Implementing Gaussian process modelling in predictive maintenance of mining machineries.

Area of Science:

• Engineering
• Data Science
• Materials Science

Background:

• Mining machinery is critical, with increasing complexity and reliability demands.
• Sensor data enables predictive maintenance, but current methods have limitations.
• Machine availability and reliability are paramount in modern mining operations.

Purpose of the Study:

• To explore the potential of Gaussian processes for predictive maintenance in the mining sector.
• To address limitations in current sensor-driven predictive maintenance schemes.
• To validate the efficacy of Gaussian process models using real-world mining asset data.

Main Methods:

• Utilized Gaussian processes, a powerful predictive modeling technique.
• Applied the models to three case studies with sensor data from mining assets.
• Validated the models' ability to capture temporal patterns and generate predictions.

Main Results:

• Gaussian process models successfully predicted sensor data degradation trends.
• Accurate remaining useful lifespan predictions were generated.
• Simultaneous monitoring and prediction of multiple machine conditions were achieved with uncertainty estimates.

Conclusions:

• Gaussian processes offer a robust solution for predictive maintenance in mining.
• The models provide credible predictions and valuable uncertainty quantification.
• This approach can significantly facilitate maintenance decision-making for mining assets.