A Mathematical Programming Framework for

Stope and Production Scheduling Optimization: A

Stochastic Integrated Approach
Emmanuel J. A. Appianing
(Ph.D. Natural Resources Engineering)

Ph.D. Thesis Defense

Laurentian University,
JanuarySudbury,
2018 Ontario, Canada
December 06, 2024
 2
Presentation Outline
• Introduction
• Research Problem
• Summary of Literature Review
• Scope of work
• Research Objectives
• Research Methodology
• SMILP Model Framework and Formulation
• Implementation of the Formulation – Case Study
• Results and Discussions
• Scientific Contributions
• Conclusions
• Future Research Work
 3
Introduction
1. Mine planning can be conducted using two distinct
approaches:

i. Deterministic; and
ii. Stochastic

2. The deterministic approach relies on historical data

from past mining activities, if future trends will mirror
previous patterns.

3. In contrast, the stochastic approach acknowledges the

uncertainty inherent in mining processes by
incorporating variability into the planning model, based
on historical data.
 4
Introduction (Cont’d)

4. Unlike deterministic methods, which overlook

fluctuations, stochastic mine planning recognizes
the inherent unpredictability of mining activities,
making it a more comprehensive and realistic
approach.

5. Current stochastic approaches may result in local

optimal solutions. However, an integrated model
that generates a global optimal solution may add
significant value to stochastic mine planning.
 5
Impact of grade uncertainty in mine planning
❑ For instance;
Cu price (pr) = $ 4030/tonne; Recovery (r) = 90%,
Processing cost (pc) = $6/tonne; Mining cost (mc) =$2/tonne;
Block Tonnage (B) = 18650 tonnes
Marginal COG = 0.23%Cu

Economic Block Value evaluation considering grade uncertainty

 6
Impact of grade uncertainty in mine planning (Cont’d)

$26,673 -$37,300 -$37,300 $87,552 $35,467 -$37,300 -$37,300

4 −$37,300 +149692
Expected Economic Block Value = = $70.29
7
 7
Problem Definition

a) Conventional geostatistical techniques, like ordinary

kriging, can produce biased and overly smoothed
estimates, leading to unrealistic production schedules.

b) In contrast, simulation-based methods offer a more

effective approach by generating multiple equiprobable
models of the orebody, allowing for better assessment
and management of grade uncertainty risks in mine
planning.
Problem Definition (Cont’d) 8

c) The study presents a Stochastic Mixed Integer Linear

Programming (SMILP) model, which is an advanced
version of an initial Mixed Integer Linear Program (MILP)
model.

d) This enhanced model encompasses a variety of mining

operations, including primary (decline) development,
ventilation development, operational development, ore
pass development, stope design and extraction, as well
as backfilling and stockpile management, all aimed at
maximizing the net present value (NPV).
 9
Problem Definition (Cont’d)
Primary development constraint

Primary development schedule

Fig. 1: Schematic representation of the integrated underground mining optimization problem

 10
Problem Definition (Cont’d)
• The research focuses on three main mine planning areas:
a) Development layout planning and scheduling,
• Primary development (Decline option)
• Ventilation development
• Operating development
• Ore pass development

b) Stope design planning and simulation,

c) Production scheduling,
• Mining/extraction
• Processing
• Backfilling
• Stockpiling
 11
Summary of Literature Review

a) Majority of studies so far illustrate the benefits of

incorporating grade uncertainty in open-pit production
scheduling optimization.

b) Notwithstanding the complexities associated with the

development and implementation of underground
mining methods, some efforts have been made in
incorporating grade uncertainty in underground
mining production scheduling optimization.

c) However, there exist room for improvement to the

current underground stochastic mine planning
optimization approaches.
 12
Summary of Literature Review (Cont’d)
Summary of research gaps in this area include:
I. Stope design layout without production schedule and
development schedules approach,
II. Production schedule without stope design layout and
development consideration approach,
III. Production scheduling with only development schedule
(operational) consideration approach,
IV. Stope design layout without stochastic consideration
approach, and
V. Production scheduling without stochastic consideration
approach.
 13
Summary of Literature Review (Cont’d)

1. Generally, stockpiles can be considered as

buffers of material for future use for controlling
capacity and blending requirements.

