[Introduction]
Keeping the climate uniform at low costs is essential for the storage of food inside a storage room or container. Hence, at every place in the storage facility the temperature must be close to a prescribed value. For the ultrafiltration of water it is important to guarantee that the ultrafiltration is uniform. Hence, every fraction of water should be equally purified, again at low costs.
These two problems may seem to differ a lot. However, from a mathematical control point of view they have great similarities. Namely, in both applications one wants to identify and control the flow and dispersion inside a closed room in order to satisfy the user demands, and from existing physical theory one knows that the mathematical models for above problems will be similar. [Research]
The above motivates the combined study of both problems, as this project proposes. The aim of the project is first to gain a better system theoretical insight into systems in which flows and dispersion play a dominant role. Secondly, we aim to develop a number of algorithms for identification and control of these systems.
Specific attention will be given to the design of on-line algorithms (software sensors) for the identification of these systems from experimental data. Because distributed parameter systems are mostly ill-posed estimation problems due to the limited number of measurement points, additional information is used in the form of assumptions about state/parameter values. This so-called regularization approach will be further investigated for the specific applications in this study. Furthermore, recently so-called particle filters (or also called ensemble filters), have been introduced which avoids the laborious linearization step in classical Extended Kalman filtering. Nowadays, it receives more and more attention in data assimilation studies of very large scale systems, as meteorological/hydrological/geophysical systems. In this study especially the ensemble filtering approach, combined with regularization, will be further worked out for the specific applications.
Finally, we want to demonstrate the practical viability by applying the algorithms to a pilot or full scale plant. |
