Machine Learning in Mining
0 likes256 views
The document discusses the integration of machine learning within the mining industry's value chain, highlighting safety incidents and the importance of data management. It emphasizes the advancements in predictive maintenance and automation, aiming to enhance productivity and reduce operational inefficiencies. Additionally, it outlines the industry's commitment to sustainability and reducing greenhouse gas emissions in line with long-term environmental targets.
![Fatigued, “Nappy Valley”, Maintainers
270 people,29 severe incidents (2009 – 2011)
Within the feature Diff. to the popn
Employees 263 people < 5% of pop
Tenure with xxxx 4.5 years (mean) 35%
Tenure in current
position
2.3 years (mean) 42%
Age 37 (mean) -5%
Maintainers 234 people 334%
Task RelevantTraining 180 people (approx) -
Carers and Parental
Leave
172 people 128% - 274%
Fatigued 40% of shifts -
Nappy Valley 223 people 128%
Voluntary additional
shifts
172 people 334%
[Withheld] [Withheld] [Withheld]
People involved in
severe incidents
27 people 92%
Low severe incidents
reported
183 people 79%
Hazard reporters 24 people -72%
Safety interaction
reporters
61 people -48%
Incidents involving
working at heights
5 855%
Billions of data elements processed
Some truths would never find a home
Protected in PDF
distribution](https://image.slidesharecdn.com/coertduplessis-maxmine-machinelearninginminingpresentationvupload-211110114045/85/Machine-Learning-in-Mining-4-320.jpg)
![Scale can varies – one [general] algorithm for all size dig units
800 Tonnes
48 cubic meter bucket
3 MW engine power
300 Tonnes
18 cubic meter bucket
1.1 MW engine Power](https://image.slidesharecdn.com/coertduplessis-maxmine-machinelearninginminingpresentationvupload-211110114045/85/Machine-Learning-in-Mining-14-320.jpg)
Machine Learning in Mining
- 1. Coert du Plessis 10 November 2021 Machine Learning in Mining’s Value Chain
- 3. People who had safety events (not blue) Machine learning taught me cruel, haunting lessons in the early days…
- 4. Fatigued, “Nappy Valley”, Maintainers 270 people,29 severe incidents (2009 – 2011) Within the feature Diff. to the popn Employees 263 people < 5% of pop Tenure with xxxx 4.5 years (mean) 35% Tenure in current position 2.3 years (mean) 42% Age 37 (mean) -5% Maintainers 234 people 334% Task RelevantTraining 180 people (approx) - Carers and Parental Leave 172 people 128% - 274% Fatigued 40% of shifts - Nappy Valley 223 people 128% Voluntary additional shifts 172 people 334% [Withheld] [Withheld] [Withheld] People involved in severe incidents 27 people 92% Low severe incidents reported 183 people 79% Hazard reporters 24 people -72% Safety interaction reporters 61 people -48% Incidents involving working at heights 5 855% Billions of data elements processed Some truths would never find a home Protected in PDF distribution
- 5. Exploration & Geoscience Planning & Scheduling Extraction Processing Logistics Marketing Rehabilitation How we look at the mining value chain Supporting Capabilities Exploration Resource Characterisation Resource Dev. Planning Planning & Scheduling Production Management Drill & Blast Load & Haul Supporting Services Crushing & Blending Beneficiation / Processing Rail & Road Logistics Quality Management Marine Logistics Sales & Marketing Monitoring & Compliance Rehabilitation & Sale Port Operations Geotechnical Engineering Product development Asset Management & Maintenance Work Management People (HR and Contractors) Travel & Accommodation Facilities (Non Processing) Finance Procurement Safety, Risk, Security Technology & infield comms Data & Information Management Improvement Corporate Affairs & Legal Tenancy Community