2. Also, the presence of a stockpile allows

optimization to extract extra ore at early stages of
the mine-life and the extra ore reduce the
chances of short falls at later years.
Summary of Literature Review (Cont’d) 14

3. The use of stockpiling in the optimization process has

been extensively used in open pit mining but less
considered in underground mining

4. Develop a risk-based optimization framework using

stochastic mixed integer linear programming (SMILP)
that effectively integrates grade uncertainty into the
optimization of long-term production scheduling, by
developing a stochastically simulated equally probable
representations of a deposit for underground production
scheduling optimization considering: Primary, ventilation,
operational, and ore pass development schedules, stope
design layout, backfilling, and stockpile management
 15
Research objectives

1) Propose and develop a mathematical programming

framework based on MILP for: production scheduling
optimization taking into consideration primary (decline),
ventilation, operational, and ore pass developments,
backfilling, and stockpile management.

2) Extend the MILP optimization framework to incorporate

grade uncertainty as stochastic variable for an integrated
stochastic stope and production scheduling optimization.
This mathematical programming implementation is
referred to in this research as the SMILP framework.
 16
Research objectives (Cont’d)

3) Implement techniques and methodologies to

maximize the NPV of the SMILP model subject to
economic, operational and technical constraints.

4) Develop computer codes/tools for the formulated

MILP and SMILP frameworks for risk-based mine
planning.
 17
Scope of Work
I. Generating a strategic production schedule for designed stopes based
on a kriged block model using a MILP model.

II. Incorporating primary development, ventilation development,

operational development, ore pass development, backfilling, and
stockpile management.

III. Extending the MILP framework to a SMILP model to include grade

uncertainty using equally probable orebody realizations from
Sequential Gaussian Simulation (SGS).

IV. Evaluating the impact of the SMILP formulation for risk-based

production scheduling optimization.
 18
Research Methodology
• Fig. 2: Schematic
illustration of research
methodology and
workflow
 19
SMILP Framework and Formulation
• The SMILP framework is developed based on:
a) Development planning, stope design layout and production
scheduling optimization approach

• Primary development schedule (Decline)

• Ventilation development schedule Development
consideration
• Operational development schedule
• Ore pass development schedule

• Mining/extraction schedule
• Processing schedule Production
scheduling
• Backfilling schedule
consideration
• Stockpiling management schedule
 20
Assumptions
• To store ore that exceeds the current
processing plant capacity and is
reclaimed in the future.

• No physical or chemical modifications

should ore stay longer in surface
stockpile.

• Primary point from where all material

mined (ore and waste) exit the mine.

• Managed on a first-in first-out basis with

no mixing.

• Ore pass material to plant or stockpiled

at the surface ore stockpile

Fig. 3: Material flow from mine to processing plant

 21
Model Layout

Fig. 4: A schematic representation of stopes and development layout

 22
SMILP Framework and Formulation (Cont’d)
• Objective function (Maximize NPV): (a)

 (( Rkt , p , s  xkt ,m ) + ( Rkt , so , s  ukt , so )) − (Qkt  ykt ) − (Ckt  d cdt ) 

 
 −( H k  d vd ) − ( Ek  d od ) − ( Z k  d ops ) − ( Fk  f k ) 
S T K t t t t t t t t
1
Max   
S s =1 t =1 k =1  (1 + i )t 
 
 
Interest Rate
 23
SMILP Framework and Formulation (Cont’d)
• Objective function (Min. cost of uncertainty associated
with targeted ore grade and ore tonnage): (b)
Grade tonnage
Ore deviation variable deviation variable

1 S T  ( pnCTo ,+  odevs ,+ ) + ( pntCTo ,−  odevs ,− )  1 S T  ( pnCTg ,+  gdevs ,+ ) + ( pntCTg ,−  gdevs ,− ) 

t t t t t t t t

Min    + Min   
S s =1 t =1  (1 + GDR)t  S s =1 t =1  (1 + GDR)t 

Geological discounted rate

Penalty cost (GDR)
 24
SMILP Framework and Formulation (Cont’d)
• Objective function (Min. cost of uncertainty associated
with stockpile ore grade): (C)
Grade tonnage deviation
variable

1 S 
T ( SPpnCT t
 jgdev t
) + ( SPpntCT t
 jgdev t
s ,− )

Min  
g ,+ s ,+ g ,−

S s =1 t =1  (1 + GDR)t 

Geological discounted rate

(GDR)
Penalty cost
 25
SMILP Framework and Formulation (Cont’d)

Min cost of uncertainty

• Objective function: (a) + (b) + (c) associated with targeted ore
  ( Rkt , p , s  xkt ,m ) + ( Rkt , so, s  ukt , so ) − (Qkt  ykt ) − (Ckt  d cdt ) 
tonnage

and ore grade for
Max NPV    direct processing
  −( H k  d vd ) − ( Ek  d od ) − ( Z k  d ops ) − ( Fk  f k )
t t t t t t t t
 
