Research Of Ink Presetting Algo-Rithm Based On Fuzzy ART-BP Hybrid Neural Network Of Multicolour Offset Press

Since the improvement of quality of printing requirements, the printing enter-prise is facing more and more fierce competition, so many enterprises have introducedthe ink preset system to improve the quality and efficiency of printing. Based on thissystem every ink key opening value of printing machine is preset before printing ma-chine getting to work. So the ink amount of first piece of printed matter is close toprinting sample as far as possible. However, the performance of this system is not sogood in practical application. On the other hand, many domestic printing enterprisesstill use the traditional ink preset technology. Preset ink key opening value is preset byoperating workers according to his printing experience before the machine getting towork. This kind of technology is not only poor efficiency but also complex in actualoperation. And the good work experience of operating workers is necessary. So it cannot adapt to the high speed and efficiency of modern printing.Using artificial neural network to ink preset, which takes into account that it hasmore the samples, and the samples data is also great differences in the practical appli-cation, using a neural network can’t realize incremental learning of input data, andpoor convergence, predict accuracy is also low, therefore, this paper according to theinput vector and certain combination rules, the whole sample into a number of sam-ples of independent son and establish the multiple baryons BP neural network, namelythe Fuzzy ART-BP hybrid neural network, and I have completed this the system byC++Builder platform. This technology introduces the Fuzzy ART (Fuzzy Adaptiveneural network) and BP (Back Propagation) neural network integrated applicationmethod. The nonlinear mapping relationships are established between the digital andpicture information of print, the opening of the ink key value and it’s influencing fac-tor. the influencing factor of opening of the ink key value are considered comprehen-sive and the ink amount of each ink area are preset accurately, so the quality and effi-ciency of printing is improved.The good effect of this technology is demonstrated by the experiment. In the ex-periment, the boot adjust time is shorted, the waste of human and material is reducedand the quality and efficiency of printing is improved. In addition, this technology ismore suitable for the big or medium-sized enterprise which has mass data of printthan the system based on BP neural network, because of its ability of adaptive identi-fication of input data. It improve the prediction accuracy of ink preset technology,better connected the prepress and printing, and improve the automation degree ofprinting process.