Abstruct-Unbalance of magnetically levitated rotors causes undesirable synchronous vibrations which may lead to saturation of the magnetic actuator. To avoid this problem we propose a generalized narrow-band notch filter which is inserted into the multivariable feedback without destabilizing the closed loop. The parameters of the generalized notch filter strongly depend on the inverse sensitivity matrix evaluated at rotational speed. This matrix is easily measured a priori and stored in a look-up table. It is shown that a decentralized notch is feasible for weakly gyroscopic rotors. The proposed notch filter approach has advantages in terms of runtime complexity and analytical verification of closedloop stability. Results from the implementation of the proposed unbalance compensation in industrial magnetic bearing systems NOMENCLATURE Complex unit J-7.Vector of unbalance compensation signals.Vector of filtered sensor signals. Vector of sensor signals (displacement). Vector of synchronous unbalance forces. Rotational speed, notch center frequency. Complex frequency variables. Real parameter matrices of the notch filter.Complex notch parameter matrixTR + jT J .

No abstract

Abstruct-Active magnetic bearings (AMB's) are an example for industrial applications of digital signal processors (DSP's). The complex control functions involved include multichannel control in a delay-time sensitive system. DSP's are well suited for this challenging application. Hardware and software issues are discussed. Unbalance compensation and other additional functions of the controller are presented. An industrial milling spindle with AMB's and DSP control is discussed in detail. A section on AMB research projects with DSP control indicates future trends.

No abstract

The present paper proposes a refined nonlinear axial bearing model not too complex for real-time applications, which may potentially improve the estimator quality of modulation-based self-sensing. Measurements show that the AC component (ripple) of the bearing current is an underlinear function of the AC excitation voltage, which causes the bearing admittance amplitude |Y(jω, i ac)| to considerably increase at low amplitudes of i ac. The non-consideration of this effect might explain the poor dynamic performance observed in former implementations of modulation-based self-sensing. The model proposed in this paper includes a dynamic (rate dependent) hysteresis model of the Bertotti type, where the magnetic H field is a nonlinear function of the time derivative of flux density dB/dt. Eddy currents are modelled by a linear lumped RL ladder network (Cauer type). The proposed model gives encouraging results and is able to reproduce the observed phenomena.