| his work considers the problem of vibration cancellation in magnetic bearings. This problem, and the problem of vibration cancellation for rotating machinery in general, are complex because they require accurate models for two distinct types of vibration: periodic vibration and random vibration. These models can be difficult to develop because vibration data typically consists of a mixture of both types of vibration, and hence, it is necessary to be able to discern between and identify both types of vibration. Moreover, these models must, in general, be developed with no a-priori knowledge. As a result of the rigorous demands upon obtaining system models, no active vibration cancellation algorithms have been developed to date which are capable of handling unknown and arbitrary periodic and random vibration.;In this work, the problem of obtaining distinct models is addressed through a new approach termed "mixed spectrum adaptive cancellation" (MSAC). MSAC utilizes a combination of the minimum variance spectrum and the Levinson-based autoregressive spectrum for modeling. The key to this approach is the minimum variance spectrum which discerns between periodic and random vibration through its unique convergence properties. The minimum variance spectrum allows the direct development of a periodic vibration model and an estimate of the noise correlation sequence. An autoregressive random vibration model can then be easily obtained using Levinson's method in association with Akaike's order selection method. The resulting vibration models are then used to supervise an adaptive, open loop vibration cancellation controller based on a least-mean-square algorithm.;Based on the results of the computer simulations, this technique is shown to be capable of identifying and canceling sinusoids with as little as... |
