SIGNAL+DRAHT | Ausgabe 12/2019

Transparent failure diagnostics for railway switches using Bayesian networks

Dezember 2019 | Thorsten Neumann, Daniela Narezo Guzmán, Jörn Christoffer Groos

The application of modern methods of artificial intelligence (AI) for diagnostic purposes requires careful consideration as to the reliability of the results, especially in the case of safety-critical systems. As an established AI method, Bayesian networks show some very positive characteristics in this regard due to their transparency and the high degree of traceability of the diagnostic conclusions algorithmically reached with their help. This paper discusses and demonstrates the theoretical and practical potential of Bayesian networks for fault diagnosis in the case of switch failures using concrete numerical examples...