1

ISABE-2009-1263

Controlled Excitation of Rotor Blades via Macro Fiber Composites in an Axial Compressor
J. Siemann Institute of Turbomachinery and Fluid-Dynamics, Leibniz University Hannover, Appelstr. 9, 30167 Hannover R. Nollau HAWK-Gttingen, Faculty of Natural Science and Technology, Von-Ossietzky-Str. 99, 37085 Gttingen B. Grber MTU Aero Engines GmbH, Dachauer Str. 665, 80995 Mnchen J. Seume Institute of Turbomachinery and Fluid-Dynamics, Leibniz University Hannover, Appelstr. 9, 30167 Hannover

ABSTRACT This paper presents the controlled excitation of rotor blades in an axial compressor for the investigation of blade vibrations through simulated aerodynamic flutter. Three rotor blades of the first stage of a low-speed axial compressor were equipped with piezo-electric actuators (macro-fibre composites, MFC) and strain gages. For experimental investigations, the MFCs are actuated in phase and inverted. Via strain gages, the vibrations of the blades are measured. Gain and phase angle are evaluated and an FFT-analysis is performed. Further investigations with controlled actuation used a controller specifically developed within the present project. For this latter experimental setup, the values of the controlled and the actuating time signals were recorded. The measured results were compared with the output of models of the uncontrolled and the controlled configuration. The investigations showed that with the experimental setup used, a controlled excitation of the blades is possible. Furthermore, an adequate model for the controlled as well as the uncontrolled excitation was validated. Symbols

NOMENCLATURE

A
D D h K & m n p t u V x y

Subscripts

amplitude hub diameter damping blade height transfer factor mass flow rotational speed Laplace operator time voltage amplification control variable actuating variable total pressure ratio phase angle angular frequency

0 Amp BE Bl Br CBM

hub amplifier blade end blade measuring bridge composite blade model controlled system real ... reduced ... reference ... strain gage

CS
real red ref sg

Copyright 2009 by the American Institute of Aeronautics and Astronautics Inc. All rights reserved

2
INTRODUCTION Blade vibration and flutter have an increasing impact on the operation of aircraft engine compressors. This is mainly due to two factors: The demand of reducing engine weight for decreasing the specific fuel consumption leads to more slender blades on one hand and higher pressure ratios per stage on the other hand. The consequences are more flexible blades with higher aerodynamic loading. This in turn leads to stronger vibrations or even flutter, i.e. self-excited blade vibrations due to an interaction between blades and unsteady aerodynamic forces. Additionally, the new blisk designs intensify the problem of vibration due to the decrease of mechanical damping in a blisk rotor. Hence, there is a strong and increasing interest in understanding aeroelastic phenomena. Previous investigations on the subject of rotor blade vibrations and active control have been carried out based on computer simulations, e.g. Montgomery et al. [Mon2005] who gave a good overview of the application of modern CFD tools for predicting unsteady aerodynamics and aeroelasticity. Further investigations were done in wind tunnels where the examined blades were more easily accessible than in a rotation frame of reference, e.g. Belz and Hennings (2004) and Watanabe et al. (2008). The actuation of rotor blades so far was realized mainly by the two following techniques. Christensen and Santos (2004) and (2006) implemented a shaft-based actuation via electro-magnetic bearings. Very recent investigations by Charleux et al. (2006) use piezo-electric actuators and sensors. Their test rig consists of four rotor blades rotating in a vacuum chamber. However, to the present authors knowledge, so far no investigations with active actuated rotor blades have been carried out on real rotating turbomachinery. The aim of this present study is to demonstrate a controlled actuation of rotor blades via piezo-electric actuators (in this case macro-fibre composites, MFCs) in a rotating turbomachine. The actuation is utilized for active vibration control. As a demonstrator machine, the twostage low-speed axial compressor with IGV at the Institute for Turbomachinery and Fluid-Dynamics of Leibniz University Hannover is used. After a successful implementation of the instrumentation, the actuation, and the controller and its exprerimental verification, further investigations are planned on a blisk fan operating in the M2VP facility at the DLR Institute of Propulsion Technology in Cologne. EXPERIMENTAL SETUP The test facility used for the present research is a closed loop low-speed axial-compressor with IGV and two stages, Fig. 1. The main specifications are given in Table 1.

