The modeling and control of freeze dryers

Two models of the freeze drying process were constructed and used to study various operational policies in order to determine the policy that would provide the shortest drying times. The sublimation model which accounts only for the removal of frozen water, is found to predict shorter times for the free water removal phase than the sorption-sublimation model which properly accounts for the removal of frozen and bound water. This study shows that the policy which produces the shortest free water removal phase, also produces the shortest overall drying time. This policy, predicted by both models, holds the chamber pressure at its lowest value, heats the upper surface by radiation and the lower surface by conduction, the heating plates operate at different temperatures such that the melting and scorch constraints are both encountered and held during free water removal phase, and uses a low condenser temperature. During the free water removal phase, at least 80% of the total amount of heat supplied to the sample, is transferred through the frozen layer. The sorption-sublimation model provides sorbed water as well as temperature profiles in the dried layer during the free water removal phase and during the terminal drying phase. This study, by developing bound water profiles, incorporates an important factor for operational policies for quality retention in freeze drying. The temperature and sorbed water data predicted by the sorption-sublimation model during drying, can be combined with kinetic data on product deterioration to determine operational policies which may minimize not only the drying time but also the deteriorative changes. The economics of the freeze drying process were also studied, and it is shown that the policy that produces the shortest drying time, will operate the dryer most economically for a given sample thickness. The economic analysis suggests that large sample sizes would increase dryer capacity, and models which account for the removal of frozen water only, predict erroneous economic results --Abstract, pages iii-iv