0 Votes +0 Votes -

2 vues13 pagesFeb 12, 2017

© © All Rights Reserved

PDF, TXT ou lisez en ligne sur Scribd

© All Rights Reserved

2 vues

© All Rights Reserved

- MEMB343 Mechanical Vibrations Presentation Slides
- UT Dallas Syllabus for opre7346.001 06s taught by Alain Bensoussan (axb046100)
- Ratio Control Technology
- Ali et al (2007)
- Design of Type-III Control Loops Using Explicit Analytical PID Tuning Technique
- MEEG_5113b.ppt
- Dugdale 1988
- Thesis
- Andersen_2002_Post-Critical Behavior of Beck's Column With a Tip Mass
- Centac Vibration Procedure%2Epdf
- Study Guide 2013
- vibration
- Midya Thesis
- w25-29
- SkopjeSE40EEE_Goerguelue
- Brinkgreve&Kappert&Bonnier2007
- Dynamic Analysis of Steel Frame Using Manufactured Viscous Damper
- MV Motor Publication Jan 2012(2).pdf
- A Frequency Domain Technique
- Lab Manual

Vous êtes sur la page 1sur 13

journal homepage: www.elsevier.com/locate/ymssp

from experimental rotor unbalance response

Yuanping Xu a, Jin Zhou a,n, Long Di b, Chen Zhao a

a

College of Mechanical and Electrical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China

b

Rotating Machinery and Control Laboratory (ROMAC), University of Virginia, Charlottesville, VA 22904-4743, USA

a r t i c l e in f o abstract

Article history: Active magnetic bearings (AMBs) support rotors using electromagnetic force rather than

Received 14 July 2015 mechanical forces. It is necessary to accurately identify the AMBs force coefficients since

Received in revised form they play a critical role in the rotordynamic analysis including system stability, bending

9 May 2016

critical speeds and modes of vibrations.

Accepted 10 June 2016

Available online 29 June 2016

This paper proposes a rotor unbalance response based approach to identifying the

AMBs stiffness and damping coefficients during rotation. First, a Timoshenko beam finite

Keywords: element (FE) rotor model is created. Second, an identification procedure based on the FE

Active magnetic bearings model is proposed. Then based on the experimental rotor unbalance response data from

Dynamic parameters identification

1200 rpm to 30,000 rpm, the AMBs dynamic force parameters (stiffness and damping) are

Finite element method

obtained. Finally, the identified results are verified by comparing the estimated and ex-

Stiffness and damping

perimental rotor unbalance responses, which shows high accuracy.

& 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Active magnetic bearings generate forces through magnetic fields rather than mechanical forces as in lubricated fluid

films or contact of rolling element bearings; therefore, the special advantage of AMBs is that there is no contact between

bearing and rotor, and this permits operation with no lubrication, no mechanical wear, long life, lower costs and high

attainable rotating speed [1,2]. Another attractive advantage of AMBs is that the dynamic force parameters, stiffness and

damping, are closely related to the feedback controller parameters, which can be changed easily [3], such that the ro-

tordynamics can be controlled and changed actively through the bearings.

For either AMBs or traditional mechanical bearings, it is vital to accurately obtain the dynamic force bearing parameters

since these coefficients for a rotor system are the foundation for the rotor dynamics analysis including system stability,

bending critical speeds, modes of vibrations, and response of rotating dynamic systems. The force coefficients of the tra-

ditional mechanical bearing are commonly modeled as stiffness and damping, which are also used for the AMBs. Compared

with traditional bearing system, the AMBs system is open loop unstable and feedback control is needed for levitation. Apart

from the rotor dynamics analysis, from the control design point of view, it is also vital to accurately predict the AMBs

dynamics parameters to optimize the control strategies design since the AMBs stiffness and damping are closely related to

the feedback controller parameters.

Humphris et al. [4] investigated the stiffness and damping properties of a magnetic bearing with two control algorithms

n

Corresponding author.

