Pregled bibliografske jedinice broj: 1218399
Prediction of weld geometry of AlMg3 sheets using artificial neural networks with different input data sets
Prediction of weld geometry of AlMg3 sheets using artificial neural networks with different input data sets // Proceedings of the 10th International Scientific and Expert Conference
Slavonski Brod, Hrvatska, 2022. str. 109-117 (predavanje, međunarodna recenzija, cjeloviti rad (in extenso), znanstveni)
CROSBI ID: 1218399 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Prediction of weld geometry of AlMg3 sheets using artificial neural networks with different input data sets
Autori
Jurišić, Živko ; Krnić, Nikša
Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni
Izvornik
Proceedings of the 10th International Scientific and Expert Conference
/ - , 2022, 109-117
Skup
10th International Scientific and Expert Conference TEAM2022
Mjesto i datum
Slavonski Brod, Hrvatska, 21.09.2022. - 22.09.2022
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
weld geometry prediction ; artificial neural networks ; robotic GMAW process ; aluminium alloys
Sažetak
Predicting the geometry of welded joints, mechanical properties and welding process parameter using scientific methods and mathematical models is an integral part of a comprehensive research in welding practice. In this paper, besides the experiments of robotic metal inert gas welding of 1 mm thick aluminium alloy (AlMg3) sheet, the method of artificial neural networks was used to predict the weld geometry. In the application of neural networks, the approach with four different input data for network structure was used: 1. nominal values of welding current, arc voltage and instantaneous arc power ; 2. measured values of above mentioned process parameters ; 3. measured values with standard deviation values and 4. root mean squared values of above mentioned measured parameters. The selected models, based on the calculation of mean relative error, showed that the neural network method can be applied with high reliability in predicting the geometric characteristics of welded joints of selected material. Also, for practical reasons, it can be used nominal values for predicting weld geometry, but application of measured data in combination with standard deviation or root mean squared values gives significantly more accurate results of weld geometry prediction.
Izvorni jezik
Engleski
Znanstvena područja
Strojarstvo
