In this page you can see the project’s latest News
Check out to our recent publication!
Our recent paper entitled «Self-tuning Model Predictive Control for Wake Flows,» has been published in the Journal of Fluid Mechanics. In this work we apply model predictive control with automatic hyperparameter optimization and robustness to noise to control wake flows. For further details, check out the following link.
Check the contribution of Prof. Stefano Discetti to VKI-ULB Lecture Series of Machine Learning for Fluid Mechanics
From January 29th to February 2nd, 2024, we had the opportunity to participate in the VKI-ULB’s machine learning lecture series. The event explored the integration of machine learning methods in fluid mechanics, focusing on analysis, modeling, control enhancement, and deeper insights into Fluid Dynamics. Prof. Stefano Discetti showcased his expertise in model predictive control in a lecture where the participants learned about the advantages of Model Predictive Control in Fluid Mechanics applications.
___________________________________________________________________________
New preprint now available in arXiv!
Manifold learning is a powerful tool to unravel the underlying structure of complex datasets. In our recent study, we show that it is possible to automatically distill actuation manifolds of wake flows, i.e. manifolds describing the behaviour under control actions!
The method employs isometric mapping as an encoder and k-nearest neighbour regression as a decoder. It is validated using the fluidic pinball system, as an established benchmark for flow control. The approach discovers an easily interpretable five-dimensional manifold. Read more about the paper here.
___________________________________________________________________________
Dissemination activity STEM for Girls UC3M
We are extremely pleased to have participated in the STEM for Girls event with a workshop entitled «Unveiling the charms of controlling water». We presented an exciting experiment at our water tunnel facility. Led by Alicia Rodríguez, Andrea Meilán, Luigi Marra, Stefano Discetti and Juan Alfaro, our team explained the fascinating theory behind fluid-flow control in a brief and interactive presentation. We are immensely proud to have contributed to such an engaging and educational experience, fostering curiosity and interest in STEM fields among young minds. Stay tuned for more updates on our research endeavors and future outreach initiatives!
___________________________________________________________________________
Dissemination activity for 2023 Science and Innovation Week
Our team members participated in «Science and Innovation Week» and contributed with a wonderful experiment in our water tunnel facility. Luigi Marra, Stefano Discetti, Andrea Meilán, Alicia Rodríguez and Juan Alfaro, explained the theory behind the fluid-flow control to all the attendants in a brief and interactive presentation, pointing out the importance in real life of doing this research. The experimental activity consisted of visualizing the flow around a triangular arrangement of three rotating cylinders, called fluidic pinball. Kids and families were able to control the flow and do some magic, by moving the cylinders immersed in water and simulating the different configurations given during the presentation. In the end, all of them were able to visualize stunning images.
___________________________________________________________________________
PREDATOR at the 2nd Spanish Fluid Mechanics Conference
Our team member Luigi Marra had the opportunity to participate in the 2nd Spanish Fluid Mechanics Conference, organized in Barcelona from July 2nd to July 5th, 2023. Luigi presented his work (co-authored by Andrea Meilán Vila and Stefano Discetti) on self-tunable MPC,...
Luigi Marra participates in Math 2 Product Conference
Last week our team member Luigi Marra participated in the "Math 2 Product" conference (M2P 2023) in Taormina, Sicily. He gave a talk on an innovative control framework based on Model Predictive Control (MPC) for fluid flows that can automatically select its parameters...