  
 (1 + i ) t
 
  
   
   ( pntCTot,+  odevst ,+ ) + ( pntCTot,−  odevst ,− )  
1 S T K 
Max    
S s =1 t =1 k =1   (1 + GDR) t
 
  
 1   ( pntCTgt,+  gdevst ,+ ) + ( pntCTgt,−  gdevst ,− )   
− +   
+ t
    
K (1 GDR )
   ( SPpntCT t  jgdevt ) + ( SPpntCT t  jgdev t )   
 +  g ,+ s ,+ g ,− s ,−
 
  

 
 (1 + GDR)t 
 

 

Min cost of uncertainty associated

with targeted stockpile ore grade
 26
SMILP Framework and Formulation (Cont’d)
• Constraints

Fig. 5: Constraint dependencies

 27
Constraints – Mining
K

  k ,s k ,s k  m,ub , t {1,..., T }, s {1,..., S};


k =1
( o + w )  y t
  T t
Mining
K
capacity
  k ,s k ,s k  m,lb , t {1,..., T }, s {1,..., S};

k =1
( o + w )  y t
  T t

t
b -  yst  0, s  Gk ( S ), k  {1,..., K };
t
k
t =1

t Precedence
y
t =1
t
k - bkt  0, k  {1,..., K }; relations

bkt - bkt +1  0, t {1,...T -1}, k {1,..., K};

 28
Constraints – Processing

K
 SO
t 
  k , s k ,m  k ,so ,s k ,so
k =1 
(( o  x t
) +
so =1
( o  u t
) − odevs , +   Tpr ,ub t {1,..., T }, s {1,..., S};

t

Processing
K
 SO
t 
capacity
  k ,s k ,m  k , so, s k , so
k =1 
(( o  x t
) +
so =1
( o  u t
) + odevs,− 

 T t
pr ,lb
t {1,..., T }, s {1,..., S};

 k ,m  1, k {1,..., K };
x t

t =1
Processing variable logic

xkt ,m + ukt ,so  ykt , t {1,..., T }, k {1,..., K }; Mining, Processing and

Stockpile variables relation
 29
Constraints – Grade blending
K
 SO
t 
 ( g k , s − grpr ,ub )  (ok ,s  xk ,m ) +  ( g k , so , s − grpr ,ub )  (ok ,so ,s  uk ,so ) − gdevs , +   0,
k =1 
t t

so =1
t t t


t  {1,..., T }, k  {1,..., K }, s {1,..., S}; Grade capacity
control – Direct
K
 t SO
t  processing
 ( gr − g
 pr ,lb k , s
k =1 
)  ( ok ,s  x t
k ,m ) + 
so =1
( gr t
pr ,lb − g t
k , so , s )  ( ok , so , s  u t
k , so ) + gdevs , −   0,

t {1,..., T }, k {1,..., K }, s {1,..., S}

 (( g
k =1
k ,s − grsit ,ub )  (ok , s  ukt , si ) − jgdevst , + )  0,

t  {1,..., T }, k  {1,..., K }, s {1,..., S}

Grade capacity control -
Stockpile
K

 (( gr
k =1
t
si ,lb − g k , s )  (ok , s  ukt , si ) + jgdevst , − )  0,

t  {1,..., T }, k  {1,..., K }, s  {1,..., S};

 30
Constraints - Stockpile management
K

  k ,s k ,si  si,ub , t  1,..., T  , s  1,..., S ;


k =1
( o  u t
)   T t
Stockpile capacity

t K t −1 K

 (ok ,so,s  u )   (ok , s  uki , si ), t  1,..., T  , s  1,..., S  ;

i Stockpile ore
k , so
i =1 k =1 i =1 k =1
accumulation

t K t −1 K

 (o
i=1 k =1
k , so , s g t
k , so , s u i
k , so )  (o
i =1 k =1
k ,s  g k , s  uki ,si ), t  1,..., T  , s  1,..., S  ;

Stockpile metal content

T accumulation
u
t =1
t
k , si  1, k  {1,..., K };
Stockpile variables logic
T

 k ,so  1, k {1,..., K};

u t

t =1
 31
Constraints - Primary Development

(d )
C
  c ,ub t {1,..., T };
t t
cl d c L
c =1 Primary development

(d )
C
capacity
  c ,lb t {1,..., T };
t t
cl d c L
c =1
t
b −  d st ,l  0
t
c ,l
l  (1,..., L), s  Dc ( L);
t =1
Primary
t

 c,l c,l  0, l {1,..., L}, c {1,..., C};

d t

t =1
- b t development
precedence

bct ,l - bct ,+l1  0, l {1,..., L}, t {1,..., T -1}, c {1,..., C};

t
b -  d ct ,l  0, l  {1,..., L}, c  Dc ( L), v  Dv ( L); Primary and ventilation
t
v ,l
t =1
development precedence
T

 c  1, c {1,..., C};
d t

t =1
Primary development variable control
 32
Constraints - Ventilation Development

