Self-tuning model predictive control for wake flows

 

Description:  In the article, the application of model predictive control (MPC) is proposed, wherein control parameters are not manually selected but identified through a Bayesian optimization method. The selected control plant is the chaotic wake of the fluidic pinball at Reynolds number Re_D = 150. A plant model is obtained using the SINDYc technique. Control is applied for drag reduction and lift stabilization, with the challenge posed by the presence of measurement noise in the control sensors. The effect of noise is mitigated using the Local Polynomial Regression technique.

 

Authors: Luigi Marra, Andrea Meilán-Vila, and Stefano Discetti.

DOI: 10.1017/jfm.2024.47

 

Code repository: https://github.com/Lmarra1/Self-tuning-model-predictive-control-for-wake-flows

 

Funding: The authors acknowledge the support from the research project PREDATOR-CM-UC3M. This project has been funded by the call "Estímulo a la Investigación de Jóvenes Doctores/as" within the frame of the Convenio Plurianual CM-UC3M and the V PRICIT (V Regional Plan for Scientific Research and Technological Innovation).