is exaggerated by the seam
dip.
o Pillars with high width to height ratios are thought to deform in a stable manner
and of utmost importance is the rate of deformation. In this case, the maximum
measured rate of roof to floor convergence was about 2.3mm per week.
o Remnant pillars within the chequerboard goaf appeared to be able to carry
load. In addition, pillars confined by slurry or rubble appeared to have
suffered less damage.

Esterhuizen (1997) studied the effect of jointing on coal pillar strengths and, using the
previously mentioned scan-line survey results, Madden concluded that the pillars in this
area could have a strength reduction factor of 0.7 to 0.75 applied to them. This would
reduce the safety factor in this area to between 1.20 and 1.89 from 1.72 and 2.70
respectively if the more conservative value is used.

According to Oldroyd (2000):
o Increased floor movement had occurred, in the form of dilation cracks. In order
for the floor to dilate, there must be compressional features present to
accommodate the horizontal movement. This is illustrated in Figure 4.16.
63
o Theoretically, pillars with a width to height ratio of five and more should have
an ever-increasing load deformation graph and thus controlled pillar
deformation. Oldroyd stated that there would be periods of faster closure and
then some slowing. The total picture is quite complex as each pillar would have
its own characteristic graph and be on a different part of it at any time. Periods
of more rapid movement would, in all probability, tend to ripple though the
panel.

He also stated that a general breaking up of the roof would lead to a softening of the
loading system and could result in an increased rate of closure.

Figure 4.16. Mngeni floor failure mechanism (after Oldroyd, 2000).

4.4.6 Remedial work

Numerous steps were taken to safeguard workers and monitor conditions, while allowing
coal to be mined in-bye of this pillar deterioration area:
The convergence monitoring stations mentioned earlier were read on a four to six week
cycle by members of IRED while the shift bosses at the shaft measured them every
week.
One to two thirds of each split on the goaf side of the travelling road was filled with
rubble to about half the pillar height ( Figure 4.12).
Three four-pole cluster stick packs were installed in each split on either side of the
travelling way through this area. These were replaced when damage to them made it
necessary.
Most of the damaged pillars adjoining the travelling and belt roads have been wrapped
with weld mesh and strapped with 13mm hoist rope tensioned to 5tonnes. A number of
mine poles were installed between the cables and mesh to improve contact between the
mesh and pillar sides (Figure 4.17). Difficulty installing bolts in the pillar side was
experienced, as the holes would close immediately after drilling.

Floor
dilation
Floor
thrusting
Floor
heave
64

An electronic closure meter was developed on the mine which provides a flashing light
alarm if more than 2mm/day of total closure occurs. In conjunction with this,
evacuation drills were carried out on a regular basis for people working in-bye of this
area.
Roof cracks were being monitored by means of steel wedges to see if they were
opening up.
The rescue bay in-bye the pillar deterioration area was checked t o ensure it was fully
functional.
To monitor roof behaviour beneath a deforming beam, the installation of 1.8m long
roof tell-tales was recommended.

The chequerboarded area gradually deteriorated and by June 2002 had goafed over most of
the area. Goafing of the chequerboarded pillars has led to a significant reduction in pillar
loading and is considered one of the main reasons for the reduction in the convergence rate
Figure 4.15).

4.4.7 Conclusions

Chequerboarding of the eastern portion of the panel was the prime cause of the pillar
deterioration as the increased load on the pillars caused pillar dilation as well as
overloading of the core of the pillars. This is borne out by the reduction in the convergence
rate since the chequerboarded area goafed.

Furthermore, the floor is weak in places and has been displaced along bedding planes.
Observations indicate that the pillars have both punched into the floor and slid slightly
down dip towards the bottom of the chequerboarded area. Displacement was indicated by
the opening up of floor cracks and floor ride ridges. No signs of floor material squeezing
from beneath the pillars was noticed. The zones of shattering in the pillars indicate that the
Figure 4.17. Mngeni pillar wrapping.
13mm
diameter
wire rope
Weld mesh
100mm
aperture
Mine poles 125mm
diameter
NOT TO SCALE
65
floor cracks developed quickly. The floor damage can best be described as open topped
floor heave but with occasional closed ridge features probably caused by pillar ride.
Floor heave was more prevalent in bords than intersections.

The massive sandstone roof was the primary driving mechanism and the greatest damage
was close to the edge of the chequerboarding area which was where the greatest roof
deflection would be expected.

The high width to height ratio of the pillars in this area ensured that the pillar deterioration
occurred in a slow and controlled manner.

It is possible that accumulations of water on the floor, in faces on the eastern half of the
panel, may have further weakened the strata.

4.5 Other ZAC cases

Three other cases of pillar failure associated with floor damage have been recorded at ZAC.
Unfortunately, there is not a great deal of information available and they are mentioned
more for the sake of completeness. Another reason for excluding these from the analysis is
to avoid what would otherwise be a large bias towards the ZAC experience.

4.5.1 North shaft, panel NW2N

This panel was laid out on 14m centres with a pillar design safety factor of 1.6 in a trial to
optimise percentage extraction as an alternative to stooping:
Cover depth: 50 to 66m (average 58)
Pillar centres: 14m
Bord width: 6.67m
Pillar width: 7.33m

Due to observed pillar damage and roof fracturing as well as floor damage, (Ha ttingh, 1987
and Oldroyd and Buddery, 1988), the pillar centres were increased to 18m and the area was
visited frequently by rock engineering staff. Other remedial actions taken and observations
made include :
o Timber breakerlines (described as being const ructed from heavy mine poles)
constructed along the belt road and a second means of egress was established.
o Roof to floor convergence was monitored by the mine and the rock engineering
department and was found to be relatively steady at about 1mm per day.
o While there was no significant deterioration of the roof, extensive pillar
damage was noted.

On a subsequent visit, Buddery (1987) noted that convergence rates had increased to
between 1.68 and 1.75mm per day. He stated that consideration should be give n to
destroying two rows of pillars near the problem area to arrest a spread of the failure.

66
The experience gained with the failure of the pillars in this panel was published by Oldroyd
and Buddery (1988). Some of their findings are considered relevant t o this study and are
listed below:
o The floor consisted of layers of siltstone, fine to medium grained, micaceous sandstone
and carbonaceous shale with occasional coal stringers.
o The pillars were structurally weak with consequently low stiffness and were co upled
with a stiff and competent sandstone roof. In stooping operations elsewhere on the
mine, leaving of partially extracted pillars due to faults or floor heave caused goaf
hang-ups.
o Convergence rates eventually reached as much as 3.3 to 4.7mm/day. A sta tion
monitored by the mine showed movement of up to 10.6mm/day. By this stage, floor
heave had become a problem with floor ride type features about 120 to 150mm high.
o At the same stage the area was described as talking continuously with new roof and
floor cracks and increased spalling being noticed on each visit.
o Up to 1.09m total convergence was estimated to have occurred. This is equivalent to
41% of original pillar height.

4.5.2 South shaft, section 3

With the exception of four rows of pillars, this panel was stooped even though pillar failure
was occurring.

The mining geometry at this site was as follows:
Cover depth: 74.4 to 87.4m (average 80.9)
Pillar centres: 17m
Bord width: 6.5m
Pillar width: 10.5m

Increased convergence of between 7.2 and 24mm/day, as well as visibly increasing pillar
spalling and, in the worst area, serious floor and pillar cracks, led to stooping being
stopped after about four rows of pillars had been extracted (Buddery, 1988a). In this area
the seam dip was very steep, by ZAC standards, at 11. The floor consisted of 0.21m of
carbonaceous shale underlain by 1.75m of laminated, micaceous, sandstone and shale.

Stooping re-commenced 4 rows of pillar back and the following comments were made after
a rock-engineering appraisal (Buddery, 1988b):
o The pillars and floor, and not the roof, were deteriorating locally.
o Convergence of 10.6mm/day was measured close to the stooping line but none 6
rows of pillars behind it. If the rate of convergence at any station should double
over any two shift period, all other stations should be read to check if the trend was
similar throughout the area. If this was so, then management would need to be
informed immediately.

A possible eventuality could be that the pillars and floor would soften to such an extent
that the goaf would not break off, leading to overruns.

67

4.5.3 South shaft, chequerboarding

During a routine visit, it was noticed that the remnant pillars (after turbo chequerboarding)
had failed (Latilla, 1998). Pillar and panel geometries were as follows:
Cover depth: 133.8m
Pillar centres: 25m
Bord width: 6.5m
Safety factor (Salamon)- before chequerboarding: 2.17
- after chequerboarding: 0.60
Pillar width before chequerboarding: 8.5m
Estimated remnant pillar width: 4.5m
Pillar width/height ratio - before chequerboarding: 6.10
Width/height ratio of remnant pillar: 4.79
Load on remnant pillar 6.77MPa

Amongst others, the following observations were made:
o Roof to floor convergence (estimated from buckled mine poles) was of the order of
0.2 to 0.4m and floor heave was also observed.
o The bottom of a set of mine poles as well as a ventilation wall were displaced down
dip, most likely as a result of sliding floor plates.

4.6 Conclusions from the study of the failed cases

In section 8 comparisons are made between the failed and stable cases. It is in this section
that the main conclusions are made for both types of cases.

5 STABLE CASES

While the study of failed cases is self-evidently important, it is considered equally
desirable to investigate a number of stable cases in order that a comparison be made
between the two types. This is in order to improve prediction of floor and pillar behaviour.
Studying only failed cases could lead to a one-sided appraisal of the problem.

Stable cases are defined as instances where, although the soft adjacent st rata showed signs
of damage, there were no signs of associated pillar deterioration.

In the following section the stable cases are studied to try and determine why no failure
occurred in these panels even though soft floor was present. Mining geometries a nd
available geotechnical data are given for each site.

68
5.1 Matla 1, 5 seam, Panel R14 West

Cover depth: 34.2 to 36.4m
Mining height: 2.2 to 2.4m
Pillar centres: 12m
Bord width: 5.9 to 6.1m
Safety factor (Salamon): 2.4 to 3.0
Pillar width: 5.9 to 6.1m
Pillar width/height ratio: 2.5 to 2.8

Panel R14 West, developed to the north west of the failed R14 South panel, was stable at
the time of the failure event and indications are that it remains so to this day. Cover depth
and mining height in these two panels were similar at around 34.9 and 2.3m respectively. In
addition, the centres in panel R14West were 12m as opposed to 11m in the failed area. The
lithology of the immediate floor in panel R14West is shown in Table 5.1 (Latilla, 1996a).

Table 5.1. Matla 1, panel R14West, floor lithology.
Borehole Thickness
(m)
Description Slake durability
index
H85/57 0.32 Siltstone and mudstone -
H85/98 0.42 0.16m Sandy shale
0.26m Shale / sandstone, interlaminated
78.0
42.1
Weighted SD for
0.5m of floor 47.66
(Very poor)

The thickness of soft floor mined in this panel was on average 0.08m whereas in the failed
panel it was 0.31m. These thicknesses were obtai ned from coal quality control channel
samples taken by the mines Geology department and are considered a fair indication of the
relative strength of the floor rock in this area, considering that both panels were mined with
the same equipment, namely a Joy 12CM6 continuous miner and shuttle cars. The
percentage of the total thickness of possible soft floor mined dropped from 48 to 17%
between the failed and stable panels. In other words, almost half the thickness of floor
strata classified as potentially soft, was mined in the failed panel.

Only minor floor heave was reported in this panel (section 4.2.5), it is assumed to have
been of the closed ridge variety.

5.2 Matla 2, 5 seam, Panel M12 North

Cover depth: 57 to 63.7m (average 59.3)
Mining height: 1.7 to 2.0m (average 1.8)
Pillar centres: 15m
Bord width: 6.2 to 6.5m (average 6.4)
Safety factor (Salamon) 2.6 to 3.1 (average 2.9)
Pillar width 8.5 to 8.8m (average 8.6)
Pillar width/height ratio: 4.4 to 5.0 (average 4.7)
69

This is a seven-road bord and pillar panel, of which eight of the twelve boreholes along it,
or close to it, recorded the presence of soft floor (Latilla, 1996a). This panel was last
inspected in the early part of 2001 when the pillars, roof and floor were found to be in good
condition.

Little geotechnical work has been done on the 5 seam horizon in this area as the major
focus of previous investigations was the likely behaviour of the overlying strata above the
4 and 2 seam shortwalls (Latilla, 1993, 1996b and 1997b). Geotechnical results are
summarised in Table 8. Moisture content tests (Appendix A) indicated that the total
moisture content (saturated) was 2.09 to 4.03%. These tests were carried out on samples
obtained from the 5 seam at the Etingweni shaft, the closest accessible workings to the
currently closed Matla 2 number 5 seam.

Two different types of floor heave were observed, one being low, narrow compression
humps, generally less than 0.1m high and 0.5m wi de, while others were the more classic
form of floor heave where the floor cracked and slabbed in plates. The floor damage can
best be described as predominantly the closed ridge type. Pillar conditions throughout the
panel were good to excellent with the worst damage being occasional cracked pillar corners
and 0.1m of spalling.

Table 5.2. Matla 2, 5 seam, geotechnical results.
Borehole
number
R62/99 Good: IS = 94.6
The weak roof sandy shale is locally much thicker
in this area and is supported. The roof consists of
0.77m carbonaceous sandy shale overlain by 0.82
carbonaceous and glauconitic shaly sandstone
and 1.10m fine-grained sandstone.

0.82m shale / sandstone (IS unit rating 33
very good)
R62/100 Good: IS =108.7

0.27m highly laminated sandstone and shale,
micaceous and carbonaceous with 0.06m clay
band at base. This weak unit is mined down to
expose 1.25m massive, medium to coarse-grained
sandstone.
0.24m very poor - SD = 58.9
(Very poor)
0.15m shale /sandstone, soft and
carbonaceous underlain by 0.09m shale
and 0.30m sandstone / shale.
R62/101 Very good: IS = 161.7

0.14m sandy shale, carbonaceous and micaceous,
mined down to reveal 1.17m massive, medium to
coarse grained sandstone.
0.30m very poor - SD = 57.5 & DS = 0.4
underlain by 0.58m good - SD = 4.9 &
13.3
(Very poor)
0.30m sandy shale, carbonaceous and
micaceous underlain by 0.24m fine-
grained sandstone and 0.34m shale /
sandstone.
R62/167 Very good: IS = 116.1 (includes the weak shale
roof which is marginally stronger locally)
0.23m clayey sandstone, carbonaceous and
micaceous, overlain by 1.26m massive coarse-
grained sandstone.

0.34m sandy shale (I S= 9.7 poor)
underlain by 0.40m sandstone / shale (IS
= 21.4 good)
R62/170 Good: IS = 108.7
0.27m highly laminated sandstone and shale,
micaceous and carbonaceous, with 0.06m clay
band at base. This weak unit is mined down to
expose 1.25m massi ve, medium to coarse-grained
sandstone.

0.25m shale (IS = 5 very poor)
underlain by 0.16m sandstone / shale (IS
= 13 poor). Beneath this is a further
0.2m coal and shale (IS = 7 / 9 very
poor) Total weak floor thickness 0.61m.
Averages
Very good: IS=118 Weighted SD for 0.5m of floor 34.5
(Very poor)
70
5.3 ZAC, Mngeni, panel MN

Floor conditions were cause for concern during development of this panel, which at the
time was mining into virgin ground. The floor was broken to a depth of 0.15m (Latilla,
1995).

Cover depth: 100.1m
Pillar centres: 22m
Bord width: 5.98m
Safety factor (Salamon) before chequerboarding: 2.93
after chequerboarding: 0.55

Pillar width before chequerboarding: 6.02m
after chequerboarding: 10.44m
Pillar width/height ratio before chequerboarding: 6.23
after chequerboarding: 4.06

After turbo chequerboarding (Figure 4.13), severe floor heave was reported in the goaf
as was the fact that numerous mine poles had broken towards the centre of the panel.
Fractures and cracks were opening up in the pillars and hairline roof cracks were forming
as the extraction line approached. As a result, it was decided to change to normal
chequerboarding for a couple of rows of pillars due to concerns of a failure similar to that
at Welgedacht, where additional pillar loading caused the failure of the weak floor (section
4.3). This proved successful and the panel was chequerboarded back beyond this point
without incident.

A floor plate was observed which had lifted along the up-dip side of a sliped pillar. The
height of the resultant step was about 0.3m and appeared similar to the floor ride features
observed in the pillar deterioration area of panel MEN1 (section 4.4) which lies 40m to
the east of this one. An open crack, 0.01m wide and at least 0.5m deep was associated with
this floor step. Floor damage was largely confined to bords in this area.

Although no geotechnical tests were carried out in this area, borehole Mg9 falls within the
area of interest and the floor is described as consisting of 0.62m laminated sandy shale and
shaly sandstone underlain by 0.19m coal and a further 0.51m of shale. This represents a
potential soft floor some 1.32m thick. The seam itself is 2.53m thick overlain by 10.49m of
massive, coarse-grained feldspathic sandstone. Borehole Mg19 ( Table 4.6) is fairly close to
this area.

The floor heave in this area has been classified as the open topped type.

Moisture content tests (Appendix A) indicated that the total moisture content (saturated)
was 1.53% made up of 0.45% existing and 1.08% absorbed.

71
5.4 ZAC, Maye, poor roof case

Cover depth: 98 to 111m
Seam thickness: 1.47 to 2.04m
Pillar centres: 17m
Bord width: 5.67 to 5.89m
Safety factor (Salamon): 2.1 to 3.5
Pillar width (narrowest side): 11.11 to 11.33m
Pillar width/height ratio: 5.44 to 7.07

This case has been included as it was assumed at the time of mining in this area that the
coal pillars were punching into the roof. As such, this case is considered similar to the
other cases of weak floors being punched by the pillars, onl y in this case, the problem has
been turned on its head (Latilla, 1997a).

Of most interest in this case is that a number of roofbolt bearing plates (washers) were
observed to have buckled (Figure 5.1). As the thickness of soft roof was 1.2m, consisting
of laminated, micaceous, sandy shale, overlain by 10m of massive sandstone, the 1.5m
long, M16 rebar bolts used would have anchored in the sandstone. Tests on intact washers
showed that they collapse at 105kN at which stage the deformation was 7.5 to 9mm. The
average bolt spacing throughout the area was 1.30m (minimum 1.26 and maximum 1.37m).

A detached block of sandy shale, 1.2m thick by 1.3 x 1.3m in area, with a density of
2500kg/m
3
would weigh 49.74kN. The load on the roofbolts has been estimated as being
105kN which is roughly double that of the rock suspended from the bolts. It may be
reasonable to speculate that a portion of the excess stress responsible for overloading the
roofbolts is due to horizontal st resses and buckling of the soft roof beam.

Structurally the area was reasonably complex with a 1m fault and 3m thick dyke running
diagonally through it. In addition, the panel was stopped against another (2m) fault. A
borehole about 45m ahead of the fi nal face positions indicated a 7m thick dolerite sill
above the soft roof.

