In fixed-block railway signaling systems, Durmuş, Takai and Söylemez (2014) modeled the field equipment (railway blocks, point machines, signals, route reservation) as ordinary safe Petri nets and constructed a classical discrete-event-systems (DES) diagnoser proving diagnosability of five single fault types on a small example yard with four blocks and one point machine. Two limitations were explicitly left open: the diagnoser design assumes a single fault occurs, so configurations with multiple simultaneous faults of different types are not covered; and the monolithic diagnoser construction suffers state-space explosion, so extension to larger, realistic networks was unexplored. The open problem is to extend this model-based fault-diagnosis methodology to simultaneous distinct-type faults and to realistically sized fixed-block installations while retaining formal diagnosability guarantees (bounded detection delay) and the safety-relevant classification of diagnoser outcomes into alarm-only versus immediate-safe-state conditions.

Background

In a fixed-block railway signaling system, the track is divided into sections called railway blocks, each protected by a signal at its entry, and only one train may occupy a block at a time; block occupancy is detected by devices such as track circuits (an electrical circuit through the rails that is shorted by a train's axles). Before a train is admitted, a route through a sequence of blocks must be reserved by the interlocking, the safety logic that prevents conflicting movements; where routes diverge or join, a point machine (railway switch) physically moves the rails, and it must reach its commanded position within a prescribed time (the source paper uses a 7-second limit following Turkish State Railways practice). Failures of this equipment — a point machine failing to reach its position, a signal displaying a wrong color while a block is occupied — can lead a driver to pass a signal at danger, so their timely detection is safety-critical; the standards for safety-related software in this domain (IEC 61508 and the railway-specific EN 50128) explicitly recommend formal methods such as Petri nets for the highest integrity levels.

A discrete event system (DES) is a system whose state changes only at discrete events, such as a signal changing color or a point machine completing its movement. The source paper models the signaling equipment by ordinary safe Petri nets — bipartite graphs of places (drawn as circles, holding tokens that encode the local state) and transitions (bars, whose firing moves tokens between places), with at most one token per place. Some transitions are observable to the monitoring system and some are unobservable, and special unobservable fault transitions model each failure mode; the 2014 paper defines five fault types F1F_1–F5F_5 (two point-machine position failures, point-machine indication failures, and signal color faults).

For fault diagnosis, the paper adopts the classical diagnoser framework of Sampath and co-workers: the diagnoser is a deterministic observer automaton built from the Petri-net model, whose states are sets of (marking, fault-label) pairs, where a label records which fault types may have occurred on the paths reaching that marking. A state is FiF_i-certain when every pair in it carries FiF_i, meaning fault type FiF_i has definitely occurred. The system is diagnosable for FiF_i if, after any occurrence of FiF_i, the diagnoser reaches an FiF_i-certain state within a bounded number of observable events; the standard necessary and sufficient condition is the absence of FiF_i-indeterminate cycles — cycles of uncertain diagnoser states that can be traversed forever without ever becoming FiF_i-certain. Beyond detection, the authors manually classify each fault-certain diagnoser outcome as an alarm condition (it suffices to inform the traffic control centre) or a safe-state condition (all trains must be stopped immediately, e.g. a signal wrongly showing green while its block is occupied).

On this basis the 2014 paper verified diagnosability for the example yard (blocks T001, PMT01, T002, T003; one point machine PM1; signals RS301, LS301, RS203, LS203, VS01–VS03), presenting the diagnoser in detail only for one of the four possible route reservations. Two limitations were acknowledged: the design assumes that only a single fault occurs, so nothing is said about diagnosability when faults of different types are present simultaneously (for instance a point-machine failure together with a signal color fault); and the monolithic construction of the diagnoser for the full net suffers the well-known state-space explosion, so the approach had not been explored on larger, more complex networks. A companion paper by the same authors applied the same DES style to decision-making strategies rather than diagnosis, and the authors' own follow-up (Durmuş et al., 2016) proposed assessing diagnosability module-by-module — building a small diagnoser per component-type Petri-net module (route, point machine, signal, block) instead of one global diagnoser, following the modular diagnosability theory of Contant, Lafortune and Teneketzis — but verified only a single three-state signal-module diagnoser for one sample field, imposed its Petri-net assumptions (boundedness, deadlock-freedom, absence of unobservable cycles with identical observations) without verifying them, provided no formal composition operator matching the interlocking-table-based linking of modules, and did not address multiple simultaneous faults at all. The general DES literature offers related but distinct threads — decentralized diagnosis protocols, intermittent and repeated faults, and symbolic diagnoser constructions, encoded as binary decision diagrams (BDDs — compact shared graph representations of Boolean functions), that mitigate state explosion for bounded labeled Petri nets — none of which has been assembled into a fault-diagnosis methodology demonstrated on realistic fixed-block signaling installations under compound faults. No published work is known to resolve both limitations together, and the wider railway-signaling fault-diagnosis field has increasingly turned toward data-driven, multi-source fusion methods rather than model-based DES diagnosers.

