top of page
Image by SJ Objio
Title: Advanced Predictive Analytics for Mining Equipment Reliability and Maintenance
Organization

Vale - The University of Queensland

Period

June 2015 - May 2017

Status

Completed

Est. Budget

USD 1.5M

Est. Business Benefit

USD 12.0M

Indicative Value-to-cost

8.0x

Keywords

Predictive maintenance, mobile equipment, alarm rationalisation, reliability analytics, brake failure prediction

Benefit basis

Estimated value from avoided high-severity failures, reduced unscheduled downtime, better alarm quality, and improved maintenance planning.

Business Challenges 
Mining operations face high maintenance expenditure, production losses from unscheduled downtime, and safety risks associated with equipment failures. Maintenance costs commonly represent a substantial portion of operating costs, and critical failures can stop production or create hazardous events. The project focused on identifying practical predictive-analytics methods to reduce unplanned maintenance delays, distinguish real alarms from spurious alarms, and identify proactive diagnostic signals for mobile-equipment failure modes.
Expanded Technical Solution
The project developed an advanced predictive analytics framework for mining mobile equipment, with a deep dive into catastrophic haul-truck brake failure and alarm rationalisation. The solution combined historical maintenance records, onboard event data, sensor readings, fault codes, operating conditions, inspection information, and failure history. Classification, anomaly detection, survival analysis, and alarm-filtering methods were used to identify precursor patterns and improve maintenance response quality.
Technical Work Packages/Methods
  • Created an integrated failure-analysis dataset combining maintenance work orders, fault/event codes, sensor telemetry, inspections, and operating context.
  • Labelled historical events into real failures, near misses, normal operating events, and spurious alarms.
  • Used supervised classification and anomaly-detection methods to identify precursors to high-severity failures.
  • Applied reliability analytics to estimate failure probability, time-to-failure windows, and maintenance intervention priorities.
  • Developed alarm-performance metrics such as precision, recall, false-positive rate, lead time, and maintenance-actionability.
Key Deliverables
  • Predictive-maintenance framework for mobile-equipment failure modes.
  • Alarm rationalisation method for distinguishing real and spurious alarms.
  • Analytical model for identifying proactive diagnostic indicators.
  • Decision-support outputs for maintenance planners and reliability engineers.
Results and Value Created
The completed work demonstrated a practical path for reducing unscheduled maintenance and improving reliability decision-making. By improving failure precursor detection and reducing low-value alarms, the approach supported better maintenance timing, improved safety risk management, and reduced production loss from unplanned failures. The estimated benefit reflects avoided downtime, reduced catastrophic failure risk, lower emergency maintenance cost, and improved mobile-equipment availability.
bottom of page