Fig. 1: Drawing of the low-speed axial-compressor (NGAV) Table 1: Main specifications of the two-stage low speed compressor
design speed mass flow

3000 rpm 16.5 kg/s

& red m

total pressure ratio hub diameter blade height

1.035 760 mm 140 mm 20 30 20

D0 hBl

Number of Blades IGV Number of Blades Rotor Number of Blades Stator

The two-stage low speed compressor is equipped with NACA 65-(14)10 profiled blades with a thickness-to-height ratio of 0.043. For transferring measuring signals, the rotor is equipped with a slip ring of eleven conductor paths, see Fig. 2.

Copyright 2009 by the American Institute of Aeronautics and Astronautics Inc. All rights reserved

Fig. 4: Pressure side with strain gages Fig. 2: Slip-ring Besides, the two-stage low speed compressor is equipped with roller bearings and its aerodynamic loading is extremely low. Therefore, flow induced vibrations of the blades as well as the entire machine are rather minor. The actuation of the rotor blades is accomplished by piezoelectric actuator foils glued to the suction side of the blades. Three of the 30 blades of the stage 1 rotor are prepared with these actuators. Strain gages are mounted on the pressure side of the same blades. Each of the three blades is equipped with one full bridge and one single strain gage (metal foil, resistive). Suction and pressure side of the equipped blades are shown in Fig. 3 and Fig. 4. The full bridges are made up of four sensors, with two of them are oriented in the direction of the main strain axis and two perpendiculat to it. Full bridge strain gages were chosen for reducing the sensitivity to interference. A disadvantage of this type of strain gages is that they need more conductor paths on the slip ring than a quarter bridge strain gage to transmit the output signal. For that reason both types of strain gages full and quarter bridges are tested to get a direct comparison of the signal quality. Due to the limited number of conductor paths on the slip ring, not all strain gage signals can be transferred simultaneously. Therefore, only four strain gages, which are sketched in Fig. 5 are used in the present study, whereas mainly the two full bridge strain gages were examined in the present investigations. The designations of the strain gages given in Fig. 5 are also used in the following illustrations.

Fig. 3: Suction side with piezoactuators

Fig. 5: Sketch of used strain gages

Copyright 2009 by the American Institute of Aeronautics and Astronautics Inc. All rights reserved

4
The positions of the strain gages were chosen based on a previously executed numerical simulation of the eigen modes of the blades. Also, because of the limited number of conductor paths on the slip ring, the actuation signal is transmitted over only one conductor path and is split behind the slip ring. So, all piezoactuators are provided with the same amplitude, merely the phase can be shifted by 180 by switching the polarity at the output contacts of the slip ring. For the first investigations determining the real eigen frequencies of the blades and their phase shift, a conventional signal generator is used. It generates a sinusoidal signal oscillating between 0 V and 10 V (stepped sinus). For further investigations, the sinus signal is generated by a control unit which has been developed within the present project. Fig. 6 presents the block diagram of the control unit. The input values to this control unit are the measured control variable x and the angular frequency . The control variable gets separated into the amplitude Areal and the phase angle real in the signal processing. With the resulting functions

f1 ( Areal , real ) = wcos x

Areal cos( real ) 2 A f 2 ( Areal , real ) = wsin x real sin(real ) 2

the amplitude Areal and the phase angle real are:

Areal = 2

f12 + f 22

and

real = arctan

f2 f1

In combination with the reference values Aref and ref, the signal gets processed by two PIcontrollers. The output amplitude signal yA* is limited by a subsequent limiter, hence an overcharge of the high-voltage amplifier is prevented.

FMFC
sin ( t + y )

f1 (x1 , x2 ) f 2 (x1 , x2 )

sin( t + ) w =U

sin ( t )

cos ( t )

Fig. 7: Sketch of the measuring chain without (black) and with the control unit (red)B For the measurement analysis, the relative amplitude and the phase shift between actuating voltage and strain gage signals are determined and Fast Fourier Transformations (including Hanning windowing) are performed.

Fig. 6: Flow chart of the developed control unit, x being the control variable and y the actuating variable

Copyright 2009 by the American Institute of Aeronautics and Astronautics Inc. All rights reserved

5
MODELLING Before experiments on the rotating machine, simulations were done. Therefore, the complete data acquisition was analysed and modelled. The frequency of 327 Hz which was determined st via numerical simulation as the 1 flap eigen frequency. The model for the opened control loop is given in Fig. 8. Its total static transfer factor from the input to the output was determined to be 0.012. The appropriate relative amplitude and the phase shift are shown in Fig. 9. occurs according to the displacement of blade 2. The controller works is according to Fig. 6.

