| Sažetak | U ovom diplomskom radu, za odabrani brodski sustav, provedena je analiza pouzdanosti mapiranjem analize stabla grešaka (eng. Fault Tree Analysis, FTA) u Bayesovu mrežu (eng. Bayesian Network, BN). FTA i BN metode su opisane, te su navedene njihove prednosti i ograničenja, uz detaljan opis postupka grafičkog i numeričkog mapiranja FTA u BN. Kombinacijom FTA i BN dobije se detaljnija, fleksibilnija i preciznija procjena sigurnosti i pouzdanosti sustava integracijom determinističkih i vjerojatnosnih metoda. Za odabrani brodski sustav, izmjenjivač topline (eng. Shell & Tube), definirana je granica sustava, zadatci, podsustavi i komponente, te je dan shematski prikaz. Koristeći bazu podataka OREDA, utvrđena su četiri potencijalna kritična kvara odabranog sustava: abnormalno očitanje instrumenta (eng. Abnormal instrument reading, AIR), vanjsko curenje procesnog medija (eng. External leakage – Process medium, ELP), strukturni nedostatak (eng. Structural deficiency, STD), odstupanje parametara (eng. Parameter deviation, PDE). Nadalje, kritični kvarovi su prikazani kao glavni događaji i za njih je konstruirano stablo grešaka i izračunate vjerojatnosti pojave svih događaja unutar stabla. Mapiranjem stabla grešaka u BN formulirane su tablice uvjetne vrijednosti prema vrsti logičkih vrata. Na taj način je utvrđeno koje greške najviše doprinose razvoju kritičnog kvara. Korištenjem programskog alata Netica dobiveni su faktori poboljšanja za kritične kvarove. Utvrđeno je da kritična greška AIR ima najveći intenzitet kvarova, čemu najviše pridonosi njena podgreška opći kvar instrumenta (IS). Uvođenjem alarma povezanog s greškom IS, moguće je umanjiti vjerojatnost njene pojave s 26,9 % na 4,83 %. |
| Sažetak (engleski) | In this thesis, a reliability analysis of a selected ship system was carried out by converting Fault Tree Analysis (FTA) into a Bayesian Network (BN). Both the FTA and BN methods are explained, highlighting their benefits and drawbacks, and providing a detailed guide on the graphical and numerical mapping process from FTA to BN. This combination allows for a more comprehensive, adaptable, and precise evaluation of system safety and reliability, integrating both deterministic and probabilistic approaches. For the selected ship system, a Shell & Tube heat exchanger, the system boundaries, tasks, subsystems, and components were outlined, and a schematic diagram was included. Using the OREDA database, four potential critical failures were identified: abnormal instrument reading (AIR), external leakage of the process medium (ELP), structural deficiency (STD), and parameter deviation (PDE). These critical failures were treated as primary events, leading to the construction of a fault tree, with probabilities calculated for all events within the tree. By mapping the fault tree into a BN, conditional probability tables (CPT) were created based on the type of logic gates used. This approach pinpointed the errors that most significantly contribute to critical failures. Utilizing the Netica software tool, improvement factors for the critical failures were determined. It was discovered that AIR had the highest failure rate, primarily due to the sub-error of general instrument failure (IS). By implementing an alarm system linked to the IS error, the probability of its occurrence could be reduced from 26.9% to 4.83%. |
