Interval Global Optimization in Solvent Design

This paper discusses the use of an interval analysis based global optimization approach for the systematic design of ink cleaning solvent (commonly referred to as blanket washes) for lithographic printing. Very often the end-user (for example the printing press operator) has available to them blanket washes from different manufacturers. These blanket washes have to be mixed in different proportions (often by trial and error) until a blend is obtained that is suited to a particular cleaning operation. The trial and error approaches are costly, time-consuming and may not necessarily yield the solvent blend with the desired performance attributes. For example, the solvent blend may have a drying time that is too slow for the intended use. Another major performance objective is minimizing the effect of the solvent blend on the surface characteristics of the blanket. Surface characteristics of the blanket are affected through the swelling of the rubber blanket, which supports the printing paper. It should be noted that depending on the performance criteria and process constraints, a different solvent blend would be obtained. These criteria may vary according to the printer, the blanket wash sources, and whether or not the blanket wash blend would be dispensed by an automated process or by a manual process. Another thing to note is that in newspaper printing, the ink residue on the blanket has to be removed (using blanket wash) after the morning newspaper edition, in time for the afternoon or evening edition.The simultaneous consideration of associated process constraints, property requirements, and environmental restrictions makes the blanket wash design a rather difficult problem. To address this, we present a framework for designing cleaning solvent blends that meet thermo-physical property requirements and environmental restrictions. The resulting mathematical program is a mixed-integer optimization problem involving (a) the selection of solvents from a set of pure component solvents (the discrete problem) and (b) finding the blend composition (the continuous problem).The framework has been used to solve an industrially relevant problem of designing optimal blends for blanket wash applications in the printing industry taking into account solvent power, viscosity and surface tension.