An ML Framework for Safety in Mining Industry

Latest benchmarking on safety performance by ICMM shows passive safety systems do not prevent accidents from happening, but significantly reduce accidents being fatal. Machine Learning based solutions can address this gap by actively predicting the accidents and demonstrating the trends and complex interrelation between leading indicators for actionable insights. This session shares insights into a framework tailored to the mining industry requirements for developing such Machine Learning solutions in eliminating the accidents and hence improving the safety performance in this industry.

Key Takeaways

• Advanced machine learning solutions enable mining companies to unfold non-linear correlation of leading indicators from various categories including system, people, equipment, process, and culture.
• In problem formulation, ensure prioritizing accidents, identifying key players and data sources. At this stage assume availability of all required data.
• The biggest challenge is data collection! Start small by (1) adjusting the solution to the organization maturity level, (2) using leading indicators already known by users, however, providing a roadmap to the ultimate solution, and (3) carefully selecting model KPIs for this type of problems.
• Bridge between model KPIs and business KPIs to be monitored and improved in cycles and demonstrate the learned trends among leading indicators using advanced libraries designed for interpreting Machine Learning solutions.

————————————————————————————————————————————————————

Bio

Dr Soheila Ghane – Data Science Lead | AutoGrab| Australia

Soheila Ghane is Principal Data Scientist at BHP. She also collaborates with the business leaders in devising strategies for ML/AI adoption, education, and maturity assessment. She also leads a team of data scientists in discovering new data driven opportunities, developing and productionizing best practices, and tailoring the solutions’ metrics to the industry and business requirements.

Soheila received her PhD degree from The University of Melbourne in the field of computer science and differentially private data analysis. She published papers in top-tier journals including IEEE Transactions and IEEE IoT journal. In her research, she proposed robust and efficient algorithms for preserving individuals’ privacy while sharing and analysing their sensitive data in applications such as transportation.