Razvoj sistema za hitro ugotavljanje obdelovalnosti jekla na osnovi iskrenja in metod podatkovnega rudarjenja : doktorska disertacija

Munđar, Goran

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Razvoj sistema za hitro ugotavljanje obdelovalnosti jekla na osnovi iskrenja in metod podatkovnega rudarjenja : doktorska disertacija
Authors:	ID Munđar, Goran (Author) ID Župerl, Uroš (Mentor) More about this mentor... ID Kovačič, Miha (Mentor) More about this mentor...
Files:	PhD-Goran_Mundar_-_za_vezavo2.pdf (13,66 MB) MD5: 8B9D1BAB7C03D36767C88804E7BD8EF0
Language:	Slovenian
Work type:	Doctoral dissertation
Typology:	2.08 - Doctoral Dissertation
Organization:	FS - Faculty of Mechanical Engineering
Abstract:	Doktorska disertacija obravnava razvoj inovativnega sistema za hitro ugotavljanje obdelovalnosti nizko legiranih jekel s povečano obdelovalnostjo na podlagi analize iskrenja med postopkom brušenja. Obdelovalnost materiala, opredeljena kot relativna enostavnost mehanske obdelave, se tradicionalno določa z dolgotrajnimi in stroškovno zahtevnimi standardnimi preizkusi, kot je ISO 3685. Za odpravo teh omejitev smo v raziskavi uporabili iskrenje, ki nastaja med brušenjem, kot vir podatkov o materialu. Eksperimentalni del raziskave vključuje načrtovanje in izvedbo brušenja vzorcev različnih jekel, pri čemer je bil razvit in implementiran sistem strojnega vida za zajem slik isker. Na osnovi slikovnih podatkov smo najprej s klasičnimi metodami izluščili numerične značilke isker, ki so bile uporabljene kot vhodni podatki za razvoj različnih napovednih modelov: umetne nevronske mreže (ANN), adaptivni nevronsko-mehki inferenčni sistem (ANFIS), genetsko programiranje (GP) in odločitvena drevesa (DT). Poleg tega smo za neposredno analizo slik brez predhodne ekstrakcije značilk uporabili globoke konvolucijske nevronske mreže (CNN), kot sta ResNet-50 in MobileNet-v2. Rezultati potrjujejo, da razviti modeli omogočajo napovedovanje obdelovalnosti z visoko natančnostjo, primerljivo s standardnimi testi, ob bistveno nižjih stroških in krajšem času izvajanja. Najboljši modeli so dosegli povprečno absolutno odstotno napako (MAPE) med 3 % in 4 % ter koren srednje kvadratne napake (RMSE) med 16 in 19. Razvit sistem strojnega vida uspešno deluje v realnem času ter omogoča hitro in avtomatizirano analizo značilk isker med brušenjem. Disertacija predstavlja izvirni znanstveni prispevek, saj uvaja robustno, hitro in stroškovno učinkovito alternativo klasičnim metodam določanja obdelovalnosti, primerno za uporabo v industrijskem okolju.
Keywords:	Obdelovalnost jekla, brušenje, iskrenje, strojni vid, podatkovno rudarjenje, genetsko programiranje, konvolucijske nevronske mreže
Place of publishing:	Maribor
Place of performance:	Maribor
Publisher:	[G. Munđar]
Year of publishing:	2026
Number of pages:	XIII, 164 str.
PID:	20.500.12556/DKUM-93681
UDC:	[621.923.01:669.15]:004.8/.9(043.3)
COBISS.SI-ID:	270662403
Publication date in DKUM:	27.02.2026
Views:	142
Downloads:	23
Metadata:
Categories:	KTFMB - FS
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:	CC BY-NC-ND 4.0, Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Link:	http://creativecommons.org/licenses/by-nc-nd/4.0/
Description:	The most restrictive Creative Commons license. This only allows people to download and share the work for no commercial gain and for no other purposes.
Licensing start date:	13.07.2025

Secondary language

Language:	English
Title:	The development of a system for fast detection of steel machinability based on spark testing and data mining methods
Abstract:	The doctoral dissertation addresses the development of an innovative system for rapid determination of steel machinability of low-alloy steels with enhanced machinability based on spark analysis during the grinding process. Machinability, defined as the relative ease of mechanical processing, is traditionally determined by time-consuming and costly standard tests such as ISO 3685. To overcome these limitations, this research utilizes sparks generated during grinding as the data source. The experimental phase includes the design and execution of grinding tests on various steel samples, combined with the development and implementation of a machine vision system for spark image capture. From these images, classical image-processing techniques were used to extract numerical spark features as inputs for predictive models: artificial neural networks (ANN), adaptive neuro-fuzzy inference system (ANFIS), genetic programming (GP), and decision trees (DT). Additionally, deep convolutional neural networks (CNN), including ResNet-50 and MobileNet-v2, were employed for direct image analysis without manual feature extraction. Results confirm that the developed models predict machinability with high accuracy comparable to standard tests, while significantly reducing cost and processing time. The best-performing models achieved Mean Absolute Percentage Error (MAPE) between 3 % and 4 % and Root Mean Squared Error (RMSE) between 16 and 19. The machine vision system operates in real time, enabling rapid and automated spark feature analysis during grinding. The dissertation presents an original scientific contribution by introducing a robust, fast, and cost-effective alternative to classical machinability tests, suitable for industrial application.
Keywords:	Steel machinability, Grinding, Spark testing, Machine vision, Data mining, Genetic programming, Convolutional neural networks

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back