(d  dvt )  Ltv ,ub , t  1,..., T  ;

V

vl Ventilation development
v =1
capacity

(d  d vt )  Ltv ,lb , t  1,..., T  ;

V

vl
v =1
t
b -  d st ,l  0, l  1,..., L , s  Dv ( L);
t
v ,l
t =1
t
Ventilation development
d
t =1
t
v ,l - b  0, l  {1,..., L}, v  1,...,V  ;
t
v ,l precedence

bvt ,l - bvt ,+l1  0, l {1,..., L}, t  1,..., T -1 , v  1,..., V ;

t Ventilation and
b -  d  0, l  1,..., L , a  Da ( L), v  Dv ( L);
t
a ,l
t
v ,l Operational development
t =1 precedence
T

 v  1, v {1,...,V };
d t

t =1
Ventilation development variable control
 33
Constraints - Operational Development

(d )
A

a  d t
a  La ,ub , t  {1,..., T };
t

a =1
Operational development
capacity
(d )
A

a  d t
a  La ,lb , t  {1,..., T };
t

a =1
t
b -  d st ,l  0, l {1,..., L}, s  Da ( L);
t
a ,l
t =1

t
Operational
d
t =1
t
a ,l - bat ,l  0, l  {1,..., L}, a {1,..., A}; development
precedence
bat ,l - bat +,l1  0, l {1,..., L}, t {1,..., T -1}, a {1,..., A};
t
Operational development
b -  d at ,l  0, l {1,..., L}, k {1,...Kl }, a  Da ( L);
t
k ,l
t =1 and mining precedence
T

 a  1, a {1,..., A}; Operational development variable control

d t

t =1
 34
Constraints – Ore pass Development

(d  d pt )  Ltp ,ub , t  1,..., T  ;

P

p
p =1 Ore pass development
capacity
(d  d pt )  Ltp ,lb , t  1,..., T  ;
P

p
p =1

t
b -  d st ,l  0, l  1,..., L , s  Dp ( L);
t
p ,l
t =1
t
Ore pass development
 d tp,l - bpt ,l  0, l  (1,..., L), p  1,..., P ;
t =1
precedence

btp ,l - bpt +,l1  0, l  (1,..., L), t  1,..., T -1 , p  1,..., P;

t
b −  d at ,l  0, p  1,..., P , a  Da ( L), l  (1,..., L) ; Ore pass and operational
t
p ,l
t =1 development precedence
T

 p  1, p {1,..., P}; Ore pass development variable control

d t

t =1
 35
Constraints – Backfilling

(d )
K

fv  f k
t
 V f ,ub , t  {1,..., T };
t

k =1
Backfilling capacity

(d  f kt )  V ft ,lb , t  {1,..., T };

K

fv
k =1

bkt + f kt  1, t {1,..., T }, k {1,..., K}; Stope status

t
f -  ( xkt ,m + ukt , si )  0, t  {1,..., T }, k {1,..., K };
k
t
Backfilling precedence
t =1

bkt  bkt +1 + f kt +1 , t {1,..., T -1}, k {1,..., K}; Stope sequence

f
t =1
k
t
 1, k  {1,..., K }; Backfilling variable logic
 36
Constraints – Geotech stability

Fig. 6: Schematic illustration of non-adjacent stope

extraction and backfilling

bkt + btj  1, t {1,..., T }, k {1,..., K}, j  Ak ( J ); Non-adjacent stope

f t + bt  1, t {1,..., T }, k {1,..., K }, j  Ak ( J ); extraction and backfilling
k j
 37
Constraints – Active Level and Stope Duration
t
1 D t t 
  D  a ,l l   0, t {1,..., T }, l {1,..., L}, a  Ll ( D);
t =1 
d - lc

Operational
d =1 development activity
L

 l lc , t {1,..., T };
lc t

l =1
 N t Maximum allowable
activity – operational

t
 1 M

 
t =1  M
x
m =1
t
k ,m - lelt   0, t {1,..., T }, l {1,..., L}, k  Ll ( M );

Extraction activity

 l le , t {1,..., T };
le
l =1
t
 N t Maximum allowable
activity – extraction

 k  N xd
b t

t =1
Maximum allowable
extraction duration
 38
Constraints – Non-Negativity

ykt , xkt ,m , ukt ,si , ukt ,so , f kt , dct , dvt , dat , d p  0

b , b , b , b , b , lc , le  0 and integers
t
k
t
c ,l
t
v ,l
t
a ,l
t
p ,l
t
l
t
l

odevst ,+ , odevst ,− , gdevst ,+ , gdevst ,− , jgdevst ,+ , jgdevst ,−  0

 39
Implementation of the Formulation

• The mathematical equations are formulated in

MATLAB 2023b environment.