Prior to changing over to 1.5m long bolts, large portions of this soft roof area were bolted
with 1.2m long bolts operating in beam creation mode. A number of roof falls occurred i n
this area, some just above the 1.2m M16 full column resin bolts. Shallow angled slips were
common in the soft roof as were striations, both indicating a previously active tectonic
environment. Roof fracturing was frequently observed in roadways and start ed to open up
soon after loading out and often before support could be installed. This indicates that
horizontal stresses were influential in driving roof failure.

There was water present in most roof falls but inspection of the sandstone / shale interfa ce
did not reveal any excess of water or weathering present. Often there was a thin (5cm) coal
band at the top of the soft roof but it is considered unlikely that gas or water pressure
would have caused the roof falls due to the large number of holes drill ed through the
interface for roofbolt installations.

Geotechnical results are available for two boreholes in the Maye reserve block. One of
these has a massive sandstone roof and as such is in no way representative of conditions in
the area of interest. The other, Ma9, is quite far from the area of interest (950m SE) and
has 0.43m shale overlain by 0.37m sandy shale and 0.05m coal in the immediate roof.
Impact splitting unit ratings for this unit range between 4 and 15 (very poor to moderate).
72
Above the soft roof is 1.6m massive sandstone with a unit rating of 50 (Excellent). The
seam floor consists of 1.25m shaly sandstone / sandstone and is rated as Good.

Figure 5.1. Portion of Maye poor roof area.

5.5 Conclusions of study of stable cases

As was the case for the failed cases, the conclusions drawn for the stable cases are
presented as a comparison between the failed and stable sites in section 8.

6 NUMERICAL MODELLING

Numerical modelling has been carried out to simulate the stress and deformation state for
each of the case studies. Elastic three and two-dimensional methods were used with
MAP3D utilised to model the overall three-dimensional stress distributions and PHASE
2
for
analysis of bords and pillar foundations.
73

A major limitation of elastic modelling is that no failure can take place. This means that for
all the case studies it is only possible to model the stress and deformation state prior to
failure. This explains t he reason for the modelled deformations being an order of magnitude
smaller than those measured in the field.

6.1 Rock mass properties

The scarcity of geotechnical data from each site has made it necessary to refer to the
literature when assigning various properties to the rock types modelled for both the
MAP3D and PHASE
2
packages.

Table 6.1 shows a range of rock mass properties as published by various authors
(references in brackets). Where actual values are not available this table has been referred
to for material properties when setting up the various numerical models.

6.2 Three dimensional stress analysis

The MAP3D software package was used to attempt to model realistic pillar loads in both
the failed and stable cases. The use of a three-dimensional analysis was considered
important due to the presence of high extraction areas as well as dolerite intrusions close to
failed zones. A two-dimensional program would not be able to realistically model the
complex geometries of some of the sites. To illustrate this point, an example of the basic
model for the Mngeni case at ZAC is shown in Figure 6.1.

Rock properties assigned and used for MAP3D are shown in Table 6.2.

Based on local knowledge and experience, a variety of horizontal to vertical stress ratios
were used. These are summarised in Table 6.3.

Underground observations reveal that gutteri ng often only occurs on one side of a roadway
where horizontal stresses are present. This is typically due to the orientation of the heading
with respect to the principal horizontal stress direction as illustrated in Figure 3.3
However, this is difficult to replicate in the MAP3D model without using a great number of
mining steps. The damage bias could also indicate that the principal stress vector in the
horizontal or near horizontal plane (
x
/
) is inclined at a slight angle. Stress measurements
carried out on some Highveld collieries reveal that
x
/

is indeed angled at about 5 off
horizontal and that the minor principal stress is roughly half that of the principal horizontal
stress (Minney, 2002).

As a result of this,
x
/

has been inclined 5 from horizontal in both the Emaswati and ZAC
cases. Setting
x
/

at a slight angle models the zones of elevated stress as observed
underground, where it takes the form of guttering or pillar top damage. The position of the
damage observed underground has been simulated by the model when using this technique.

74
Table 6.1. Summary of rock mass properties
Rock type Rock mass property
UCS
(MPa)
UTS
(MPa)
Brazilian
tensile
strength
(MPa)
Shear
strength
(MPa)
Youngs
modulus
(GPa)
Density
(kg/m
3
)
Moisture
content
(%)
Internal
friction
angle
()
Sandstone 75 (1) 5 (1) 1.8-8.2
(2)
15 (1) 13 (1) 2480
(1)
0.5-3.9
(2)
20-30
(4)
Fine grained
sandstone
68-155 (2) 7.5-8.2
(2)

Coarse grained
sandstone
19.6-34.0
(2)
3.0-4.7
(2)

Shaly sandstone 77.1-102.9
(2)
5.2-5.7
(2)

Shale/sandstone,
interlaminated
63.2-76.0
(2)
5.5 (2) 2.05 (2)
Shale 75 (1)
34.6 (2)
5 (1) 1-4.7
(2&3)
7 (1) 15 (1) 2480
(1)
20-30
(4)
Siltstone 70 (1) 6 (1) 8 (1) 1 (1) 2480
(1)

Mudstone 40 (1) 5 (1) 2.1-6.2
(3)
8 (1) 7 (1) 2480
(1)

Dolerite 190 (1) 14 (1) 20 (1) 100 (1) 3000
(1)

Coal 25 (1) 5 (1) 8 (1) 5 (1) 1500
(1)
15-18
(4)
5 seam 1.2-3.6
(3)

4 seam 16.5-21.4
(2)
1.4-3.0
(3)

2 seam 18.5-31.2 1.2-2.6
(3)

Key to authors:
1. van der Merwe and Madden (2002)
2. Latilla (1997)
3. Chen (1994)
4. York (2002)

Table 6.2. MAP3D Material properties.
Material Youngs Modulus
E (MPa)
Poissons ratio

Density (kg/m
3
)
Host rock stratified sandstone 12 000 0.20 2 500
Dolerite 100 000 0.25 3 000
Coal Highveld coalfield 4 000 0.25 1 500
Coal KwaZulu-Natal and
Swaziland
2 400 0.25 1 500

Table 6.3. MAP3D horizontal to vertical stress ratios.
Mine Vertical Principal
horizontal
Minor
horizontal
Emaswati 1 2 1
Matla, 5 seam 1 1 1
ZAC 1 2 1
Welgedacht 1 1 1
75

Figure 6.1. MAP3D. Model geometry. Mngeni failed case. Perspective view.

Modelled deformations are very much smaller than those measured at the case study sites.
This is due to the models reflecting only elastic movement while in actual fact there is
considerable inelastic movement in cases of this type. This movement probably takes place
along jointing in the roof and floor as well as pillar dilation. As a result of this, it is
considered to be of no purpose in reporting total displacement in any of the cases where
actual roof to floor convergence has not been measured.

The following outputs have been reported for each failed case model:
o Vertical stress (
zz
) on a window positioned 0.5m into the seam floor. In all
but the ZAC cases this is a horizontal window. In the ZAC case the window is
inclined to match the seam dip of 6.
o Vertical stress (
zz
) on a vertical window placed along the li ne of pillars closest
to the centre of the failed area.
o Total displacement (U
t
) on a vertical window placed along the line of pillars
closest to the centre of the failed area in the Emaswati and ZAC, Mngeni shaft,
panel MEN1case.

For the stable cases, onl y the horizontal window showing the vertical stress (
zz
) has been
reported.
Chequerboarded
Faults
Bord and
pillar
workings
Goaf
76

Output type with units, contour unit scales, modelled stress orientations, and view
orientations are shown for each figure. As the position of these keys are always in the same
position in the MAP3D output figures, it is considered necessary only to identify them once
as shown in Figure 6.2.

6.2.1 Emaswati, main haulage south

The model geometry for this case study was taken to be the following: cover depth of 65m;
pillar centre distance of 16m; bord width of 6m and mining height of 2.5m. The model
included surrounding dolerite dykes and an underlying dolerite sill.

Vertical stress is fairly evenly distributed over the central pillars and shows a skewed stress
increase close to the pillar edge whi ch can be explained by the horizontal stress direction
(Figure 6.2). Figure 6.3 shows the same output for a vertical window and clearly illustrates
x
/
. Significant tensile stress zones are present in the floor of bords and absent in
intersections. This would indicate potential buckling of the floor beams.

The average pillar stress (
zz
) is 3.5MPa and vertical stress in the pillar base is 5MPa close
to the pillar edge (Figure 6.3). The floor is displaced 1.5mm up into the roadways while the
maximum displacement in the roof is in the order of 4mm. The total modelled closure of
4mm (Figure 6.4) is about ten times less than the maximum measured roof to floor
convergence of 45mm. This illustrates the limitations of elastic modelling of this type of
problem.

No significant increase in vertical stress was noted towards the centre of the large
horizontal window in this model.

As no records of detailed floor and pillar damage mapping have been found it is not
possible to compare modelled zones of higher stress with actual damage underground.
What is interesting is the distribution of tensile stress in the floor ( Figure 6.2) where it is
concentrated in the bords. A similar phenomenon was modelled for the Welgedacht case
(Figure 6.7) where floor heave was documented as being most prevalent in bords instead of
intersections. It is possible that the floor behaviour was similar at Emaswati.

77

Figure 6.2. MAP3D. Emaswati, Main haulage south, horizontal window
zz
.

Figure 6.3. MAP3D. Emaswati, Main haulage south, vertical window,
zz
.

View
orientation
Output type
with units
Stress scale
Modelled
stress
orientations
78

Figure 6.4. MAP3D. Emaswati, Main haulage south, vertical window, U
t
.
6.2.2 Matla 1, 5 seam, panel R14South

This panel was modelled with a cover depth of 32.5m, 6m bord widths, 11m pillar centres
and a mining height of 2.5m. Also included is a dolerite dyke on one side of the collapsed
area.

Stress distributions are uniform on the horizontal window placed 0.5m into the floor. The
average pillar stress is 2.2MPa. Tensile stress zones are widespread throughout the floor
and are present in both bords and intersections ( Figure 6.5).

The lobes of relatively higher stress (>3MPa) close to the pillar edges in Figure 6.6 are
noteworthy in that they indicate a potential failure zone where pillar slabbing was seen to
occur. This may indicate the position beneath the pillar that squeezing of the soft floor was
initiated.

The extent of the 2.4 to 2.7MPa vertical stress contour is greater for the right hand side
pillars in the model (Figure 6.5). These pillars were placed close to the panel centre-line.
The average stress in the right hand pillars (closest to the centre of the panel) is 2MPa
while those to the left (closer to the barrier are slightly less stressed at 1.95MPa. While the
difference is admittedly slight, it may account for the increase in pillar deterioration
observed towards the centre of the panel.

Displacement scale
79

Figure 6.5. MAP3D. Matla 1, 5 seam, panel R14South, horizontal window,
zz
.

Figure 6.6. MAP3D. Matla 1, 5 seam, panel R14South, vertical window,
zz.
Panel
centreline
80
6.2.3 Welgedacht, Alfred seam stooping

This area was modelled as being on average 165m deep with 24m pillar centres and mining
height of 4m. The bord width was modelled as 6m to simulate the floor area of the pillars
and not the roof contact area (Figure 4.9). A stooped out area is included in the model. This
is defined as a void with the floor of the void being loaded by the goaf weight. The surface
topography is modelled by building a roughly equivalent shape above a set datum line to
more accurately distribute overburden loading. In addition, part of the area is underla in by
Gus seam workings. These are modelled as underlying the Alfred seam pillars and
terminate three rows of pillars from the goaf edge.

High vertical stress levels of 7.5MPa are modelled 0.5m beneath the pillars closest to the
goaf edge. Three rows from the goaf edge, the pillars on the Alfred seam are superimposed
over those on the Gus seam which is 4m below. Average stress beneath the pillars
decreases from 5.8MPa closest to the goaf to 5MPa on the first row of superimposed pillars
(Figure 6.7).

The distribution of the tensile stresses mostly in bords in Figure 6.7 could indicate why
floor heave was more common in bords than in intersections (Wevell, 2002). Vertical str ess
(
zz
) reaches in excess of 9MPa beneath the pillar edge on the side closest to the goaf
dropping to 6.5MPa three rows of pillars from the goaf. There is a general tilting of the
stress field into the floor towards the goaf ( Figure 6.8).

The jagged appearance of the graphical outputs for the Welgedacht case is due to gridding
constraints. As the MAP3D modelling was run more to obtain a broader understanding of
the stress state at the individual sites it was not considered necessary to run the model with
a fine grid setting.

Figure 6.7. MAP3D. Welgedacht, Alfred seam, horizontal window,
zz
.
81

Figure 6.8. MAP3D. Welgedacht, Alfred seam, vertical window,
zz
.

6.2.4 ZAC, Mngeni, panel MEN1

The model geometry is as follows: cover depth of 102 to 151m; seam dip of 6; pillar
centre distance of 22m; bord width of 6m and mining height of 3m. This layout is analysed
in two stages to simulate the situation before and after chequerboard extraction.

Before chequerboarding the average vertical stress on the pillars is approximately 6MPa
(Figure 6.9). After secondary extraction, this value increases to 7.9MPa in the pillars
closest to the goaf edge while three rows away it decreases to 6.6MPa ( Figure 6.10).
Small patches of tensile stress are shown in the floor of the goaf after chequerboarding.

The inclined stress contours in Figure 6.11 are due to the seam being modelled as di pping
by 6. Higher vertical stress is indicated on the goaf side of the pillars after
chequerboarding. Underground mapping has shown that all pillars within four rows of the
goaf have spalled severely and that there is no tendency for the side closest to the goaf
to be worse than any others. If pillar conditions were mapped soon after chequerboarding
and before large scale deterioration took place the indicated (by MAP3D) tendency for
more damage on the side closest to the goaf, may have been appar ent.

Displacements of about 5mm in the roof and 7mm in the floor have been indicated in bords
after chequerboarding (Figure 6.12). Measured roof to floor convergence of between 20 and
55mm is once again far greater than that modelled due to MAP3D modelling only elastic
behaviour.

82
The model was able to show the direction of the ride which was picked up during a re -
survey of underground pegs (Figure 4.14) but once again the magnitude was an order of
magnitude smaller. This is not surprising in that numerous open roof cracks and pillar
dilation cracks are present which indicates that a large degree of inelastic movement has
occurred. The elastic component of such movements may be expected to be relatively
small.

Figure 6.9. MAP3D. ZAC, Mngeni, panel MEN1, horizontal window, step 1,
zz
.
83

Figure 6.10. MAP3D. ZAC, Mngeni, panel MEN1, horizontal window, step2,
zz
.

Figure 6.11. MAP3D. ZAC, Mngeni, panel MEN1, vertical window, step 2,
zz
.
84

Figure 6.12. MAP3D. ZAC, Mngeni, panel MEN1, vertical window, step 2, U
t
.
6.2.5 Matla 1, 5 seam, R14West

This panel has been modelled as having a cover depth of 35m, pillar centres of 12m, mining
height of 2m and 6m bord width.

In this case the average pillar stress is about 2.5MPa and there are large patches of tensile
stress evident in the floor of both bords and intersections ( Figure 6.13).

The distribution of tensile stress for the stable case is very similar to that modelled for the
failed case on the same seam (Figure 6.5) and the average pillar stress is slightly higher.
Comparing the two cases, it appears that the main difference is the even distribution of
stress beneath the pillar in the stable case, whereas the stresses are higher around the
central core of the pillar in the failed case. This slightly increased stress footprint may
account for the squeezing of the pillar foundation.

6.2.6 Matla 2, 5 seam, panel M12North

This panel has been taken as having pillar centres of 15m, at a depth of 60m, the mining
height is 2m and the bord widths are 6m.

Small patches of tensile stress are visible in the floor of the bords while the maximum
pillar stress is 4MPa (Figure 6.13).
85

Compared to the other two Matla 5 seam cases, the following conclusions can be drawn:
o In this case, unlike the other two, there is very little tensile stress shown in the
output.
o In common with panel R14South, there is a zone of slightly increased stress around
the central core of the pillar.
o It appears that the conditions indicative of failure for these three cases (panels
R14South and R14West at Matla 1 and Panel M12North at Matla 2) is a
combination of higher stress around the pillar edge as well as the presence of
tensile stress zones in the floor.

Figure 6.13. MAP3D. Matla 1, 5 seam, panel R14West, horizontal window,
zz
.

86

Figure 6.14. MAP3D. Matla 2, 5 seam, panel M12North, horizontal window,
zz
.

6.2.7 ZAC, Mngeni, panel MN
The model geometry is as follows: cover depth of 100m; pillar centre distance of 22m; bord
width of 6m and mining height of 2.5m.

Before chequerboarding the average pillar stress is about 5MPa ( Figure 6.15). This
increases to 9.3MPa on the large remnant pillars and 11MPa on the small remnant pillars
(Figure 6.16).

In Figure 6.16 the effect of changing from turbo chequerboarding to normal
chequerboarding is clearly seen. The vertical stress acting over the central core of the
pillars decreases from about 10MPa down to 8MPa on the larger remaining pillars. There is
also a slight decrease in the amount of tensile floor zones in the area of large remaining
pillars.

Prior to chequerboarding the distribution of tensile stress zones is similar to that modelled
for Emaswati and Welgedacht in that there is very little tensile stress shown in
intersections.

87

Figure 6.15. MAP3D. ZAC, Mngeni, panel MN, before chequerboarding,
zz
.

Figure 6.16. MAP3D. ZAC, Mngeni, panel MN, after chequerboarding,
zz
.
Normal
chequerboarding
Turbo
chequerboarding
Turbo
chequerboarding
88
6.2.8 Summary of MAP3D results

Results obtained from all the MAP3D models are summari sed in Table 6.4.

It can be seen that the vertical stress beneath the pillars for the failed cases ranges from 2.2
to 7.9MPa while for the stable cases the range is between 2.5 and 5.0MPa. This represents
an overlap of some 44% of the values.

Lobes of higher stress indicated areas where pillar slabbing was observed to occur in the
Matla 1, R14West case.

Tensile stress zones in the floor were confined to the bords only in the ZAC and
Welgedacht cases. The actual position of floor heave was found to closely agree with this
finding at Welgedacht and the ZAC, Mngeni shaft, panel MN cases.

For the 5 seam cases, tensile stress zones were present in bords and intersections in two out
of three cases while for Emaswati, Welgedacht and the ZAC cases, the tensile stress zones
(where present) were mostly confined to the bords. As no soft floor material was modelled
with MAP3D this cannot be due to floor composition. It is considered most likely to be due
to the influence of horizontal stress, which was modelled as twice that of vertical for
Emaswati and ZAC and 1:1 for the other cases. The Welgedacht pillars are modelled
alongside a goaf area, which may have had the effect of concentrating horizontal stresses
though the floor.

The combination of slightly increased loading around the pillar core, combined with
extensive areas of tensile stress in bords and intersections, may indicate failure in the Matla
5 seam, panel R14South case.

