versão impressa ISSN 0120-6230
Rev.fac.ing.univ. Antioquia n.55 Medellín set. 2010
Active vibration control of a rotorbearing system based on dynamic stiffness
Control activo de vibraciones en un sistema rotorchumaceras basado en la rigidez dinámica
Andrés Blanco Ortega1*, Francisco Beltrán Carbajal2, Gerardo Silva Navarro3, Marco Antonio Oliver Salazar1
1 Centro Nacional de Investigación y Desarrollo Tecnológico (Cenidet). Coordinación de Mecatrónica. Interior Internado Palmira s/n col. Palmira, Cuernavaca Morelos C. P. 62490. México.
2 Universidad Politécnica de la Zona Metropolitana de Guadalajara C.P. 45640 Tlajomulco de Zúñiga, Jalisco, México.
3 CINVESTAVIPN. Departamento de Ingeniería Eléctrica. Sección de Mecatrónica. C.P. 07360 México, D.F.
This paper presents an active vibration control scheme to reduce unbalance induced synchronous vibration in rotorbearing systems supported on two ball bearings, one of which can be automatically moved to control the effective rotor length and, as an immediate consequence, the rotor stiffness. This dynamic stiffness control scheme, based on frequency analysis, speed control and acceleration scheduling, is used to avoid resonant vibration of a rotor system when it passes (runup or coast down) through its first critical speed. Algebraic identification is used for online unbalance estimation at the same time that the rotor is taken to the desired operating speed. Some numerical simulations and experiments are included to show the unbalance compensation properties and robustness of the proposed active vibration control scheme when the rotor is started and operated over the first critical speed.
Keywords: active vibration control, eccentricity identification, Jeffcott rotor
En este artículo se presenta un esquema de control activo de vibraciones para atenuar las amplitudes de vibración síncrona inducidas por el desbalance en un sistema rotorchumaceras; donde una de las chumaceras puede ser desplazada automáticamente para modificar la longitud efectiva del rotor, y como consecuencia, la rigidez del rotor. El control de la rigidez dinámica se basa en un análisis de la respuesta en frecuencia, control de velocidad y en el uso de esquemas de aceleración, para evadir las amplitudes de la vibración en la resonancia mientras el sistema rotatorio pasa (acelerado o desacelerado) a través de una velocidad crítica. Se utiliza identificación algebraica para estimar el desbalance en línea, mientras el rotor es llevado a la velocidad de operación deseada. Algunas simulaciones numéricas y resultados experimentales son incluidos para mostrar las propiedades de la compensación del desbalance y la robustez del esquema de control activo de vibraciones propuesto, cuando el rotor se opera a una velocidad por encima de la primera velocidad crítica.
palabras clave: control activo de vibraciones, identificación de excentricidad, rotor Jeffcott
Rotating machinery is commonly used in many mechanical systems, including electrical motors, machine tools, compressors, turbo machinery and aircraft gas turbine engines. Typically these systems are affected by exogenous or endogenous vibrations produced by unbalance, misalignment, resonances, material imperfections and cracks.
Vibration caused by mass unbalance is a common problem in rotating machinery. Rotor unbalance occurs when the principal inertia axis of the rotor does not coincide with its geometrical axis and leads to synchronous vibrations and significant undesirable forces transmitted to the mechanical elements and supports. Many methods have been proposed to reduce the unbalanceinduced vibration, where different devices such as electromagnetic bearings, active squeeze film dampers, lateral force actuators, active balancers and pressurized bearings have been developed . Passive and active balancing techniques are based on the unbalance estimation to attenuate the unbalance response in the rotating machinery. The Influence Coefficient Method has been used to estimate the unbalance while the rotating speed of the rotor is constant [6, 7]. This method has been used to estimate the unknown dynamics and rotorbearing system unbalance during the speed varying period . On the other hand, there is a vast literature on identification methods , which are essentially asymptotic, recursive or complex, which generally suffer of poor speed performance.
This paper presents an active vibration control scheme to reduce unbalanceinduced synchronous vibration in rotorbearing systems supported on two ball bearings, one of which can be automatically moved along the shaft to control the effective rotor length and, as an immediate consequence, the rotor stiffness. This dynamic stiffness control scheme, based on frequency analysis, speed control and acceleration scheduling, is used to avoid resonant vibration of a rotor system when it passes (runup or coast down) through its first critical speed. Algebraic identification is used for online unbalance estimation at the same time that the rotor is taken to the desired operating speed. The proposed results are strongly based on the algebraic approach to parameter identification in linear systems reported , which requires a priori knowledge of the mathematical model of the system. This approach has been employed for parameter and signal estimation in nonlinear and linear vibrating mechanical systems, where numerical simulations and experimental results show that the algebraic identification provides high robustness against parameter uncertainty, frequency variations, small measurement errors and noise [13, 14]. In addition, algebraic identification is combined with integral reconstruction of time derivatives of the output (GPI Control) using a simplified mathematical model of the system, where some nonlinear effects (stiffness and friction) were neglected; in spite of that, the experimental results show that the estimated values represent good approximations of the real parameters and high performance of the proposed active vibration control scheme, which means that the algebraic identification and GPI control methodologies could be used for some industrial applications, when at least a simplified mathematical model of the system is available .