• CPLEX Optimization Studio is integrated into

MATLAB 2023b to solve the MILP formulation.

• Simulation of grades carried using SGeMs.

• The SMILP is implemented for a gold deposit.

• Six Scenarios
 40

Case Study – Gold Deposit

(Model Implementation)
 41
Block Model

Fig. 7: Gold deposit showing grade ranges of mineralized block

 42
Table 1: Characteristics of Gold Deposit and Designed Stopes

No. Description Value

1 Total mineralized material (Mt) 2.88

2 Maximum grade value of Au (g/t) 5.34

3 Minimum grade value of Au (g/t) 2.02

4 Average grade value of Au (g/t) 3.05

5 Number of levels 8

6 Total number of designed stopes 120

7 Number of stopes per level 15

8 Stope height (m) 30

9 Stope length (m) 25

10 Stope width (m) 10-12

17
 43
Block Model and Constraints

Fig. 8: Block model with some designed stopes and development

44
45
46
 47
Spatial Description of Data

Fig. 9: Location of stopes with gold grade distribution

 48
Spatial Correlation Analysis Using Variogram

Fig. 11: Histogram for raw data and transformed normal score data of gold deposit
 49
Variogram modelling (Cont’d)

Variogram Sill
Direction Azimuth (°) hmin (m) hmax (m) Nugget
model contribution
Vertical 0.0 Gaussian 0.5 3.5 5.9 0.5
Minor 112.5 Gaussian 0.5 3.5 5.9 0.5
Major 67.5 Gaussian 0.5 3.5 5.9 0.5

ɣ (h)
ɣ (h)

Distance (m) Distance (m)

  −3h 2  
 (h) = C. 1 − exp  2  
ɣ (h)

  a 
Fig. 12: Experimental
Variograms Distance (m)
 47
Table 2: Economic and technical data

No. Parameter Value

1. UG mining cost ($/t) 96.04
2. Processing cost ($/t) 15.21
3. Backfilling cost ($/t) 15.00
4. Operational development cost ($/m) 4,500
5. Primary development cost ($/m) 6,500
6. Ventilation cost ($/m) 3,000
7. Selling price of gold ($/oz) 2,000
8. Discount rate (%) 8.00
9. Mining recovery (%) 93.00
10. Variable Mining dilution* (%) 5.00 / 10.00
13. Maximum mining capacity (Kt/period) 140
14. Minimum mining capacity (Kt/period) 0.00
15. Maximum processing capacity (Kt/period) 110
16. Minimum processing capacity (Kt/period) 0.00
17. Maximum backfilling capacity (m3/period) 50,000
18. Minimum backfilling capacity (m3/period) 0.00
19. Maximum operational development (m/period) 200
20. Minimum operational development (m/period) 0.00
21. Maximum capital development (m/period) 500
22. Minimum capital development (m/period) 0.00
23. Maximum ventilation development (m/period) 500
24. Minimum ventilation development (m/period) 0.00
27. Maximum Stope extraction duration (Periods) 2
28. Maximum active levels 7
 48
Table 3: Stockpile parameters

Parameter Value

Maximum grade value of Au (g/t) 1.40

Minimum grade value of Au (g/t) 0.10

Rehandling cost ($/t) 0.5

 52
Table 4: Summary Results
Ore tonnage Average grade Au metal NPV
Scenarios
(Mt) (g/t) (t) (M$)
Scenario 1 – MILP
2.19 2.98 6.53 7,344.00
without Stockpile
Scenario 2 – MILP
2.52 2.87 7.15 7,801.20
with Stockpiling

