Mining machinery constitutes essential assets for a mining corporation. Due to economies of scale, technological innovations and stringent quality and safety requirements, the size, complexity, functionality and diversity of industrial machinery have expanded markedly over the last two decades. This growth has increased sensitivity to machine availability and reliability. Mining operations install comprehensive maintenance units tasked with inspection, repair, replacement and inventory management for the machines in use. Leveraging the proliferation of sensor technologies integrated within the machines, maintenance units obtain rich data streams synchronously disclosing machine health and performance metrics, which enables a predictive maintenance programme. This programme performs prognostic detections of anomalies and permits timely intervention to avert catastrophic breakdowns. However, such sensor-driven predictive maintenance scheme for machinery in the mining sector is limited. The present paper utilises the Gaussian process, a powerful predictive modelling technique, to show its potential in addressing this challenge. The efficacy of this approach is validated through three case studies. Each case study is equipped with sensor data and represents a typical predictive maintenance task for mining assets. The developed Gaussian process models successfully capture meaningful temporal patterns in sensor data and generate credible predictions across all three tasks: temporal prediction of sensor data degradation trends, remaining useful lifespan prediction and simultaneous monitoring and prediction of multiple machine conditions. Furthermore, the models offer uncertainty estimates to the prediction outcomes, potentially facilitating maintenance decision-making process.