Some numerical simulations and experiments are included to show the unbalance compensation properties and robustness when the rotor is started and operated over the first critical speed.
The Jeffcott rotor system consists of a planar and rigid disk of mass m mounted on a flexible shaft of negligible mass and stiffness k at the midspan between two symmetric bearing supports (see figure 1(a) when a = b). Due to rotor unbalance the mass center is not located at the geometric center of the disk S but at the point G (center of mass of the unbalanced disk); the distance u between these two points is known as disk eccentricity or static unbalance [15, 16]. An end view of the whirling rotor is also shown in figure 1(b), with coordinates that describe its motion.
In our analysis the rotorbearing system is modeled as the assembly of a rigid disk, flexible shaft and two ball bearings. This system differs from the classical Jeffcott rotor because the effective shaft length can be increased or decreased from its nominal value. In fact, this adjustment is obtained by enabling longitudinal motion of one of the bearing supports (right bearing in figure 1.a) to different controlled positions into a small interval by using some servomechanism, which provides the appropriate longitudinal force. With this simple approach one can modify the shaft stiffness; moreover, one can actually control the rotor natural frequency, during runup or coast down, to evade critical speeds or at least reduce rotor vibration amplitudes. Our methodology combines some ideas on variable rotor stiffness  and rotor acceleration scheduling  but completing the analysis and control for the Jeffcottlike rotor system.
Figure 1 Rotorbearing system: (a) Schematic diagram of a rotorbearing system with one movable (right) bearing and (b) end view of the whirling rotor
For simplicity, the following assumptions are considered: flexible shaft with attached disk, gravity loads neglected (insignificant when compared with the actual dynamic loads), equivalent mass for the basebearing mb, linear viscous damping cb between the bearing base and the linear sliding, force actuator to control the shaft stiffness F, angular speed controlled by means of an electrical motor with servodrive and local Proportional Integral (PI) controller to track the desired speed scheduling in presence of small dynamical disturbances. The mathematical model of the four degreeof freedom Jeffcottlike rotor is obtained using Newton equations as follows.
where k and c are the stiffness and viscous damping of the shaft, Jz is the polar moment of inertia of the disk and t(t) is the applied torque (control input) for rotor speed regulation. In addition, x and y denote the orthogonal coordinates that describe the disk position and φ = w is the rotor angular velocity. The coordinate b denotes the position of the movable (right) bearing, which is controlled by means of the control force F(t) (servomechanism).
In our analysis the stiffness coefficient for the rotorbearing system is given by 
where l = a + b is the total length of the rotor between both bearings with b the coordinate to be
controlled, is the moment of inertia of a shaft of diameter D and E is the Young's modulus of elasticity (E = 2.11 x 1011N/m2) for AISI 4140 steel). The natural frequency of the rotor system is then obtained as follows 
In such a way that, controlling b by means of the control force F one is able to manipulate wn to evade appropriately the critical speeds during rotor operation.
The proposed control objective is to reduce as much as possible the rotor vibration amplitude, denoted in adimensional units by
for runup, coastdown or steady state operation of the rotor system, even in presence of small exogenous or endogenous disturbances. Note, however, that this control problem is quite difficult because of the 8th order nonlinear model, many couplings terms, underactuation and uncontrollability properties from the two control inputs (t, F).
Some experimental results were performed in open loop in a RotorKit experimental platform of Bently Nevada® The positions of the inertial disk, sensors and bearing supports in the rotorkit are shown in figure 2. The experimental results were performed to show how the natural frequency can be modified to different positions of the right bearing. Furthermore, the rotor is started and operated over the first critical speed where the speed operating condition for the rotor is given as φ = 418.9 rad/s (4000 rpm) and the acceleration for the speed ramp was 1.74 rad/s2.
Figure 2 Positions of the inertial disk, sensors and bearing supports in the rotorbearing system
The response for the Jeffcott rotor configuration (right bearing position at point B, see figure 2) is shown in figure 3 and it was recorded by the sensor S3 (see figure 2). The natural frequency for the Jeffcott rotor configuration was about wn = 2200 rpm.