Scenario 3 – SMILP
2.33 3.05 7.19 7,601.30
without Stockpiling

Scenario 4 – SMILP
2.54 3.02 7.67 8,077.86
with Stockpiling

Scenario 5 – E-Type
2.29 2.77 6.39 7,036.10
without Stockpiling

Scenario 6 – E-Type
2.51 2.93 7.35 7,360.52
with Stockpiling

• The SMILP model exhibited superior performance, reporting approximately 4%

increase in NPV compared to the MILP model.
 50

Case 2
MILP Model with Stockpiling
 51
Results and Discussions – Case 2

Fig. 11: Primary, ventilation and ore pass development schedule

Results and Discussions – Case 2 (Cont’d) 52

Fig. 12: Operational development schedule

 53
Results and Discussions – Case 2 (Cont’d)

Fig. 13: Mining schedule with estimated waste dilution

 54
Results and Discussions – Case 2 (Cont’d)

Fig. 14: Stockpiled material schedule and grade curve

 55
Results and Discussions – Case 2 (Cont’d)

Fig. 15: Direct processed material, material from stockpile and grade blended
curve schedule
 56
Results and Discussions – Case 2 (Cont’d)

Fig. 16: Stockpiled inventory profile

 57
Results and Discussions – Case 2 (Cont’d)

S5

S91

Fig. 17: Yearly operational development, stope extraction and backfilling

schedule per level
 58
Results and Discussions Case 2(Cont’d)

The MILP model successfully planned and scheduled

development, mining, processing, stockpiling and backfilling
of void stopes for a gold deposit generating an NPV of
$7,801.2M

The model completed development, mining, processing,

stockpiling and backfilling in 25 years of mine life.

A total of 2.52 Mt of material was processed.

 59

Case 4
SMILP Model with Stockpiling Strategy
• Stockpiling Strategy
 60
Implementation of the Formulation
Table 5: Risk parameters for SMILP model

Parameter Cases 2
Number of realizations (#) 50

Geological risk discount rate (%) 20

Cost of shortage in ore production ($/tonne) 1500

Cost of excess in ore production ($/tonne) 1500

Cost of shortage in grade ($/g) 500

Cost of excess in grade ($/g) 500

Cost of shortage in stockpile grade ($/g) 500

Cost of excess in stockpile grade ($/g) 500

 61
Results and Discussions – Case 4 (Cont’d)

Fig. 18: Primary, ventilation and ore pass development

schedule
 62
Results and Discussions – Case 4 (Cont’d)

Fig. 19: Operational development schedule

 63
Results and Discussions – Case 4 (Cont’d)

Fig. 20: Mining schedule with estimated waste dilution

 64
Results and Discussions – Case 4 (Cont’d)

Fig. 21: Stockpiled material schedule and grade curve

 65
Results and Discussions – Case 4 (Cont’d)

Fig. 22: Direct processed material, material from stockpile and grade blended
curve schedule
 66
Results and Discussions – Case 4 (Cont’d)

Fig. 23: Stockpiled inventory profile

 67
Results and Discussions – Case 4 (Cont’d)

Fig. 24: Yearly operational development, stope extraction and backfilling

schedule per level
 68
Results and Discussions - Case 4 (Cont’d)

The SMILP model successfully planned and scheduled

development, mining, processing, stockpiling and backfilling
of void stopes for a gold deposit generating an NPV of
$8,77.86M

The model completed development, mining, processing,

stockpiling and backfilling in 25 years of mine life.

A total of 2.54 Mt of material was processed.

 69
Table 6: Summary of Results

Average
Ore tonnage Au metal NPV
Cases grade
(Mt) (t) (M$)
(g/t)
1 - (MILP) 2.52 2.87 7.23 7,801.20
2 - (SMILP) 2.54 3.02 7.67 8,077.86

• The SMILP model exhibited superior performance, reporting

approximately 4% increase in NPV compared to the MILP
model.
 73
Comparative risk analysis – Risk Profile

Fig. 25a: Average grade

Fig. 26c: Processed ore tonnage

• SMILP model closely follows the P50 risk

profile
• Strong indication of no over estimation or
under estimation of the simulated NPV cash
flows, ore tonnages and ore grades
Fig. 25b: Cash flow
Comparative schedule results based on random realizations 74

