Drugo pozvano predavanje Znanstvenog centra izvrsnosti za znanost o podatcima i TTP2018 projektnih inicijativa

21. lipnja 2018. u 11 sati

Institut Ruđer Bošković
Centar za informatiku i računarstvo
Plava predavaona

PREDAVAČ
Dr. sc. Matej Babič
Jožef Štefan Institute

PREDAVANJE

Novi hibridni sustav strojnog učenja i prepoznavanja statističkih uzoraka za 3D mrežu vidljivosti

Abstract
Intelligent systems are an excellent tool to use for solving complex problems in the field of industrial applications. We use the mathematical method of fractal geometry and network theory when laser hardening techniques are applied. The microstructure of the robot-laser-hardened specimens is very complex; however, we can present it by using a 3D visibility network. We convert the scanning electron microscope (SEM) images of the microstructure to a 3D graph and calculate the density of the visibility network of these 3D networks. The topographical properties of the hardened specimens are analysed using the algorithm for the construction of a visibility network in a 3D space. We develop a new hybrid system of machine learning for predicting carbide content of the hardened specimens by using multiple regression, neural networks, and a genetic algorithm. We find the statistical significance of the relationship between attributes of the hardened specimens, the topological properties of visibility graphs, and carbide content of the hardened specimens.