E-mail addresses: ypxu@nuaa.edu.cn (Y. Xu), zhj@nuaa.edu.cn (J. Zhou).

http://dx.doi.org/10.1016/j.ymssp.2016.06.009

0888-3270/& 2016 Elsevier Ltd. All rights reserved.

Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240 229

Rotor

AMB AMB AMB AMB

at static state; it was shown that varying amounts of position and velocity feedback affected the stiffness and damping

characteristics of the bearing, respectively. Williams et al. [5] proposed the estimated equivalent stiffness and damping

coefficients based on a theoretically derived frequency dependent feedback controller transfer function. However, the

theoretical estimated equivalent stiffness and damping coefficients procedures do not consider the time lags in the digital

control system. Lim et al. [3] experimentally obtained the stiffness and damping of a one-axis electromagnetic suspension

system under the shaker excitation. Similarly, under the shaker excitation, Lim [6] identified the parameter of five-degree-

of-freedom hybrid magnetic bearings for blood pump applications with proportional-integral-derivative (PID) controllers.

Tsai [7] applied the wavelet transform algorithm to identify the magnetic damping and magnetic stiffness coefficients.

Kozanecka [8] applied an unbalance excitation with the rotating vector force and the dynamic coefficients of the bearing

under nominal operating conditions are presented by calculating unbalance response, but did not consider the differential

location between sensors and actuator of AMBs. Although numerous studies have attempted to estimate these AMBs

coefficients directly or indirectly, few papers obtain AMBs coefficients under rotating condition. This paper seeks to address

this issue, by proposing a rotor unbalance response based approach to identifying the AMBs stiffness and damping coef-

ficients Fig. 1.

The identification method using unbalance response has been widely adopted in rolling element bearings and sliding

bearings [9,10], however few works have been reported on its application to AMBs. Compared with other identification

method such as under the shaker excitation, this method does not need additional devices and unbalance force excitation is

the simplest form of external excitation. More importantly, this method permits identification under the operating con-

dition. Before the parameters identification, a precise rotor model is indispensable. For the rotor adopted in the AMBs

system, its theoretical dynamic characteristics such as resonance and mode shape obtained from the FE model are not

always coincident with the experimental data since the rotor is often assembled by several components with interference

fits. For example, a typical rotor in AMBs system includes rotor shaft, front/rear AMB lamination stacks, thrust AMB

230 Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240

lamination stacks, sensor rings and motor (Fig. 2), which are assembled with interference fits and shrink effects, con-

tributing to the bending stiffness of the rotor shaft. Previous work [11] by us has already detailed the rotor modeling, model

updating and updated model verification. Here, the updated FE rotor model is employed directly to calculate the AMBs

dynamic parameters. Before the test rig staring operation, we add two unbalance mass to the rotor. Then using the ex-

perimental unbalance response data, we identify the AMBs stiffness and damping coefficients during rotation ranging from

1200 rpm to 30,000 rpm. In order to make the results reliable, the verification by comparison experimental unbalance

response amplitude and phase data with the calculated unbalance response data using the identified coefficients is carried

out. The verification shows that this identified method is accurate and reliable.

The remainder of the paper is organized as follows. Section 2 describes the rotor AMBs test rig employed in this paper

and the mathematical modeling of AMBs rotor. Section 3 presents the identification method based on rotor unbalance

responses. Section 4 describes the unbalance responses and experimental results. Section 5 describes the verification of

identified results. Conclusions are drawn in Section 6.

The experimental test rig for this study is a five degree of freedoms (DOFs) rotor AMBs system designed and built as a

research platform at Nanjing University of Aeronautics and Astronautics, pictured in Fig. 1. The rotor is supported by two

radial and two thrust AMBs and is 0.468 m long and weighs around 2.4 kg. A 1.5 kW AC asynchronous induction motor rotor

is located in the middle of the rotor between the front AMB and rear AMB. Two laminated silicon steel sheets are mounted

at the end of rotor shaft for the two radical support AMBs and one laminated silicon steel sheet is assembled in the middle

for the driven motor. The specific details of the rotor are listed in Table 1. The radical and axial air gaps are 0.3 mm and

0.5 mm, respectively. Two rolling element ball bearings are assembled at the radical support AMB casings to prevent da-

mage to AMBs when the rotor is out of control.

