The reliability of transport maintenance depot systems is critical for infrastructure sustainability in developing economies. Existing reliability assessments often lack methodological rigour in accounting for hierarchical operational data and contextual factors specific to such regions. This case study aims to methodologically evaluate the reliability of a national network of transport maintenance depots. Its objectives are to develop a robust multilevel modelling framework for reliability analysis and to identify key operational and managerial determinants of system performance. A longitudinal case study was conducted, integrating technical audits, maintenance logs, and operational performance data. System reliability was modelled using a three-level hierarchical linear model: Reliability₈₉₊ = ₀₉₊ + X₈₉₊ + u₀₉₊ + v₀₊ + e₈₉₊, where levels represented individual components, depot facilities, and regional zones. Inference was based on robust standard errors. The multilevel analysis revealed that depot-level management practices accounted for 34% of the variance in system reliability. A one-standard-deviation improvement in spare parts inventory turnover was associated with a 15. 2% increase in mean time between failures (95% CI: 11. 8% to 18. 6%). The methodological framework successfully quantified the disproportionate influence of depot-level management over component-level factors on overall system reliability. This underscores the importance of mid-level organisational processes in engineering system sustainability. Implement reliability-centred maintenance protocols prioritising inventory management. Allocate training resources based on the identified depot-level variance. Institutionalise the multilevel regression methodology for continuous system monitoring and resource planning. system reliability, multilevel regression, maintenance engineering, infrastructure management, hierarchical linear model This study provides a novel, generalisable multilevel modelling methodology for decomposing variance in engineering system reliability across operational hierarchies, offering a superior alternative to conventional single-level analyses.
Uwimana et al. (Fri,) studied this question.