Cost–Reliability Trade-offs in Additive Manufacturing Maintenance: A Digital Twin, Kaplan–Meier, and Monte Carlo Approach

Additive Manufacturing (AM) has matured into a production technology in which long, high-value build cycles magnify the operational and economic consequences of even brief disruptions. This paper advances an integrated maintenance framework that unifies stochastic degradation modeling, condition-based maintenance (CBM), and a digital-twin prognostics layer to jointly minimize lifecycle cost and enforce mission-time reliability. Degradation is represented as a drift–diffusion (Wiener) process, enabling closed-form first-passage benchmarks and efficient numerical propagation. CBM decisions are posed as single-threshold stopping rules acting on filtered state estimates; inspection cadence and preventive thresholds are governed by low-dimensional, saturating control laws equipped with explicit guardrails that bound short-horizon failure risk and preserve a fixed safety margin to the functional limit. The digital twin fuses noisy, discrete measurements via Kalman filtering to produce remaining-useful-life (RUL) summaries that drive real-time adaptation of both cadence and trigger. We articulate a policy taxonomy spanning four archetypes (fixed cadence/trigger (ΔT,M), adaptive cadence/fixed trigger (ΔT _k ​,M), fixed cadence/adaptive trigger (ΔT,M _k ​), and fully adaptive (ΔT _k ​,M _k ​)) and evaluate their cost–reliability trade-offs within a Monte Carlo simulation–optimization program that accounts for inspection overhead, preventive/corrective asymmetry, downtime accrual, and production alignment to build boundaries. The framework is analytically transparent, operationally implementable, and readily extensible to richer sensing and learning-based prognostics, offering a principled pathway to reliability enhancement, downtime mitigation, and lifecycle cost optimization in industrial AM.