Title Na kvarove tolerantan sustav upravljanja sporohodnim dizelskim motorom za brodsku propulziju Title (english) Fault Tolerant Control System for Large Diesel Engine in Marine Propulsion Author Radovan Antonić Mentor Zoran Vukić (mentor)Committee member Ljubomir Kuljača (član povjerenstva)Committee member Zoran Vukić (član povjerenstva)Committee member Ante Munitić (član povjerenstva)Committee member Branko Novaković (član povjerenstva)Committee member Sejid Tešnjak (član povjerenstva)Granter University of Zagreb Defense date and country 2002-10-29, Croatia Scientific / art field, TECHNICAL SCIENCES Universal decimal classificationUDC ) 004 - Computer science and technology. Computing. Data processing Abstract U radu se istražuju tehnike, metode i postupci temeljeni na novim tehnologijama (neizrazita logika, ekspertni sustavi, umjetne neuronske mreže, i dr.) s kojima se može znatnije unaprijediti dijagnostika brodskog dizelskog motora u eksploataciji i predlažu rješenja za sustav upravljanja (SU) sa sposobnošću prilagodbe na kvarove tijekom operativnog rada. Brodski dizelski proces je po svojoj prirodi veoma kompleksan i čine ga niz međusobno ovisnih i povezanih podsustava s velikim brojem
... More komponenti. Pouzdanost i raspoloživost dizelskog propulzijskog sustava ključna je za ostvarenje osnovne misije broda. Zbog toga je veoma važno primijeniti takav SU koji će omogućiti osnovnu misiju broda i u najtežim situacijama (kvarovi, vanjski poremećaji, situacije nužde i sl.) uz dopušteno degradiranje performansi sustava. Tako rigorozne zahtjeve ne može ispuniti konvencionalni SU primjenjivan do sada u eksploataciji broda, već samo SU s mogućnostima tolerancije kvarova i dinamičke prilagodbe strukture. Praksa i iskustvo su pokazali da, najčešće nedostaju potrebna znanja o ponašanju sustava u nenormalnim situacijama, situacijama kvarova i težih poremećaja, da bi se pravovremeno i adekvatno reagiralo. Cilj i osnovni doprinos ovog rada jest iznalaženje učinkovitih tehnika i principa za prikupljanje, formalizaciju i predstavljanje heurističkih znanja eksperata i operatera, izgradnju baze znanja – pravila, implementaciju tih znanja u SU i dijagnostike sporohodnog dizelskog motora kao dijela propulzijskog procesa broda, te proširenje konvencionalnih metoda upravljanja novima, koje će biti u stanju pravovremeno otkriti i dijagnosticirati kvar i obaviti prilagodbu strukture za nastavak rada sustava s kvarom uz što manje degradiranje njegovih performansi. Veliki dio istraživačkog rada uložen je u svrhe kvalitetne identifikacije dizelskog motora , tj. pronalaženje modela ponašanja u uvjetima različitih kvarova njegovih komponenti i podsustava. Pri tome su korištene različite metode i tehnike (matematički modeli, heuristička znanja eksperata i operatera, eksperimenti i snimanja na realnom motoru na probnom stolu, simulacije i dr.). Analizirane su i prezentirane neke od mogućnosti unapređenja postojećih SU i dijagnostike dizelskog motora (fuzijom informacija, novim tehnikama procesiranja senzorskih signala, optimalnim razmještajem i umrežavanjem senzora, analitičkom redundantnošću, ...).
