Nonlinear parameter estimation for real-time analytical distillation models

Analytical process model based control is already a practical reality. The design of such controllers requires model parameter estimation for the nonlinear analytical process description, both under steady-state and dynamic conditions. Steady-state models of different degrees of complexity were examined and compared with experimental data from a pilot scale distillation column. While it is desirable to use a rigorous transport phenomena type model for understanding distillation dynamics, reduced order models offer several advantages. For real-time work, the use of dynamic optimization based on transient data provides a better scheme of parameter estimation. This algorithm, when used with the semirigorous model, provided the requisite speed and flexibility of utilizing noisy data without significant loss of accuracy. Nonlinear model predictive control of the top product composition of a pilot scale distillation column, using the proposed parameter estimation procedure, is cited as an example.