Intelligent multi-resolution modelling: Application to synthetic jet actuation and flow control

A novel "directed graph" based algorithm is presented that facilitates intelligent learning and adaptation of the parameters appearing in a Radial Basis Function Network (RBFN) description of input output behavior of nonlinear dynamical systems. Several alternate formulations, that enforce minimal parameterization of the RBFN parameters are presented. An Extended Kalman Filter algorithm is incorporated to estimate the model parameters using multiple windows of the batch input-output data. The efficacy of the learning algorithms are evaluated on judiciously constructed test data before implementing them on real aerodynamic lift and pitching moment data obtained from experiments on a Synthetic Jet Actuation based Smart Wing.