(507d) Integration of Controllability and Diagnosticability of Reactive Distillation Column As From Conceptual Design Step. Application to the Production of Ethyl Acetate

Increasing attention has been directed towards reactive distillation processes as a successful process integration example. Reactive distillation means the simultaneous implementation of reaction and distillation in a counter currently operated column, where chemical equilibrium is superimposed on vapor liquid equilibrium. Conversion can be increased far beyond what is expected by the equilibrium due to the continuous removal of reaction products from the reactive zone, reducing capital and investment costs and contributing for sustainable development due to a lower consumption of resources. Reactive distillation process is although a complex system; the combination of separation and reaction zones leads to strong interactions and nonlinearities, hindering the process control and diagnosis. When the conceptual design is optimally obtained based on a static analysis, regarding only economics and environmental criteria, the result may be a unit very difficult to control. It is thus appropriate to consider controllability and diagnosticability criteria as from conceptual design steps.

An entire methodology for the conceptual design of a reactive distillation column that considers controllability and diagnosticability aspects is proposed and applied to the production of ethyl acetate from esterification of acetic acid and ethanol.

A tool of feasibility and pre-design of reactive columns has been developed in the LGC (Laboratoire de Génie Chimique, Toulouse, France) based on a sequential approach with development and application of different types of software (Thery et al. 2005). First, a combination of algorithms developed in Fortran and MatLab^® connected to a thermodynamic properties server (Simulis^®) is used for feasibility and synthesis calculations. Then, the column configuration is defined in Aspen Plus^®and process simulation at steady state provides more operational parameters, by considering the heat phenomena. The main interest is that these tools lie on sequential and progressive introduction of process complexity; from a minimal set of information concerning the physicochemical properties, successive refinements are considered. The techniques applied on the feasibility analysis, the synthesis and the conceptual design steps lead to the project of a column configuration and the necessary operating parameters to attain the process objectives. Different configurations may be feasible; so that the engineer should choose the best configuration according to his own aims (lower investment, lower operating costs, better operability, etc.).

The objective now is to choose the column with best operability and to identify the structural parameters that provide the best controllability. By considering an indices-based analysis method, some criteria to quantify and qualify the system controllability are identified at early design steps. The first steps of the controllability methodology are obtained by steady state simulations, with the calculation of the Sensitivity Matrix, the Singular Value Decomposition, the Condition Number and the Intersivity Index (Moore 1992). Then, the dynamic simulation is studied, by considering the properly chosen control loops with the Integral of the Squared Errors calculations. At this moment, the importance of an experimental validation is highlighted so as to provide realistic hydrodynamic parameters, to understand the sensitive parameters such as heat losses and to adapt values for the catalyst behavior in function of the system.

Conventional distillation columns have two degrees of freedom, so two different variables have to be controlled to ensure good separation and the respect of production quality. In the case of reactive distillation columns, a desired conversion rate is also of control interest.  The additional degree of freedom given by the feed ratio on a double-feed column configuration allows the introduction of a third control loop with a measured variable inside the reaction zone that addresses well the reaction conditions. The three control loops are defined so as to show high sensitivity, good balances and small interferences. The application of the methodology for the production of ethyl acetate identifies the system specific sensibilities and the best control configurations. It is found that the addition of theoretical stages between the two feed positions ameliorates the column controllability and that homogeneous catalysis results on more reliable operability than heterogeneous catalysis.

Once the column configuration is proposed and the process dynamic simulation is mastered, the objective of process diagnosis is to rapidly detect abnormal operating conditions so as to avoid reduction on its performance and to promote proper operation with more security to its surroundings. The classification technique LAMDA (Learning Algorithm for Multivariable Data Analysis), based on fuzzy logic and developed at the LAAS (Laboratoire d’Analyse et d’Architecture des Systèmes, Toulouse, France), is considered and the calculations are made with the software P3S^®(Kempowsky et al. 2003). The simulation data with regard to potential operating faults lead to a learning procedure, when the knowledge of human experts is included; the algorithm is applied and the historic measures are differentiated into several classes, which are to be interpreted as physical operating situations by the human expert. After the learning procedure, a recognition step is lead: the data is analyzed, the classes already established can be identified and the most relevant sensors are identified automatically by the Membas procedure based on their ability of optimizing the fault discrimination (Molina 2005). The composition sensors were verified to give strong information of the process diagnosis. However, the complexity on using online composition analyzers can be overcome by the use of a higher quantity of temperature sensors to give the same situation recognition degree. The diagnosticability analysis at the design phase of processes is important so as to preview the installation of relevant sensors at specific column locations. Some sensor placements are not always possible when the column has a fixed structure and the lack of measures would jeopardize the system operation.

In conclusion, some indicators of controllability and diagnosticability have been identified to be considered as from the design step of reactive distillation columns. The work results on important contributions to the study of the ethyl acetate production by reactive distillation, due to the fact that few studies address the dynamic behavior of this system.

References:

Kempowsky T., Aguilar-Martin J., Subias A. et Le Lann M.V. Classification Tool based on Interactivity between Expertise and Self-Learning Techniques, IFAC-Safeprocess, Washington D.C., USA, 2003.

Molina A. O. Méthodologie pour le placement des capteurs a base de méthodes de classification en vue du diagnostic. PhD thesis, Laboratoire d’Analyse et d’Architecture des Systèmes du CNRS, France, 2005.

Moore C. Chapter 8, Practical Distillation Control. Van Nostrand Reinhold, edited by William L. Luyben, 1992.

Théry R., Meyer X. M., Joulia X. et Meyer M. Preliminary design of reactive distillation columns. Chemical Engineering Research and Design, 83 (A4), 379, 2005.