Modelling Manila rail transit reliability with dynamic Bayesian networks

• Main Authors: Buquiron, Gabriel D., Roasa, Justin Andre R.
• Format: text
• Language: English
• Published: Animo Repository 2018
• Subjects: Mathematics
• Online Access: https://animorepository.dlsu.edu.ph/etd_bachelors/18577
• Institution: De La Salle University

the Manila Rail Transit (MRT) Line 3. In the aim of modelling the reliability of the MRT 3, ten of its components were modeled after a renewable and repairable indices provided in the literature. The relationship of future and past component reliability were then represented as a vector autoregressive (VAR) process with a dynamic Bayesian network without the need of a fault tree analysis or reliability block design. To gather further knowledge of the correlation between components, an incremental association structural learning algorithm was applied between the reliabilities, where the association between reliabilities learned from this algorithm is represented with an undirected graph together with the represented VAR process in the dynamic Bayesian network. To further the possible inference, a maximum likelihood estimation parameter learning algorithm was used to derive the conditional probabilities of the reliabilities of the system.