Title: Dynamic and Global Optimisation of Processes

Partners

• Slovak University of Technology in Bratislava, Faculty of Chemical and Food Technology, Department of Inform. Eng. and Process Control (M. Fikar, R. Paulen, M. Jelemenský, L. Petáková)
• Institut National Polytechnique de Lorraine (INPL) - Ecole Nationale Supérieure des Industries Chimiques (ENSIC) (M. A. Latifi, M. Daroux, F. Lesage)

Date: 2012-2013

Project Code

APVV SK-FR-0004-11

Objectives, project description, background, proposed research

A majority of chemical, pharmaceutical, and biotechnological products is manufactured in processes that operate in an inherently transient manner rather than at some nominal steady-state. Examples include batch and semi-batch processes, and periodic processes that operate at a cyclic steady state, such as chromatographic and absorption processes. The economic design of these processes is in essence an example of optimal process operation. In particular, batch processes are invariably suitable for a large number of applications in pharmaceutical, biotechnological industries, polymerization, special chemistry and agricultural chemistry, or in production of electronic materials. Even in product design or formulation, which is the new emerging paradigm in chemical engineering, most of the new products are manufactured in batch processes, i.e. transient in nature. All of these considerations motivate research on design procedures that optimize the performance of inherently transient processes, particularly batch processes. The most important aspects of this research are closely tied to economical competitiveness, environmental and process safety. To understand and master these issues, design and optimal operation of dynamic processes is of the utmost importance.

This research project deals with unsteady-state operation of dynamic processes that are described by a detailed mathematical models, typically with non-linear ordinary differential equations. The optimization of performances of such processes consists in the determination of optimal profiles of decision variables (temperature, pressure, flow, heat, ...) or optimal parameter values of the dynamic model which optimize (minimize, maximize) a given performance index (time of operation, yield, energy consumption,...), over a time horizon, under specified constraints (safety, environment, process physical limits,...). This kind of problems are known as dynamic optimization (or open-loop optimal control). Some selected problems include determination of optimal control in batch processes, estimation of optimal kinetic parameters in chemical reactions based on experimental data, optimal input design for parameter estimation, determination of optimal control trajectory during set-point change, security analysis of processes, model based predictive control based on continuous model, etc.

Cooperation between LRGP (ENSIC-INPL, Nancy) and Slovak University of Technology in Bratislava started about 17 years ago. During that time Slovak researchers did research in cooperation with Nancy in the framework of European projects (TEMPUS), post-doctoral programs (CNRS, Elf Aquitaine), programs for high quality foreign researchers (Ministére de la Recherche, Région de Lorraine), and bilateral Stefanik projects. As the result of the cooperation in the present time (since 2005), we mention here ERASMUS mobilities of the students from Department of Information Engineering and Process Control FCFT STU. In 2012, it will be already the sixth diploma work finished at LRGP. The work realized during these stays resulted in development of methods and algorithms for dynamic optimization applied to refinery processes, distillation and membrane processes, small size waste-water treatment plants. The development of a dynamic optimization software DYNO was started. This software is currently distributed for exclusively academic purposes and it is being used in France, Slovak Republic, India, Malaysia, USA (NASA). The development of a package for dynamic optimization within MATLAB environment – DYNOPT has started in 2004. DYNOPT exploits the approach based on simultaneous discretization of both state and control variables using the orthogonal collocation on finite elements. This software is currently distributed for exclusively academic purposes and it is being used in France, Slovak Republic, India, USA, Canada.

Objectives of the cooperation are as follows:

1. to enrich the knowledge already acquired so far by developing rigorous methods of dynamic and global optimization of processes. In fact, it is well known for several years that solutions of dynamic optimization in chemical engineering and process control exhibit numerous local optima. Therefore, due to the non-convexity, of these problems, it is almost impossible to find the true global optimum and the results obtained are only local in nature. This sub-optimality can have serious consequences on economical, environmental operation, safety, and clearly shows the importance of the development of global optimization methods, 
2. to further develop the softwares DYNO, DYNOPT, namely (i) incorporate methods of global optimization, (ii) improve theirs robustness and speed, (c) broaden its application area by testing it on problems from other domains (biotechnology, mechanical engineering, ...) as well as by incorporating the methods for dynamic and global optimization of hybrid process (processes with switching dynamics),
   
3. and finally, to apply methods of global dynamic optimization in experimental control of real plants (batch polymerization processes).