Zaznavanje bazena taline pri obločnem navarjanju z žico

Kuster, Boris (author), Bračun, Drago (mentor), Klobčar, Damjan (comentor)

Pri izdelavi 3D tiskanih kovinskih izdelkov s postopkom obločnega navarjanja z žico je težko doseči dobro dimenzijsko natančnost zaradi nestacionarnosti varilnega postopka, kar se odraža v valovitosti sten izdelkov in razliki med željeno in dejansko višino navarjenega sloja. Napaka se povečuje z večanjem števila navarjenih slojev. Z namenom povečanja dimenzijske natančnosti postopka smo razvili sistem za snemanje varjenja z visokohitrostno kamero ter algoritem za robustno in hitro zaznavo bazena taline iz dobljenih slik. Analizirali smo možnosti uporabe dobljenih podatkov o bazenu taline kot povratno zanko za krmiljenje varilnega aparata.

• Language: Slovenian
• Keywords: aditivne tehnologije, obločno navarjanje z žico, zaznavanje bazena taline, strojni vid, konvolucijske nevronske mreže, optimizacija postopka
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FS - Faculty of Mechanical Engineering
• Place of publishing: Ljubljana
• Publisher: [B. Kuster]
• Year: 2021
• Number of pages: XXII, 62 str.
• PID: 20.500.12556/RUL-125651
• UDC: 004.946:681.518.52:621.9.04(043.2)
• COBISS.SI-ID: 58915331
• Publication date in RUL: 30.03.2021

KUSTER, Boris, 2021, Zaznavanje bazena taline pri obločnem navarjanju z žico [online]. Master’s thesis. Ljubljana : B. Kuster. [Accessed 22 March 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=125651

Secondary language

• Language: English
• Title: Melt pool detection at wire arc welding

Abstract

Achieving good dimensional accuracy for 3D printed objects with the Gas metal arc welding process (Wire-arc additive manufacturing) is difficult because of the non-stationarity of the welding process, which results in object wall waviness and differences between the desired and actual layer height of the welded material. This error is cumulative and rises in proportion to the number of welded layers. To improve the dimensional accuracy of Wire-arc additively manufactured objects, we design a system to record welding using a high speed camera, and design an algorithm for robust and fast melt pool detection from the images. We analyze the possibilities of using this data as a feedback loop for controlling the welding machine parameters in real time.

• Keywords: additive manufacturing, gas metal arc welding, melt pool detection, computer vision, convolutional neural networks, process optimization