The use of artificial neural networks to adjustment of the sheet delivery devices in gathering machines

The publication presents the results of research of possibility using the different variants of artificial neural networks for solution the problem of prediction parameters in edition adjustment of the sheet delivery devices in gathering machines. Currently, this is a significant problem faced by the printing industry. This is due to the fact that the parameters in the printing devices change automatically during their operation. The question of setting up and changing of parameters of the thickness signatures is insufficient research and has a random character. Lack of knowledge of the displacement values of these parameters forces long machine downtime in the printing industry. In the article, prognoses were made of the setting parameter for assembling signatures in book collators in gathering machine. These prognoses were made using artificial neural networks. Performance of three neural network architectures were analysed. It was also suggested which of the examined architectures can be used on real objects, and which should be further improved.