6.2.9 Three-dimensional modelling conclusions

As no stress measurements have been obtained at any of the case study sites it is not
possible to comment on the accuracy of the MAP3D modelled outputs. The limited roof to
floor convergence data indicate that the numerical models underestimate the actual
deformations by as much as ten times. That realistic appearing outputs have been produced
is clear in that stresses modelled are what would generally be expected. However their
value in determining why some pillars failed while others did not, is not establis hed.

The ability of the model to indicate where damage is likely to appear in the floor and, to a
lesser extent, the pillars, is worthy of note and indicates the value of three -dimensional
modelling in assisting with understanding complex problems of this nature.

No obvious factor stands out as a simple failure predicting tool. To achieve this, it would at
least be necessary to determine the actual stress state at a number of the case study sites in
order to calibrate the models. As the failed sites are generally not accessible anymore, this
option is impractical.

Therefore, due to the resource intensive nature of three-dimensional modelling, it was
decided not to pursue this avenue further.

89
Table 6.4. Summary of MAP3D results.
Model Window

Output Floor beam Beneath pillar
Maximum Minimum Average Maximum Minimum

Emaswati,
main south
(failed)
Horizontal zz (MPa) 0. 6 -0. 3 3. 5 4. 8 2. 4
Vertical Ut (mm) 1. 5 1. 0 1 0

Matla 1,
R12South
(failed)

Welgedacht
(failed)
Goaf edge
Horizontal zz (MPa) 0. 5 5. 8 7. 5 5. 5
1
st
row
superimposed
pillars

ZAC, Mngeni,
MEN1
(failed)
before
c/boarding
Vertical Ut (mm) 5 3 3
After
c/boarding
Horizontal zz (MPa) 1. 0 4 7. 9 >10 5. 0
Vertical Ut (mm) >10 9 7
3 rows from
goaf edge
Vertical Ut (mm) 5 4 4 4

Matla 1, R14
West
(stable)

Matla 2,
M12North
(stable)

ZAC, Mngeni,
MN
(stable)
Before
c/boarding
After c/
boarding
Horizontal zz (MPa) 1. 2 -1. 2 9. 3 Large
11. 0 Small
>12
>12
8. 4
10. 8

6.3 Two dimensional stress analysis

The PHASE
2
two-dimensional plastic finite element programme (Rocscience, 2003) was
chosen to further analyse stresses and displacements of soft floor bands of varying
thicknesses in bords and beneath pillars. The program was also found to be relatively quick
and easy to use which makes it suitable for running numerous comparative geometries and
material types if required.

In all cases, 4 noded quadrilaterals are used for the mesh and the model boundaries are set
as free to move along the y (vertical) axis with the surface being unrestrained. The top and
bottom of the soft floor band is unrestrained along the x (horizontal) axis.

As PHASE
2
is a two-dimensional program it is usual to adjust either the bord width or
pillar width to achieve realistic pillar loading. This can be done by modelling the pillar
90
width as a function of the percentage extraction. A sensitivity study on the Emaswati case
revealed that the principal stresses in the soft floor beneath the pillar were similar for
adjusted and actual pillar widths ( Figure 6.17 and Figure 6.18). With this in mind it was
decided to ignore the adjustment for the other cases and to define the bord and pillar
widths as actually mined.

For the purpose of this analysis, soft floor is defined as transversely isotropic while the
host rock and coal is defined as isotropic.

Transversely isotropic material

The soft floor was defined as anisotropic as it breaks much more readily along the
horizontal plane. The determination of the degree of anisotropy was based on a number of
axial and diametral point load tests carried out at Matla Colliery (Latilla, 1996) from which
UCS values were estimated according to the method described by Brady and Brown (1985).
The results obtained are summarised in Table 6.5:

As with many other geotechnical tests, there is a bias towards testing better rock qualities
as the weaker rocks tend to break up before a correct sized sample can be obtained. As a
result of this, it is likely that shale is even more anisotropic than predicted. For the purpose
of this study the diametral strength has been defined as 30% of the axial strength.

Table 6.5. UCS from point load (MPa).
Lithology A
Diametral
B
Axial
A/B
Sandstone / shale, laminated 21.9 69.1 0.32
Sandstone, massive, coarse grained 45.5 64.6 0.70
Sandstone, bioturbated 36.0 50.5 0.71
Sandstone, stratified 67.2 75.7 0.89
Shaly sandstone, stratified 62.7 57.5 1.09

Table 6.6. PHASE
2
transversely isotropic material.
Input parameter Soft floor
Stress state Field stress only
Shear modulus -G (MPa) 400
Angle between E
1
and E
2
() 90
Youngs Moduli E
1
(MPa)
E
2
(MPa)
1000*
300
Poissons ratios
1

2

0.3
0.3
Tensile strength (MPa) 6*
Friction angle () 20
Cohesion (MPa)
#
10.6

*Value quoted by van der Merwe and Madden (2002) for siltstone.
#
As it is possible to derive a value for cohesion when the tensile strength and friction angle
are known (Figure 7.2) this approach has been used in this case.

91
Isotropic materials

The host rock (sandstone) and coal are assumed to be isotropic and the input parameters
shown in Table 6.7 have been assigned. Coal in KwaZulu-Natal and Swaziland has been
modelled as weaker than that of the Highveld and Witbank coalfields. Experience suggests
that even though on a small or laboratory scale the seams have similar strengths, the degree
of jointing in KwaZulu-Natal and Swaziland is such that pillars are considerably weaker.
Table 6.7. PHASE
2
isotropic materials.
Host rock
(sandstone)
Coal
(Highveld)
Coal (KwaZulu-
Natal &
Swaziland)
Mudstone
Stress state Field stress only Field stress
only
Field stress only Field stress
only
Youngs Modulus E
(MPa)
12 000 4 000 2 400 7 000
Poissons ratio

0.2 0.25 0.25 0.25
Failure mode Mohr Coulomb Mohr Coulomb Mohr Coulomb Mohr
Coulomb
Tensile strength
(MPa)
5 2.4 2.4 2.1
Friction angle () 20 16.5 16.5 25
Cohesion* (MPa) 8.75 8.5 8.5 3.4
*see Figure 7.2.

6.3.1 Emaswati Colliery, Main haulage south

In this panel the cover depth was modelled as 65m, the pillar centre distance was 16m, bord
width 6m and the mining height was 2.5m. The KwaZulu-Natal and Swaziland coal strength
is used (Table 6.7) and the soft floor is defined as 0.5m thick. Horizontal stress is assumed
to be twice that of vertical.

The contour values and vertical to horizontal stress ratios are shown in the same place on
all of the PHASE
2
output figures, these are identified once in Figure 6.23.

Results for principal stress (
1
) show that the soft floor carries slightly lower loads than the
underlying host rock. The maximum and minimum stresses in the soft floor beam in the
bord are 3.4 and 2.2MPa, respectively. The corresponding values within the soft floor band
beneath the pillar are 4.7 and 3.4MPa (Figure 6.18).

A sensitivity analysis was run to evaluate using pillar sizes adjusted to model percentage
extraction. In this case the percentage extraction is about 60% so the pillars were adjusted
to 4m wide with 6m bords. A comparison of Figure 6.17 (modelled according to extraction
percentage) and Figure 6.18 (modelled with actual pillar width) reveals little difference in
the maximum and minimum principal stresses in the soft floor beam beneath the pillars
(area of interest) for both cases. As a result of this it was decided to rather model all the
other cases with actual pillar widths.

The minor in-plane principal stress (
3
) in the floor beam shows a tensile lobe under the
pillar corner (A in Figure 6.19), while Figure 6.20 illustrates the direction and magnitude
92
of the total displacement (U
t
), in this case 1mm on the floor and 3mm at mid height of the
pillar.

Roof to floor convergence at Emaswati was measured at between 10 and 45mm. In an
attempt to replicate these values a number of runs were carried out with weaker f loor
material. An example is shown in Figure 6.21 where the following values were used for the
floor rock: a shear modulus of 100MPa, Youngs moduli of 250 and 75MPa for E1 and E2
respectively and tensile strength of 1.5MPa. The floor deformation increased to 1.5mm
from the 1mm modelled when using the values in Table 6.6.

-2.00000
-0.71429
0.57143
1.85714
3.14286
4.42857
5.71429
7.00000
-
7
0
.
5
-
6
9
-
6
7
.
5
-
6
6
-
6
4
.
5
-
6
3
12 13.5 15 16.5 18 19.5 21 22.5 24 25.5
Figure 6.17. PHASE
2
1
(MPa), sensitivity analysis,
adjusted for percentage extraction.

93
-2.00000
-0.71429
0.57143
1.85714
3.14286
4.42857
5.71429
7.00000
-
7
4
-
7
2
-
7
0
-
6
8
-
6
6
-
6
4
-
6
2
15 18 21 24 27 30 33 36 39
Figure 6.18. PHASE
2
1
(MPa),

using actual pillar
width.

-4.00000
-2.71429
-1.42857
-0.14286
1.14286
2.42857
3.71429
5.00000
-
6
9
.
2
-
6
8
.
6
5
-
6
8
.
1
-
6
7
.
5
5
-
6
7
-
6
6
.
4
5
18.9 19.35 19.8 20.25 20.7 21.15 21.6 22.05 22.5 22.95 23.4 23.85 24.3
Figure 6.19. PHASE
2
, Emaswati, Main haulage south,
3
(MPa).

A
94
0.00000
0.00050
0.00100
0.00150
0.00200
0.00250
0.00300
0.00350
-
6
9
.
2
-
6
8
.
6
5
-
6
8
.
1
-
6
7
.
5
5
-
6
7
-
6
6
.
4
5
18.55 19.1 19.65 20.2 20.75 21.3 21.85 22.4 22.95 23.5 24.05
Figure 6.20. PHASE
2
, Emaswati, Main haulage south, U
t
(m).

0.00000
0.00051
0.00103
0.00154
0.00206
0.00257
0.00309
0.00360
-
6
9
.
5
-
6
8
.
6
-
6
7
.
7
-
6
6
.
8
-
6
5
.
9
-
6
5
-
6
4
.
1
14.6 15.5 16.4 17.3 18.2 19.1 20 20.9 21.8 22.7 23.6 24.5
Figure 6.21. PHASE
2
. Emaswati, Main haulage south, U
t
(m), sensitivity analysis.

95
6.3.2 Matla 1, 5 seam, panel R14South

Soft floor of 0.5m thick, overlain by 6m bords and 5m pillars is analysed in this model. The
coal strength for the Highveld coalfield as shown in Table 6.1 is used and the cover depth
is taken as 33m. The percentage extraction in this case is 79.3%.

Once again, the soft floor exhibits slightly lower values for
1
than the underlying host
rock. The maximum and minimum values are 1.1 and 0.4MPa in the floor beam beneath the
bords while the corresponding values beneath the pillar are 1.5 and 1.1MPa ( Figure 6.22).

A tensile stress zone (-0.3MPa) is observed from the
3
output in the middle of the floor
beam as well as another running right through the soft floor beam about 1m from the pillar
edge (A and B, respectively in Figure 6.23).

-0.65000
0.01429
0.67857
1.34286
2.00714
2.67143
3.33571
4.00000
-
3
9
.
5
-
3
7
-
3
4
.
5
-
3
2
-
2
9
.
5
11.5 14 16.5 19 21.5 24 26.5 29
Figure 6.22. PHASE
2
, Matla 1, 5 seam, panel R14South, (MPa).

96
-2.00000
-1.00000
0.00000
1.00000
2.00000
3.00000
4.00000
5.00000
-
3
6
.
7
-
3
6
.
2
-
3
5
.
7
-
3
5
.
2
-
3
4
.
7
-
3
4
.
2
14.1 14.6 15.1 15.6 16.1 16.6 17.1 17.6 18.1 18.6 19.1

Figure 6.23. PHASE
2
, Matla 1, 5 seam, panel R14South,
3
(MPa).

6.3.3 Welgedacht Colliery, Alfred seam stooping

For this situation, a more complex model is required, in order to simulate the goaf and
underlying Gus seam workings (Figure 6.24). The Alfred seam pillars are 18m wide with
6m bords, while the cover depth is 165m. A goaf, 60m into the roof, is defined on the
Alfred seam as are superimposed bords on the underlying Gus seam. To, as far as possible,
simulate the actual situation the floor of the goaf is subjected to a load equivalent to the
weight of goafed material. The lower coal strength for KwaZulu-Natal and Swaziland is
used and the soft floor is defined as being 0.6m thick. The percentage extraction (before
stooping) on the Alfred seam was 43.8%.

Figure 6.24 shows an overview of the
1
output for this model while Figure 6.25 is an
enlarged view showing the bord closest to the goaf edge. In this bord the principal stress
(
1
) was found to be between 2.1 and 6MPa in the floor beam and between 4.7 to 6MPa in
the soft floor beneath the pillars. The minor in-plane principal stress in the same bord (
3
)
is tensile over about two thirds of the floor beam ( -0.1MPa) with a single zone of 1MPa
near the base of the soft floor beam, it compressive elsewhere. The range for the whole
soft floor beam is 1 to 2.4MPa (Figure 6.27).

Above the first row of superimposed bords,
1
in the soft floor beam varies from 2.1 to
5.5MPa (Figure 6.26) while
3
ranges from 0.1 to 2.3MPa (Figure 6.28).

The influence of the underlying Gus seam workings as identified in 4.3.4 was analysed by
comparing the principal and minor principal stresses in the floor beam and it was found
that:
A
B
Vertical /
horizontal
stress
distribution
Contour
value key
97
o There is no dramatic difference for the principal stress values in the bords close to
or three rows from the goaf edge.
o The minor in-plane principal stress however shows a zone of tensile stress of
1MPa at the base of the soft floor beam in the road nearest the goaf, while the
lowest tensile stress value above the Gus bord is ten times less at only 0.1MPa.

This would indicate that floor instability (or buckling) could be expected closer to the goaf
edge rather than above the Gus seam workings abutment. As greater damage was observed
underground the closer one got to the goaf edge, this would appear to be borne out by the
results of the modelling.

0.00000
1.71429
3.42857
5.14286
6.85714
8.57143
10.28571
12.00000
-
1
7
6
-
1
6
8
-
1
6
0
-
1
5
2
-
1
4
4
-
1
3
6
-
1
2
8
106 115 124 133 142 151 160 169 178 187 196

Figure 6.24. PHASE
2
1
(MPa), overview.

Goaf
Alfred seam bords
Gus seam bord
Area enlarged in
Figure 6.25
98
0.00000
1.71429
3.42857
5.14286
6.85714
8.57143
10.28571
12.00000
-
1
7
1
-
1
6
9
.
5
-
1
6
8
-
1
6
6
.
5
-
1
6
5
-
1
6
3
.
5
136.5 138 139.5 141 142.5 144 145.5 147 148.5 150 151.5

Figure 6.25. PHASE
2
1
(MPa), road nearest goaf.

0.00000
1.71429
3.42857
5.14286
6.85714
8.57143
10.28571
12.00000
-
1
7
5
.
5
-
1
7
4
-
1
7
2
.
5
-
1
7
1
-
1
6
9
.
5
-
1
6
8
-
1
6
6
.
5
184.5 186 187.5 189 190.5 192 193.5 195 196.5 198 199.5
Figure 6.26. PHASE
2
, Welgedacht, Alfred seam, 1(MPa), first row of superimposed
bords.

99

-4.00000
-2.28571
-0.57143
1.14286
2.85714
4.57143
6.28571
8.00000
-
1
7
1
-
1
6
9
.
5
-
1
6
8
-
1
6
6
.
5
-
1
6
5
-
1
6
3
.
5
-
1
6
2
136.5 138 139.5 141 142.5 144 145.5 147 148.5 150 151.5 153
Figure 6.27. PHASE
2
, Welgedacht, Alfred seam, 3(MPa), road nearest goaf.

-4.00000
-2.28571
-0.57143
1.14286
2.85714
4.57143
6.28571
8.00000
-
1
7
5
.
5
-
1
7
4
-
1
7
2
.
5
-
1
7
1
-
1
6
9
.
5
-
1
6
8
-
1
6
6
.
5
183 184.5 186 187.5 189 190.5 192 193.5 195 196.5 198 199.5
Figure 6.28. PHASE
2
, Welgedacht, Alfred seam, 3(MPa), first row of superimposed
bords.
100
6.3.4 ZAC, Mngeni, panel MEN1

The portion of this panel covered in this case study, was initially developed and then a
portion of it had chequerboard extraction carried out on it due to pit roo m constraints. As a
result, this analysis was done with two mining stages to simulate chequerboard pillar
extraction on the eastern side of the panel.

Soft floor of 0.22m was recorded in a borehole close to the failed area but underground
observations indicated that failure extended 1.15m into the floor. For this reason 1.2m of
soft floor lying beneath the 16m wide pillars and 6m bords is used in the analysis. The
cover depth varies from 102 to 151m due to seam dip. For this model, the average depth of
127m is used, as is the weaker coal strength of KwaZulu-Natal and Swaziland. The
percentage extraction (before chequerboarding) was 47.1%.

In the second stage two pillars and a portion of another are mined. This stage is described
as after chequerboarding in the following figures. The load is split 54% to 46% between
stages 1 and 2, based on area mined.

Principal stress (
1
) in the floor beam increases dramatically after chequerboarding with a
range of 2.8 to 3.8MPa before (Figure 6.29) and 5.5 to 7.5MPa after (Figure 6.30). The
minor in-plane principal stress in the floor beam (
3
) is 0.1MPa both before and after
chequerboarding. While a tensile zone of 0.1MPa is only shown under the soft floor beam
beneath the bord before chequerboarding it extends beneath the pillar as well after
chequerboarding (Figure 6.34 and Figure 6.35). The implication of this is that the floor is
trying to flow towards the goaf.

The floor damage described in the case study (4.4.5) is in broad agreement with the
modelled behaviour. The floor damage often took the form of open cracks with zones of
shattering in the pillar sides. Another common form of floor damage is where the floor
plates appear to have ridden up over each other. Provided there is cohesion on the parting
between the host rock (tensile) and the weak overlying floor material, damage could be
expected.

Total deformation is a maximum of 2.3mm in the floor beam as well as in the soft floor
beneath the pillar before chequerboarding. This increases to 5.2mm in the floor beam after
chequerboarding (Figure 6.37).

A distance of three rows of pillars from the chequerboarded area the values for
1
,
3
and
U
t
beneath the pillar are almost the same as in the bord closest to the pillar extraction
(Figure 6.33, Figure 6.36 and Figure 6.38). In the soft floor beam, the range for
1
and
3
remains much the same as for the bord closest to the goaf.

101
0.00000
1.00000
2.00000
3.00000
4.00000
5.00000
6.00000
7.00000
-
1
3
7
-
1
3
3
-
1
2
9
-
1
2
5
-
1
2
1
-
1
1
7
51 54 57 60 63 66 69 72 75 78 81 84
Figure 6.29. PHASE
2
1
(MPa), before chequerboarding.

