"background": "Industrial machinery fleet management in South Africa faces significant challenges due to ageing assets, variable operational conditions, and limited data for predictive maintenance. Traditional reliability models often fail to account for site-specific heterogeneity and the propagation of uncertainty in risk assessments, leading to suboptimal maintenance scheduling and capital allocation. ", "purpose and objectives": "This case study aimed to develop and evaluate a novel Bayesian hierarchical modelling framework to quantify risk reduction in industrial machinery fleets. The objective was to provide a robust, data-driven methodology for prioritising maintenance interventions and capital renewal across heterogeneous asset groups. ", "methodology": "A case study methodology was employed, analysing operational and failure data from a large national fleet. A Bayesian hierarchical model was constructed, formalised as y{ij \ (\, \), \\, \, \ \ (\\, \\, \), where yij is the time-to-failure for the i-th asset in group j. Posterior distributions were estimated using Markov chain Monte Carlo sampling, with inference focusing on credible intervals for group-level shape and scale parameters. ", "findings": "The model successfully quantified risk reduction from proposed maintenance strategies. A key finding was that implementing the model's prioritisation schedule was associated with a projected 22% reduction in the annual probability of critical fleet-wide failure over a five-year horizon. The hierarchical structure revealed substantial variability in baseline hazard rates between different machinery classes, which had been obscured in previous aggregate analyses. ", "conclusion": "The Bayesian hierarchical model provides a statistically rigorous framework for risk-informed decision-making in machinery fleet management. It effectively synthesises sparse and heterogeneous data, offering a transparent mechanism for propagating uncertainty into operational and strategic plans. ", "recommendations": "Practitioners should adopt hierarchical modelling techniques to account for operational heterogeneity when assessing fleet-wide risk. Further work should integrate real-time condition monitoring data into the model's likelihood
Nkosi et al. (Mon,) studied this question.