2 2 1+ 2DCBM2TCBM2 p + TCBM 2p

2 2 1+ 2DCBM3TCBM3 p + TCBM 3 p

VAmp

(1+ T
KCBM 2 1+ 2DCBMTCBM p +TCBM p2
VAmp

Amp1

p ) (1 + TAmp2 p)

(1+ T

Amp1

p) (1 + TAmp2 p )
Fig. 10: Quantitative model of the closed loop

Fig. 8: Model of the opened control loop

The time signal of the modelling of this closed control loop is shown in Fig. 11. The actuating variable leads by about 90. Additionally, the model showed that the control loop needs nearly 1 sec to reach steady state.
phase normal actuating value controlled value

Fig. 9: Relative amplitude and phase shift for the model of the opened control loop The model of the closed loop is presented in Fig. 10. The control Fig. 11: Time signal of the closed loop modelled at the eigen frequency of blade 2

Copyright 2009 by the American Institute of Aeronautics and Astronautics Inc. All rights reserved

2 2 1+ 2DCBM1TCBM1 p + TCBM 1 p

Kcs2 /VAmp

Kcs3 / VAmp

Kcs1 / VAmp

6
EXPERIMENTAL RESULTS In the first investigations, the characteristics of the real opened control loop were determined and compared with the results of the model described above. For this purpose, measurements with two full-bridge strain gages on blades 1 and 2 were taken. These measurements were performed for rotational speeds of 2000 rpm and 3000 rpm. All three blades which are equipped with piezo-electric actuators are actuated in phase as well as inverted. Further experiments for comparing the full-bridge strain gages with the quarter bridges were performed in the non-rotating machine. In Fig. 12, the relative amplitude and the phase shift of the full-bridge strain gages at 2000 rpm and in-phase actuation are shown. It is peculiar, that the eigen frequencies of the two blades diverge. Whereas the eigen frequency of blade 2 is at 326 Hz with a relative amplitude of roughly 0.25 and therefore agrees well with the model, the eigen frequency of blade 1 deviates strongly from the model with 312 Hz and a relative amplitude of 0.17. The phase shift changes by 180 while running through the eigen frequency of the corresponding blade, as expected. With increasing the rotational speed to 3000 rpm, the eigen frequencies of both blades increase by approximately 20 Hz, Fig. 13. For blade 1, this results in an eigen frequency of 336 Hz, up by 24 Hz. The eigen frequency of blade 2 increases by 18 Hz up to 344 Hz. The relative amplitude at the eigen frequency rises with increasing rotational speed as well. The relative amplitude of blade 1 rises to approximately 0.33 and for blade 2 it increases roughly to 0.24. The phase shift is nearly the same and it occurs at the eigen frequencies.

Fig. 13: Amplitudes and phase shift at 3000 rpm Basic cause for the different eigen frequencies of the blades are geometrical variations which arise from manufacturing tolerances and slight differences in assembly and thus clamping conditions. By comparison with the modelled phase shift a difference of 120 is observed. This phase shift originates from the measuring signal amplifier. It is a systematic offset not considered in the model. Furthermore, repeated measurements showed that the eigen frequencies of one blade range at the same rotational speed vary by about 2 Hz. This is founded to be due to the blade clamping, which changes a little with every startup of the machine. By analysing the measurements with an additional quarter bridge
Copyright 2009 by the American Institute of Aeronautics and Astronautics Inc. All rights reserved

Fig. 12: Amplitudes and phase shift at 2000 rpm by equiphased actuation

7
strain gage at blade 2 (Fig. 14), the relative amplitudes of the quarter bridge are nearly 25% of the amplitudes of the full bridge. Phase shift and eigen frequency take on the same values. used the full-bridge strain gage signal of blade 2. A comparison of the controlled actuation in Fig. 15 with the modelling results shown in Fig. 11 shows good agreement. The amplitudes have the same values and the phase shift between control and actuating value amount to approx. 90, as well.
phase normal actuating value controlled value