- 6. Exploration & Geoscience Planning & Scheduling Extraction Processing Logistics Marketing Rehabilitation We just covered safety, and the delicate balance in knowing Exploration Resource Characterisation Resource Dev. Planning Planning & Scheduling Production Management Drill & Blast Load & Haul Supporting Services Crushing & Blending Beneficiation / Processing Rail & Road Logistics Quality Management Marine Logistics Sales & Marketing Monitoring & Compliance Rehabilitation & Sale Port Operations Geotechnical Engineering Product development Asset Management & Maintenance Work Management People (HR and Contractors) Travel & Accommodation Facilities (Non Processing) Finance Procurement Safety, Risk, Security Technology & infield comms Data & Information Management Improvement Corporate Affairs & Legal Tenancy Community Supporting Capabilities
- 7. Exploration & Geoscience Planning & Scheduling Extraction Processing Logistics Marketing Rehabilitation And we have great coverage at this event Supporting Capabilities Exploration Resource Characterisation Resource Dev. Planning Planning & Scheduling Production Management Drill & Blast Load & Haul Supporting Services Crushing & Blending Beneficiation / Processing Rail & Road Logistics Quality Management Marine Logistics Sales & Marketing Monitoring & Compliance Rehabilitation & Sale Port Operations Geotechnical Engineering Product development Asset Management & Maintenance Work Management People (HR and Contractors) Travel & Accommodation Facilities (Non Processing) Finance Procurement Safety, Risk, Security Technology & infield comms Data & Information Management Improvement Corporate Affairs & Legal Tenancy Community Gurav Chris Coert Greg Chelsea Eun-jung Alex Alex Edin Justin Matt Mitin Fred
- 8. Exploration & Geoscience Planning & Scheduling Extraction Processing Logistics Marketing Rehabilitation Exploration is puuuuurrrrrfect for Machine Learning… almost…. You also need to handle the firehose of data… Exploration Resource Characterisation Geotechnical Engineering
- 9. Exploration & Geoscience Planning & Scheduling Extraction Processing Logistics Marketing Rehabilitation Predictive maintenance is all the rage… is knowing enough? Exploration Resource Characterisation Resource Dev. Planning Planning & Scheduling Production Management Drill & Blast Load & Haul Supporting Services Crushing & Blending Beneficiation / Processing Rail & Road Logistics Quality Management Marine Logistics Sales & Marketing Monitoring & Compliance Rehabilitation & Sale Port Operations Geotechnical Engineering Product development Asset Management & Maintenance Work Management People (HR and Contractors) Travel & Accommodation Facilities (Non Processing) Finance Procurement Safety, Risk, Security Technology & infield comms Data & Information Management Improvement Corporate Affairs & Legal Tenancy Community Supporting Capabilities
- 10. BHP’s Maintenance Centre of Excellence (MCoE)
- 11. Exploration & Geoscience Planning & Scheduling Extraction Processing Logistics Marketing Rehabilitation A deeper dive into the nuance of sensitive data… dig unit hang time Exploration Resource Characterisation Resource Dev. Planning Planning & Scheduling Production Management Drill & Blast Load & Haul Supporting Services Crushing & Blending Beneficiation / Processing Rail & Road Logistics Quality Management Marine Logistics Sales & Marketing Monitoring & Compliance Rehabilitation & Sale Port Operations Geotechnical Engineering Product development Asset Management & Maintenance Work Management People (HR and Contractors) Travel & Accommodation Facilities (Non Processing) Finance Procurement Safety, Risk, Security Technology & infield comms Data & Information Management Improvement Corporate Affairs & Legal Tenancy Community Supporting Capabilities
- 12. Automated Dig Unit Time Usage Classification Aim: Simplicity • Less sensors. • Less failures. • Less maintenance • Foster Trust
- 13. The Basic Challenge ACTIVITY DEFINITION Cycling Productive work Hang Waiting for truck / not doing anything Clean Up Non-productive work Repositioning Moving to a new location 9-DOF IMUs Dual feed GPS Using independent sensors Input Output
- 14. Scale can varies – one [general] algorithm for all size dig units 800 Tonnes 48 cubic meter bucket 3 MW engine power 300 Tonnes 18 cubic meter bucket 1.1 MW engine Power
- 15. Calibration can be fun… installation in remote Africa Inertial measurement units (IMUs) can be installed in any location or orientation. Anticipate and automate calibration.
