DATA ANALYTICS—REAL VALUE BEING DELIVERED FOR THE MINING INDUSTRY
through Commercialization
Damien Duff VP Geoscience and Geotechnical R&D October 3rd, 2016 Mira Geoscience- Earth Modeling Forum
Centre for Excellence in Mining Innovation 129 Projects 28 Institutions
$33.2M Miners
$15.6M SMEs $6.6M Researchers
National & Global Reach Global recognition - Vale, Rio Tinto, Glencore Kidd, KGHM, LKAB, xxx Iamgold, Agnico Eagle, Newcrest, Glencore SINO, Shell, Nexen. © CEMI 2015
Centre for Excellence in Mining Innovation Established 2007, Not for Profit Corporation, Sudbury, Ontario
Vision To be the leading source of innovation for the global mining industry.
Mission To advance the innovation that the mining industry will use to: find more ore, mine ore more effectively and safely, generate more value from mines, have a more benign impact on the environment and, a more beneficial impact on communities.
© CEMI 2015
Centre for Excellence in Mining Innovation 5 Strategic Pillars: Industry-focused Programs FINDMINE: Orebody Discovery Increase Ore Discovery Rate: Geophysics & Geochemistry DEEPMINE: Rock Stress & Heat Management Reduce Technical Risk VALUEMINE: Improve Mine Productivity Lean Mining, ROCE & ROI SUSTAINMINE: Environment & Sustainability Reduce environmental impact, Improve social license BIZMINE: Mining Business Program Support, Business Cases, Commercialization Services
© CEMI 2015
Canada’s Ultra-Deep Mining Network
TO HELP THE MINING INDUSTRY TO ADOPT COMMERCIALLY VIABLE R&D PROJECT RESULTS, AND ACCELERATE THE DEPLOYMENT OF PROVEN INNOVATIVE TECHNIQUES AND TECHNOLOGIES.
27 projects
45 network members
$35
million in private and public funds
Managed by CEMI
Addressing the Challenges of Ultra-Deep Mining
Managed by CEMI
“Data is the New Gold”!
~ 67% mining executive respondents say that big data AND analytics were having a significant and positive impact on their revenues (Absolutdata after Teradata-McKinsey survey)
CONTEXT • Collecting more data than we ever have (in the Tb range at times) • More installed sensors and more self-monitoring capability built into our mobile and fixed-plant equipment fleets • Opportunities for maintenance scheduling and associated cost reduction strategies • Opportunities for more awareness of process flows and their performance
Opportunities for “real-time” situational awareness and enhanced management decision support
Data Management/Analytics projects at CEMI 1. SUMIT― Smart Underground Monitoring and Integrated Technologies for deep mines 2. Energy audits at older mine sites 3. NHEA ― Natural Heat Exchange Areas Optimizing their effectiveness
4. Geothermal capacity of Mine water 5. GeoHazard Assessment in deep mines
SUMIT Smart Underground Monitoring and Integrated technologies for deep mines “Innovation leading to novel technologies and reduced investor risk” CEMI ANNUAL GENERAL MEETING (October 6th, 2016, Sudbury)
Data management in the SUMIT program
Step Change in Data Management Mira Geoscience
Geoscience INTEGRATOR: Initial Design Requirements
Data Management/Analytics at CEMI • SUMIT― Smart Underground Monitoring and Integrated Technologies for deep mines • Energy audits at older mine sites
Energy audit by [Data] analysis of a single mine electricity meter Michelle Y. Levesque Dean L. Millar Mining Innovation, Rehabilitation and Applied Research Corporation (MIRARCO) & Bharti School of Engineering, Laurentian University, Sudbury, ON April 15, 2015
www.mirarco.org
You can’t manage what you can’t measure … but how can we measure component electricity use at mine sites without submeters? FOURIER ANALYSIS: the analysis of a complex waveform expressed as a series of sinusoidal functions, the frequencies of which form a harmonic series.