0.00000
2.00000
4.00000
6.00000
8.00000
10.00000
12.00000
14.00000
-
1
3
7
-
1
3
3
-
1
2
9
-
1
2
5
-
1
2
1
-
1
1
7
42 45 48 51 54 57 60 63 66 69 72 75 78
Figure 6.30. PHASE
2
1
(MPa), after chequerboarding.

Enlarged in
Figure 6.31
Enlarged in
Figure 6.32
102
0.00000
0.71429
1.42857
2.14286
2.85714
3.57143
4.28571
5.00000
-
1
3
2
.
6
-
1
3
1
.
7
-
1
3
0
.
8
-
1
2
9
.
9
-
1
2
9
-
1
2
8
.
1
68.85 69.7 70.55 71.4 72.25 73.1 73.95 74.8 75.65 76.5 77.35 78.2 79.05

Figure 6.31. PHASE
2
1
(MPa), before
chequerboarding.

0.00000
2.00000
4.00000
6.00000
8.00000
10.00000
12.00000
14.00000
-
1
3
2
.
3
-
1
3
1
.
6
-
1
3
0
.
9
-
1
3
0
.
2
-
1
2
9
.
5
-
1
2
8
.
8
-
1
2
8
.
1
47.1 47.9 48.7 49.5 50.3 51.1 51.9 52.7 53.5 54.3 55.1

Figure 6.32. PHASE
2
1
(MPa), after
chequerboarding.

103

0.00000
2.00000
4.00000
6.00000
8.00000
10.00000
12.00000
14.00000
-
1
3
1
.
9
-
1
3
1
.
2
-
1
3
0
.
5
-
1
2
9
.
8
-
1
2
9
.
1
-
1
2
8
.
4
-
1
2
7
.
7
2.5 3.2 3.9 4.6 5.3 6 6.7 7.4 8.1 8.8 9.5 10.2
Figure 6.33. PHASE
2
1
chequerboarding.

-2.10000
-1.37143
-0.64286
0.08571
0.81429
1.54286
2.27143
3.00000
-
1
3
2
.
5
-
1
3
1
.
7
5
-
1
3
1
-
1
3
0
.
2
5
-
1
2
9
.
5
-
1
2
8
.
7
5
-
1
2
8
68.9 69.55 70.2 70.85 71.5 72.15 72.8 73.45 74.1 74.75 75.4 76.05 76.7
Figure 6.34. PHASE
2
3

-0.1MPa
-1MPa
1MPa
-0.5MPa
104
-2.10000
-0.50000
1.10000
2.70000
4.30000
5.90000
7.50000
9.10000
-
1
3
1
.
3
-
1
3
0
.
6
5
-
1
3
0
-
1
2
9
.
3
5
-
1
2
8
.
7
-
1
2
8
.
0
5
-
1
2
7
.
4
47 47.75 48.5 49.25 50 50.75 51.5 52.25 53 53.75 54.5
Figure 6.35. PHASE
2
3
-2.10000
-0.50000
1.10000
2.70000
4.30000
5.90000
7.50000
9.10000
-
1
3
2
.
8
-
1
3
2
-
1
3
1
.
2
-
1
3
0
.
4
-
1
2
9
.
6
-
1
2
8
.
8
-
1
2
8
4.1 5 5.9 6.8 7.7 8.6 9.5 10.4 11.3 12.2 13.1 14
Figure 6.36. PHASE
2
3
chequerboarding.

-0.1MPa
-1MPa
105
0.00000
0.00190
0.00380
0.00570
0.00760
0.00950
0.01140
0.01330
-
1
3
1
.
3
-
1
3
0
.
6
5
-
1
3
0
-
1
2
9
.
3
5
-
1
2
8
.
7
-
1
2
8
.
0
5
-
1
2
7
.
4
47 47.75 48.5 49.25 50 50.75 51.5 52.25 53 53.75 54.5

Figure 6.37. PHASE
2
t
(m), after chequerboarding.

0.00000
0.00190
0.00380
0.00570
0.00760
0.00950
0.01140
0.01330
-
1
3
2
.
8
-
1
3
2
-
1
3
1
.
2
-
1
3
0
.
4
-
1
2
9
.
6
-
1
2
8
.
8
-
1
2
8
0.9 1.8 2.7 3.6 4.5 5.4 6.3 7.2 8.1 9 9.9 10.8
Figure 6.38. PHASE
2
t
(m), 3 pillars away from
chequerboarding.

106
6.3.5 Matla 1, 5 seam, panel R14West

In this stable case, the pillars are 6m wide with 6m bords at a depth of 35m. The Highveld
coal strength is used and the soft floor is defined as 0.4m thick. As for panel R14South, a
horizontal to vertical stress ratio of 1:1 is used. Percentage extraction was 75%.

Low values of
1
result (maximum), as would be expected with this stress environment. No
tensile stresses are indicated in the floor beam. The maximum deformation is 3mm and is
evident near the pillar side.

Compared to the failed case (R14South)
1
is

similar beneath the pillars and higher (1 to
1.3MPa as opposed to 0.4 to 1.1MPa) in the bords. The main difference however is the
absence of tensile stresses anywhere in the floor beam. This is felt to be significant as a
potential indicator of floor behaviour.

6.3.6 Matla 2, 5 seam, panel M12North

This stable case is defined as being 60m deep with 6m bords and 9m wide pillars, the soft
floor thickness is taken as 0.3m. A 1 to 1 vertical to horizontal stress ratio is used. The
percentage extraction was 64%.

As for the previous stable case, low values of
1
result, as well as no tensile component of
3
.

As was the case for the other stable 5 seam panel ( 6.3.5), the values for
1
are higher in the
bords as well as beneath the pillars than for the failed case. Once again, there is no tensile
stress modelled at any point in the floor beam.

0.47500
0.80071
1.12643
1.45214
1.77786
2.10357
2.42929
2.75500
-
4
0
-
3
8
-
3
6
-
3
4
-
3
2
-
3
0
12 14 16 18 20 22 24 26 28 30
Figure 6.39. PHASE
2
1
(MPa).
107

-0.30000
0.00000
0.30000
0.60000
0.90000
1.20000
1.50000
1.80000
-
3
9
.
4
-
3
8
.
8
-
3
8
.
2
-
3
7
.
6
-
3
7
-
3
6
.
4
-
3
5
.
8
15.1 15.7 16.3 16.9 17.5 18.1 18.7 19.3 19.9 20.5 21.1
Figure 6.40. PHASE
2
3
(MPa).

0.90000
1.41429
1.92857
2.44286
2.95714
3.47143
3.98571
4.50000
-
6
6
-
6
4
-
6
2
-
6
0
-
5
8
-
5
6
16 18 20 22 24 26 28 30 32 34 36

Figure 6.41. PHASE
2
1
(MPa).

108

-0.45000
0.06429
0.57857
1.09286
1.60714
2.12143
2.63571
3.15000
-
6
4
.
1
5
-
6
3
.
5
-
6
2
.
8
5
-
6
2
.
2
-
6
1
.
5
5
-
6
0
.
9
18.8 19.45 20.1 20.75 21.4 22.05 22.7 23.35 24 24.65 25.3 25.95

Figure 6.42. PHASE
2
3
(MPa).

6.3.7 ZAC, Mngeni, panel MN

This case was stable during development but floor heave occurred during subsequent
chequerboarding. The problem is analysed in two stages and the load has been split as 41%
in stage 1 and 59% in stage 2. This is determined from the area mined in each step. The
magnitude of the horizontal stress has been assumed to be twice that of the vertical, based
on underground observations elsewhere at ZAC where damage of soft roof material was
experienced. No stress measurements have been carried out at ZAC and the 2:1 horizontal
to vertical stress ratio is only an estimate.

In the second stage (chequerboarding) every second pillar is completely mined, as is a 6m
slice off the remaining pillars.

The panel is defined as being 100m deep with 16m pillars and 6m bords. The remaining
pillars after chequerboarding are 10m wide. The reduced coal strength for KwaZulu-Natal
and Swaziland is used (Table 6.7). The percentage extraction before chequerboarding was
47.1%.

After chequerboarding the principal stress (1) in the floor beam increases from between
1.6 and 2.2MPa to between 3.7 and 4.9MPa. Under the pillars the maximum value for the
principal stress increases from 2.2 to 4.9MPa ( Figure 6.43 and Figure 6.44).

The minimum value for
3
in the floor beam is tensile (-0.2MPa) after chequerboarding
(Figure 6.46), while before chequerboarding it ranges from 0.3 to 0.6MPa.

109
The roughly doubled values for
1
are expected during chequerboard extraction. Of
significance however, is the fact that there are only tensile stresses in the floor beam after
chequerboarding which is when the pillars are expected to fail. In general the remaining
pillars at ZAC fail between two weeks and six months after chequerboarding. The
appearance of tensile stresses in the floor beam at this stage could arguably lead to the
classification of this particular site as being both a stable and a failed case.

0.45000
1.10000
1.75000
2.40000
3.05000
3.70000
4.35000
5.00000
-
1
1
0
-
1
0
6
-
1
0
2
-
9
8
-
9
4
-
9
0
52 56 60 64 68 72 76 80 84 88
Figure 6.43. PHASE
2
1

0.45000
1.64286
2.83571
4.02857
5.22143
6.41429
7.60714
8.80000
-
1
1
0
-
1
0
6
-
1
0
2
-
9
8
-
9
4
-
9
0
52 56 60 64 68 72 76 80 84 88
Figure 6.44. PHASE
2
1
110

-0.30000
0.02857
0.35714
0.68571
1.01429
1.34286
1.67143
2.00000
-
1
0
5
.
6
-
1
0
4
.
8
-
1
0
4
-
1
0
3
.
2
-
1
0
2
.
4
-
1
0
1
.
6
-
1
0
0
.
8
69.3 70 70.7 71.4 72.1 72.8 73.5 74.2 74.9 75.6 76.3 77 77.7

Figure 6.45. PHASE
2
3

-0.30000
0.30000
0.90000
1.50000
2.10000
2.70000
3.30000
3.90000
-
1
0
5
.
6
-
1
0
4
.
8
-
1
0
4
-
1
0
3
.
2
-
1
0
2
.
4
-
1
0
1
.
6
-
1
0
0
.
8
83.3 84 84.7 85.4 86.1 86.8 87.5 88.2 88.9 89.6 90.3 91 91.7
Figure 6.46. PHASE
2
3

0.3MPa
0.6MPa
-0.2MPa 1.1MPa
111

6.3.8 ZAC, Maye, soft roof problem

At Maye shaft it is suspected that the pillars were punching into the soft roof. Six metre
wide bords are analysed with 11m pillars at a depth of 105m. The percentage extraction
was 58%. The soft floor material used for all the previous PHASE2 models is defined as
1.2m thick and lying on top of the pillars. A two to one horizontal to vertical stress ratio is
used as is the weaker coal strength for KwaZulu-Natal and Swaziland.

While
1
was relatively high in the roof beam,
3
ranged from 0.15 to 2.7MPa (Figure
6.48).

This site is the only stable case where there is a tensile component of
3.
This occurs about
one third of the way across the bord, in the soft roof beam. Beam flexure due to gravity
may be the cause of this occurrence.

0.40000
1.77143
3.14286
4.51429
5.88571
7.25714
8.62857
10.00000
-
1
0
9
.
5
-
1
0
8
-
1
0
6
.
5
-
1
0
5
-
1
0
3
.
5
-
1
0
2
13.5 15 16.5 18 19.5 21 22.5 24 25.5 27 28.5
Figure 6.47. PHASE
2
1
(MPa).

112
-0.30000
0.30000
0.90000
1.50000
2.10000
2.70000
3.30000
3.90000
-
1
0
9
.
5
-
1
0
8
-
1
0
6
.
5
-
1
0
5
-
1
0
3
.
5
-
1
0
2
13.5 15 16.5 18 19.5 21 22.5 24 25.5 27 28.5
Figure 6.48. PHASE
2
3
(MPa).

6.3.9 Summary of two-dimensional modelling results

Results obtained from all the PHASE
2
models are summarised in Table 6.8.

From the table it is clear that o
3
is tensile somewhere in the floor beam in all the failed
cases and not in the stable ones, with the exception of the ZAC, Mngeni, MN case after
chequerboarding and the Maye shaft soft roof case. The finding that the minor in-plane
principal stress is tensile in the floor beam of failed cases is considered worthy of further
investigation. These additional studies are not considered to fall within the scope of this
project report.

However, the results are consistent enough to warrant the inclusion of PHASE
2
modelling
as a tool for evaluating the likely performance of pillar foundations in seams with a known
propensity to foundation failure.

6.4 Numerical modelling conclusions

The following main conclusions have been drawn from the numerical modelling phase of
this project:
o The PHASE
2
modelling suggests that it may be used to identify potential failures
when the soft floor is modelled as being very weak. The presence of tensile stresses
in the minor in-plane principal stress outputs appears to agree very well with the
failed cases (Table 6.8).
o The encouraging results obtained from the PHASE
2
modelling indicate that it could
be included in the design guidelines to assist with the decision making process.

113
Table 6.8. Summary of PHASE
2
results.
Model Output Floor beam Base of pillar (foundation)
1 3
(MPa)
Ut (mm)
Centre
of floor
beam
Maximum Minimum Under
pillar
edge
Maximum Minimum

Emaswati, main
south
1 2.2 3.4 2.2 3.1 4.7 3.1
(failed) 3 0.4 0.4 -0.5 -0.5 3.4 -0.5
Ut 1 1 0.8 0.8 0.8 0

Matla 1, R14South 1 0.4 1.1 0.4 1.5 1.5 1.1
(failed) 3 -0.3 1.3 -0.3 1.3 1.3 0.8

Welgedacht
(failed)
Goaf edge
1 2.5 2.1 6.0 6.0 6.0 4.7
3 -0.1 2.9 -1.0 2.4 3.7 1.6
1
st
row
superimposed bords
1 2.1 5.5 2.1 5.5 5.5 3.8
3 -0.1 2.4 -0.1 2.5 3.3 2.5

ZAC, Mngeni,
MEN1
(failed)
before c/boarding
1 2.8 3.8 2.8 3.8 3.8 3.3
3 -0.1 1.0 -0.1 1.0 1.7 1.0
Ut 0.1 2.4 0.1 2.4 2.4 0.1
After c/boarding 1 5.5 7.5 5.5 6.5 7.5 6.5
3 -0.1 1.5 -0.1 1.5 3.1 1.5
Ut 0.5 5.2 0.5 5.2 5.2 0.5
3 rows from goaf
edge
1 5.5 7.5 5.5 7.5 7.5 6.5
3 -0.1 2.3 -0.1 2.3 3.1 1.5
Ut 0.5 4.3 0.5 4.3 0.5 0.5

Matla 1, R14 West
(stable)
1 1.0 1.5 1.0 1.5 1.5 0.9
3 1.0 1.3 1.0 1.0 1.0 0.7

Matla 2,
M12North
(stable)
1 1.7 2.0 1.7 1.5 1.5 1.2
3 1.7 2.1 1.7 1.6 1.6 1.1

ZAC, Mngeni, MN
(stable)
Before c/boarding
1 1.6 2.2 1.6 2.2 2.2 1.9
3 0.3 0.6 0.3 0.6 1.1 0.4
After c/ boarding 1 3.7 4.9 3.7 4.9 4.9 4.3
3 0.2 1.1 -0.2 1.1 1.7 1.1

ZAC, Maye
(roof beam)
1 4.2 4.9 4.2 5.5 5.5 5.5
3 0.2 2.6 -0.2 3.1 3.1 2.9

o Numerical modelling, especially the complex three-dimensional type, is manpower
and time intensive. Being elastic models, they are only able to model the situation
prior to failure and this may lead to results, that in the case of deformation, are an
order of magnitude smaller than those measured underground. The MAP3D
modelling did not suggest that it would be able to distinguish between weak
foundation cases where the pillars fail and those where the pillars are stable.
114
o The MAP3D models do however enable better understanding of the stress state and
in some cases indicate where pillar or floor damage may be expected to occur.
o As no in situ stress field measurements have been determined for any of the case
study sites it is impossible to verify the outputs of the numerical models.

7 FLOOR BEARING CAPACITY

According to the literature, the most useful geotechnical factors for use in determining
floor bearing capacity are indirect tensile strength, nat ural moisture content and angle of
internal friction. These properties are not routinely determined for South African coal
measures rocks but values shown in Table 6.1, obtained from various sources, have been
used where necessary.

The bearing capacities of both stable and failed cases are determined using two widely
recognised methods, those of Terzaghi and Vesic. This is to evaluate the use of civil
engineering analysis techniques in predicting the behaviour of soft floor material beneath
coal pillars.

7.1 Terzaghis Method

According to Hill (1994), Terzaghis method is widely used for determining the bearing
capacity of foundations (Stacey and Page, 1986).

The foundation strength (q
u
) is given by:
l q c u
bN qN cN q + + = Equation 3
Where:
c = cohesion (MPa)
q = Normally taken as zero
*

= specific weight of the material (MN/m
3
)
b = half the pillar width (m)

N
c
, N
q
and N
l
are bearing capacity factors which depend on the angle of friction of the
material. Approximate values of these factors can be read off Figure 7.1.
*
Unless failure is expected to take place some distance into the floor.

According to Chugh, Atri and Dougherty (1988) cohesi on (c) may be approximated from
indirect tensile strength (T
o
) and internal friction angle () according to the graphical
method shown in Figure 7.2. They further state that =20 may be used with confidence
when determining the ul timate bearing capacity (UBC) of coal mine floor foundations. A
single value of 20 for is frequently stated in the literature ( 3.3.1) and it is not clear to
the author how this parameter can be classified as useful in determining foundation
strength.

Whether this method of determining cohesion holds true for Southern African conditions is
not certain, and the approach suggested by Stacey and Page (1986) is also used. In this
method, cohesion is taken as 0.16 x DRMS (design rock mass strength) in MPa.
115

DRMS has been determined for each case study site in Table 7.1 using the mining rock
mass classification of Laubscher (1990). The reduction factor to determine DRMS from
RMS has been taken as 63% where jointing and weathering are present and 85% where
weathering is not expected to be significant.

The foundation strength (q
u
) is calculated using an angle of internal friction () of 20.
Values for N
c
, N
q
and N
l
are determined from Figure 7.1 to be 23, 7.5 and 1 respectively for
=20. Rock density is assumed at 2 480kg/m
3
(0.0243MN/m
3
) for all cases. The rock
mass properties used in this study are shown in Table 7.2.

Where the foundation strength is greater than the average pillar stress stable conditions are
indicated and vice versa.

Foundation strengths determined by this method are very much higher than the average
pillar stress as can be seen in Table 7.2. Using Terzaghis method the foundations for both
the failed and stable cases would all be predicted to be stable. This patently was not the
case, and it is considered that this method is not suited to the analysis of performa nce of
soft coal seam floors.

