Purpose This paper develops and validates a preventive maintenance (PM) optimization framework that jointly accounts for (1) interval-dependent maintenance effectiveness and (2) heterogeneous failure severity in repairable systems. Design/methodology/approach This study extends the NHPP-based virtual age (ARA8) framework of Said et al. by incorporating severity-weighted estimation and interval-dependent maintenance effectiveness. Failure events are weighted using downtime to reflect their operational impact, while preventive maintenance effectiveness is modeled as a function of the maintenance interval. The model is estimated from haul-truck maintenance data using global metaheuristic algorithms (GA, SA, DE). In a second stage, preventive maintenance intervals are optimized using metaheuristics (GA, SA, PSO) to maximize long-run system availability under practical feasibility constraints. Findings Severity-aware estimation leads to higher inferred baseline failure intensity and lower restoration factors compared with OEM assumptions, resulting in more conservative reliability forecasts. The optimized schedules adjust PM intervals within their feasible bounds, yielding stronger restoration effects and improving long-run availability relative to static OEM policies. Research limitations/implications Failure severity is proxied solely by downtime, which may not fully capture broader operational impacts such as safety or cascading effects. Future work should explore multidimensional severity measures and extend the framework to incorporate stochastic maintenance execution and age-based or condition-based policies. Practical implications The framework shows that maintenance effectiveness depends not only on timing but also on the operational impact of failures. The results guide planners to treat severity-aware schedules as conservative, risk-sensitive baselines and to focus resources and precision on the most influential preventive maintenance types. Originality/value To the best of my knowledge, this study is the first to jointly integrate severity weighting and interval-dependent maintenance effectiveness within an NHPP-based virtual age framework, and to embed the resulting effectiveness functions directly into an availability optimization. The approach balances interpretability with empirical fidelity, providing risk-sensitive maintenance schedules grounded in real operational data.
Uthman Said (Fri,) studied this question.