Artificial intelligence approach to the modeling of rolling loads in technology design for cold rolling processes

The paper presents an attempt to apply artificial neural networks (ANNs) to the prediction of the influence of various frictional conditions on rolling forces and torques. Training of the network was done using experimental data, which consist of the results of load measurements during cold rolling of aluminum alloys in different lubrication conditions. The properties of the lubricant became the input variables for the neural network. Accurate prediction of the rolling forces and torques during cold rolling under varying frictional conditions is the main ability of the model. The artificial neural network was validated using data, which were not used during the training procedure. Next, the predictions of the artificial neural network were compared with the finite element calculations of rolling under varying friction conditions. This validation confirmed the good predictive ability of the ANN model.