0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
0 5 10 15 20 25 30 35 40
Angl e of fri cti on (Degrees)
B
e
a
r
i
n
g

c
a
p
a
c
i
t
y

f
a
c
t
o
r
s
Nc
Nq
Nl

Figure 7.1. Foundation bearing capacity factors.
Table 7.1. Design rock mass strength for failed and stable cases.
CASE Soft floor lithology c rating
(MRM
value)
Soft floor
MRMR
Soft floor
UCS
(MPa)
RMS
(MPa)
DRMS
adj.
(%)
DRMS
(MPa)
Emaswati
(fai led)
Sandst one and shale,
int er laminat ed
8 29 63 13 85 11. 1
Matla 1, R14S
(fai led)
Ver y soft sandy shale 6 20 35 5 63 3. 2
Welgedacht
(fai led)
Sandst one and shale,
int er laminat ed
8 25 63 11 63 6. 9
ZAC, Mngeni
MNE1
(fai led)
Shale and sandy shale
under lain by sandst one and
coal
6
to
8
27 35
to
75
7
to
14
85 6. 0
to
11. 9
Matla 1, R14W
(stable)
Sandy shale under lain by
int er laminat ed shale and
sandst one
6 23 35 6 63 3. 8
Matla 2, M12N
(stable)
Shaly sandst one 8 27 77 15 85 12. 7
ZAC,
Mngeni, MN
(stable)
Laminat ed sandy shale and
shaly sandst one
8 28 56 11 85 9. 4
ZAC,
Maye - roof
(stable)
Shaly sandst one and
sandst one
8 22 77 11 63 9. 4
116

Table 7.2. Foundation strength by Terzaghis method.
Values taken from Table 6.1 are marked *.

Figure 7.2. Determination of cohesive strength.
CASE Soft floor
lithology
(2m)
UCS*
(MPa)
DRMS
(MPa)
Tensi l e
strength
*
(MPa)
Cohesi on A
(MPa)
from Err or !
Refer ence
sour ce not
found.
=20
Cohesi on
B
(MPa)
from
0. 16 x
DRMS
Half
pillar
width
(m)
Foundation st rength
(MPa)
Average
pi l l ar
stress
(MPa)
Cohesio
n A
=20
Cohesion
B
=20

Emaswati
(fai led)
Sandst one
and shale,
int er laminat
ed
63 11. 1 5. 5 9. 7 1. 8 4. 85 224 43 4. 42
Matla 1,
R14S
(fai led)
Ver y soft
sandy shale
35 3. 2 2. 9 4. 9 0. 5 2. 55 113 12 3. 78
Welgedacht
(fai led)
Sandst one
and shale,
int er laminat
ed
63 6. 9 5. 5 9. 7 1. 1 9. 05 225 28 7. 50
ZAC,
Mngeni
MNE1
(fai led)
Shale and
sandy shale
under lain
by
sandst one
and coal
35
to
75
6. 0
to
11. 9
2. 9
to
5. 5
4. 9
to
9. 7
1. 0
to
1. 9
7. 8 115
to
225
25
to
46
7. 37
Matla 1,
R14W
(stable)
Sandy shale
under lain
by
int er laminat
ed shale and
sandst one
35 3. 8 2. 9 4. 9 0. 6 2. 95 113 15 3. 76
Matla 2,
M12N
(stable)
Shaly
sandst one
77 12. 7 5. 5 9. 7 2. 0 4. 30 224 47 4. 84
ZAC,
Mngeni, MN
(stable)
Laminat ed
sandy shale
and shaly
sandst one
56 9. 4 4. 2 7. 4 1. 5 8. 01 172 37 4. 72
ZAC,
Maye - roof
(stable)
Shaly
sandst one
and
sandst one
77 9. 4 5. 5 9. 7 1. 5 5. 56 224 36 6. 50
90-
Shear
stress
T
o

Normal
stress
-T
o
3T
o

c
117
7.2 Vesics Equation

Vesics equation was modified by Speck (1981) to include undercl ay water content rather
than underclay triaxial strength. From this modified equation it is possible to determine a
heave factor which is equivalent to a factor of safety and defined as:

OS
MBC
HF = Equation 4

Where:
MBC = Mine-floor bearing capacity (MPa)
*

OS = Overburden stress applied to the mine floor via the coal pillar (MPa).

*
Based on natural water content correlation with triaxial compressive strength.

The mine floor bearing capacity (MBC) is determined as follows:

RF UWC N MBC
mu
) 167 2070 ( = Equation 5

Where:
UWC = Underclay natural water content (%).
RF = Reduction factor (0.15)
N
mu
= A bearing capacity factor based on underclay natural water content and
determined as follows:

( ) ( )
( )
|
.
|
\
|
+
|
.
|
\
|
+
(
+
|
.
|
\
|
+
(
+ + +
(
(
+ + + +
|
.
|
\
|
+
|
.
|
\
|
=
1
*
1
*
1
* *
1
*
1
1
*
1
2
*
1 1
* *
c
N
H c
rN
H c
N
H c
N
H
r
c
N r r
H c
N
H
r
c
N r
H c
N
c
N r
mu
N
| | | |
| | |

Equation 6

Where:
N
c
*
= Equivalent to Vesics N
c
*
= 6.17
H
= P/4H (equivalent to Vesics term)
P = coal pillar width (m)
H = underclay thickness (m)
r =
| | 15 . 0 167 2070 UWC
S
uc
Equation 7
Where:
S
uc
= 60% of the average cohesive strength of claystone from Specks laboratory
test data. (248psi in the original reference)

The failed and stable case studies have been analysed according to Vesics modified
formula and the results are tabulated in Table 7.3. Where no test data are available Table
6.1 has been referred to, such cases are marked (*). The modified formula was developed in
the USA with data based on Imperial units. Due to the use of constants derived with
Imperial units, it was decided to use Imperial units in this case, up to the determination of
MBC, and then convert to SI units.

118
Overburden stress is equivalent to average pillar stress (APS) and has been determined as
follows:

2
2
25
w
HC
APS OS = = ( kPa) Equation 8

Where:
H = cover depth (m)
C = pillar centre distance (m)
w = pillar width (m)

Table 7.3. Heave factors for failed and stable cases from Vesic's modified formula
CASE Fl oor
moi st ure
content
(%)
Soft
fl oor
thi ck-
ness
(ft)
Pi l l ar
wi dth
(ft)
r H Nmu MBC
(psi )
MBC
(MPa)
Cover
depth
(m)
Pi l l ar
centres
(m)
Pi l l ar
wi dth
(m)
OS
(MPa)
HF
OS
MBC

Emaswati
(fai led)
2. 05* 1. 64 31. 83 0. 96 4. 85 6. 06 1570 10. 83 65 16 9. 7 4. 42 2. 45
Matla 1,
R14S
(fai led)
10. 15 1. 61 16. 73 4. 41 2. 60 7. 22 406 2. 80 32. 5 11 5. 1 3. 78 0. 74
Welgedacht
(fai led)
2. 20 1. 97 59. 38 0. 97 7. 54 6. 05 1545 10. 65 165 24. 4 18. 1 7. 50 1. 42
ZAC,
Mngeni
MNE1
(fai led)
2. 27 0. 72
to
3. 77
51. 64 0. 98 3. 42
to
17. 93
6. 04
to
6. 13
1531
to
1554
10. 56
to
10. 71
151 22 15. 74 7. 37 1. 43
to
1. 45
Matla 1,
R14W
(stable)
2. 05* 1. 38 19. 36 0. 96 3. 51 6. 09 1578 10. 88 36. 4 12 5. 9 3. 76 2. 89
Matla 2,
M12N
(stable)
3. 06 0. 98 28. 22 1. 06 7. 20 6. 38 1491 10. 28 63. 7 15 8. 6 4. 84 2. 12
ZAC,
Mngeni, MN
(stable)
1. 53 4. 33 52. 56 0. 91 3. 03 6. 02 1638 11. 29 100. 1 22 16. 02 4. 72 2. 39
ZAC,
Maye - roof
(stable)
2. 05* 3. 94 36. 45 0. 96 2. 31 6. 13 1588 10. 94 111 17 11. 11 6. 50 1. 68

The heave factor is in essence a factor of safety with any value lower than one signifying
a potentially unstable mine floor and vice versa. As can be seen from Table 7.3, this
appears not to be the case for southern African conditions. However if a safety factor cut
off point of 1.5 is assumed, the modified Vesics formula (Speck, 1981) appears to predict
pillar failure for all but the Emaswati case and this may have been different if actual
moisture content was known for the floor strata in this case.

7.3 Conclusions for Terzaghis and Vesics methods

The following conclusions have been reached with respect to the two floor bearing capacity
methods assessed:
119
o Terzaghis formula, in its current form, does not appear to predict foundation
behaviour for soft coal mine floor material. Further work is needed to determine the
actual value for cohesion and angle of internal friction for weak coal floors.
o Vesics equation, as modified by Speck, predicted stability successfully in 75% of
failed cases and 100% of stable cases. However, the number of case studies is small
and considerable additional effort will be required in evaluating the applicability of
this formula for South African conditions. With more accurate determination of
moisture content, this approach may be suitable for the prediction of coal mine
foundation stability.
o Vesics modified equation shows promise as a tool to be used in determining the
likely behaviour of soft foundations and may be included in the design guidelines.

8 CASE STUDY COMPARISONS

The approach taken to analyse the case studies is simply to compare as many aspects as
possible from the failed and stable cases to try and identify any common factors or
relationships.

8.1 Identification of similarities

The three following tables show ranges of values and all similarities consider ed important
for the failed (Table 8.1) and stable cases (Table 8.2) as well as a comparison between the
two types of case (Table 8.3). The case study summaries in Appendices B and C have been
used for this purpose. The reference number in the left hand column of each of the three
tables is for ease of cross-reference.

This study centred around identifying common factors which would allow the separation of
the two types of cases. Of all the varying factors studied, no clear separation between the
failed and stable cases was noted. The best relationships were those where the overlap was
the smallest.

As the stable cases are, by definition, those in which the floor exhibited signs of damage
but the pillars remained stable, an argument can be made that there is no value in
comparing the physical characteristics of the floor or the associated panel parameters, as
they all represent the failed floor condition, i.e. the failed and stable cases both have
failed floor strata present. However, it is considered likely that differences in some of
these characteristics or parameters may be influential in determining the degree of
deterioration of the floor and whether or not the pillars will f ail as a result.

120
Table 8.1. Summary of failed cases.
Ref. Aspect Range Remarks
1 Ave. mining ht. (m) 2.4 to 3.8 3 cases range only 2.4 to 2.77m
2 Pillar w/h ratio 2.1 to 5.6 3 cases range only 4.1 to 5.6
3 Pillar safety factor
(Salamon)
1.51 to 2.3 3 cases the range is 2.2 to 2.3
4 Panel width (m) 116 to 271 Width is over 200m in 3 cases
5 Seam dip () 0 to 6 2 or less in 3 cases
6 Roof lithology Massive sandstone Variable thickness but always
present
7 Floor lithology Sandy shale/ interlaminated
sandstone/shale
Variable thickness but the floor
is invariably described as being
shaly
8 Soft floor thickness (m) 0.49 to 1.15 3 cases range is 0.49 to 0.6m
9 Geological structure Dykes and faults present Dolerite present in 3 of the
cases
10 Cover depth (m) 32.5 to 165 Large variation
11 Pillar centre distance (m) 11 to 24.4 Large variation
12 Ave. pillar stress (MPa) 3.8 to 7.5 Large variation
13 Pillar joint spacing (m) 0.6 to 2m Large variation
14 CMRR 60.3 to 64.7 Only 2 tested
15 Average Impact Splitting
rating
98 to 215 Large range moderate to very
good overall roof conditions
16 W/h ratio of soft floor 10.2 to 30.2 Large variation
17 Slake durability index
Weighted SD for 0.5m of
floor
14.8 to 62.7
7.5 to 58.7
Large range
18 Floor rating (FR) 4.5 and 20.5 Only 2 tested
19 Class of floor heave 2 x type I. Type II and
new.
2 x open topped and 2 x
closed ridge
Large range
20 Max. floor heave (m) 0.1 to 2.0 Large range
21 Convergence rate mm/day 0.24 to 7.5 Large range, only measured at
2 sites
22 Surrounding workings Virgin / bord and pillar /
stooped
Large range
23 Pillar width (m) 5.1 to 18.1 Very large range
24 Moisture content (%) 1.36 to 10.38 Large range
25 Panel width / cover depth
(m)
1.21 to 3.56 Large range
26 MAP3D numerical
modelling results
Tensile stresses in bords
only in three out of four
cases
Slight agreement
27 PHASE2 numerical
modelling results
Minor in-plane stress (
3
) is
tensile in all cases
Very good agreement
28 Vesics modified
equation (HF)
0.74 to 2.45 Predicted failure in 75% of
cases

121
Table 8.2. Summary of stable cases.
Ref. Aspect Range Remarks
1 Average mining height
(m)
1.47 to 2.57 This represents the lower range
of the failed cases
2 Pillar width to height
ratio
2.5 to 7.07 In 3 cases it is over 4.4
3 Pillar safety factor
(Salamon)
2.1 to 3.5
(0.55 after chequerboarding)
Fairly good agreement
4 Panel width (m) 74 to 205 In 3 cases less than 130m
5 Seam dip () 0 to 5 1 or less in 3 cases
6 Roof lithology Generally sandy shale
overlain by massive
sandstone
In 3 cases, sandy shale is
present beneath the massive
sandstone beam
7 Floor lithology Sandy shale to shaly
sandstone
Split 50% - 50% between sandy
shale and shaly sandstone
8 Soft floor thickness (m) 0.3 to 1.32 Large range evenly split
between very thick >1m and
thinner <0.42m
9 Geological structure Faults and / or dykes often
present
Dyke present in 3 cases
10 Cover depth (m) 34.2 to 111 Large range
11 Pillar centre distance (m) 12 to 22 Large range
12 Average pillar stress
(MPa)
3.3 to 6.5
(22.2 after chequerboarding)
In 3 cases it is between 3.3 and
5.0 MPa
13 Pillar joint spacing (m) None to 1m Large range
14 CMRR 72 Only 1 test
15 Average Impact Splitting
rating
88 to 116.05
118 to 221
Good roof in 2 out of 3 tested
cases
16 Soft floor w/h 9.3 to 28.7 Large range
17 Slake durability index
Weighted SD for 0.5m
floor
6.2 to 81.6
9.2 to 47.7
Large range
18 Floor rating (FR) 17.5 Only 1 test
19 Class of floor heave Type I
2 x closed ridge and 1 x
open topped

20 Max. floor heave (m) 0.1 to 0.3 Fair agreement
21 Convergence rate
(mm/day)
Not measured
22 Surrounding workings Generally in virgin ground
23 Pillar width (m) 5.9 to 16.02 Large range
24 Moisture content (%) 1.53 to 2.09
25 Panel width / cover depth
(m)
1.03 to 3.68 Large range
26 MAP3D numerical
modelling results
Tensile stress zones in bords
in two out of four cases
No apparent relationship
27 PHASE2 numerical
modelling results
No tensile values for minor
in-plane stress (
3
) all cases
(if Mngeni MN
chequerboarding is
excluded)
Very good agreement
28 Vesics modified
equation (HF)
1.68 to 2.89 Indicated stability in all cases

122
Table 8.3. Failed vs. stable cases comparison.
Ref. Aspect Failed Stable Remarks
1 Ave. mining ht.
(m)
2.4 to 3.8 1.47 to 2.57 Only 0.17m overlap
2 Pillar w/h ratio 2.1 to 5.6 2.5 to 7.07 No apparent relationship
3 Pillar safety
factor
(Salamon)
1.51 to 2.3 2.1 to 3.5 Only 0.2 overlap
4 Panel width (m) 116 to 271 74 to 205 For stable cases, 3 out of 4
were less than 130m. Slight
agreement
5 Seam dip () 0 to 6 0 to 5 More stable cases where
seam is relatively flat
6 Roof lithology Massive sandstone Generally sandy
shale overlain by
massive sandstone
Behaviour of softer band
between top of seam and
massive sandstone roof to
be further investigated
7 Floor lithology Sandy shale to
interlaminated
sandstone/shale
50/50 split sandy
shale/shaly
sandstone
No discernable difference
between floor litho types
8 Soft floor
thickness (m)
0.49 to 1.15 0.3 to 1.32 No apparent relationship
9 Geological
structure
Dykes and faults
present
Faults and /or
dykes often
frequent
No simple relationship
between presence of dykes /
faults and performance of
floor / pillars.
10 Cover depth (m) 32.5 to 165 34.2 to 111 No apparent relationship
11 Pillar centre
distance (m)
11 to 24.4 12 to 22 No apparent relationship
12 Average pillar
stress (MPa)
3.8 to 7.5 3.3 to 6.5 In 3 of the stable cases it is
over 4.4 MPa but even then,
there is no strong
relationship
13 Pillar joint
spacing (m)
0.6 to 2m None to 1 No apparent relationship
14 Coal mine roof
rating (CMRR)
60.3 to 64.7 72 Insufficient data at present
15 Impact splitting
rating of roof
98 to 214 88 to 116.05 Roof stronger in failed
cases
16 Width to height
ratio of soft
floor
10.2 to 30.2 9.3 to 28.7 No apparent relationship
17 Slake durability
index
Weighted SD
for 0.5m floor
14.8 to 62.7
7.5 to 58.7
4.6 to 31.6
9.2 to 47.7
Lower range for stable
cases but large overlap of
values
18 Floor rating
(FR)
4.5 and 20.5 17.5 No apparent relationship
19 Class of floor
heave
2 x type I. Type II
and Newcastle /
2 x closed ridge
and 2 x open
topped

Type I /
2 x closed ridge
and 1 x open
topped.
More type I than any other
but fairly evenly split
between open topped and
closed ridge types.
20 Max. floor
heave (m)
0.1 to 2.0 0.01 to 0.3 Overlap only 0.2m
21 Convergence
rate mm/day
0.24 to 7.5 Not measured No comparison possible
123
Ref. Aspect Failed Stable Remarks
22 Surrounding
workings
Virgin ground in 2
cases and adjacent
to secondary
extraction in
another 2.
Generally in virgin
ground
No apparent relationship
23 Pillar width (m) 5.1 to 18.1 5.9 to 16.02 No apparent relationship
24 Moisture
content (%)
1.36 to 10.38 1.53 to 2.09 Limited data but there
appears to be a relationship
when compared on a seam
for seam basis
25 Panel width /
cover depth (m)
1.21 to 3.56 1.03 to 3.68 Similar range for stable and
failed cases
26 MAP3D
numerical
modelling
results
Tensile stresses in
bords only in 75%
of cases
Tensile stresses in
bords only in 50%
of cases
No apparent agreement
between failed and stable
cases.
May be better when
compared on a seam for
seam basis with tensile
stresses in bords and
intersections only modelled
for the number 5 seam.
27 PHASE2
numerical
modelling
results
Tensile stresses
present in floor
beam in all cases
No tensile stresses
present in floor
beam in all cases

Very good agreement
28 Vesics
modified
equation (HF)
0.74 to 2.45 1.68 to 2.89 Indicated failure or stability
in 87% of cases

8.2 Case studies analysed by heave classification

Weighted Slake Durability values for 0.5m of seam floor have been determined (where
possible) in an attempt to standardi se the floor rating system. During this exercise the floor
heave type was also noted down for each case and the relationship shown in Table 8.4
emerged.