Figure 3 Unbalance response for a Jeffcott Rotor (right bearing position at point B)
Experimental results for the bearing position at points A (l = 0.4 m) and C (l = 0.5 m), are shown in figures 4 and 5, respectively. Here, the openloop responses show that a smaller length l = 0.4 m leads to a higher natural frequency and a bigger length l = 0.5 m leads to a smaller natural frequency. Hence to get a minimal unbalance response, the rotor length should start at l = 0.4 m and then abruptly change to l = 0.5 m. This change of the bearing position must occur exactly when the response for l = 0.4 m crosses the response for l = 0.5 m.
Figure 4 Rotor unbalance response for the right bearing position at point A
Figure 5 Rotor unbalance response for the right bearing position at point C
In order to control the speed of the Jeffcott like rotor system, consider w = z6 = f equation (3), under the temporary assumption that the eccentricity u is perfectly known and that c « 0 to simplify the analysis. Then, the following local PI controller is designed to track desired reference trajectories of speed w*(t) and acceleration scheduling w*(t) for the rotor:
Speed control with trajectory planning
In order to control the speed of the Jeffcott like rotor system, consider w = z6 = φ equation (3), under the temporary assumption that the eccentricity u is perfectly known and that c « 0 to simplify the analysis. Then, the following local PI controller is designed to track desired reference trajectories of speed w*(t) and acceleration scheduling w*(t) for the rotor:
The use of this controller yields the following closedloop dynamics for the trajectory tracking error e1 = w - w* (t):
Therefore, selecting the design parameters [a1, a0] so that the associated characteristic polynomial for equation (9) is a Hurwitz polynomial, one guarantees that the error dynamics is asymptotically stable. The prescribed speed and acceleration scheduling for the planned speed trajectory is given by
where wi and wf are the initial and final speeds at the times t. and t respectively, passing through
the first critical frequency, and σ(t, ti, tf) is a Bezier polynomials, with σ(t, ti, tf) = 0 and σ(t, ti, tf) = 1, described by
with γ1 = 252, γ2 = 1050, γ3 = 1800, γ4 = 1575, γ5 = 700, γ6 = 126, in order to guarantee a sufficiently smooth transfer between the initial and final speeds.
The fundamental problem with the proposed feedback control in equation (8) is that the eccentricity u is not known, except for the fact that it is constant. The Algebraic identification methodology is proposed to online estimate the eccentricity u, which is based on the algebraic approach to parameter identification in linear systems .
Algebraic identification of eccentricity
Consider equation (3) with perfect knowledge of the moment of inertia Jz and the shaft stiffness k, and that the position coordinates of the disk (x, y) and the control input t are available for the identification process of the eccentricity u.
Multiplying equation (3) by t and integrating by parts the resulting expression once with respect to time t, one gets
Solving for the eccentricity u in equation (12) leads to the following online algebraic identifier for the eccentricity:
where δ is a positive and sufficiently small value.
Therefore, when the denominator of the identifier of equation (13) is different to 0, at least for a small time interval [O, δ] with δ < 0, one can find from equation (13) a closedform expression to online identify the eccentricity.
An adaptivelike controller with algebraic identification
The PI controller given by equation (8) can be combined with the online identification of the eccentricity in equation (13), resulting the following certainty equivalence PI control law
Note that, in accordance with the algebraic identification approach, providing fast identification for the eccentricity, the proposed controller (14) resembles an adaptive control scheme. From a theoretical point of view, the algebraic identification is instantaneous . In practice, however, there are modeling errors and other factors that inhibit the algebraic computation. Fortunately, the identification algorithms and closedloop system are robust against such difficulties .
Some numerical simulations were performed using the parameters listed in table 1.
Table 1 Rotor system parameters
Figure 6 shows the identification process of eccentricity and the dynamic behavior of the adaptivelike PI controller given by equation (14), which starts using the nominal value u = 0 µm. One can see that the identification process is almost instantaneous. The control objective is to take from the rest position of the rotor to the operating speed wf = 300 rad/s.
Figure 6 Algoritmos secuenciales vs. Algoritmos inteligentes
The desired speed profile runs up the rotor in a very slow and smooth trajectory while passing through the first critical speed. This control scheme is appropriate to guarantee stability and tracking. The resulting rotor vibration amplitude (system response when t = 0) is shown in figure 7, for three different and constant positions of the right bearing (i.e., b = 0.25 m, 0.30 m, 0.35 m), using the PI controller.
