ADAPTIVE EVOLUTION ALGORITHM FOR PRODUCTION SCHEDULING IN DYNAMIC ENVIRONMENT

V doktorski disertaciji obravnavamo problem razporejanja proizvodnje, ki v proizvodnih podjetjih predstavlja enega glavnih problemov, saj lahko prihaja do vsakodnevnih sprememb zaradi novih naročil, okvar strojev, zamud v nabavi, itd. Ker je razporejanje že v osnovi zahtevno opravilo, te spremembe lahko privedejo do zastojev proizvodnje, kar pa si podjetja ne morejo privoščiti. V okviru doktorske disertacije smo razvili evolucijski algoritem, ki temelji na evoluciji in uporabi demutacij, selektivnih mutacij in prilagodljive kriterijske funkcije in se uporablja za reševanje razporejanja proizvodnje (job shop scheduling) po kriteriju minimalnega izvršnega časa. V raziskavi smo pokazali, da z uporabo demutacij, selektivnih mutacij in prilagodljive kriterijske funkcije dosežemo učinkovitejši algoritem za reševanje razporejanja proizvodnje. Predlagani algoritem smo testirali na podatkih iz realnega okolja in na znanih problemskih instancah. Rezultate predlaganega algoritma smo prav tako primerjali z obstoječimi sorodnimi algoritmi iz literature in pokazali, da je algoritem konkurenčen omenjenim algoritmom glede na doseganje najkrajšega izvršnega časa. V okviru raziskave je bil razvit tudi sistem za izvoz podatkov iz informacijskega sistema Perftech Largo v algoritmu razumljivo skripto in kasnejša objava razporedov na internetnem portalu, ki omogoča vsem vpletenim v sam proces proizvodnje (tako planerju kot izvajalcem), natančen vpogled v trenutno stanje proizvodnje.Scheduling represents one of the biggest issues in production process, because of daily changes that are a consequence of new orders, machines\u27 malfunctions, delays in purchasing, etc. Since scheduling is demanding regardless of those changes, this can lead to delays in production which are unaccaptable for companies in today\u27s world. Within this thesis, we developed an evolution algorithm, based on evolution and the usage of de-mutation, selective mutations and adaptive fitness function, and is used for solving production scheduling (job shop scheduling) based on the criteria of the shortest execution time. In our reseserch we proved that using demutations, selective mutations and adaptive fitness function we can achieve more efficitent algorithm to solve production scheduling. Suggested algorithm was tested on data from real environment and on known benchmark instances. Results of our suggested algoritm were also compared to an existing related algorithms from literature and they show that our algorithm is compatitive. A system for data export from information system Perftech Largo was developed within our research to help users create a script for our algorithm. We also created an internet portal for companies to publish their production schedules. This way all people involved into process can get a detailed insight about the current status of production