As can be seen from Table 8.4, there appears to be a relationship between type of floor
heave and the weighted slake durability index for 0.5m of immediate floor. While the
number of cases is admittedly small, it appears that open topped floor heave can be
expected for a weighted SD of between 7.5 and 23.1 while an SD of 34.5 to 58.7 indicates
closed ridge type floor heave.
The failed and stable cases are fairly evenly split between the two types of floor heave and
so the determination of floor heave type is not considered influential in pr edicting
foundation stability.

124
Table 8.4. Relationship between slake durability and floor heave type.
Case study Weighted SD for 0.5m floor Class of floor heave
Matla 1, 5 seam, panel R14
South - failed
58.7 Very poor Closed ridge
Welgedacht, Alfred seam
stooping - failed
Closed ridge
Matla 1, 5 seam, Panel R14
West - stable
47.7 Very poor Closed ridge (assumed)
Matla 2, 5 seam, Panel M12
North - stable
34.5 Very poor Mostly Closed ridge
ZAC, Mngeni shaft, panel
MEN1 - failed
7.5 Good Mostly open topped
Emaswati Colliery, main
haulage south - failed
23.1 Poor Open topped
ZAC, Mngeni, panel MN -
stable
9.2 Good Open topped

A further examination of the case studies revealed the relationships shown in Figure 8.1
when class of floor heave is considered as the primary factor.

Table 8.5. Case studies compared by heave classification.
Ref. Aspect Closed ridge type
floor heave
Open topped
type floor
heave
Remarks
2 Pillar w/h ratio 2.1 to 5 4.1 to 6.3 Open topped heave with
greater width / height ratio
4 Panel width (m) >97 >183 Wider panels more prone to
open topped heave
5 Seam dip () Mostly flat 1 to 6 Closed ridge heave more
likely for flat seams
6 Roof lithology Often shaly
immediate roof
Only massive
sandstone
Open topped floor heave
associated with massive
sandstone roof
8 Soft floor thickness
(m)
0.3 to 0.6 0.22 to 1.32 Larger range for open topped
heave
13 Pillar joint spacing
(m)
>1.43 0.5 to 1 Open topped heave more
likely where pillars are
highly jointed
14 Coal mine roof rating
(CMRR)
64.7 to 72 60.3 Insufficient data at present -
relationship may exist
15 Impact splitting
rating of roof
71.9 to 116.05 98 to 269 Stronger roof types
experience open topped
heave
17 Weighted SD for
0.5m floor
34.5 to 58.7 7.5 to 23.1 Two distinct ranges- slaking
floor likely to exhibit closed
ridge type heave
18 Floor rating (FR) 4.5 and 17.5 20.5 Weaker floors experience
closed ridge type heave
limited data
24 Moisture content (%) 1.36 to 10.38
avg. 5.1
1.53 to 2.27
avg. 1.9
Closed ridge heave more
likely with greater moisture
content
125

Geotechnical aspects and heave classification
-20
0
20
40
60
80
100
120
CR CR CR CR OT OT OT RF
Floor heave classification
S
l
a
k
e

d
u
r
a
b
i
l
i
t
y

i
n
d
e
x

&

m
o
i
s
t
u
r
e

(
%
)
0
50
100
150
200
250
I
m
p
a
c
t

s
p
l
i
t
t
i
n
g

r
a
t
i
n
g
SD - floor avg SD moist x 10
IS - roof Linear (avg SD) Linear (SD - floor)

Figure 8.1. Relationship between slake durability and class of floor heave.

8.3 Case study findings

With only four stable and four failed case studies there is insufficient data to establish
reliable correlations or trends. No single factor has been identified in this study, which
clearly defines the likelihood of failure of coal pillars standing on weak floor rock. From
the preliminary analysis contained in Table 8.3, the following aspects appear to be
associated with this type of pillar failure:
o Pillars with a Salamon safety factor in excess of 2.3 were stable.
o Mining height - taller (not necessarily less squat) pillars being more prone to
foundation damage.
o Panel width and influence of seam dip.
o Maximum height of floor heave.
o The tendency for more deterioration towards the centre of panels (Matla 1, 5 seam
and Emaswati) has been analysed by numerical modelling. Higher pillar loads were
only modelled for the Matl a case and it is likely that deformation of the strong roof
beam is responsible for the damage observed. The greatest deflection would be
expected towards the centre of the panel. It is expected that the beam would behave
as a Voussoir beam due to the presence of joints in the roof.
o The roof (as tested by the impact splitting method) appears stronger in the failed
cases. This suggests that a strong, stiff roof beam is instrumental in driving the
floor and pillar failures.
o It appears that the weighted slake durability index for the immediate 0.5m of floor
may show promise in predicting the type of floor heave which may be expected
(i.e. open topped or closed ridge). However, the failed and stable cases are
fairly evenly distributed between the two heave t ypes and thus they have no
apparent usefulness for predicting pillar failure.
126
o A number of other aspects appear to indicate the type of floor heave that may be
expected (Figure 8.1). While this is interesting, the pr actical value of this
knowledge is questionable in the light of the previous statement.

9 SUMMARY OF FINDINGS OF THIS RESEARCH PROJECT

9.1 Previous work on foundation stability

Some of the work carried out on this problem in South Africa, Australia and the USA was
reviewed to determine if there were any common factors which could be considered when
analysing the case study data.

The most significant findings of the literature survey, roughly in perceived order of
importance, are as follows:
o The following seams have floors which are prone to failure or damage: the 5 seam
of the Witbank and Highveld coalfields, the Free state number 3 seam as well as the
Alfred and Main seams in KwaZulu-Natal.
o Natural moisture content, slake durability and Duncan swell appear to be the most
useful data for the assessment of soft floor behaviour in South Africa. They have
the added advantage of being relatively easy and inexpensive to obtain.
o The presence of water (saturated floor rock) or dolerite intrusions (burning) is
recorded in a significant number of cases of floor failure both here and overseas.
o Pillar settlement and floor heave as experienced in the United States and Australia
tend to be more of an economic consideration than a safety hazard as they lead to
surface subsidence and underground roadway closures.
o In southern Africa, pillar failures resulting from foundation instability can have
significant economic consequences, but they also potentially pose more of a threat
to the safety of underground personnel.
o Average pillar width to height ratios are 2.96 and 4.29 for the South African cases
and US / Australian cases, respectively. For the failed case study sites the range of
width to height ratios is 2.1 to 5.6.
o While the use of numerical modelling has been evaluated in a number of studies, a
clear set of guidelines has not been established to assist with the design of pillars
standing on soft floors. The numerical modelling portion of this study also came to
a similar conclusion (9.3).

9.2 Case studies

Although the number of stable and failed case studies is admittedly low, a number of
promising relationships are suggested. They are listed below in their perceived order of
significance:
o Pillars with a safety factor in excess of 2.3 (as calculated using Salamons formula)
were found to be stable.
o Mining height - taller (not necessarily less squat) pillars being more prone to
failure as a result of foundation damage. This is possibly due to taller pillars
having a greater propensity to buckle in the near vertical axis.
127
o Impact splitting ratings and composition of the seam roof, failure being associated
with stronger roof conditions. Deformation of a thick and strong roof beam may be
influential in loading of the floor through t he pillars. It could also be that the floor
damage is only the outward manifestation of already occurring pillar failure and
that the floor does not fail first
o Panels with a width of 130m or less seem less prone to failure.
o Relatively flat seams (dip 1 or less) are less prone to failure.
o Where floor heave is low (typically less than 0.1m) the pillars appear to be more
stable.
o It appears that the weighted slake durability index for the immediate 0.5m of floor
may show promise in predicting the type of floor heave which may be expected
(i.e. open topped or closed ridge). This is an interesting finding and while not
relevant at present to predicting stability, it may be useful when applied to a larger
database.
o A number of other aspects appear to indicate the type of floor heave that may be
expected and may also be found to be more significant if a larger number of cases
is studied in future.

9.3 Numerical modelling

Both three and two dimensional numerical modelling was carried out to simulate the stress
conditions for the failed and stable case study sites. While the lack of in situ stress field
measurements precluded an analysis of the absolute values obtained from the numerical
modelling outputs their value as tools for gaining a better understanding of the problems is
recognised. As may be expected from elastic modelling, deformations were an order of
magnitude smaller than those measured in the field.

The most important findings from the modelling are as follows:
o The two-dimensional PHASE
2
model was successful in indicating zones of tensile
minor principal stresses in the floor beams of the failed cases.
o The three-dimensional MAP3D model was, in some cases, able to indicate where
damage is likely to appear in the floor and, to a lesser extent, the pilla rs.

9.4 Civil engineering based stability analyses

Two civil engineering based methods were used to see if they could be used to predict
foundation performance for soft floors beneath coal pillars. The methods used were
Terzaghis formula and Vesics modi fied formula. While the number of case studies is
small, some promising findings were made:
o Reasonably good results were obtained using Vesics equation as modified by
Speck (1981). It was capable of predicting 75% of failed cases and 100% of stable.
o Terzaghis method did not produce any meaningful results.

128
10 STABILTY RATING

In the following section, a simple stability rating is established as an aid to designing
pillars for seams with known foundation failure problems. The development of this rating
has been based on the findings of the research described in this research report.

10.1 Discussion of findings

Even though the small number of case studies is insufficient to allow for determining
trends and definitive design procedures, there were sufficient relationships identified to
enable a set of basic design guidelines to be produced.

The data was initially sorted by various data sets and the only complete separation of
failed and stable cases resulted from sorting by safety factor. The split between stable and
failed is illustrated in Figure 10.1.
The following relationships are considered significant:
o The following seams have floors, which are prone to failure or damage, the 5 seam
of the Witbank and Highveld coalfields, the Free state number 3 seam as well as the
Alfred and Main seams in KwaZulu-Natal. For these seams:
o No pillar failures may be expected where the safety factor exceeds 2.3 ( Figure
10.2). Note that during this analysis only the Salamon and Munro (1967) safety
factor formula has been used.
o Where the panel width is 150m or less, the chances of failure is appreciably
lower (Figure 10.3).
o Failure is unlikely where the mining height is less than 2.3m (Figure 10.2).
o Flat seams appear slightly less prone to failure ( Figure 10.4).
o There appears to be a weak relationship between the impact splitting rating of
the roof and pillar stability. Stronger roof types feature more in failed cases
(Figure 10.5).
o There is no apparent relationship between the weighted average slake durability
rating for 0.5m of floor and pillar failure (Figure 10.6).

It appears that when the average slake durability rating of the immediate 0.5m of soft floor
exceeds 30, closed ridge floor failure will occur.

Numerical modelling, in particular the PHASE
2
two-dimensional programme, was found to
be useful in predicting behaviour of the soft floor.

Further work is indicated with respect to gathering geotechnical data before any of the two
civil engineering approaches to foundation bearing capacity can be properly assessed.
However, Vesics equation as modified by Speck (1981) appears the more promising of the
two looked at in this project.

129

Relationships plotted for stable or failed cases
0
50
100
150
200
250
300
F (Welg) F (Z Mng) F (Mat 1) F (Emas) S ( Z Mng) S (Z Maye) S (Mat 1) S (Mat 2)
Case studies
I
S

a
n
d

a
v
g
.

S
D

r
a
t
i
n
g

&

p
a
n
e
l

w
i
d
t
h
0
1
2
3
4
5
6
7
S
a
f
e
t
y

f
a
c
t
o
r

/

m
i
n
i
n
g

h
e
i
g
h
t

/

d
i
p

&

f
l
o
o
r

h
e
a
v
e
panel width IS - roof avg SD safety factor mining height dip max heave
Stable / failed cut off line
safety factor = 2.5

Figure 10.1. Relationships plotted for stable and failed cases.

safety factor vs mining height
1.5
2
2.5
3
3.5
4
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
safety factor
M
i
n
i
n
g

h
e
i
g
h
t

(
m
)
failed stable

Figure 10.2. Safety factor vs. mining height.
130

Safety factor vs panel width
0
50
100
150
200
250
300
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
safety factor
P
a
n
e
l

w
i
d
t
h

(
m
)
failed stable

Figure 10.3. safety factor vs. panel width.

safety factor vs seam dip
0
1
2
3
4
5
6
7
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
safety factor
S
e
a
m

d
i
p

(
d
e
g
)
failed stable

Figure 10.4. Safety factor vs. seam dip.

131

Safety factor vs roof IS rating
0
50
100
150
200
250
2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3
safety factor
R
o
o
f

I
S

r
a
t
i
n
g
failed stable

Figure 10.5. Safety factor vs. roof impact splitting rating.

Safety factor vs avg. SD rating
0
10
20
30
40
50
60
70
2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3
Safety factor
A
v
e
r
a
g
e

S
D

r
a
t
i
n
g
failed stable

Figure 10.6. Safety factor vs. average weighted slake durability rating.

132
10.2 Stability rating

With no strong relationships evident, and a number of aspects showing weak agreement, it
was decided to investigate the use of a rating based approach to the floor performance. This
method works on the premise that as the number of factors identified with failure increases
so too does the likelihood of instability.

As parameters such as safety factor and panel width are outcomes of the method they were
not used in the development of the proposed Floor Stability Rating (FSR) as shown in
Table 10.1.

The following characteristics were used:
o Mining height as this aspect appears to have the strongest relationship it is
weighted as double that applied to any of the other three aspects. Note that mining
height and seam thickness are not necessarily the same. It may be necessary to add
the thickness of soft floor likely to be mined. As this aspect has the strongest
perceived relationship it carries a weighting double that of the other factors as
shown in Table 10.1.
o Weighted average slake durability rating for 0.5m of floor. While no apparent
relationship exists between this parameter and pillar failure it is influential in
determining the type of failure expected. It is also considered a useful and simple
characteristic to determine relati ve weakness of the floor material.
o Impact splitting rating for the roof.
o Seam dip.

Even though PHASE
2
numerical modelling gave very encouraging results, in that the minor
in-plane stress (
3
) was found to be tensile in all the failed case studies, it is not used for
this stability analysis as it cannot always be assumed that persons needing to design pillars
on soft foundations will have access to the programme.

Table 10.1. Floor Stability Rating matrix.

Range of values
Mining height Thickness (m) <2.3 2.3 - 2.8 2.8 - 3.2 >3.2
Description Low seam Medium seam High seam
Rating value 2 4 6 8
Dip Dip () <1 1 - 3 3.1 - 5 >5
Description Flat Slight dip Moderate dip Steeply
dipping
Weighted average SD
rating for 0.5m

0.5m floor SD
rating
<10 10 - 30 31 - 50 >50
Description Low slaking
potential.
Good floor
Moderate
slaking
potential.
Moderate to
poor floor
Moderate
slaking
potential. Very
poor floor
High slaking
potential.
Extremely
poor floor
Roof IS rating

IS rating for 2m of
immediate roof
< 88 89 - 113 114 - 130 >130
Description Moderate to
good roof
Good roof Very good roof Very good to
excellent roof

133
Applying the FSR to the case studies (and using average values where actual values were
not available shown in bold type in Table 10.1) the following values were obtained:

Table 10.2. Floor Stability Rating values for case studies.
Case Study Values from stability rating matrix Floor
Stability
Rating
Mining height Dip SD IS
ZAC, Maye - stable 2 1 2.5 1 6.5
Matla 2, 5 seam - stable 2 1 3 2 8
Matla 1, 5 seam - stable 4 1 3 2 10
ZAC, Mngeni - stable 4 3 1 2.5 10.5
Matla 1, 5 seam - failed 4 1 4 2 11
Emaswati - failed 4 2 2 3 11
ZAC, Mngeni - failed 4 4 1 4 13
Welgedacht - failed 8 2 2.5 2.5 15

The FSR values were then plotted against the safety factor and panel width as shown in
Figure 10.7. As can be seen there is better correlation between the FSR and safety factor
(as determined with the Salamon and Munro (1967) formula) than with the panel width.
The curves shown in Figure 10.7 represent the best fit attainable and are linear with respect
to safety factor and exponential for panel width.

Floor Stability Rating vs safety factor and panel width
y = 50.839e
0.1063x
R
2
= 0.5458
y = -0.1741x + 4.278
R
2
= 0.7602
0
50
100
150
200
250
300
6 7 8 9 10 11 12 13 14 15 16
Floor Stability Rating
P
a
n
e
l
w
i
d
t
h

(
m
)
0
0.5
1
1.5
2
2.5
3
3.5
S
a
f
e
t
y

f
a
c
t
o
r
panel width safety factor Expon. (panel width) Linear (safety factor)

Figure 10.7. Safety factor and panel width plotted against FSR.

134
Floor Stability Rating vs FSR/safety factor
y = 0.6407e
0.1805x
R
2
= 0.9539
0
2
4
6
8
10
12
5 7 9 11 13 15 17
Floor Stability Rating (SR)
S
t
a
b
i
l
i
t
y

r
a
t
i
n
g

/

s
a
f
e
t
y

f
a
c
t
o
r
FSR / SF Expon. (FSR / SF)
F
F
F
S
S
S
S
F
Stable/failed cut-off point

Figure 10.8. Floor Stability Rating design chart.
The curve shown in Figure 10.8 represents the best fit and in this case is exponential.

10.3 Design procedure

The following rudimentary design guidelines are suggested as an interim measure pending
further work on the subject.
o These guidelines are only applicable to the following seams:
Witbank and Highveld coalfields; number 5 seam.
Kwa-Zulu Natal and Swaziland; Alfred and Main seams.
o Where the FSR exceeds 10, the safety factor of pillars must be designed so that the FSR
/ safety factor ratio does not exceed 4. This may be determined with the following
simple formula:
4
FSR
SF
FSR
= Equation 9
Where:
SF
SR
= New safety factor adjusted for stability rating.
FSR = Floor Stability Rating as determined from the matrix ( Table 10.1).

Panel widths should not exceed 150m with the continuous inter-panel barrier pillars
being at least as wide as the pillars in the adjacent panel. Where the pillars in two
adjacent panels are of differing sizes, then the inter -panel barrier pillar must be the
width of the wider of the two pillar sizes. Note that the above equation is only valid for
FSR 10.
o Where the FSR is between 7 and 10 further investigation must be carried out. This
should take the form of modelling the problem with PHASE
2
, or alternatively, using the
modified Vesics formula (Speck, 1981) to analyse foundation stability.
135
o Where the minor in-plane stress is tensile at any point in the floor when
PHASE2 is used then FSR must be taken as 11.
o Where the modified Vesics formula indicates a heave factor of less than 1.5
then FSR must be taken as 11.
o A FSR of less than 7 indicates no likely foundation problems and normal design
methods can be adopted.

