Title Predikcija korektivnih mjera u stabilizaciji sustava proizvodnog procesa Title (english) Corrective Measures Prediction in the Stabilization of the Production Process System Author Mladen Šverko Mentor Nikola Tanković (mentor)Mentor Darko Etinger (komentor)Committee member Tihomir Orehovački (predsjednik povjerenstva)Committee member Nikola Tanković (član povjerenstva)Committee member Darko Etinger (član povjerenstva)Committee member Goran Oreški (član povjerenstva)Granter University of Pula Defense date and country 2021-06-23, Croatia Scientific / art field, SOCIAL SCIENCES Abstract Razina kompleksnost industrijskih kontrolnih sustava neprestano se povećava tijekom posljednjih desetljeća. To je uglavnom slučaj zbog primjene novih tehnologija s krajnjim ciljem postizanja boljih rezultata produktivnosti poboljšanjem kontrole procesa, dostupnosti sustava, pouzdanosti i fleksibilnosti. S tim ciljem, u eri koncepta Industrije 4.0, granice su pomaknute prema Big data, industrijskom Internetu stvari (IIoT), strojnom učenju i heterogenim mrežama. Kao posljedica navedenog, sustavi nadzora,
kontrole i prikupljanja podataka (SCADA) upravljaju s više podataka, imaju više funkcija i postaju složeniji za upotrebu. To možda nije problem za industrijske grane u kojima razina automatizacije omogućuje kontrolu procesa bez ljudi. Ali u slučaju industrijskih grana s kontinuiranim procesima poput industrije čelika s ulančanim pogonima kontinuiranog lijevanja, valjaonica, galvanizacijskih linija i valjačkih stanova, gdje se očekuje određena stohastička komponenta, operateri su i dalje nezamjenjivi. Za takva postrojenja kompetencija i iskustvo operatera presudni su u održavanju stabilnosti i kontinuiteta procesa u vremenu. Ovaj se rad fokusira na podršku operateru primjenom nadziranih modela strojnog učenja više-izlazne linearne regresije i modela potpornih vektora koji se koriste za predikciju najučinkovitijeg odgovora operatora za određeno stanje procesa u stvarnom vremenu, putem analize skupa podataka sa aktivnostima svih operatora u odnosu na kontrolirani proizvodni proces te izdvajanja najučinkovitijeg operatora koji će se koristiti kao referenca.
Abstract (english) Complexity of industrial control systems has been increasing constantly over past decades. This is mostly the case due to implementation of new technologies with the ultimate goal of achieving higher productivity results by improving process control, system availability, reliability and flexibility. Following this goal, in the era of Industry 4.0 concept, boundaries have been pushed toward Big data, Industrial internet of things, machine learning and heterogeneous networks. As a consequence, supervisory control and data acquisition (SCADA) systems manage more data, have more functions and become more complex to use. This might not be the issue for some industries where level of automation allows unmanned process control. But in the case of industries with continuous stream processes like the steel industry with chained processes of continuous casting, rolling mills, galvanizing lines, and wire rods mills where a certain stochastic component is expected, operators are still irreplaceable. For those plants, an operator's competence and experience is crucial in maintaining process stability and continuity over time. This work focuses on operator decision support by implementing supervised machine learning models of multi-output
linear regression and support vector machine used for predicting the most efficient operator’s response for a given process state in real time, achieved by analysing dataset with all operator activities to the controlled production process, and allocating the most efficient operator to be used as a reference.
Keywords
Informacijski sustavi i informatologija
izrada informatičkih projekata
scada sustav
hmi
stroj potpornih vektora
više-izlazna regresija
kontinuirani proizvodni proces
Keywords (english)
Information systems and informatology
development of IT projects
scada system
hmi
support vector machine
multi-output regression
continuous production process
Language croatian URN:NBN urn:nbn:hr:137:385426 Study programme Title: Master in Informatics; specializations in: Master in Informatics, Master in Informatics, specialisation: teaching, Master in Informatics, Master in Informatics, specialisation: teaching Type of resource Text File origin Born digital Access conditions Open access Terms of use Created on 2021-07-16 10:54:08
