Abstract
{ "background": "Municipal infrastructure asset systems in Nigeria face significant reliability challenges, yet conventional reliability assessment methods often fail to account for the hierarchical structure of these systems and the substantial epistemic uncertainties present in sparse inspection data.", "purpose and objectives": "This study develops and validates a novel Bayesian hierarchical modelling framework to diagnose the system-level reliability of municipal infrastructure networks, explicitly integrating component-level data with network topology to provide robust probabilistic reliability estimates.", "methodology": "A three-level Bayesian hierarchical model was constructed, where the reliability of individual assets is modelled at the first level, their aggregation into subsystems at the second, and overall system reliability at the third. The core system reliability function is given by $R{system}(t) = \\prod{i=1}^{k} \\left[ 1 - \\Phi\\left(\\frac{\\ln(t) - \\mui}{\\sigmai}\\right) \\right]^{wi}$, with parameters $\\mui$ and $\\sigma_i$ estimated via Hamiltonian Monte Carlo sampling. The model was applied to condition assessment data from water distribution and road networks in three Nigerian municipalities.", "findings": "The model quantified a high posterior probability (0.92) that the overall system reliability for the studied networks fell below the target threshold of 0.85. A key finding was the dominant influence of a small proportion (approximately 15%) of critical subsystem assets, whose failure contributed to over 60% of the predicted system unreliability. Parameter estimates showed robust convergence with $\\hat{R} < 1.01$ for all major parameters.", "conclusion": "The proposed Bayesian hierarchical model provides a statistically rigorous tool for infrastructure reliability diagnostics, successfully capturing the multi-scale nature of asset systems and formally propagating uncertainty from component to system level.", "recommendations": "Municipal asset managers should adopt probabilistic reliability frameworks to prioritise investments in critical subsystems identified by hierarchical modelling. Further research should integrate real-time sensor data to transition from diagnostic to predictive reliability analysis.",