|ABBRACCIAVENTO FRANCESCO||Cycle: XXXIV |
Section: Systems and Control
Tutor: PIRODDI LUIGI
Advisor: FORMENTIN SIMONE Major Research topic
:ADVANCED DATA ANALYTICS FOR DYNAMICAL SYSTEMSAbstract:
Dynamical system identification has represented a crucial research topic in control theory for several decades. In the last years, the spread of Big Data and IoT technologies has introduced a new perspective, leading to more heterogenous systems with complex structures and interconnections. In order to achieve a complete characterization of such complex systems, standard identification tools are no longer powerful enough, if they are not combined with modern big data analytics. As a matter of fact, such analytical tools are needed to examine large and varied data sets in order to uncover hidden patterns and obtain important insights.
The purpose of this thesis consists in merging classical system identification techniques with big data analytics approaches, with the aim of refining the models derived by a physical description (or a-priori knowledge) of the system of interest via parameter optimization, structure selection, additional experiment design and validation tools typical of learning approaches. The outcome of this overall procedure can be useful for control design, filtering and simulation applications.
The effectiveness and the flexibility of the introduced theory will be validated on a wide range of real case studies from different domains, such as anomaly detection in industrial processes, human resources management and quantitative finance using feedback control.