engleski

Simulation of grain size behavior in microstructure of AA5251 aluminum ingots by neural networks

Computer simulation for prediction of the grain size in aluminum ingots using artificial neural networks is presented in this work. A feed forward neural network model with back-propagation learning algorithm and regularization has been developed to predict the grain size in the microstructure of AA5251 aluminum ingots. Both casting speed and temperature, meniscus level, master alloy AlTi5B1 addition in the form of ingots, speed of master alloy AlTi5B1 addition in a form of wire and cooling water flow are taken as casting process parameters. The artificial neural network was trained on data measured during the vertical DC (direct-chill) casting process, to be able to describe complex dependencies between the microstructure and casting parameters. The results of simulations show satisfactory agreement with the practical experience.

Computer simulation; Grain size; AA5251 aluminum alloy ingots; DC casting; Neural networks

nije evidentirano