Keywords: active vibration control, rotor, rotating machine, model reduction, magnetic actuator.

This paper promotes an active vibration control concept for rotating machines. The concept proposes an auxial electromagnetic actuator for controlling the resonance vibrations of heavy rotors supported by conventional bearings (journal or roller bearings). In particular, the applications that are not potential active magnetic bearing applications, are considered. The experiments performed on a smaller-scale test environment encouraged us to continue the work and to inspect larger-scale machines.

The present study was numerical; its objective was to approximately evaluate the above mentioned vibration control concept applied on an electrical machine. Another objective was to outline a design procedure for an active system. The object was a 4-MW two-pole electrical motor. The motor had super-critical rotor e.g. rotor's nominal rotational frequency was higher than its first bending mode. The motor had a 3000-kg rotor supported by journal bearings at its ends. The actuator was located inside the bearing span of the rotor. The selected location may not be optimal, but realistic taking into account the assembly of the machine. The actuator supports were coupled to the frame of the machine.

The first step in the study was the reduction of the finite element (FE) model of the motor. This was done in order to achieve a compact model that may be used smoothly in simulations evaluating the active control system. Next, the actuator was modelled by using simple equations presented in active magnetic bearing literature. Then, the system was identified using band-limited white noise excitation. The identification was disturbed by other excitations due to machine operation. This would be the case in practice; the excitations due to operation would appear in the output, but the identification algorithm could not observe them as inputs. The disturbances originate from rotor unbalance, bearings, electromagnetic forces etc. Based on these identification results, the feedback and feedforward control systems were designed. The results are shown at four different speeds of rotation: 25%, 50%, 75% and 100% of nominal speed.

The simulated responses without an active control system were used as a benchmark results. In the actively controlled cases, both active control systems were working together. The largest reductions were obtained at lower speeds, because the rotational frequency was close to resonance frequencies.

The active control systems decreased the responses at the frequencies below 80Hz, but caused amplification above. The active control systems created also harmonic excitations due to the non-linear behaviour of the actuator (saturation and slew rate limits).

The active control systems increased the responses at the speed of 75%. This was due to the stability problems with the feedforward controller. The identification algorithms did not describe the phase characteristics accurately enough at this speed. The feedback controller worked normally, but the feedforward controller was about loosing its stability. This behaviour caused an increase in the responses.

The active systems reduced the responses in the rotor and in the bearings. The amount of decrease was dependent on the dynamic amplification of the system. This was found quite natural since the concept was meant for resonance vibration control. The force available may not be sufficient for forced-vibration control. Thus, the active control systems studied were more useful at resonance regions than elsewhere.

The model reduction was effective; the number of DOF of the reduced model was less than 0.5 per mil of the original FE model, and the natural frequencies up to 100Hz were described accurately. The gyroscopic rotor and the cross-couplings due to the journal bearings were taken into account in the reduced model. This was done with Modysol software, because these effects may not be included in the reduced model by using common FE software.

The disturbances due to the rotation interfered the identification procedure used in the study. This behaviour was predicted, because the input and the output of the identification system did not correlate. The behaviour caused the feedforward control to suffer stability problems. These consequences clearly address for advanced identification methods or a way to compensate the effects of the disturbances.

The active control systems worked as expected: the simple feedback controller provided general damping increase into the system, and the feedforward controller compensated the disturbance at the frequency of rotation. The normalised adaptation scheme was chosen because, it provided roughly constant convergence rate trough the operation range. Limited, and non-linear, force production capacity of the actuator created undesirable harmonics in the responses. Prevention of these excitations should be taken into account in the controller synthesis.

This paper showed a path to design an active control system in a rotating machine. Though the procedures shown were relatively simple, the design contains substantial amount of work and practical difficulties to be won.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £135 +P&P)