www.mirarco.org
Signals can be represented in the time or frequency domains Signal 1: Frequency = 10 , Amplitude = 1
Single-Sided Amplitude Spectrum of Signal 1
FFT
1 0.8
1
0.6 0.8
0.2 Mag
Amplitude
0.4
0
0.6
-0.2 -0.4
0.4
-0.6
IFFT
-0.8
0.2
-1 0
0.2
0.6 0.4 Time (seconds)
0.8
1
0 0
20
60 40 Frequency (Hz)
80
100
Even complicated signals can be represented in time or frequency domains Signal 1: Frequency = 10 , Amplitude = 1
Single-Sided Amplitude Spectrum of Signal 1 1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 Time (seconds)
0.8
1
0.5 0 0
20
40 60 Frequency (Hz)
80
100
Even complicated signals can be represented in time or frequency domains Signal 1: Frequency = 10 , Amplitude = 1
Single-Sided Amplitude Spectrum of Signal 1 1 Mag
Amplitude
2 0 -2 0
0.8 0.6 0.4 Time (seconds) Signal 2: Frequency = 2 , Amplitude = 0.7 0.2
0.5 0 0
1
80 60 40 Frequency (Hz) Single-Sided Amplitude Spectrum of Signal 2 20
100
1 Mag
Amplitude
2 0 -2 0
0.2
0.6 0.4 Time (seconds)
0.8
1
0.5 0 0
20
60 40 Frequency (Hz)
80
100
Even complicated signals can be represented in the time or frequency domains Signal 1: Frequency = 10 , Amplitude = 1
Single-Sided Amplitude Spectrum of Signal 1 1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 0.8 Time (seconds) Signal 2: Frequency = 2 , Amplitude = 0.7
0.5 0 0
1
20
40 60 80 Frequency (Hz) Single-Sided Amplitude Spectrum of Signal 2
100
1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 Time (seconds) Combined signal
0.8
0.5 0 0
1
20
40 60 80 100 Frequency (Hz) Single-Sided Amplitude Spectrum of Combined signal
1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 Time (seconds)
0.8
1
0.5 0 0
20
40 60 Frequency (Hz)
80
100
How do we extract the individual signals from the combined signal? Signal 1: Frequency = 10 , Amplitude = 1 Amplitude
2 0 -2 0
0.2
0.4 0.6 0.8 Time (seconds) Signal 2: Frequency = 2 , Amplitude = 0.7
1
Amplitude
2 0 -2 0
0.2
0.4 0.6 Time (seconds) Combined signal
0.8
1
0.2
0.4 0.6 Time (seconds)
0.8
1
Amplitude
2 0 -2 0
Separating the signals can be done in the frequency domain
Combined signal Amplitude
2 0 -2 0
0.2
0.6 0.4 Time (seconds)
0.8
1
Separating the signals can be done in the frequency domain
Combined signal
Single-Sided Amplitude Spectrum of Combined signal 1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 Time (seconds)
0.8
1
0.5 0 0
20
60 40 Frequency (Hz)
80
100
Separating the signals can be done in the frequency domain
Combined signal
Single-Sided Amplitude Spectrum of Combined signal 1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 Time (seconds)
0.8
1
0.5 0 0
20
60 40 Frequency (Hz)
80
100
Separating the signals can be done in the frequency domain
Single-Sided Amplitude Spectrum of Signal 2
Signal 2: Frequency = 2 , Amplitude = 0.7 1 Mag
Amplitude
2 0 -2 0
0.2
0.6 0.4 Time (seconds) Combined signal
0.8
0.5 0 0
1
100 80 60 40 Frequency (Hz) Single-Sided Amplitude Spectrum of Combined signal 20
1 Mag
Amplitude
2 0 -2 0
0.2
0.6 0.4 Time (seconds)
0.8
1
0.5 0 0
20
60 40 Frequency (Hz)
80
100
Separating the signals can be done in the frequency domain
Single-Sided Amplitude Spectrum of Signal 2
Signal 2: Frequency = 2 , Amplitude = 0.7 1 Mag
Amplitude
2 0 -2 0
0.2
0.6 0.4 Time (seconds) Combined signal
0.8
0.5 0 0
1
100 80 60 40 Frequency (Hz) Single-Sided Amplitude Spectrum of Combined signal 20
1 Mag
Amplitude
2 0 -2 0
0.2
0.6 0.4 Time (seconds)
0.8
1
0.5 0 0
20
60 40 Frequency (Hz)
80
100
