REPUBLIC OF SERBIA

MINISTRY OF DEFENCE

www.mod.gov.rs

MINISTRY OF DEFENCE

Material Resources Sector

Defensive Technologies Department

Military Technical Institute

www.vti.mod.gov.rs

USING NEURAL NETWORK AND GENETIC PROGRAMMING IN DETERMINATION OF Impact Toughness of Welded Joint Components

BAHRUDIN HRNJICA

University of Bihać, Bihać, Bosnia and Herzegovina, bahrudin.hrnjica@unbi.ba 

fadil islamović

University of Bihać, Bihać, Bosnia and Herzegovina, fadil.islamovic@unbi.ba

ZIJAH BURZIĆ

Military Technical Institute, Beograd, Serbia, zijah.burzic@vti.vs.rs

DŽENANA GAČO

University of Bihać, Bihać, dzenana.gaco@unbi.ba

Abstract: In this paper different artificial intelligence methods were used in determination of impact toughness. The impact toughness was measured on welded joint components of V-notch specimens. The experimental research was performed in laboratory conditions by testing V notch specimens at different temperatures using instrumented Charpy pendulum. Experimental result was input data set for models training using artificial neural network (ANN) and genetic programming, (GP). In order to test the models, testing dataset was generated by performing additional experiment, and the results were compared. The performance analysis shows that using artificial intelligence methods e.g. ANN and GP can obtain high quality impact toughness models and should be considered as basic tool in most experimental researches.

Keywords: impact energy, genetic programming, artificial neural network, GPdotNET, ANNdotNET, modeling, impact toughness, welded joint.