نوع مقاله : پژوهشی
تازه های تحقیق
عنوان مقاله English
نویسنده English
In multi-asset systems, adequate remaining useful life (RUL) prediction accuracy does not necessarily translate into satisfactory system-level performance because dependence among prediction errors can cluster maintenance activations without increasing total demand. This study uses system dynamics to isolate the effect of cross-asset RUL error dependence on tail risk in shared maintenance resources. A 120-asset baseline system is simulated for 280 cycles under five dependence levels , with total predictive-maintenance demand fixed at 120 activations. In the baseline, increasing dependence reduces active arrival cycles from 96 to 35 and increases maximum arrivals per cycle from 4 to 15. Mean predictive queue length falls by 3.9%, whereas its 95th percentile rises by 46.1% and peak capacity pressure by 147.7%. The robustness program tests alternative delay forms, repair-capacity/MTTR variation, with failures while waiting, system sizes from 30 to 1,000 assets, Student-t copulas with tail dependence, asset heterogeneity, explicit spare-parts constraints, and an independent production-capacity mapping. Across these tests, dependence-driven tail congestion remains broadly persistent, although its magnitude is condition-dependent: very small systems and severe spare-parts bottlenecks attenuate the incremental effect, while adequate production reserve can absorb downstream output loss. The results identify temporal concentration as a system-level risk mechanism and show that prognostic evaluation should complement marginal accuracy with cross-asset dependence, tail-oriented queue metrics, and resource-resilience measures.
کلیدواژهها English
Copyright © Mohammad Ebrahimi
License
This article is released under the Creative Commons Attribution (CC BY 4.0) license. Anyone is free to copy, share, translate, and adapt this article for any purpose, whether commercial or non-commercial, as long as proper citation is given to the authors and original publication.