Figure 7 Unbalance response R for different and constant positions of the movable bearing: (dashed line), (solid line) and (dashdotted line)
The purpose of these simulations is to illustrate how the position of the bearing truly affects the rotor vibration amplitudes for the desired speed profile. The nominal length of the shaft is l = 0.60 m. A smaller length l = 0.55 m leads to a higher natural frequency and a bigger length l = 0.65 m leads to a smaller natural frequency (see figure 7). Hence to get a minimal unbalance response, the rotor length should start at l = 0.55 m and then abruptly change to l = 0.65 m. This change of the bearing position must occur exactly when the response for l = 0.55 crosses the response for l = 0.65, in order to evade the resonance condition, because the rotor speed is different from the natural frequency of the rotorbearing system.
Position control of the bearing support
It is evident from equations (5) and (6) that controlling the position of the movable (right) bearing b applying the control force F and according to a prespecified speed profile w*(t) the modification of the rotor amplitude response to the unbalance is possible. As a matter of fact this methodology is equivalent to a dynamic stiffness control for the Jeffcottlike rotor system, enabling smooth changes on coordinate b.
To design a controller for position reference tracking, consider equation (4). Then, one can propose the following Generalized Proportional Integral (GPI) controller for asymptotic and robust tracking to the desired position trajectory b* (t) for the bearing position and velocity, which employs only position measurements of the bearing. For more details on GPI control see .
where b is an integral reconstructor of the bearing velocity, which is given by
The use of the GPI controller given yields the following closedloop dynamics for the trajectory tracking error e2 = b- b* (t):
Therefore, selecting the design parameters (β0, β 1, β 2,} such that the associated characteristic polynomial for equation (17) be Hurwitz, one guarantees that the error dynamics be globally asymptotically stable. The desired trajectory planning b* (t) for the bearing position and velocity is also based on Bezier polynomials similar to equation (10).
Results and discussionThe proposed methodology for the active vibration control of the transient runup or coast down of the rotorbearing system consists of the following steps:
- Define the trajectory planning for the speed trajectory profile w*(t) to be asymptotically tracked by the use of the adaptivelike PI controller with the algebraic identifier of the eccentricity, i.e., lim tèα w*(t) = w*(t).
- Establish an appropriate smooth switching on the position of the movable bearing b*(t) to be asymptotically tracked by the application of the GPI controller, i.e., lim tèα b(t) = b*(t). The switching time has to be at the crossing point leading to minimal unbalance response in figure 7.
Figure 8 shows the unbalance response of the rotorbearing system when rotor speed PI controller with algebraic identification of eccentricity and GPI control of the bearing position are simultaneously used. Note that the switching of the bearing position leads to small transient oscillations due to inertial and centrifugal effects on the overall rotor system.
Figure 8 Rotor vibration amplitude response using active vibration control (solid line)
First of all, the speed trajectory planning and control torque shown in figure 6 are similarly used. The smooth switching for the bearing position is implemented in such a way that the runup of the rotor system starts with the position bi = 0.25 m (i.e., l = 0.55 m) and changes to bf= 0.35 m (i.e., l = 0.65 m) exactly at the crossing point shown in the corresponding response in figure 7. The switching time occurs when w = 170.6 rad/s, that is, t = 23.9 s. The desired position of the bearing b(t) is illustrated in figure 9 together with the applied control force F. A comparison of the openloop response and the closedloop response in figure 8 results in important unbalance reductions about 64%.
Figure 9 Response of the bearing support using GPI controller: (a) position of the movable bearing and (b) control force
The active vibration control of a Jeffcottlike rotor through dynamic stiffness control and acceleration scheduling is addressed. The control approach consists of a servomechanism able to move one of the supporting bearings in such a way that the effective rotor length is controlled. As a consequence, the rotor stiffness and natural frequency are modified according to an offline and smooth trajectory planning of the rotor speed/ acceleration in order to reduce the unbalance response when passing through the first critical speed. The vibration control scheme results from the combination of passive and active control strategies, leading to robust and stable performance in presence of the synchronous disturbances associated to the normal operation of the rotor and some small parameter uncertainties. Since this active vibration control scheme requires information of the eccentricity, a novel algebraic identification approach is proposed for online estimation of the eccentricity. From a theoretical point of view, the algebraic identification is practically instantaneous and robust with respect to parameter uncertainty, frequency variations, small measurement errors and noise. The proposed active vibration control scheme, used to reduce unbalanceinduced synchronous vibration, is restricted to use in small rotating machinery (e.g., tools machines, motors and generators).
