Integrated Decision Framework for Railway Signaling and Train Protection: Modeling Latency, Detection Uncertainty, and Capacity-Safety Trade-Offs Under Faults

Abstract

This article presents an engineering-oriented framework that treats railway signaling as an end-to-end decision pipeline and quantifies how uncertainty propagates through occupancy detection, train localization, interlocking logic, radio block center messaging, onboard supervision, and operational recovery procedures to determine risk-relevant metrics such as probability of separation violation, probability of spurious braking, time-to-restrict and time-to-restore distributions, and expected capacity degradation during degraded modes. A scenario-based quantitative study is developed around three representative architectures that span fixed-block signaling with track circuits, fixed-block signaling with axle counters, and a moving-block style supervision concept consistent with modern CBTC or ETCS-level architectures, and for each architecture the study compares baseline thresholds and governance against a reliability-governed strategy that uses nuisance-constrained decision limits, drift-aware plausibility checks, redundant evidence fusion, and staged interventions. Results show that the tail of recovery time and the tail of nuisance restriction duration dominate operational cost, while dangerous exposure is dominated by rare false-clear and localization integrity failures that become most consequential when decision latency is long and when degraded-mode rules are ambiguous or inconsistently applied. The paper provides copy-ready tables and full prompts for data-driven figures suitable for Techne submission, emphasizing applied engineering interpretation rather than purely theoretical safety discussions.