Fig. 15: Time signal at 347 Hz with closed control loop

CONCLUSION AND OUTLOOK The first measurements with the opened control loop yield the conclusion that the eigen frequency and the amplitude highly depend on the clamping condition of the blade. Hence, it is assumed that the shift in eigen frequency and the increase in amplitude are caused by a changing of the clamping condition in the machine when increasing the rotational speed, as previously reported by Kielb and Abhari (2003) in their turbine study. Variations in the eigen frequencies between the different blades are caused by manufacturing tolerances as well as by the clamping conditions. The investigations with the closed control loop show that a control of the blade flutter via the developed control unit developed is possible. The control unit showed a stable control mode by controlling based on the fullbridge strain gages as well as on the quarter bridges. Both systems, controlled and uncontrolled, can be completely visualized and simulated by the model developed. At the DLR in Gttingen further investigations will be done at a blisk rotor. In this type of design the blades possess eigen frequencies which are closer up to

Fig. 14: Amplitudes and phase shift at 0 rpm While measuring with the quarter bridge setup, different problems caused by electrical stray fields occurred in the strain gage signal. The reason for this failure is that electrical stray fields of the actuation signal are induced in the long arm of the quarter bridge. This long arm is the completion to a full bridge behind the slip ring. Thus, the induced signal is superimposed on the measured signal of the quarter bridge, which is therefore not detectable. Further investigations with inverted actuation of the three blades showed similar results to those described above. These results depend on the different eigen frequencies of the blades. They do not interact with each other. Therefore no resonance or cancellation appears. When implementing the control unit, the phase angle between actuation voltage and strain-gage signal can be adjusted in any order. For the time signal shown below in Fig. 15, the controller

Copyright 2009 by the American Institute of Aeronautics and Astronautics Inc. All rights reserved

8
each other. So an interaction of the blades can be investigated better. Special attention should be paid when using quarter bridges and the shielding of their leads to avoid interferences by stray electrical fields of the highvoltage actuation. ACKNOWLEDGEMENTS The investigations of this paper were done within the scope of the project LuFo IV BliDes Compressor flutter in collaboration with the MTU Aero Engines Munich, the DLR in Cologne and Gttingen and the HAWK Gttingen. Thanks go to everyone who contributed to this project, including the initiators Dr. Kahl of MTU and Dr. Hennings of DLR. Special thanks go to Dr. Isabel Goltz, who did a lot of preliminary work at the Institute of Turbomachinery and FluidDynamics (TFD). Further thanks go to Mr. Bhmer from the HAWK Gttingen and Mr. Belz from the DLR Gttingen, who contributed essentially and the technical staff at the TFD, particularly Mr. Wichmann, Mr. Weber and Mr. Hruschka. REFERENCES Belz, J. and Hennings, H. (2004): Aeroelastisches Verhalten von Verdichtergittern, Abschlussbericht des AG-Turbo 2 Vorhabens Charleux, D., Gibert, C., Thouverez, F. et al. (2006): Numerical and Experimental Study of Friction Damping in Blade Attachments of Rotating Blade Disks, International Journal of Rotating Machinery, Vol. 2006, pp. 1-13 Christensen, R.H. and Santos, I.F. (2004): A Study of Active RotorBlade Vibration Control Using Electromagnetic Actuation Part I and II, Proceedings of the ASME Turbo Expo 2004 Christensen, R.H. and Santos, I.F. (2006): Active Rotor-Blade Vibration Control Using ShaftBased Electromagnetic Actuation, Journal of Engineering for Gas Turbines and Power, Vol.128, pp. 644-652 Dowell, E.H. (Editor) (2004): A Modern Course in Aeroelasticity
Copyright 2009 by the American Institute of Aeronautics and Astronautics Inc. All rights reserved

Kluwer Academic Publishers, Dordrecht, Boston, London Goltz, I., Bhmer, H., Nollau, R., et al. (2009): Piezo-electric actuation of rotor-blades in an axial-compressor, ETC Graz 2009 Kielb, J.J. and Abhari, R.S. (2003): Experimental Study of Aerodynamic and Structural Damping in a Full-Scale Rotating Turbine, J. of Engineering for Gas Turbines and Power, Vol. 125 pp. 102 Montgomery, M., Tartibi, M., Eulitz, F., et al. (2005): Application of Unsteady Aerodynamics and Aeroelasticity in Heavy-Duty Gas Turbines, Proceedings of the ASME Turbo Expo 2003, pp. 635-649 Nollau, R. (2009): Modellierung und Simulation technischer Systeme, Springer Verlag, Berlin, Heidelberg, New York Watanabe, T., Kazawa, J., Uzawa, S. et al. (2008): Numerical and Experimental Study of Active Flutter Suppression with Piezoelectric Device for Transonic Cascade, Proceedings of ASME Turbo Expo 2008: Power for Land, Sea and Air, Berlin, GT2008-51467