Table 7: SMILP schedule comparison to MILP and E-type schedules based on random
realizations
Au
Ore Tonnage Average Grade NPV
Realization (R#) Schedule Type Metal
(Mt) (g/t) (M$)
(t)
MILP 2.58 3.34 8.11 7302
E-Type 2.46 3.25 7.87 7054
R1 SMILP 2.57 3.27 8.41 7398
E-Type comparison to SMILP (%) -4.28 -0.61 -6.42 -4.65
MILP comparison to SMILP (%) 0.39 2.14 -3.57 -1.29
MILP 2.41 3.31 8.12 7285
E-Type 2.40 3.24 7.88 6991
R30 SMILP 2.58 3.28 8.47 7329
E-Type comparison to SMILP (%) -7.06 -1.22 -6.97 -4.61
MILP comparison to SMILP (%) -6.67 0.91 -4.31 -0.59
MILP 2.52 3.25 7.87 7081
E-Type 2.49 3.16 7.63 6836
R40 SMILP 2.56 3.20 8.20 7223
E-Type Comparison to SMILP (%) -2.83 -1.25 -6.95 -5.35
MILP Comparison to SMILP (%) -1.66 1.56 -4.02 -1.97
Table 7: Sensitivity Analysis 1 – Au Price 75

Au Price NPV Ore Tonnage Runtime

Run #
($/oz) ($) (Mt) (hrs)
1 2,450 8,635.53 2.98 5.9
2 2,150 8,354.33 2.73 5.1

3 (Base) 2,000 8,077.86 2.54 4.2

4 1,850 7,403.13 2.46 3.3

5 1,550 6,142.92 2.39 2.8
Table 8: Sensitivity Analysis 2 – Mining Cost 76

Ore
Mining NPV Runtime
Run # Tonnage
Cost ($/t) ($) (hrs)
(Mt)

1 126.04 7,070.89 2.38 2.2

2 111.04 7,346.91 2.43 3.8

3 (Base) 96.04 8,077.86 2.52 4.2

4 81.04 8,102.32 2.61 5.3

5 66.04 8,297.11 2.85 5.1
 77
Sensitivity Analysis 3 – Number of Realizations

8,280
8258.18
8,260

8,240 8239.18
8,220

8,200 8194.68
NPV ($)

8,180

8,160
8139.32
8,140 8140.21
8,120

8,100 R² = 0.8023
8077.86
8,080

8,060
0 10 20 30 40 50 60
Number of Realizations

Fig. 23: Number of realizations and corresponding NPVs for Scenario 4

 78
Sensitivity Analysis 4 – GDR Parameter
8,140

8,130 8118.4
R² = 0.8722 8115.00
8,120

8,110

8,100 8091.35
NPV ($)

8,090
8077.86
8,080

8,070

8,060

8,050 8048.00

8,040
0 10 20 30 40 50 60 70
GDR (%)

Fig. 24: GDR and corresponding NPVs for Scenario 4

 74
Conclusions
1. The multi-objective SMILP optimization framework
with stockpile management maximizes the NPV of
the mining project in the presence of grade
uncertainty.
2. Minimization of geological risk cost by placing higher
penalties for ore grade and ore tonnage deviations
from production targets.

3. The SMILP optimization framework with SGS block

model realizations performed better than MILP
optimization framework implementation with Kriged
and E-type block models.
 75
Conclusions (Cont’d)

4. The implementation of the SMILP model

contributes significantly to decrease the effect of
grade uncertainty during mining.

5. The proposed SMILP framework was verified in

terms of both feasibility and risk assessment
through systematic workflow towards promoting
robust risk-based strategic underground mine
planning with stockpile management.
 81
Scientific Contributions
1. This is an innovative integrated stochastic mixed integer
linear programming (SMILP) model that incorporates grade
and tonnage uncertainties to simultaneously.

2. The research developed a stochastic programming

framework that further expand the boundaries of uncertainty-
based integrated mine planning and optimization.

3. The SMILP framework provides a systematic workflow to

evaluate a stochastic underground mining production
schedule and its relationship with successful stope
extraction in the presence of grade uncertainty.
 82
Scientific Contributions (Cont’d)

4. The presented SMILP has successfully developed

an innovative alternative for primary development
(decline) in underground mining.

5. The robust SMILP formulation simultaneously

integrates the following in the underground mine
planning framework:
 83
Scientific Contributions (Cont’d)

• Primary development • Stope extraction scheduling;

requirement;
• Ventilation development • Ore processing scheduling;
requirement;
• Operational development • Backfilling strategy; and
requirement
• Ore pass development • Stockpile management
requirement; strategy.
• Mineral extraction
sequencing;
 78
Recommendation for Future Work
a) A (3D) stope boundary optimization constraint where
mineralized stopes are automatically generated from
the equiprobable blocks (realizations).

b) Detailed rock support and reinforcement strategy

could be included to eliminate the assumptions
adopted through the stope status constraints defined
to monitor geotechnical and ground quality
conditions and active levels.

