MPC for advanced process control and optimization:
The aim of this line is the development of advanced control techniques that cope with constrained problems, mainly Model Predictive Control (MPC), and that together with the traditional dynamic objectives, take into account additional objectives like managing complex systems, zone control, or economic optimization of large scale processes. This last point copes with important plant objectives like, maximizing the production, reducing costs, controlling prices, etc.
In the last years, the theoretic progress achieved in this line has been remarkable. In particular, control formulations have been proposed, capable of ensuring important properties like stability, recursive feasibility and economic optimality. The drawback of all these formulations is that they can hardly be applied to industrial plants.
The aim of this research line is therefore to re-formulate the theoretical results in literature in such a way they can be easily applied to the industries. That is, designing controllers with low computational burden and simple applicability, taking also into account the theoretic advantages provided by such properties like stability, feasibility, economic optimality, particularly in case of changing economic objectives or operation points.
More details on this research topic can be found here.
In medium and large scale industries, like chemical, food, or steel production, etc., there is a big number of control and controlled variables. The multivariable nature of these processes demands the usage of statistical techniques that make possible to improve them continuously. The data driven multivariate analysis techniques provides the instruments to quickly understand complex systems, without changing them, and at the same time makes possible to evaluate the performance of chemical or biochemical processes and of the controllers involved in them, to optimize processes and products, providing the basis for a decision making layer. It is here where the usage of Process Analytical Technology (PAT) software takes place. PAT is mainly based on multivariate methods like the design and analysis of experiments, batch/ semi-batch /continuous processes monitoring, quality prediction, fault detection and statistical process control.
On-line monitoring of a process is the main instrument that provides Quality Assurance and Validation, which represents a requirement of standards and regulations, such as the FDA regulations of the pharmaceutics industry. Moreover, these tools can be applied to different fields, like biofuel, petrochemical, biotechnology, food and manufacture industries.
On-line process re-identification:
The aim of this research line is to use and propose advanced identification methods that make possible to obtain dynamic models suitable for the model based advanced control methods. Moreover, it is desirable that these methods can be applied without opening the control loop. This is what is known as on-line identification (or re-identification), and sets out the problem of doing at the same time, two a priopri opposite operations like controlling and exciting the plant for identification. In order to obtain a dynamic model of a process, it is necessary a sufficient excitation, for making the involved variables varying sufficiently. There are different strategies that make possible a satisfactory identification. One of them is subspace identification. However, other classic techniques, like for instance mean square identification, can be considered. Maintaining stability, while exciting for identification, is the main objective of this research line. This implies, doing identification without stopping the process, which is highly desirable in process industries.