New Double Cycle Hybrid Method of Machine Learning Using Laser Heat Treatment Pattern Recognition with the Topological Properties of a Network
M. Babič
The aim of this paper is to present a new double cycle hybrid method of machine learning using pattern recognition with topological properties of a network in the robot laser hardened process with different angles of the laser beam. The microstructure of laser hardening is very complex. In this article we use the mathematical method visibility network to determine the complexity of these specimens. We use an artificial neural network (ANN) and multiple regression to predict the martensite microstructure of materials after laser hardening. Finally, we use a hybrid method of these methods of intelligent systems – a hybrid intelligent system (HIS). Relations between the microstructure and angles of the robot laser beam were formulated based on the mechanical stability of the austenite.
Keywords: Laser, robot, microstructure, machine learning, visibility network, heat treatment intelligent systems, neural network, multiple regression, double cycle hybrid