Stabilization of Neural Network Models for VIV Force Data Using Decoupled, Linear Feedback

The hydrodynamic forces on an oscillating circular cylinder are predicted using neural networks under flow conditions where Vortex-Induced Vibrations (VIV) known to occur. derived network approximators then incorporated in a dynamical model that allows prediction of the motion given and initial conditions. Using experimental data, minimum-least-squares compensator is tuned includes linear stiffness damping su-perimposed with constant force offset. decoupled, i.e., equations in-dependent for each degree freedom. By applying simulated experiments can be performed. These show which cancels components any bias hydrody-namic effectively stabilizes VIV motion. To support this time-domain analysis per-formed along phase-plane investigations. Maximum Lyapunov exponent also shown.