Separating the signals can be done in the frequency domain Signal 1: Frequency = 10 , Amplitude = 1
Single-Sided Amplitude Spectrum of Signal 1 1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 0.8 Time (seconds) Signal 2: Frequency = 2 , Amplitude = 0.7
0.5 0 0
1
20
40 60 80 Frequency (Hz) Single-Sided Amplitude Spectrum of Signal 2
100
1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 Time (seconds) Combined signal
0.8
0.5 0 0
1
20
40 60 80 100 Frequency (Hz) Single-Sided Amplitude Spectrum of Combined signal
1 Mag
Amplitude
2 0 -2 0
0.2
0.4 0.6 Time (seconds)
0.8
1
0.5 0 0
20
40 60 Frequency (Hz)
80
100
How can this be applied to an underground mine? 1.2 1.1
Normalized electricity
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 Jan 19
Jan 20
Jan 21
Jan 22
Jan 23
Jan 24
Jan 25
Jan 26
A pattern resembling that of a mine hoist emerges if we look closer Normalized energy
1.5 1 0.5 0 -0.5 0
5
10
20
15 Time (minutes)
25
30
35
A pattern resembling that of a mine hoist emerges if we look closer Normalized energy
1.5 1 0.5 0 -0.5 0
5
10
15
20
25
30
35
20
25
30
35
Normalized energy
Time (minutes)
1 0.5 0 -0.5 0
5
10
15 Time (minutes)
A pattern resembling that of a mine hoist emerges if we look closer Normalized energy
1.5 1 0.5 0 -0.5 0
5
10
15
20
25
30
35
20
25
30
35
Normalized energy
Time (minutes)
1 0.5 0 -0.5 0
5
10
15 Time (minutes)
We can get the fingerprint of the mine hoist to know which frequencies to filter Norm. magnitude
Norm. energy
1.5 1 0.5 0 -0.5 0
5
10
15 20 Time (minutes)
25
30
35 Norm. magnitude
Norm. energy
1.5 1 0.5 0 -0.5 0
5
10
15 20 Time (minutes)
25
30
35 Norm. magnitude
Norm. energy
1.5 1 0.5 0 -0.5 0
5
10
15 20 Time (minutes)
25
30
35
1.5 1
actual signal synthesized signal
0.5 0 0
0.2 0.4 0.6 0.8 Norm. frequency (π rads/sample)
1
1.5 1
actual signal synthesized signal
0.5 0 0
0.2 0.4 0.8 0.6 Norm. frequency (π rads/sample)
1
1.5 1
actual signal synthesized signal
0.5 0 0
0.2 0.4 0.6 0.8 Norm. frequency (π rads/sample)
1
We can train a Neural Network to recognize periods when the hoist is operating 1
0.1
0.8
0.05
0.6 0
5
10
25
30
0 0
35
0.1
0.8
0.2 0.4 0.6 0.8 Normalised frequency (π rads/sample)
1
0.2 0.4 0.6 0.8 Normalised frequency (π rads/sample)
1
0.2 0.4 0.6 0.8 Normalised frequency (π rads/sample)
1
0.2 0.4 0.6 0.8 Normalised frequency (π rads/sample)
1
0.05
0.6 0
5
10
1
15 20 Time (minutes)
25
30
35
0.8 0.6 0
5
10
1
15 20 Time (minutes)
25
30
35
0.8
Normalised magnitude
Normalised electricity (kWh)
1
15 20 Time (minutes)
0 0 0.1 0.05 0 0 0.1 0.05
0.6 0
5
10
15 20 Time (minutes)
25
30
35
0 0
Now we know when the hoist was operating and what frequencies to filter
0.5
0 1
0.75
0.5
0.25 Jan 19
Jan 20
Jan 21
Jan 22
Jan 23
Jan 24
Jan 25
Jan 26
Probabilty of hoist running
Normalized total electricity
1
The hoist consumed 4% of the total electricity at the mine 0.75
Normalized hoist electricity
0.5
0.5
0 0.25
0
Jan 19
Jan 20
Jan 21
Jan 22
Jan 23
Jan 24
Jan 25
Jan 26
Probabilty of hoist running
1
We can repeat the process for different equipment with the filtered signal 1.2 1.1
Normalized electricity (kWh)
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 Jan 19
Jan 20
Jan 21
Jan 22
Jan 23
Jan 24
Jan 25
Jan 26
Such as: ventilation, pumps, compressors… until all electricity is characterized 1.2 1.1
Normalized electricity (kWh)
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 Jan 19
Jan 20
Jan 21
Jan 22
Jan 23
HUGE OPPORTUNITY FOR SAVINGS!
