doktorska disertacija
Marko Hrelja (Avtor), Miran Brezočnik (Mentor), Jože Balič (Komentor)

Povzetek

Izboljševanje obstoječe proizvodnje in obdelovalnih sistemov zahteva nenehno posodabljanje in integracijo najnovejših tehnologij v proizvodne sisteme. Proizvodnih spremenljivk je čedalje več, s tem pa se povečuje množica podatkov, ki jo moramo obdelati, tu pa velikokrat klasične analitične metode optimizacije odpovedo. Zaradi tega smo prisiljeni bolje izkoristiti razpoložljive proizvodne vire, zato pa moramo poseči po naprednejših pristopih reševanja problemov. Za reševanje zahtevnih problemov čedalje pogosteje uporabljajo različna področja umetne inteligence, še zlasti strojnega učenja. Pregled do sedaj opravljenih raziskav je pokazal, da so obstoječi razviti sistemi precej ozko usmerjeni. V disertaciji predlagamo popolnoma nov pristop k modeliranju CNC-obdelav s pomočjo novega gravitacijskega iskalnega algoritma (GSA), ki spada med metode skupinske inteligence. Razviti inteligentni sistem deluje na osnovi osnovnih Newtonovih fizikalnih zakonov oziroma na osnovi interakcij med masnimi telesi v prostoru. Za primerjavo in potrditev ustreznosti rezultatov doktorske disertacije smo uporabili tudi metodo modeliranja z rojem delcev (PSO). Primerjava je pokazala, da je GSA algoritem primeren za modeliranje obdelav z odrezovanjem, saj so odstopanja od eksperimentalnih podatkov v sprejemljivih mejah. Dobljeni modeli so dobro opisali postopek odrezovanja materiala s struženjem, ki smo ga uporabili kot postopek odrezovanja. Posebej velja omeniti, da je GSA algoritem v najslabšem primeru vsaj dvakrat hitrejši od enakovrednega PSO algoritma. Dobljen model CNC-obdelave smo nato uporabili za večkriterijsko optimiranje obdelovalnih parametrov: optimalne hrapavosti obdelane površine, rezalnih sil in časovne obstojnosti orodja. Vsaka izmed omenjenih odvisnih spremenljivk prispeva k optimalnemu delovanju CNC-obdelovalnega stroja, kar znižuje stroške proizvodnje. Večkriterijsko optimiranje smo izvedli s pomočjo NSGA-II algoritma. Za optimiranje smo morali določiti tudi omejitve. Te smo določili s pomočjo teoretičnih izračunov in jih preverili s pomočjo eksperimentalnih podatkov. Zaradi obsega dela smo se omejili na struženje, hkrati pa so v delu predstavljene osnove prilagoditev za uporabo metod na ostalih obdelovalnih strojih, saj je predlagan pristop univerzalen.

Ključne besede

CNC-obdelovalni stroji;struženje;odrezovanje;skupinska inteligenca;optimizacija z rojem delcev;gravitacijski iskalni algoritem;genetski algoritmi;večkriterijska optimizacija;

Podatki

Jezik: Slovenski jezik
Leto izida:
Tipologija: 2.08 - Doktorska disertacija
Organizacija: UM FS - Fakulteta za strojništvo
Založnik: M. Hrelja]
UDK: 004.832.021:621.941-52(043.3)
COBISS: 277945856 Povezava se bo odprla v novem oknu
Št. ogledov: 2202
Št. prenosov: 322
Ocena: 0 (0 glasov)
Metapodatki: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Ostali podatki

Sekundarni jezik: Angleški jezik
Sekundarni naslov: Modeling and optimizing CNC machine processes using swarm intelligence
Sekundarni povzetek: Improving of the current production and machining systems demands for constant upgrading and integrating latest technologies into manufacturing processes. Increasing manufacturing variables affect also increase of machining information, which has to be processed, and with increased information classic analytical methods of optimization fail. Due to demanding market, complex product design, tendency for more successful manufacturing management with minimal production costs, it is of the essence to use optimal manufacturing resources. Therefore, an advanced approach for problem solution has to be used. For such advanced problems, most commonly are used different approaches of artificial intelligence, especially machine learning. Review of the researches made so far, has shown, that current developed systems are much specialised, and not flexible. In presented dissertation we propose a completely new approach for modelling CNC-machining processes using new Gravitational search algorithm (GSA), which is in essence swarm intelligence method. Developed intelligent system works on basic Newtonian laws, on elementary mass objects interaction in search space. For GSA modelling verification particle swarm optimisation (PSO) method was used. Comparative method PSO has shown, that the GSA-algorithm is appropriate for machining parameters processing, as the deviations are minimal. Acquired models described the cutting procedure by turning successfully. Important fact is also, that the GSA method has shown an improvement in terms of data processing time, as the procedure reduces processing time for a minimum of half. Results have shown, that the processing time of the GSA method is at least twice as fast, comparing to the PSO method. Appropriately designed model of CNC-machining can be used for optimizing machining parameters for the purpose of achieving optimal surface roughness, cutting forces, and increasing of the tool life. Each of the dependent variables adds a share to optimal functioning of the machine tool, which reduces production costs. Multi-objective optimisation is most efficiently made by using NSGA-II algorithm, with most accurate results. For the purpose of multi-objective optimisation constraints have to be determined. These were based on theoretical calculations and confirmed with experimental measurements. Due to dissertation comprehension, machining processes have been limited to turning, yet in the dissertation were presented also required basic differences for successful adaptation of the optimization for other machining processes.
Sekundarne ključne besede: intelligent manufacturing system;CNC-machine tool;cutting;turning;swarm intelligence;particle swarm optimization;gravitational search algorithm;genetic algorithms;multi-objective optimization;NSGA-II algorithm;Inteligentni obdelovalni sistemi;Disertacije;
URN: URN:SI:UM:
Vrsta dela (COBISS): Doktorska disertacija
Komentar na gradivo: Univ. v Mariboru, Fak. za strojništvo
Strani: XI f., 144 str.
ID: 8701092