Presentation of smart method for cold rolling force prediction using Artificial Neural Networks

Neural networks can learn automatically complex relationships between data. This feature makes this technique very useful in the modeling of processes which mathematical modeling is difficult or impossible. In this study, Artificial Neural Network (ANN) is investigated as a tool for cold rolling’s mean force estimation and its performance is evaluated. Two types of the feed forward networks, levenberg-marquardt and bayesian, are considered for cold rolling force prediction of 1100 aluminium sheet. Measurable variables and effective parameters on cold rolling force (width sheet, back tension, front tension, initial and final sheet's thickness) are five input parameters of the network. Moreover, mean rolling force is used as the output parameter of the network. Our experiments suggest that ANN is an effective method for cold rolling force prediction. Particularly, the Bayesian network demonstrates high performance in this case.

کلیدواژه ها:

Artificial neural network’s ، Bayesian method ، Cold rolling force ، Levenberg-marquardt method