Chemical Engineering Journal 88 (2002) 163–174 Application of model predictive control and dynamic analysis to a pilot distillation column and experimental verification M. Alpbaz a,∗ , S. Karacan a , Y. Cabbar b , H. Hapo ˘ glu a a Faculty of Science, Chemical Engineering Department, Ankara University, 06100 Tandogan, Ankara, Turkey b General Director and Chairman of Turkish Sugar Factories Cooperation, Yeni¸ sehir, Ankara, Turkey Received 12 February 2001; received in revised form 26 November 2001; accepted 28 November 2001 Abstract The steady-state and dynamic behaviour of a binary packed distillation column have been simulated using a stagewise approach. The model solutions have been obtained employing modified equilibrium including efficiency and non-equilibrium transfer unit. This work compares an experimental and a theoretical analysis of the steady-state and dynamic behaviour of a packed distillation column using a 0.08m ID pilot-scale tower distilling a mixture of methanol–water. The packed distillation column was divided into four stages based on McCabe–Thiele method. For control studies, reflux ratio was chosen as a manipulated variable, so the effect of the perturbation on reflux ratio to the overhead temperature was examined. Theoretical and experimental results were compared in order to see the validity of the stagewise approximation. The application of two types of model based control system was considered theoretically, viz PID and dynamic matrix control (DMC). First-order plus dead time model and convolution model were used for step test in the control applications. Performance of these control systems were tested using control performance criterion. These control systems were also compared with open-loop dynamic behaviour and each other. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Pilot distillation; Dynamic matrix control; PID 1. Introuction Packed columns are modelled normally by employing one of the two alternative approaches. The first one consists of dividing the packing into various mass transfer sections, each being treated as equivalent to a theoretical stage in a plate column. The corresponding algorithms for stagewise systems can then be applied. This method is of course only as convenient as the ability to obtain an independent char- acterization of the height equivalent of a theoretical plate (HETP). This is a widely used approach, principally due to its ease of application and the considerable quantity of relevant data available in the literature [1]. The second technique involves the use of the average vapour-phase mass fluxes of each component and solving the resulting sets of differential equations directly [2,4]. The Abbreviations: DMC, dynamic matrix control; ID, internal diameter; IV, vapour flow; IMC, internal model control; ISE, integral of the square of the error; LF, feed flow; LI, liquid flow; LO, liquid out; LOK, liquid flow from the small reboiler to the big reboiler; LOB, liquid flow from the big reboiler to the small reboiler; MPC, model predictive control; NP, prediction horizon; NC, control horizon ∗ Corresponding author. Tel./fax: +90-312-223-2395. E-mail address: alpbaz@science.ankara.edu.tr (M. Alpbaz). latter may be achieved by using a finite difference approach, polynomial estimation techniques or orthogonal collocation on finite element methodologies. Karacan et al. [22,23] in- vestigated the steady-state and dynamic properties of a pilot plant packed distillation column experimentally and theo- retically. In the theoretical work, a back-mixing model was adapted to simulate the dynamic properties of the continuous packed distillation column with a thermosiphon reboiler and solved by orthogonal collocation on finite element. These all involve a discretization of the basic continuum problem in which the infinite set of numbers describing the unknown functions is replaced by a finite number of unknown param- eters, and this process, in general, necessitates some form of approximation. The finite difference method is generally the simplest to apply and thus is favoured by many workers. In this work, the models are described as a set of ordinary differential equations in which the height of the column is divided into a number of stages. Peters [10] first suggested the use of the concept of the HETP and Chilton and Colburn [11] proposed the height of transfer unit (HTU) approach in the investigation of the mass transfer operations. Rubac et al. [12] suggested the use of vaporization efficiencies in the modelling of packed columns. These va- porization efficiencies involve an integrating concept which 1385-8947/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII:S1385-8947(01)00304-7