Pored toga obavljena je strukturna analiza procesa dizelskog motora u sklopu dizelskog propulzijskog kompleksa broda u svrhe iznalaženja inherentnih redundantnosti informacija u pojedinim podsustavima, što će biti temelj moguće rekonfiguracije sustava u slučaju kvara. To je osnova za primjenu predložene sheme fuzije senzorskih i drugih informacija, čime se povećavaju mogućnosti detekcije, lokalizacije i prilagodbe na kvar. Temeljem navedenoga, predložene su i primjerima ilustrirane, najpovoljnije sheme za detekciju i identifikaciju kvarova DM korištenjem podataka iz sustava nadzora i heurističkih znanja eksperata. Konačno, predložena je struktura SU s više modela procesa i odgovarajućih upravljačkih algoritama - regulatora s hibridnim pristupom (kombinacija konvencionalnih, neizrazitih i neuro-neizrazitih), koja će omogućiti prilagodbu i toleranciju kvarova dizelskog motora u operativnom radu u svim režimima. Istaknuti su zahtjevi kojima SU (HW i SW struktura) mora udovoljiti u svrhe uspješne integracije u jedinstveni integrirani sustav upravljanja broda - ISUB.
Dobar dio istraživačkih rezultata ovog rada prezentiran je na simpozijima u zemlji, a jedan dio objavljen je u međunarodnim časopisima. Less Abstract (english) In the work, new techniques, methods and procedures based on new technologies (fuzzy logic, expert systems, artificial neural networks, etc.) have been investigated and new approaches and solutions for control and diagnostics system of diesel engine process have been suggested. The system should possess fault tolerance and reconfiguration possibilities. A marine diesel engine process is very complex in its nature, and includes a set of interrelated and dependent subsystems with a
... More large number of components. The reliability and availability of diesel propulsion system is most important in the realization of the ship's main mission. Therefore, it is very important to implement such a control system which will make the mission possible with acceptable performance degradation even in the most critical situations (faults, external disturbances, emergency, etc.). Such rigorous requirements can only be fulfilled by fault tolerant control system with dynamic reconfiguration possibilities. The practice and experience have showed that the knowledge about the system's behaviour in abnormal situations most often missed for adequate actions. The aim and the basic contribution of this work is to investigate effective techniques and methods for expert knowledge elicitation, formalization and representation suitable for computing usage, implementation of such knowledge schemes to diesel engine control and diagnostic system. According to the identification models of diesel engine process and experts heuristic knowledge, conventional control methods and algorithms have been extended with new ones (fuzzy and neuro-fuzzy approach) with fault tolerance and reconfiguration possibilities. A great deal of the investigation work was dedicated to the identification of diesel engine within propulsion process as a whole i.e. in finding behavioural models in different faulty conditions. In that phase, different methods and approaches were used (mathematical descriptions, heuristic knowledge of experts and operators, experimental results in testing large diesel engines on test bed, simulation and etc.) to derive more appropriate models in all circumstances. Some of the possibilities of enhancing existing diesel engine control systems have been analysed and presented (with sensor information fusion, new techniques of signal processing, sensor networking, analytical redundancy,..). For that purpose, the structural analysis of diesel engine within ship propulsion system has been carried out. In some of the engine subsystems and vital parts redundant information were found which should be used to increase fault detection, identification and accommodation. On the basis of investigation and experimental and simulation results, some of the most suitable fault detection and identification schemes of diesel engine using supervision system and expert knowledge base have been suggested and illustrated with examples. Finally, the control system structure with multiple process models and appropriate control algorithms - controllers (combination of conventional, fuzzy and neuro-fuzzy algorithms - hybrid approach) with adequate switching scheme and fault accommodation possibilities have been suggested. The most important requirements that the control system (HW and SW structure) has to satisfy in order to be successfully integrated within ship's integrated control system as a whole have been emphasised. A great deal of the investigation results in this work has been presented at domestic symposiums and some of them have been published in international journals. Less Keywords
Brodski dizelski motor
baza znanja
kvar
detekcija
identifikacija
ekspertni sustav
neizrazita logika
neuronska mreža
sustav upravljanja tolerantan na kvar
Keywords (english)
Marine diesel engine
knowledge base
fault detection
fault identification
fault tolerance
exspert system
fuzzy logic
neural network
control system
reconfiguration
Language croatian URN:NBN urn:nbn:hr:168:154565 Study programme Title: Electrical Engineering and Computing Type of resource Text File origin Born digital Access conditions Closed access Terms of use Created on 2019-04-29 13:50:44