- 16. Calibration of each sensor is contextual… E.g. the GPS antenna may be in different locations Even small placement differences affect behaviour signal classification
- 17. And sometimes it is hard to simply tell the time A computer that knows the time… Logger System Time GPS Time IO Board Counter Combined Time
- 18. But if you add a dash of passionate developer, a main course of quality high resolution data….
- 19. Rapid learning, labelling & signal validation, relearning Before Improvement Effort 65% After Improvement Effort 96% Clean Up Cycling Hang Labels MaxMine Labels MaxMine
- 20. Data Quality and Algorithm Quality cannot be trusted… ever
- 21. We now run a fully automated classification system 24/7 with almost no human input or error correction
- 22. In parallel, fully automate testing with machines Tesla has Dojo trainer, MaxMine has Canary – Test & Learn MaxEdge Testing Components Permutations of hardware versions (for different equipment types and software builds) CanaryTestand Train unit Bench test unit for MaxMine hardware Confirms edge software before deployments (In-house Developed) CanaryMaxStation Each Station is 12 canary modules with canary units driving test scenarios and system configs Canary replicates real field environment from historic data, including equipment data stream, GPS, etc Integrated to our Continuous Integration pipeline MaxEdge Compute MaxEdge IO Board Off the Shelf industrial Touchscreen
- 23. Exploration & Geoscience Planning & Scheduling Extraction Processing Logistics Marketing Rehabilitation Achieving impact in with the operators is truly rewarding… Exploration Resource Characterisation Resource Dev. Planning Planning & Scheduling Production Management Drill & Blast Load & Haul Supporting Services Crushing & Blending Beneficiation / Processing Rail & Road Logistics Quality Management Marine Logistics Sales & Marketing Monitoring & Compliance Rehabilitation & Sale Port Operations Geotechnical Engineering Product development Asset Management & Maintenance Work Management People (HR and Contractors) Travel & Accommodation Facilities (Non Processing) Finance Procurement Safety, Risk, Security Technology & infield comms Data & Information Management Improvement Corporate Affairs & Legal Tenancy Community Supporting Capabilities
- 24. Haul Speed 25% less tonnes FuelBurn 40% more diesel Fill Factor 45% less tonnes Abusive Shifts 140% more wear & tear Operator Performance Is Variable!
- 25. When we treat each operator as a human, smiles are easy
- 27. 27 The first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency. - Bill Gates
- 28. … one more thing …
- 29. “Our ambition is to reach net zero emissions by 2050. Our 2030 targets are to reduce our emissions intensity by 30% and our absolute emissions by 15%.” “A medium-term target to reduce operational greenhouse gas emissions by at least 30 percent from adjusted FY2020 levels by FY2030.”
- 30. Exploration & Geoscience Planning & Scheduling Extraction Processing Logistics Marketing Rehabilitation …one more big thing… Sustainability Exploration Resource Characterisation Resource Dev. Planning Planning & Scheduling Production Management Drill & Blast Load & Haul Supporting Services Crushing & Blending Beneficiation / Processing Rail & Road Logistics Quality Management Marine Logistics Sales & Marketing Monitoring & Compliance Rehabilitation & Sale Port Operations Geotechnical Engineering Product development Asset Management & Maintenance Work Management People (HR and Contractors) Travel & Accommodation Facilities (Non Processing) Finance Procurement Safety, Risk, Security Technology & infield comms Data & Information Management Improvement Corporate Affairs & Legal Tenancy Community Supporting Capabilities
- 31. Efficiency of diesel fuel combustion in mobile equipment is only ~30%, making the focus on increasing productivity essential 100% 30% 15-18% Of the 100% of fuel volume burned, only 30% is converted to energy And, only 15-18% is used for productive work (the balance doing unproductive work)
- 32. Carbon and greenhouse equivalent emissions from surface load and haul equipment presents an imperative for a substantial in reducing overall mining emissions 80/20 Load and haul in open pit mines produce ~84% of operational mining emissions
- 33. Unlock Your Mine’s Full Potential maxmine.com.au