Modeling and Control of an Industrial Distillation Plant Modeling and Control of an Industrial Distillation Plant

M and control of distillation columns is a research area that evokes considerable interest. On the one hand, substantial savings are possible with good controllers, on the other hand, distillation columns are often used as test cases for new methods of model based controller design. Strongly nonlinear behavior and high order are the characteristics that present challenges when designing controllers for this multivariable system. Many case studies deal with well-known binary systems that may even show ideal behavior for their vapor-liquid equilibrium. The industrial reality, however, also uses distillation columns for separating highly non-ideal multicomponent mixtures, using complicated column sequences, as it is the case for the system investigated in this thesis. The work presented here concentrates on problems which arise when an existing industrial distillation plant has to be modeled and controlled, and it points out possible solutions. A rigorous dynamic model using the description of physical processes is developed for two columns in series that are coupled by a recycle stream. It is shown that steady-state balances and physical reasoning can be used to check the available plant data. For making simulated column profiles match the measured profiles, a new solution is used: Temperature dependent vaporization efficiencies model the non-ideal behavior of the second column. This leads to a flexible description that is able to match simulation results with plant measurements in different operating points. The model is linearized numerically. Because of the size of the system (1586 state variables, 260 algebraic variables), the columns are linearized individually, leading to models that show good agreement