c) Continuous testing of the SMILP framework should

be undertaken with additional case studies.
 85
Publications from the Research
• Journals
– Appianing E. J. A, Ben-Awuah E, Pourrahimian Y. “Life-of-mine optimization for
integrated open stope development and production scheduling using a mixed-integer
linear programming framework”. Mining Technology. 2023;132(2):106-120.
doi:10.1080/25726668.2023.2182285.
– Appianing E. J. A, Mbadozie O, Ben-Awuah E, & Hooman A. , “Open Stope Mine
Production Scheduling Optimization with Stochastic Mixed Integer Linear Programming:
Considering Grade Uncertainty and Stockpiling Strategy” (In review)

• Conferences
– Emmanuel J. A. Appianing & Eugene Ben-Awuah (2021), “A Mathematical Programming
Framework for Underground Open Stope Production Planning Optimization, CIM Virtual
Convention 2021
– Emmanuel J. A. Appianing, Obinna Mbadozie & Eugene Ben-Awuah (2024), “A Mixed
Integer Linear Programming (MILP) Framework for Open Stope Mining Production
Scheduling Optimization Considering Stockpiling Strategy”, WAIMM Geology, Exploration
and Mining Conference – GEM 2024
– Emmanuel J. A. Appianing, Obinna Mbadozie & Eugene Ben-Awuah (2024), “A
Stochastic Mixed Integer Linear Programming (SMILP) Framework for Open Stope Mining
Production Scheduling Optimization Considering Stockpiling Strategy”, 6th WAIMM Annual
Industry Conference – WAIC 2024
 86
Acknowledgment
• External Examiner: Dr. Angelina Anani
• Supervisor: Dr. Eugene Ben-Awuah.
• Advisory committee members:
• Dr. Ming Cai
• Dr. Pawoumodom Takouda
• Dr. Marie-Helene Fillion
• Graduate coordinator: Dr. Ramesh Subramanian

• Other professors, and research group (MOL) at

Laurentian University

• Industry colleagues
 76
References
• MacNeil, J.A. and R.G. Dimitrakopoulos, A stochastic optimization formulation for the transition from open pit to underground
mining. Optimization and Engineering, 2017. 18(3): p. 793-813.
• Dimitrakopoulos, R. and S. Ramazan, Uncertainty based production scheduling in open pit mining. Society for Mining, Metallurgy,
and Exploration Transactions, 2004. 316(2004): p. 106-112.
• Dimitrakopoulos, R., C. Farrelly, and M. Godoy, Moving forward from traditional optimization: grade uncertainty and risk effects
in open-pit design. Mining Technology, 2002. 111(1): p. 82-88.
• Dimitrakopoulos, R. and S. Ramazan, Stochastic integer programming for optimising long term production schedules of open pit
mines: methods, application and value of stochastic solutions. Mining Technology, 2008. 117(4): p. 155-160.
• Vallejo, M.N. and R. Dimitrakopoulos, Stochastic orebody modelling and stochastic long-term production scheduling at the Kémag
iron ore deposit, Quebec, Canada. International Journal Mining, Reclamation Environment, 2019. 33(7): p. 462-479.
• Magagula, N.S., Stochastic characterisation of a mining production system, in Faculty of Engineering and the Built Environment.
2016, University of the Witwatersrand,: Johannesburg. p. 164.
• Noriega, R., Y. Pourrahimian, and E. Ben-Awuah, Optimisation of life-of-mine production scheduling for block-caving mines under
mineral resource and material mixing uncertainty. International Journal of Mining, Reclamation and Environment, 2022. 36(2): p.
104-124.
• Carpentier, S., M. Gamache, and R. Dimitrakopoulos, Underground long-term mine production scheduling with integrated
geological risk management. Mining Technology, 2016. 125(2): p. 93-102.
• Malaki, S., Block-cave extraction level and production scheduling optimization under grade uncertainty, in Department of Civil
and Environmental Engineering. 2016, University of Alberta: Edmonton. p. 157.
• Huang, S., et al., A stochastic mixed integer programming framework for underground mining production scheduling optimization
considering grade uncertainty. IEEE, 2020. 8(2020): p. 24495-24505.
• Mbadozie, O. and E. Ben-Awuah. Incorporating Grade Uncertainty in Oil Sands Mine Planning and Waste Optimization using
Stochastic Mathematical Programming. in 6th UMaT Biennial International Mining Conference. 2020. Tarkwa, Ghana: University
of Mines and Technology.
• Appianing, E.J.A., E. Ben-Awuah, and Y. Pourrahimian, Life-of-mine optimization for integrated open stope development and
production scheduling using a mixed-integer linear programming framework. Mining Technology, 2023. 132(2): p. 106-120.
Friendly questions are welcome.