System Identification of AMB Systems: Step Towards Automated Commissioning

Active magnetic bearings (AMBs) are mechatronic devices that provide contact-free support to rotors via electromagnetic forces. An accurate model of the system is necessary to design and evaluate controllers for reliable operation. In this paper, identification of AMB systems is studied with the aim of reducing user involvement in the model identification process. The assumed a priori information consists of the amplifier model, dynamics of AMB electronics, and time delays due to AD/DA conversions, which are relatively easy to identify compared to the rotor model and AMB force model. It is also assumed that there exists a predesigned controller that stabilizes the system, not necessarily a performant one, in order to conduct system identification experiments. To obtain frequency response data for identification purposes, three common excitation signals for rotor systems are considered: impulse signal, PRBS (Pseudo-Random Binary Sequence) signal, and stepped sine signal. Feasibility of using each signal along with advantages and disadvantages over each other in obtaining accurate data in the context of commissioning is discussed. The identification problem is cast as a nonlinear least square (NLS) optimization problem using a parametrized model of an AMB system. The resulting model is physically interpretable, which allows defining uncertainties for AMB force constants and flexible mode frequencies that are standard for the model-based controller design strategies. The presented identification procedure is applied to an experimental AMB rotor system to validate the approach. A signal-based H∞ controller is designed based on the identified model to show the applicability of the presented method for commissioning of AMB systems. Performance of the system with H∞ controller is compared to an experimentally-tuned PID controller to demonstrate the superiority of a model-based controller, hence the importance of having an accurate plant model.