A model-free learning approach for active damping of a rotating cylindrical shell using ADP control with rotating-frame state reconstruction technique

Abstract

This study presents a novel approach for achieving model-free active vibration control of rotating cylindrical shells, addressing the challenges posed by incomplete knowledge of physical parameters and model structure. A technique exploiting rotational symmetry is introduced to reconstruct fixed-frame state variables from surface strain measurements, wherein the issue of sensor placement optimization is also addressed. An optimal control methodology based on adaptive dynamic programming (ADP) is then applied where feedback control solutions are computed directly from input-to-state observation data. To achieve prescribed damping levels for targeted flexural vibration modes, a data-based cost-function selection technique is also proposed. Numerical and experimental investigations involving a steel cylinder with fixed-frame electromagnetic actuator confirm the efficacy of the approach in achieving at least 15 % of critical damping, both with and without rotation. The results show significant promise for real-world vibration control problems, including application in machining operations on thin-walled cylinders.