We will continue with our Seminar Series on Fri, 29 April 2022.
On site: Salón de Grados Leganés
Andrea Iannelli is currently a postdoctoral researcher with the group of Prof. Roy Smith in the Automatic Control Laboratory at ETH Zurich. He was born in Ascoli Piceno (Italy), and completed the Bachelor (2011) and Master degrees (2014) in Aerospace Engineering at the University of Pisa (Italy). During the master studies (2013), he was a visiting researcher at San Diego State University (US), where he worked on fluid-structure interaction solvers for aeroelastic analysis of unconventional aircraft configurations. In April 2019 he completed his PhD at the University of Bristol (UK) in the TASC research group, funded by the H2020 project FLEXOP. During his PhD, Andrea focused on the reconciliation between robust control modelling and analysis techniques (Linear Fractional Transformation, structured singular value, Integral Quadratic Constraints, Dissipativity) and dynamical systems approaches (bifurcation theory, numerical continuation), with application to the study of dynamic instabilities in uncertain aeroelastic systems. His basic research interests include: robust control; system identification; data-driven control theory; optimization; and dynamical systems theory. During his PostDoc he has been developing and demonstrating theoretical and practical advances in these subjects with particular emphasis on the use of data to make reliable predictions and decisions. The interplay between robust control and system identification techniques on one hand, and more recent learning-based algorithms on the other, is central in his research, owing to its potential to provide principled approaches for addressing complex engineering problems. Of special interest is the design of intelligent autonomous systems for a sustainable society, especially in the fields of energy and transportation systems and industry 4.0: e.g. sustainable aerospace systems; industrial robots; energy management in smart cities.
“The Balanced Mode Decomposition Algorithm for Data-Driven LPV Low-Order Modelling of Aerospace Systems”
The talk presents the Balanced Mode Decomposition (BMD) algorithm, a novel approach for data-driven model order reduction of dynamical systems having time-varying properties. The working assumption is that the only information available on the system comes from input, state, and output trajectories obtained, for example with a high-fidelity simulator, at different operating points. The goal is to obtain an input-output low dimensional linear model which approximates the system across its operating range. Time-varying features of the system are retained by means of a Linear Parameter-Varying (LPV) representation made of a collection of state-consistent linear time-invariant reduced-order models. This is achieved by combining the problem formulation of the Dynamic Mode Decomposition (DMD) algorithm with the concept of balanced truncation. Specifically, the BMD formulation hinges on the idea of replacing the orthogonal projection onto the Proper Orthogonal Decomposition modes, used in DMD-based approaches, with a balancing oblique projection constructed from data. This has a twofold benefit: the input-output information captured in the lower-dimensional representation is generally bigger compared to other projections onto subspaces of same or lower size; a parameter-varying projection, more accurate than parameter-independent ones adopted in other recent works, is possible while also achieving state-consistency. This fully data-driven approach has prospective applications in various aerospace domains where: LPV models strike a better balance between accuracy and complexity than fully nonlinear or linearized time-invariant models; and low-order models are required (e.g. control design, online detection and decision making). Results are shown on a morphing wing for Airborne Wind Energy applications by comparing BMD against two recent competitor algorithms. Clear improvements are registered both in terms of prediction accuracy and closed-loop performance when the models are used for Model Predictive Control.
The talk is based on the following publications:
- Iannelli, Fasel, Smith – “The Balanced Mode Decomposition Algorithm for Data-Driven LPV Low-Order Models of Aeroservoelastic Systems”. Aerospace Science and Technology, vol. 115, 2021. Link
- Iannelli, Fasel, Yogarajah, Smith – “A Balanced Mode Decomposition Approach for Equation-Free Reduced-Order Modeling of LPV Aeroservoelastic Systems”. AIAA SciTech Forum 2021. Link
The seminars will begin at 13 CEST and will take place in the Auditorium Salon de Grados (Padre Soler) campus of Leganés.
No previous registration is required.