Jan 24
Jan 25
Jan 26
Data Management/Analytics at CEMI • SUMIT― Smart Underground Monitoring and Integrated Technologies for deep mines • Energy audits at older mine sites • NHEA ― Natural Heat Exchange Areas • Optimizing their effectiveness
environmental sensor door
model
state
17MWth 3MWth
An analytical model [left] reflects, to some degree, NHEAthe currently flow through displacing the 17 broken rock and the physics of heat transfer. MW refrigeration plant. A computational fluid dynamics model [right] gives Additional higher 3 MW fidelity displaced on the airflow paths and the physics. with decision support Graphics: Courtesy MIRARCO
Managed by CEMI
Data Management/Analytics at CEMI • SUMIT― Smart Underground Monitoring and Integrated Technologies for deep mines • Energy audits at older mine sites • NHEA ― Natural Heat Exchange Areas • Optimizing their effectiveness • Geothermal Capacity of Mine water
Managed by CEMI
The Solution- 2 Geothermal Resources from Open Loop Systems • Few studies on the geothermic potential of waste mine water • Desirable ground temperature • Abundant sources of underground water • Extensive underground excavations: No drilling required • Best Case study (Spring Hill) done after mine closed • Unanswered questions • Performance assessment unclear
Managed by CEMI
Geothermal Mine Water Database
Research Objectives • Practical determination of the performance of an open loop geothermal system installed in a working Canadian mine • Determination of the acceptable geothermal energy resource obtainable from abandoned and working mines in Canada • Identification of the optimum technique for using this geothermal energy in mining infrastructure and/or mineral processing operations • …
Managed by CEMI
Geothermal Mine Water Assessment •
Analyze the design aspects of open loop systems in Canadian Mines Develop computer simulation model to verify theories and computations Water
Mine Operation
Flow
(GPM)
Mine 1
Water Temperature (0C)
Heat
Gain
Annual
CO2
Savings
Reduction
($/yr)
(t/yr)
(kW)
N.G.
E
N.G.
E
1000
16.5
1,160
-
305,360
1427
252
375
16.7
430
-
115,023
537
95
320
22
430
2,805
113,772
513
91
398
16.7
465
125,930
188,092
554
305
950
20.9
1,578
167,173
415,116
1,897
2,692
290
18.9
360
21,000
205,000
437
620
133
16.6
160
8,000
88,000
190
270
515
16.7
600
31,000
342,000
738
1,048
675
14
730
30,000
414,000
913
1,296
560
15.8
640
30,500
362,000
787
1,116
Mine 11
800
11.5
800
24,500
455,540
1,025
1,455
Mine 12
719
13.8
760
288,000
136,000
1,102
209
Mine 13
116
13.6
125
46,220
21,800
177
34
Mine 2 Mine 3
•
Mine 4 Mine 5 Mine 6 Mine 7 Mine 8 Mine 9
Mine 10
Calculations demonstrate that economic savings, and sometimes significant ones, are possible at all operations visited, were the geothermal capacity of the mine water to be captured and utilized for heating purposes. Managed by CEMI
Data Management/Analytics at CEMI • SUMIT― Smart Underground Monitoring and Integrated Technologies for deep mines • Energy audits at older mine sites • NHEA ― Natural Heat Exchange Areas • Optimizing their effectiveness • Geothermal Capacity of Mine water • GeoHazard Assessment in deep mines
Managed by CEMI
Geohazard hazard (x, y, z, t) = f (rock quality, geometry, stress, seismicity, geology, …)
courtesy Xstrata Zinc BMS#12 Mine
Multi-Disciplinary Integration: Mining January 10, 2000 January 16, 2001 November 15, 2001 July 18, 2002 September 1, 2003 November 26, 2003 July 15, 2004 May 27, 2005 June 1, 2005 April 9, 2006 April 21 2006 May 21, 2006 June 20, 2006 November 29, 2006 March 19, 2008
©2014 Mira Geoscience Ltd.
Evaluating the Results Craig Mine Ore Zone, 10 April 2008
“Efficiency of Classification” Curve
Percentage of Rockbursts
100 80 60
80% of historical fault-slip rockbursts occur at a computed hazard index of 90% or greater
40 20 0 100
80
60
40
Hazard Score
20
0
Evaluating the Results Creighton Footwall, 14 March 2009
“Efficiency of Classification” Curve
100 80 60
80% of historical fault-slip rockbursts occur at a computed hazard index of 75% or greater
40 20 0 100
80
60
40
20
0
Predictive Analytics
weights-of-evidence can be considered an optimum assessment of how well historical hazard statistically correlates to observation: a quantitative characterization of “typical” behaviour predictive analytics characterizes both “typical” behaviour and “outlier” behaviour, finding specific combinations of observations that correlate to historical hazard
MODCC OVERVIEW
MINING OBSERVATORY DATA CONTROL CENTRE
Dec 12th 2014
The space—
Revolution Mining
Mira Geoscience
Tunik Inc.
CONCLUSIONS
Value/Impact: • Decision support (real-time or not) • Cost savings • Design Optimization MODCC: • growth of data analytics for mining fostered by enhancing mining capability and capacity Managed by CEMI
CONCLUSIONS
“Data is gold but data analytics (and MODCC) is platinum”! Thanks! Email:
[email protected]
Managed by CEMI