Adaptive control of rotor vibration using compact wavelets

This paper investigates the use of dyadic wavelets for the control of multifrequency rotor vibration. A scheme for real-time control of rotor vibration using an adaptive wavelet decomposition and reconstruction of time-varying signals is proposed. Quasi-periodic control forces are constructed adaptively in real-time to optimally cancel vibration produced by nonsmooth disturbance forces. Controller adaptive gains can be derived using a model-based synthesis or from system identification routines. The controller is implemented on a flexible rotor system incorporating two radial magnetic bearings, with standard proportional-integral-derivative control employed in a parallel feedback loop for rotor levitation. An experimental investigation of controller performance is used to deduce the best choice of wavelet basis for various operating conditions. These include steady synchronous forcing, step changes in synchronous forcing and multifrequency forcing produced by a rotor impact mechanism.