Note that the above is a first pass assessment of a complex problem and due to the
limited cases studied cannot be viewed as comprehensive. It does however provide a
starting point in the design of pillars with potentially soft foundations. With more data it
may be possible to simplify the approach to allow analysis by means of the matrix only. In
this case, negative values may be assigned for high safety factors to effectively reduce the
FSR to below 10.

11 CONCLUSIONS

This project consisted of four main parts, namely: a study of a number of reports and
papers on foundation instability both local and international, analysis of four failed and
four stable case study sites, numerical modelling of case studies and evaluation of two civil
engineering methods. The following conclusions have been drawn from the research work:

i. The 5 seam of the Witbank and Highveld coalfields, the Free state number 3 seam as
well as the Alfred and Main seams in KwaZulu-Natal have floors which are prone to
failure or damage.

ii. On the affected seams, no pillar failures may be expected where the safety factor
exceeds 2.3 as determined using the Salamon and Munro safety factor formula.

iii. Where the panel width is 150m or less the chances of failure is appreciably lower.

iv. Failure is less likely when the mining height is under 2.3m.

v. Flat seams appear slightly less prone to failure and there appears to be a weak
relationship between the impact splitting rating of the roof and pillar stability, with
stronger roof types featuring more in failed cases.

vi. Numerical modelling, in particular the PHASE
2
two-dimensional programme, was
found to be useful in predicting behaviour of t he soft floor.

vii. Further work is indicated with respect to gathering geotechnical data before any of
the two civil engineering approaches to foundation bearing capacity can be properly
assessed. However, Vesics equation as modified by Speck (1981) appears the more
promising of the two looked at in this project and may be used as a decision making
tool if necessary.

viii. It was possible to produce a basic, provisional, set of design guidelines for seams
prone to foundation failure. These guidelines are based on a floor stability rating
(FSR) matrix.

136
As the number of cases studied is relatively small further work is indicated, and
recommendations are given below.

12 RECOMMENDED FURTHER WORK

The solution of this complex problem requires considerable further work in order to refine
the design of stable coal pillars on weak foundations. A larger data base than that of only
Ingwe and Eyesizwe Mining collieries is necessary and it should at least incorporate data
from the other three major coal mining groups. Many of the proposals listed below will be
difficult and expensive to carry out and any detailed additional study will probably be
beyond the resources available to a single researcher. Further work may best be carried out
under the auspices of a research organisation and should cover at least the following:
o Collection of similar data on mines of other groups to make this an industry wide
study. Data to be collected should include at least the following: seam, coalfield,
mining geometry, panel width, seam dip, presence of dolerite intrusions, roof and
floor lithology and coal pillar jointing.
o The database should be enlarged to include a third category of cases where even
though soft floor is present there is no visible damage to the floor or pillars. This
may be classified as the undamaged situation.
Determination of the following geotechnical parameters at all sites (It is in this
phase that considerable expense is likely to be incurred, as boreholes will probably
have to be drilled at all sites to be able to obtain this data):
o Slake durability, Duncan swell and moisture content of the floor.
o UCS or diametral point load and impact splitting rating of roof and floor.
o Seam dip and pillar jointing.
o Indirect tensile strength and angle of internal friction.
o Horizontal stress field.
o Analysis of gathered data to determine which combination of factors best predict
floor and pillar failure.
o Further evaluation of various numerical modelling approaches.
o Evaluation of values assigned for the FSR matrix as well as possibly adding new
factors as dictated by the additional research.
o Evaluate modifying the FSR to include safety factor and panel width and thereby
simplify the design process.

Mining on the seams that have hitherto been subject to pillar failure associated with
foundation failure produces a relatively small proportion of the total tonnes mined in
South Africa. Prior to conducting further work it would be necessary to consider the
long term future of mining on these seams while bearing in mind the safety
implications and possible economic and environmental consequences of pillar failure.

13 REFERENCES

Note: Publications marked * were referred to in other reports and papers studied by the
author during the course of the literature survey. They were not actually read by the author.

Brady, B.H.G. and Brown, E.T. (1985) Rock Mechanics for Underground Mining. George
Allen & Unwin. London. pp 331 & 332.
137

Buddery, P.S. (1987). Visit to section 2 - Wednesday 28
th
October 1987. Trans-Natal Rock
Mechanics Dept. internal memorandum. November.

Buddery, P.S. (1988a). Visit to section 3, 30 November 1988. Trans -Natal Rock Mechanics
Dept. internal memorandum. December.

Buddery, P.S. (1988b). Re: Section 3, South Shaft. Trans -Natal Rock Mechanics Dept.
internal memorandum. December.

Buddery, P.S. (1989). Fundamental principles of roofbolting in South African collieries.
SANGORM Symposium; Advances in Rock Mechanics in Underground Coal Mining.
Witbank. September.

Buddery, P.S. (1991). Major pillar collapse, Emaswati Colliery, Swaziland. Genmin Rock
Engineering Department; internal report. June.

Buddery, P.S. (1993). Geotechnical investigation of the Mgeni B and Maye reserve areas.
GENROC internal report. November.

Buddery, P.S. (1994). "Ngeni" B shaft site. GENROC internal memorandum. March.

Buddery, P.S. (1995a). 5 Seam pillar collapse, No 1 Mine. GENROC internal memorandum.
February.

Buddery, P.S. (1995b). 5 Seam pillar collapse, No 1 Mine. GENROC internal
memorandum. March.

Buddery, P.S. (1995c). Matla N
o
1 mine: 5seam collapse area. Gencor Rock Engineering
internal memorandum. June.

Buddery, P.S. and Macgregor, I.M. (1990). Recent falls of ground in the Klipriver coal
field, Natal. South African Colliery Managers Association. Papers Presented 1984 1988.
Pub. Chamber of Mines of S.A. pp 20 51.

Buddery, P.S. and Oldroyd, D.C. (1992). Development of a roof and floor classification
applicable to collieries. ISRM Symposium, Eurock 92. pp 197 202.

Buddery, P.S. (2002). Personal communication arising from review of this project report.
November.

Chen, J.F. (1994). Preliminary results on partial results. CSIR report for Matla Coal ltd.
August.

Chugh, Y.P. and Pytel, W.M. (1992). Design of partial extraction coal mine layouts for
weak floor strata conditions. Proc. Int. Workshop on Coal Pillar Mechanics and Design.
USBM IC9315. pp 32 49.

Chugh, Y.P., Atri, A and Dougherty, J. (1988). Laboratory and field characterisation of
immediate floor strata associated with Illinois No 6 and No 5 coal seams. Report for
Department of Mining Engineering, Southern Illinois University, Carbondale, IL. July. pp 1
14.

138
Chugh, Y.P., Chandrashekhar, K. and Caudle. R.D. (1988). A field geotechnical study of
the effects of weak floor strata on underground coal mining in Illinois. Key Questi ons in
Rock Mechanics, Cundall et al.(eds.). Balkema, Rotterdam. pp 681 690.

Coates, D.F. (1970). Rock Mechanics Principles. Mines Branch, Department of Energy,
Mines and Resources. Mines Branch Monograph 874. Revised.

Esterhuizen, G.S. (1997). Modification of coal pillar strength to account for the effect of
jointing. Contract research report for CSIR Miningtek as part of SIMRAC Project COL223.

Fauconnier, C.J. and Kersten, R.W.O. (1982). Rock mechanics of total extraction.
Increased Underground Extraction of Coal. SAIMM. Pp 26 111.

Federal Mine Safety and Health Review Commission. (2000). Decision in the case of
MSHA vs. Harlan Cumberland Coal co. Denver Colorado. 12
th
May.

Gale, W.J.* (1991). Strata control utilising rock reinforcement techniques and stress
control measures, in Australian coal mines. The Mining Engineer. January. pp 247 253.

Galvin, J.M.* (1995). Strata control for coal mine design. Roadway and Pillar Mechanics
Workshop.

Hattingh, K.F. (1987). Preliminary report on failing pillars in NW2N panel ZAC. Internal
report by Trans-Natal Steenkool Dienste. Rock Mechanics Department. August.

Hill, D., Gordon, N. and Madden, B. (1994). Floor heave database at Cooranbong Colliery.
ACIRL Report No CGE 1641/1. February.

Hill, D.J. (1994) Analysis and prediction of subsidence, with particular reference to first
workings panels. ACIRL report No CGE 1708/1. April.

Hill, R.W. (1989). Multiseam mining in South Africa. SANGORM, Eastern Transvaal
Branch, Symposium. Advances in Rock Mechanics in Underground Coal Mining.
September. pp 122.

Hill, R.W. (1996). Safety and environmental risks associated with shallow bord and pillar
workings. CSIR Miningtek, Final project report. Project n
o
. CE 9401. January. pp 8 49.

ISRM. (1972). Suggested met hods for determining swelling and slake-durability index
properties. Committee on Laboratory Tests, document 2 part 2, final draft. November.

Jack, B.W. and Madden, B.J. (1995). Matla Colliery, No 5 seam pillar collapse. SIMRAC
interim project report. December.

Jermy, C.A. and Ward, J.R. (1988). The application of geotechnical testing to the mining of
coal. SANGORM symposium. Rock Mechanics in Africa. November. pp 267 271.

Latilla, J.W. (1990). Addendum to report dated 9/1/90 Pillar failure in S2, main haulage
south, Emaswati. Internal memorandum, Trans Natal Coal Corporation.

Latilla, J.W. (1992). Geotechnical testing of Emaswati borehole cores western bypass
development. Genmin Rock Engineering Department internal report.

139
Latilla, J.W. (1993). Geotechnical report, shortwall project, No 2 mine, Matla Coal ltd.
Internal Genmin Rock Engineering Department Report. January.

Latilla, J.W. (1995). Routine visit to production sections. GENROC internal memorandum.
November.

Latilla, J.W. (1996a). Report on stability of 5 seam bord and pillar workings beneath
Eskom pylons, Matla No 2 Mine. IRED Internal report. July.

Latilla, J.W. (1996b). Matla Coal, No 2 mine longwall goaf induction project, geotechnical
report. GENROC internal report. April.

Latilla, J.W. (1997). Geotechnical report, Matla No 2 Mine, 2 seam shortwall project. IRED
internal report. September.

Latilla, J.W. (1997a). Poor roof, Maye, section 4. IRED Internal report. March.

Latilla, J.W. (1997b). Geotechnical report, Matla 2 mine, 2 sea m shortwall project.
September.

Latilla, J.W. (1997c). Mngeni chequerboarding, Maye poor roof, South shaft drop raise
IRED internal report. September.

Latilla, J.W. (1998). Routine visit December 1998. IRED Internal report. December.

Latilla, J.W., Oldroyd, D.C. and Wevell, E. (1997). Pillar failure associated with weak
floor rock. Three case studies. Coal Pillar Performance Workshop, SANGORM (Coalfields
Branch). August.

Latilla, J.W., van Wijk, J.J., Wevell, E. and Neal, D. (2002). Evaluation of the impact
splitting technique used for predicting geotechnical conditions in underground coal mines.
SANIRE 2002 Symposium. Re-Defining the Boundaries. Maccauvlei. September.

*Laubscher, D.H. (1990) A geomechanics classification system for the rating of rock mass
in mine design. Jl. S. Afr. Inst. Min. Metall.. Vol. 90, no. 10. pp 257-273.

Madden, B. (2000). Assessment of strata conditions, panel MEN1 at Mngeni shaft Zululand
Anthracite Colliery. Report produced for IRED. December.

Madden, B.J. and Hardman, D.R. (1992). Long term stability of bord and pillar workings.
COMA. Symposium on Construction Over Mined Areas. May. pp 38 46.

Madden, B.J. and Wagner, H. (1985). A study of in-situ strength properties of coal seams in
South African Collieries. SANGORM. Symposium on Rock Mass Characterisation.

Madden, B.J., Canbulat, I., Jack, B.W. and Prohaska, G.D. (1995). A reassessment of coal
pillar design procedures. SIMRAC Final project report. December.

Marais, M. (2002). Personal communication with D. Neal.

Mark, C. (2001). Personal communication. April.

Mark, C. and Mucho, T.P. (1991). Horizontal stress and its effects on longwall ground
control. Mining Engineering (USA). November. pp 1356-1360.
140

Mine Safety and Health Administration. (1999). Report of inves tigation. Fatal fall of roof
accident. Bowie mine. December 1998. April.

Minney, D. (2002) personal communication with D. Neal.

Molinda, G.M. and Mark, C. (1994). Coal mine roof rating (CMRR): A practical rock mass
classification for coal mines. USBM IC 9387. 83pp.

Oldroyd, D.C. (1990). Pillar failure in S2, Main haulage south, Emaswati. Trans Natal Coal
Corporation, internal report. January.

Oldroyd, D.C. (2000). Mngeni pillar overloading and continued deformation. Independent
consultants report for IRED. December.

Oldroyd, D.C. and Buddery, P.S. (1988). Two cases of progressive pillar failure at
collieries in Swaziland and Kwa-Zulu. SANGORM Regional Conference. Rock Mechanics
in Africa. November. pp 125 133.

Oldroyd, D.C. and Latilla J.W. (1989). Geotechnical testing of Emaswati Colliery core.
Genmin Rock Engineering Dept. internal report. December.

Oldroyd, D.C., Wevell, E and Kruger, A. (1996). Continuous miner burial on the Gus and
section entrapment on the Alfred seam during pillar extraction at Welgedacht. IRED
internal report. October. pp 11 21.

Ozan, T.T. (1991a). A preliminary review of factors affecting coal pillar foundation failure.
COMRO. Internal note C04/91. July.

Ozan, T.T. (1991b). Monitoring floor behaviour and stress changes in the tailgate of section
906 longwall panel at Durban Navigation Colliery. COMRO> Internal note C06/91.
October.

Ozan, T.T. and Prohaska, G.D. (1992). Monitoring the surrounding strata behaviour during
pillar extraction at section M1, panel 250 at Natal Anthracite Colliery. COMRO. Internal
note C03/92. May.

Ozbay, M.U. and Ryder, J.A. (1989). Investigations into foundation failure mechanisms of
hard rock squat pillars. Int. Symposium. Rock Mechanics and Rock Physics at Great Depth.

Peng, S.P. (1978). Coal Mine Ground Control. John Wiley and sons. pp 26 205.
Peng, S.S. and Tsang, P.* (1989). Yield pillar application under strong roof and strong
floor condition a case study. Rock Mechanics as a Guide for Efficient Utilization of
Natural Resources, Khair (Ed). pp 411 418.

Peng, S.S., Wang, Y.J. and Tsang, P. (1995). Analysis of floor heave mechanisms.
Presentation at the Annual Meeting of the Society for Mining Metallurgy and Exploration,
Inc. Denver, Colorado. March. pp 1 12.

Petzer, K.B. (1991). Geological report on the Emaswati ground failure. Internal report.
June.

Riefenberg, J. (1995). Towards a method of determining floor quality in an underground
coal mine. 14
th
Conference on Ground Control in Mining. Morgantown, WV. pp 200 208.
141

Rockaway, J.D. and Stephenson, R.W. (1982). Geotechnical evaluation of the support of
coal pillars in underground coal mines. Bulletin of the Association of Engineering
Geologists. Vol. XIX, No 1. pp 5 14.

Rocscience. (2003). Phase2 v. 5.0 Finite Element Analysis for Excavations.
http://www.rocscience.com. Cited January 2003.

Salamon, M.D.G. and Munro, A.H.* (1967). A study of the strength of coal pillars. Journal
of the SAIMM. September.

Santos, C.F. and Bieniawski, Z.T.* (1989). Floor design in underground coal mines. Rock
Mechanics and Rock Engineering 22. pp 249 271.

Seedsman, R.W. and Gordon, N. (1991). Methods to optimise mining operations affected by
weak claystone materials. NERDDC (meaning of acronym not mentioned) End of Grant
Report. Project No 1336. December.

Shakoor, A. and Sarman, R. (1992). Swelling potential classification of mudrocks based on
geotechnical properties. ISRM Symposium Eurock 92. Chester, U.K. September. pp 75
80.

Speck, R.C. (1981). The influence of certain geologic and geotechnical factors on coal
mine floor stability a case study. Proc. 1
st
Annual Conference on Ground Control in
Mining. Morgantown. WV. pp 44-49.

Stacey, T.R. and Page, C.H. (1986). Practical Handbook for Underground Rock Mechanics.
Trans-Tech Publications. pp 59-134.

Stacey, T.R. and Swart, A.H., (2001). Booklet on Practical Rock Engineering Practice for
Shallow and Opencast Mines. SIMRAC. October.

Stephenson, R.W. and Rockaway, J.D. (1981). Geotechnical evaluation of low strength
strata beneath coal seam. Proc. Int. Symposium on Weak Rock. Tokyo. September. pp 891
895.

Strata Control Technology (1993). Load Transfer Characteristics of the Spiral Bolt. Report
for Swiss Industrial Company (Australia). August. pp 16.

Su, D.W.H. and Hasenfus, G.J. (1999). Coal pillar strength and practical coal pillar design
considerations. Proc Second Int. Workshop on Coal Pillar Mechanics and Design. NIOSH
IC 9448.June..pp 155 162.

Summers, J.W., and Wevell, E. (9185). A study to determine the feasibility of high pressure
water infusion for weakening the roof. Anglo American Corporation Group Mining
Symposium. Johannesburg.

Terzaghi, K and Peck, R.B.* (1964). Theoretical soil mechanics. Engineering Practice.
John Wiley and sons, Inc. pp 137 1999.

van der Merwe, J.N. (1995). Practical Coal Mining Strata Control. Geoscientific
Technology Development, Sasol Coal Division. November. pp B.5.

van der Merwe, J.N. (1998). Personal communication.
142

van der Merwe, J.N. and Madden, B.J. (2002). Rock Engineering for Underground Coal
Mining. SAIMM. pp 9.

van Wijk, J.J. and Latilla, J.W. (2000). Pillar deterioration in panel MEN1 at Mngeni shaft.
IRED internal report. November.

Vesic, A.S.* (1975). Bearing capacity of shallow foundation. Foundation Engineering
Handbook, Winterkorn, H.D. and Fang, H. eds. Van Nostrand Reinhold Company. pp 121
147.

Wagner, H. (1974). Determination of the complete load deformation characteristics of coal
pillars. Proceedings 3
rd
ISRM Congress, Denver. Pp 1076 1080.

Wevell, E. (1996a). Umgala East Alfred Seam Section 7/8. IRED internal report. August.

Wevell, E. (1996b). Umgala Alfred east Floor Assessment. IRED internal report.
September.

Wevell, E. (2002). Personal communication.

Whitten, D.G.A. and Brooks, J.R.V. (1972). A Dictionary of Geology. Penguin Books. pp
84 & 85.

Wuest, W.J.* (1992). Controlling coal mine floor heave: An overview. USBM Information
Circular 9326.

York, G. (2002). Personal communication with D. Neal.

York, G., Canbulat, I and Jack, B.W. (2000). Coal pillar design procedures. SIMRAC Final
project report COL337. March. pp 34-102.