Problem Statement

Extend the diagnoser-based, Petri-net fault-diagnosis methodology of Durmuş, Takai and Söylemez (2014) for fixed-block railway signaling systems so that it applies beyond the single-fault, small-yard setting, retaining its formal character and safety role. Concretely: (1) multiple simultaneous faults — provide a fault model and a diagnosability notion covering configurations in which more than one fault of different types (e.g. a point-machine position failure together with a signal color fault) is present at the same time, together with a construction or verification procedure that decides diagnosability for such compound-fault situations and determines what the diagnoser can and cannot distinguish within bounded delay; and (2) scale — make the construction and the diagnosability verification feasible for fixed-block networks of realistic size, where the number of blocks, point machines, signals and routes makes the monolithic diagnoser suffer state-space explosion, without assuming away the interlocking constraints, while retaining the methodology's two safety-relevant outputs: a formal diagnosability guarantee — detection of each covered fault situation within a bounded number of observable events — and the classification of diagnoser outcomes into alarm-only conditions versus conditions requiring an immediate safe state (stopping trains), now under compound faults.

References

  1. Mustafa S. Durmuş, Shigemasa Takai, Mehmet T. Söylemez, "Fault diagnosis in fixed-block railway signaling systems: a discrete event systems approach", IEEJ Transactions on Electrical and Electronic Engineering 9(5):523–531, 2014. DOI: 10.1002/tee.22001, https://doi.org/10.1002/tee.22001
  2. Mustafa S. Durmus, İlker Ustoglu, Roman Y. Tsarev, Michael Schwarz, "Modular Fault Diagnosis in Fixed-Block Railway Signaling Systems", IFAC-PapersOnLine 49(3):459–464, 2016. DOI: 10.1016/j.ifacol.2016.07.077, https://doi.org/10.1016/j.ifacol.2016.07.077
  3. Mustafa Seçkin Durmuş, Shigemasa Takai, Mehmet Turan Söylemez, "Decision-making strategies in fixed-block railway signaling systems: A discrete event systems approach", IEEJ Transactions on Electrical and Electronic Engineering 10(2):186–194, 2015. DOI: 10.1002/tee.22052, https://doi.org/10.1002/tee.22052
  4. M. Sampath, R. Sengupta, S. Lafortune, K. Sinnamohideen, D. C. Teneketzis, "Failure diagnosis using discrete-event models", IEEE Transactions on Control Systems Technology 4(2):105–124, 1996. DOI: 10.1109/87.486338, https://doi.org/10.1109/87.486338
  5. Olivier Contant, Stéphane Lafortune, Demosthenis Teneketzis, "Diagnosability of Discrete Event Systems with Modular Structure", Discrete Event Dynamic Systems 16(1):9–37, 2006. DOI: 10.1007/s10626-006-6177-5, https://doi.org/10.1007/s10626-006-6177-5
  6. Olivier Contant, Stéphane Lafortune, Demosthenis Teneketzis, "Diagnosis of Intermittent Faults", Discrete Event Dynamic Systems 14(2):171–202, 2004. DOI: 10.1023/b:disc.0000018570.20941.d2, https://doi.org/10.1023/b:disc.0000018570.20941.d2
  7. Rami Debouk, Stéphane Lafortune, Demosthenis Teneketzis, "Coordinated Decentralized Protocols for Failure Diagnosis of Discrete Event Systems", Discrete Event Dynamic Systems 10(1–2):33–86, 2000. DOI: 10.1023/a:1008335115538, https://doi.org/10.1023/A:1008335115538
  8. João C. Basilio, Guilherme M. Ottoni Silva, "A Diagnoser-Based Approach to Diagnosis and Diagnosability of Repeated Faults of Discrete-Event Systems", IEEE Transactions on Automatic Control 70(8):5491–5498, 2025. DOI: 10.1109/tac.2025.3546163, https://doi.org/10.1109/TAC.2025.3546163
  9. Abderraouf Boussif, Mohamed Ghazel, Kais Klai, "A semi-symbolic diagnoser for fault diagnosis of bounded labeled Petri nets", Asian Journal of Control 23(2):648–660, 2021. DOI: 10.1002/asjc.2259, https://doi.org/10.1002/asjc.2259
  10. Mohamed Ghazel, Baisi Liu, "A customizable railway benchmark to deal with fault diagnosis issues in DES", 2016 13th International Workshop on Discrete Event Systems (WODES), 2016. DOI: 10.1109/wodes.2016.7497845, https://doi.org/10.1109/WODES.2016.7497845
  11. Baisi Liu, Mohamed Ghazel, "Petri Net Diagnosability Analyzer", 2016 11th International Design & Test Symposium (IDT), pp. 13–18, 2016. DOI: 10.1109/idt.2016.7843007, https://doi.org/10.1109/IDT.2016.7843007
  12. Mustafa Seckin Durmus, Ilker Ustoglu, Roman Yu. Tsarev, Josef Börcsök, "Enhanced V-Model", Informatica 42(4), 2018. DOI: 10.31449/inf.v42i4.2027, https://doi.org/10.31449/inf.v42i4.2027
  13. Haimeng Sun, Baofu Qin, Zexian Wei, Zhenpeng Lao, "A review of fault diagnosis for train signal system based on multi-source information fusion", Engineering Research Express 7(2):022503, 2025. DOI: 10.1088/2631-8695/adcaff, https://doi.org/10.1088/2631-8695/adcaff
  14. Zhuolin Hu, Tao Tang, Wanli Lu, Wei Zheng, Jidong Lv, Shigen Gao, Shuai Su, "Fault Diagnosis of the On-board Equipment in CTCS-3 Based on Timed Automata and Mutation Theory", 2019 IEEE Intelligent Transportation Systems Conference (ITSC), pp. 1013–1018, 2019. DOI: 10.1109/ITSC.2019.8916894, https://doi.org/10.1109/ITSC.2019.8916894