Modeling and inferential control of the batch acetylation of cellulose

Abstract

The batch acetylation of cellulose involves two main stages, acetylation followed by hydrolysis. Mathematical models were constructed for these processes using reaction parameters obtained front earlier publications. The two models were then combined, along with a model predicting the cellulose feedstock moisture content. The overall temperature profiles as well as the final degree of acetylation and polymerization were shown to match the data obtained from a commercial process. Front the results of the model and the observations made in a typical process, it has been established that this process is highly sensitive to disturbances in the input streams that call result in unacceptable final properties of the batch. Because the exact disturbance occurring in a batch cannot be identified and measured, inferential control is used to formulate an intrabatch control strategy. Support vector regression is used to make predictions on the final properties of the batch. These predictions, along with the model of the hydrolysis stage, are used to estimate the inputs that, when applied during hydrolysis, call counter the disturbances caused during the acetylation stage. Simulations show that this hydrolysis control strategy can improve the average batch properties of cellulose acetate required for subsequent processing. (c) 2006 American Institute of Chemical Engineers