14 BIBLIOGRAPHY

Carr, F., Martin, E and Gardner, B.H. (1984). How to eliminate roof and floor failures with
yield pillars. Coal Mining. December. pp 62 68

Triplett, T.L. and Drescher, A. (1988). Analysis of pillar punching into soft claystone in
southern Illinois. Proc. 29
th
U.S. symposium on Rock Mechanics. Minneapolis, MN.
Balkema. pp 50 65.

Nemcik, J.A., Indraratna, B. and Aziz, M. (1995). Floor failure induced by lateral stress
ahead of longwall supports. 14
th
Conference on Ground Control in Mining. Morgantown,
WV. pp 194 - 199.

Hanna, A.M. (1980). Foundation failures due to unexpected factors. Annual conference of
Canadian Society for Civil Engineers. Winnipeg. May.

a

APPENDIX A. GEOTECHNICAL TESTS

It is unlikely that further boreholes will be drilled anywhere near the areas being studied in this
project as the reserves have already been mined and no further exploration drilling is
contemplated. There is also limited underground access due to the majority of panels being
sealed off. In addition, two of the failed cases occurred at mines no longer managed by Ingwe
Collieries. Emaswati never re-opened after closure in about 1992, while the shaft where the
collapse at Welgedacht occurred is also closed.

The extreme difficulty in obtaining samples for UCS and other tests from the various sites has
dictated that use is made largely of existing data, which mostly consist of impact spl itting and
slake durability test results.

Impact splitting as well as modified Duncan swell and slake durability tests are regularly
carried out on borehole cores on Ingwe mines. These testing methods were developed with the
following philosophy being applied (Buddery and Oldroyd, 1992):
Tests should relate to the expected mode of failure of the strata.
It should be possible to test even the weakest material.
Large numbers of tests should be able to be carried out simply, quickly and at
low cost.

In achieving these goals, it was considered acceptable to sacrifice a degree of accuracy.

Impact Splitting

Impact splitting (IS) is carried out over a minimum of 2m of the immediate roof. IS consists of
dropping a chisel from a pre-determined height, aligned to strike the core parallel to the
bedding planes. The chisel is dropped every 2cm along the core. The core is laid in an angle
iron base on which a tube is mounted containing the chisel. The chisel has a 25mm wide blade
and mass of 1.5kg. The drop height is 100mm for TNW (60mm dia.) core and 64mm for NQ
(48mm dia.).

The rationale behind this test is that the main determinant of roof failure is the number of
discrete layers the strata can break into and the impact splitting test is designed to open up
poorly cemented bedding planes and other zones of weakness. Tensile stresses in beams,
amongst other factors, are inversely proportional to the square of the beam thickness (Buddery
and Oldroyd, 1992).

From the test, a mean fracture spacing is obtained for each roof unit. From this a unit rating ( r
u
)
is determined as follows:
For fs 5 unit rating = 4fs Equation A1
For fs > 5 unit rating = 2fs+10 Equation A2

Where: fs = fracture spacing (cm)

This unit rating is then weighted with respect to its height into the roof to give the weighted
rating (r
w
) for each roof unit.

b
tu hr r r
u w
) 2 ( 2 . = Equation A3

Where: hr = mean unit height above roof (m)
tu = thickness of unit (m)

The sum of the individual weighted ratings for each roof unit is the total roof rating.

Latilla et al (2002) reviewed the work of Buddery and Oldroyd (1992) to include adjustments
for coal roof beams and jointing. They proposed the following roof classification for South
African collieries.

Table A.1. Impact splitting ratings.
Unit rating Roof rating Rock classification
<9 <34 Very poor
10 - 13 35 - 51 Poor
14 - 19 52 - 75 Moderate
20 - 28 76 - 113 Good
29 - 42 114 - 167 Very good
>42 >167 Excellent

Floor rock testing

The standard methods for the determination of Duncan free swell and slake durability, as laid
down by the ISRM, were modified in order to adhere to the testing philosophy mentioned
earlier (Buddery and Oldroyd, 1992). Only the modifications are discussed here. ISRM (1972)
may be referred to for full details of these test methods. In both methods, the fluid medi um
used is tap water.

Duncan swell

For coal measures strata, it is considered necessary to measure only the swelling strain
perpendicular to the laminations, as this is much higher than that in any other direction. To
reduce cost and time, the samples are not prepared but rather chosen with roughly parallel ends.

Buddery and Oldroyd (1992) state that 90% or more of swell takes place in the first 30 minutes
so the swelling strain after 30 minutes is determined.

The swelling strain (S
30
) is calculated as follows:

% 100 .
30
30
L
d
S = Equation A4

c
Where: d
30
= swelling displacement after 30 minutes
L = initial length of the sample

On removal from the water, after 30 minutes, the sample is assigned a condition rating between
one (undisturbed) and six (totally degraded). The swell index is determined by multiplying the
swelling strain by the condition rating.

Slake durability test

This test assesses the propensity of a sample to disintegrate when subjected to two standard
cycles of drying, wetting and tumbling. The lump size for this test has been reduced from the
40 - 60g called for by the ISRM to 20 30g. This was necessary as for the core sizes normally
used, 40 - 60g lumps would only be obtained from strata that are more competent. The dr ying
periods have also been shortened from 2 6hrs to 1 - 2hrs to speed up the procedure and
because the lumps are smaller.

The slake durability index second cycle (I
d2
) is calculated as follows:

% 100 .
2
A
C
I
d
= Equation A5

Where: A = dry mass prior to testing (g)
C = dry mass after two slaking cycles (g)

Treatment of results

Conventionally, a high swell index implies poor rock while, conversely, a high slake durability
index implies good rock. To avoid confusion it was decided to present the slake durability
index as 100 - I
d2.

Latilla et al (2002) re-evaluated the work of Buddery and Oldroyd (1992) and propose the
following ranges:

Table A.2. Duncan swell and slake durability ratings.
Description Swell index Slake durability index
Floor Roof
Continuous miner /
Roadheader
Conventional drill
and blast
Good <1 <5 <10 <2
Moderate 1 3 5 15 10 20 2 - 4
Poor 3.1 15 15.1 25 20.1 - 30 4.1 - 10
Very poor >15 >25 >30 >10

d
Moisture content tests

Natural moisture is used for determining the mine floor bearing capacity using Vesics
modified equation (Speck, 1981). This test is not carried out routinely on South African
collieries but samples were collected from some accessible areas and tested. As some of the
samples were collected from dewatered areas where the floor had been mined into, a loss of
moisture content was expected.

The approach taken was as follows:
o Two batches of were chipped from large floor rock specimens for each site. These were
hand cleaned to remove any loose pieces and dust. Of the two batches, one was for
drying out and the other for moisture absorption.
o Each batch consisted of five lumps of 60 to 90g with total mass of between 300 and
450g
o Existing moisture content (as sampled) and total moist ure absorbed was determined of
reach batch.
o The existing moisture content was determined after drying each batch of samples at
105 for 48hrs.
o Moisture absorption was determined after immersing each batch of samples in water for
48hrs, the samples were then dried at 105 for ten minutes to remove surface moisture
only.
o By adding the two results, the total moisture content (saturated) for the rock was
determined.

e
APPENDIX B. FAILED CASES
CASE Emaswati Colliery, main haulage
south
South
stooping
MEN1
Area Main south haulage Matla 1, 5 seam Umgala east, Alfred seam stooping.
Sect 7&8
Mngeni shaft, panel MEN1
Year mined 1989 1981 Pre 1971 1999
Year failed 1990 to 1991 Jan 1995 to Jun 1996 June 1996 May 2000 onwards
Time span of
failure
1
st
report Jan 1990, failure Jun
1991 18 months. Final collapse
very rapid (14hrs) from reports of
accelerated pillar deterioration to
collapse.
Unknown for initial failure of
17 pillars. The extension (47
pillars) occurred within a
further 17 months.
Rapid for 10 days then slower for
a further 10 days (equipment
recovery stage). Rate increased
again after equipment recovery,
possibly due to burning pillars.
>12 months and continuing
Mining geometry
Cover depth (m) 65 32.5 165 126.5 (102 to 151)
Centre distance
(m)
16 11 24.4 22
Mining height 2.5 (1.9 to 3.1) 2.4 3.8 2.77 (2.74 to 2.8)
Bord width (m) 6.3 5.9 6.3 at floor
8.0 at roof (see remarks)
6.26 (6.11 to 6.41)
Pillar width (m) 9.7 5.1 18.1 at floor
16.4 at roof
15.74 (15.59 to 15.89)
Pillar width to
height ratio
4.1 (3.1 to5.1) 2.1 (1.7 after slabbing) 4.55 (4.3 to 4.8) 5.6
Average pillar
stress (MPa)
4.4 3.8 7.5 6.2 (4.9 to 7.5)
Pillar joint spacing
(m)
1 2 Average 1.43 0.52 to 0.68m (average 0.6).
Dip 45 to 60. Strike random.
Pillar safety factor 2.3 Salamon (2.0 to 2.6) 2.26 Salamon
(1.3 after slabbing)
1.18 to 1.51 Salamon
(1.34 average)
2.23 Squat (1.74 to 2.74)
2.20 Salamon (1.72 to 2.70)
1.20 1.89 (Esterhuizen)
Panel width (m) 214 116 No barriers, width in excess of
200m
183 solid pillars
271 including
chequerboarded area
Ratio of panel 3.29 3.56 1.21 2.14
f
south
South
stooping
MEN1
width to cover
depth
Seam dip () 1 to 2 Flat Slight (1) 4.4 to 6
Roof
Lithology 13m massive sandstone 0.4m carbonaceous shale
overlain by 0.5m sandy shale
and 0.8m massive sandstone
2m massive sandstone overlain by
strong laminated sandstone
11.15 to 11.55m massive
sandstone
Facies number 11 24 / 3 / 9 11 11
Coal mine roof
rating (CMRR)
64.7 60.3
Impact splitting
rating
98 to 191 (average 129)
Good to excellent
71.9 to 112.6 (average 98)
Moderate to good
No test results available 169 to 269 (average 214)
Excellent
RQD 75.5 to 99.3 (average 88)
Floor
Lithology 0.5m sandstone and shale,
underlain by 1.5m sandstone and
0.5m coal
0.5m sandy shale, very soft
and prone to weathering
1.2m interlaminated sandstone and
shale
Solid sandstone in one
borehole but 0.22m shale and
sandy shale underlain by
0.71m sandstone and 0.22m
coal in another
Facies number 4 / 8 / 22 2 3 3 / 8 / 22
Soft floor thickness
(m)
0.5 0.49 0.6 0 to 0.22 but suspected to
extend down 1.15m into lower
coal band
Width to height
ratio of soft floor
19.4 10.2 27.3 to 30.2
(average 28.8)
71.5 (0.22m soft floor) to
13.68 (1.15m soft floor)
(average 42.6)
Duncan free swell
index of weak floor
material
1.1 to 3.8
Slake durability
index (and rating)
for weak floor
material
23.4 (Poor) 38.7 to 84.4 (average 62.7)
(Very poor)
11.4 (Moderate) 14.8 (Moderate)
Average weighted
Slake Durability
index for 0.5m of
23.1 - Poor 58.7 Very poor 7.5 - Good
g
south
South
stooping
MEN1
floor
Moisture content
(%)
9.92 to 10.38 1.36 to 3.03 2.27
Field strength
classification
Weak Weak to very weak Weak Weak
Floor rating (FR) 4.5 20.5
Floor heave type Type I Type II New Type I
Floor heave
classification
Open topped Closed ridge Closed ridge Open topped
Maximum floor
heave (m)
0.1
Floor jointing 3 to 5m apart
0.2 1.5 to 2.0 0.4
Convergence
Rate per day (mm) Less than 0.14 during 1990. From
May 1991 it increased to between
0.24 and 0.55
Not measured Not measured Maximum 7.5, generally 1 to 2
Total convergence
(m)
Collapsed Collapsed 1 to 1.5 0.095 (at re-surveyed peg)
General
Geological
structure
Area completely enclosed by two
dolerite sills and a dyke. Seam
highly jointed and weakened by
dykes.
A dolerite dyke intersected the
area, which may have caused
further weakening of the floor.
The floor was weak and prone
to weathering.
The top portion of the seam was
very friable; this led to severe
spalling in this zone with eventual
bord widths of 8m at roof height.
Dolerite sill to south of area.
4m upthrown fault to north of
pillar deterioration area and
another small (0.3m) fault
running though the area.
Surrounding
workings
Virgin ground. Virgin ground with underlying
4 seam workings.
Surrounded by bord and pillar
workings. Goaf to north west.
Smaller 20m centres to south. No
significant barrier pillars.
Gus seam workings 2m below
Alfred seam over a portion of
failed area. Pillars superimposed.
Four lines of chequerboarded
pillars on the eastern side of
the panel and another large
chequerboarded area 40m away
to the west.
Remarks The roof above the floor appeared
to have subsided as a unit. The
entire boundary of the fall had
significant failure of slickensided
joints in the floor.
It was concluded that the failure
Seam not currently being
mined.
Severe roof deterioration
prevented further inspection of
the collapsed area.
Additional pillar loading due to
stooping caused floor failure and
associated slow pillar failure.
CM and shuttle car pushed right up
against the roof.
Numerous roof and floor
cracks as well as floor heave
observed mostly on the eastern
side of the panel (to the right
of the belt road).
Earth tremor with epicentre
h
south
South
stooping
MEN1
was in fact a widespread floor
failure, which led to pillar
settlement and resultant roof
collapse.
40km off the Zululand coast
opposite St Lucia reported just
after the period of quickest
movement.
Remedial actions
taken
Sand filling of pillar clusters
planned but not completed. Pillars
wrapped with weld mesh.
Convergence monitoring.
Double row of pillars rubble
filled for shaft protection.
Nil Convergence monitoring,
electronic convergence station
with alarm. Pillar meshing and
strapping. Rubble filling.
Travelling road roof supported.
MAP3D modelling
results
Tensile stress zones in bords Tensile stress zones in bords
and intersections
Tensile stress zones in bords Tensile stress zones in old
bords only after
chequerboarding.

i
APPENDIX C. STABLE CASES
CASE Matla 1, 5 seam, Panel R14
West
North
ZAC, Mngeni, panel MN ZAC, Maye, poor roof case
Area Panel R14West Panel M12North Panel MN Panel MWN1
Year mined 1981 1983/4 1995/6 1997
Mining geometry
Cover depth (m) 34.2 to 36.4
(average 35.3)
57 to 63.7 (average 59.3) 100.1 98 to 111
(average 104.5)
Centre distance
(m)
12 15 22 17
Mining height 2.2 to 2.4 1.7 to 2.0 (average 1.8) 2.57 1.47 to 2.04
Bord width (m) 5.9 to 6.1 6.2 to 6.5 (average 6.4) 5.98 5.67 to 5.89
Pillar width (m) 5.9 to 6.1 Average 8.6 16.02 10.44 after
chequerboarding (c/b)
11.11 to 11.33
Pillar width to
height ratio
2.5 to 2.8
(average 2.65)
4.4 to 5.0 (average 4.7) 6.23 4.06 after c/b 5.44 to 7.07
(average 6.25)
Average pillar
stress (MPa)
3.3 to 3.8
(average 3.55)
(average 6.0)
Pillar joint
spacing (m)
2 No significant jointing 0.5 to 1.0 1
(Coal UCS 25.8Mpa)
Pillar safety factor 2.4 to 3.0
(average 2.7)
(average 2.8)
Panel width (m) 130 97 205 74 to 108
Ratio of panel
width to cover
depth
3.68 1.63 2.05 1.03
Seam dip () Flat Flat 5 1
Roof
Lithology 0.24m micaceous sandy shale,
gritty at base, overlain by 0.15m
micaceous and glauconitic shaly
sandstone and 1m massive
sandstone.
0.13 to 0.30m laminated sandy
shale, mined down to expose
1m massive medium grained
sandstone, competent, no
support required.
10.49m massive sandstone 1.0 to 1.2m sandy shale with thin
coal band at top overlain by >8m
massive coarse grained sandstone
Facies number 3 3 / 9 11 2 or 3 / 11
Coal mine roof 72
j
West
North
rating (CMRR)
Impact splitting
rating
108.7 (good) 108.7 to 116.05 (good)
(average 112.4)
Overall 88 (moderate)
Soft roof 4 to 15 (poor to very
poor)
Upper roof 50 (very good)
Slake durability
index (and rating)
4.6 (Good) 6.7 to 31.6 (average 16.4) (Good to
Very poor)
Unconfined
compressive
strength (MPa)
51.3 (from point load tests) Sandstone 79.2 to 110,8
Shale 26.5 to 29.9
Floor
Lithology 0.16m sandy shale underlain by
0.26m interlaminated shale and
sandstone
0.3m shaly sandstone 0.62m laminated sandy shale
and shaly sandstone underlain
by 0.19m coal and 0.51m shale.
Total potential soft floor 1.32m
1.25m shaly sandstone and
sandstone
Facies number 2 / 3 4 4 4
Soft floor
thickness (m)
0.42m 0.3 1.32 1.0 to 1.2 (roof)
Width to height
ratio of soft floor
14.0 to 14.5 28.7 12.13 7.90 after c/b 9.3 to 11.3 (roof)
(average 10.3)
Duncan free swell
rating
3.8 (poor)
Slake durability 42.1 to 78.0
(average 60.1)
12.0 to 81.6 (average 35.9) 12.6 to 14.8
(average 13.7)
6.2 (good)
Average weighted
Slake Durability
index for 0.5m
floor
47.7 Very poor 34.5 Very poor 9.2 - Good
Moisture content
(%)
2.09 to 4.03
(average 3.1)
1.53
Floor rating (FR) 17.5
Field strength
classification
Weak Weak Weak Weak (roof)
Floor heave type Type I Type I Type 1
Floor heave
classification
Closed ridge Mostly closed ridge Open topped
k
West
North
Maximum floor
heave (m)
0.05 to 0.1 0.1 0.3
General
Geological
structure
Dyke near end of panel No dolerite activity nearby. Dyke and 2.5m fault to the west
as well as 4m fault to the east.
A few small 0.3m to 0.5m
faults run through the panel.
Two faults and 3m wide dyke in
area. Dolerite sill ahead of faces
Surrounding
workings
Virgin ground on 5 seam with
underlying 4 seam bord and
pillar workings.
Old longwall goaf area to east
of northern of panel,
otherwise solid all around on 5
seam. Underlying 4 and 2 seam
bord and pillar workings.
At time of first floor damage
noticed, this was a panel
developing in virgin ground.
Pillar damage area (panel
MEN1 lies 40m to the east of
this area.
Virgin ground. Subsequent panels
mined to the north west.
MAP3D modelling
results
Tensile stress zones in bords and
intersections
No significant tensile stress
zones
Tensile stress zones in
bords and under areas where
pillars had been extracted.

Latilla Soft Floor MSC

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Latilla Soft Floor MSC

Uploaded by

Copyright:

Available Formats

I N F L U E N C E O F W E A K F L O O R O N

You might also like