Numéro : 2805 - Year : 2024
Multi-model and machine learning for pitch optimization and control of a cycloidal propeller
Frédéric HAUVILLE, Florent BECKER, Naval Academy Research Institute (IRENav), BCRM Brest, CC600, 29240 Brest Cedex9, France
Matthieu SACHER, Guillaume FASSE, ENSTA Bretagne, CNRS UMR 6027, IRDL, 2 rue Francois Verny, 29806 Brest Cedex9, France
Grégory GERMAIN, IFREMER, LHyMar Laboratory, 62200 Boulogne-sur-Mer, France
The operation of a transverse-axis marine propeller is very different from that of conventional propellers or axial turbines. It is characterized by the rotation of several blades around a vertical axis, combined with the movement of each blade around its own axis. The advantage of these systems is that they generate a 360° vectorial thrust. The elaborate blade kinematics produce a horizontal thrust whose hydrodynamic efficiency is highly dependent on the blade control pitch law. IRENav has been interested in these vertical axis propellers for many years, and has designed an extensively instrumented, large-scale (0.8-1.6m diameter) SHIVA experimental platform, which can generate all forms of kinematics by means of independent auxiliary electric motors that ensure instantaneous pitching of each of the blades around their quarter-chord axis of rotation, synchronized with the main electric motor. Preliminary studies carried out in recent years by the Ecole Navale, ENSTA Bretagne and Ifremer seem to indicate significant and unprecedented potential for improvement in hydrodynamic efficiency. It is well known that this type of rotor features highly original pitching blade kinematics that are complex to optimize. Indeed, for both operating modes, the pitch law depends on the advance parameter, geometric parameters and the desired strategy. After a brief presentation of the SHIVA platform and a description of the meta-modeling optimization method used to optimize the performance of the vertical axis propeller, some optimizations carried out in December 2023 are detailed and analyzed.
Advanced search
Year