Reducing the defaults caused by the color differences between the laboratory and the mill in the dyehouses

Recently, economic, ecologic and on-time delivery criteria for the dyeing industry have come forward in parallel with the latest quality definition and customer demands. Particularly with the reactive dyeing, getting the right color at the first shot and the repeatability of the color are very important as increasing competition in the market is forcing every company to move fast. After the customer approves the color checking the specimens dyed with the color prepared in the laboratory according to the customers demand, mass dyeing is achieved for domestic and especially for foreign sales. But it is possible to see color differences between the mass production and the specimens dyed.Classical dye recipes are generated by computerized color measuring equipments which are working according to Kubelka-Munk theory (Beer Law) and differences in color are observed when dyeing is achieved according to these recipes in the mill. In systems based on Kubelka-Munk theory, a lot of factors that break linearity between concentration and absorption are not taken into consideration and this causes error risk to increase in recipe prediction. In recent systems like artificial neural networks in which a lot of factors that break linearity are taken into consideration, error risk becomes to decrease.In this thesis color recipe prediction was done by using artificial neural networks and fuzzy logic and the results were compared. Besides, experiments were done in laboratory to investigate the results of diffraction at various factors in laboratory conditions.