Fleet-Wide Interval-Dependent Nonparametric Modeling and Optimization of Multi-Level Preventive Maintenance Effectiveness: Application to Hybrid (AC) LHD Trucks in a Mine

Preventive Maintenance (PM) policies for repairable systems commonly assume constant effectiveness, overlooking how timing impacts restoration. This study models PM effectiveness as an interval-dependent function, [Formula: see text], embedded in a virtual age degradation model with nonhomogeneous Poisson (power-law) failures. Parameters are estimated via global-local maximum likelihood optimization. Using operational data from underground Load-Haul-Dump trucks, we calibrate PM effectiveness curves (three PM types) and simulate availability under practical interval constraints. The optimized fleet-wide schedule shortens light/moderate PM intervals and modestly adjusts major PM, delivering an 8% increase in long-run availability far above OEM baselines. Sensitivity analysis shows nonlinear, asymmetric responses of availability to interval changes. Perturbation tests show that minor deviations from the optimized intervals incur only marginal losses, indicating operational robustness. This data-driven, interpretable framework enables maintenance planners to jointly optimize timing and effectiveness for enhanced reliability in industrial fleets.