1. S. Zhou, J. Shi. "Active balancing and vibration control of rotating machinery: a survey". The Shock and Vibration Digest. Vol. 33. 2001. pp. 361371. [ Links ]
2. G. Sheu, S. Yang, C. Yang. "Design of experiments for the controller of rotor systems with a magnetic bearing". Journal of Vibration and Acoustics. Vol. 119. 1997. pp. 200207. [ Links ]
3. Y. Guozhi, Y. F. Fah, C. Guang, M. Guang, F. Tong, Q. Yang. "Electrorheological multilayer squeeze film damper and its application to vibration control of rotor system". Journal of Vibration and Acoustics. Vol. 122. 2000. pp. 711. [ Links ]
4. B. Palazzolo, S. Jagannathan, A. F. Kaskaf, G. T. Monatgue, L. J. Kiraly. "Hybrid active vibration control of rotorbearing systems using piezoelectric actuators". Journal of Vibration and Acoustics. Vol. 115. 1993. pp. 111119. [ Links ]
5. Q. Jinhao, J. Tani and T. Kwon. "Control of selfexcited vibration of a rotor system with active gas bearings". Journal of Vibration and Acoustics. Vol. 125. 2003. pp. 328334. [ Links ]
6. X. Yu. "General influence coefficient algorithm in balancing of rotating machinery"..International Journal of Rotating Machinery. Vol. 10. 2004. pp. 85¬90. [ Links ]
7. S. Lee, B. Kim, J. Moon, D. Kim. "A study on active balancing for rotating machinery using influence coefficient method". Proceedings of International Symposium on Computational Intelligence in Robotics and Automation. Espoo. 2005. pp. 659 664. [ Links ]
8. S. Zhou, S. Dyer, K. K. Shin, J. Shi, J. Ni. "Extended influence coefficient method for rotor active balancing during acceleration". Journal of Dynamics Systems Measurements and Control. Vol. 126. 2004. pp. 219¬223. [ Links ]
9. L. Ljung. Systems identification: theory for the user. Ed PrenticeHall. New Jersey. 1987. pp. 168361. [ Links ]
10. S. Sagara, Z. Y. Zhao. "Numerical integration approach to online identification of continuous systems". Automatica. Vol. 26. 1990. pp. 6374. [ Links ]
11. S. Sagara, Z. Y. Zhao. "Recursive identification of transfer function matrix in continuous systems via linear integral filter". International Journal of Control. Vol. 50. 1989. pp. 457477. [ Links ]
12. M. Fliess, H. SiraRamirez. "An algebraic framework for linear identification, ESAIM: Control". Optimization and Calculus of Variations. Vol. 9. 2003. pp. 151168. [ Links ]
13. F. BeltrânCarbajal, H. SiraRamirez, G. Silva Navarro. "Adaptivelike active vibration supression for a nonlinear mechanical system using online algebraic identification". Proceedings of the Thirteenth International Congress on Sound and Vibration. Vienna. July 2006. pp. 18. [ Links ]
14. F. BeltrânCarbajal, G. SilvaNavarro, H. Sira Ramirez, J. Quezada Andrade. "Active vibration control using online algebraic identification of harmonic vibrations". Proceedings of American Control Conference. Portland (Oregon). 2005. pp. 48204825. [ Links ]
15. J. M. Vance. Rotordynamics of Turbomachinery. Ed. John Wiley and Sons. New York. 1988. pp. 7231. [ Links ]
16. A. Dimarogonas. Vibration for Engineers. Ed. Prentice Hall. New Jersey. 1996. pp. 533536. [ Links ]
17. Z. Sandler. Robotics: designing the mechanism for automated machinery. Ed. Academic Press. San Diego (CA). 1999. pp. 162164. [ Links ]
18. K. T. Millsaps, L. Reed. "Reducing lateral vibrations of a rotor passing through critical speeds by acceleration scheduling". Journal of Engineering for Gas Turbines and Power. Vol. 120. 1998. pp. 615620. [ Links ]
19. S. S. Rao. Mechanical Vibration. Ed. Pearson Education. New Jersey. 2004. pp. 6711034. [ Links ]
20. M. Fliess, R. Marquez, E. Delaleau, H. Sira Ramirez. "Correcteurs proportionnelsintegraux generalizes". ESAIM Control, Optimisation and Calculus of Variations. Vol. 7. 2002. pp. 2341. [ Links ]
(Recibido el 13 de Junio de 2009. Aceptado el 6 de abril de 2010)
*Autor de correspondencia: + 52 +777 +362 77 70 ext. 105, fax + 52 +777 + 362 77 95, correo electrónico: firstname.lastname@example.org. (A. Blanco)