Most industrial rotors are flexible and an accurate rotor model is the fundamental of parameters identification. The rotor

model is often obtained by the finite element method. NelsonTimoshenko beam element is employed to simulate the rotor

system, and the gyroscopic moments, rotatory inertia, and shear deformation of the shaft are taken into consideration in

this study [12]. 53 NelsonTimoshenko beam finite element matrices are adopted to model the rotor according to the

geometrical and mass information, pictured in Fig. 3. The assembled parts such as lamination stacks are modeled as lumped

mass onto the corresponding nodes. Ignoring two DOFs in axial direction, each node contains 2 translational and 2 rota-

tional DOFs, therefore, the whole system possesses 216 DOFs. After assembling the governing equations for all the elements

and incorporating the boundary conditions, the linearized equations of motion for the shaft can be expressed as

Mq +( C+G) q+

Kq=f ( t ), (1)

where q and f (t ) are generalized displacement and generalized force vector in two radical directions; is the rotation

speed; M, C and K represent square symmetric mass, damping and stiffness matrices, respectively; G is the skew symmetric

gyroscopic matrix.

The FE code compiled in MATLAB for this rotor does not contain the unknown interactions of shrink fit interfaces,

inhomogeneous materials, small geometrical details, and so on, which cause potential errors to the modeling. In order to

obtain a relative accurate FE rotor model before identification, a precise FE rotor model is indispensable since any error in

the modeling of the flexible shaft will cause significant errors in the results. Previous work [11] by us has already detailed

updated the FE rotor model and validated it. For the rotor of our test rig in this study, its theoretical freefree mode shapes

plot without model updating is shown in Fig. 3. Table 2 lists the first forth bending mode modal frequencies before and after

model updating.

Table 1

Specific details of the rotor.

Label Name

B/K Magnetic isolation ring

C/J Front/rear radical AMB lamination stacks

D/I Front/rear sensor ring

E Rotor shaft

F/H Motor rotor fix stacks

G Motor rotor

Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240 231

Table 2

Experimental and theoretical (no updating, updated) modal frequencies.

modes (Hz) (Hz) (Hz)

2 1137.6 1167.5 1167.5

3 1802.7 1857.5 1855.4

4 2733.2 2805 2827.7

3.1. Theoretical equivalent stiffness and damping of one DOF AMB model

The AMB system is open loop unstable and feedback control is needed for levitation. Fig. 4 depicts a magnetic bearing

control loop system in one degree of freedom, which includes power amplifier, bearing electromagnet, controller, sensor

and rotor (flotor for non-rotating objects) [1].

The magnetic force for a single DOF AMB system can be expressed by its position stiffness Kyy and current stiffness Kiy

[1,4]:

power

amplifier

electromagnet

rotor / flotor

controller

sensor

232 Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240

The simplified control loop block diagram of one DOF AMB suspension system [5] is shown in Fig. 5, where G (s ) is the

feedback controller.

According to Fig. 5, the transfer function of one DOF AMB system can be written as

Y ( s) 1

= .

F ( s ) ms2+Kyy+Kiy G ( s ) (3)

Let s = j , where j = 1 , and the transfer function of Eq. (3) in the frequency domain can be expressed as

Y ( j) 1

= ,

F ( j) m2+Kyy+Kiy G ( j) (4)

where Y (j) is the Fourier transform of the displacement Y (s ); F (j) is the Fourier transform of excitation force F (s ). Ac-

cording to the Newton's law, the equation of motion of this magnetic bearing system can be expressed as

where m, k and c represent the rotor mass, stiffness and damping coefficients, respectively. Using the Laplace transform

Eq. (5) can be written as

Let s = j , where j = 1 , and the Eq. (6) in the frequency domain can be expressed as

Y ( j) 1

= .

F ( j) m2+k+cj (7)

Based on Eqs. (4) and (7), we can acquire the equivalent stiffness and damping coefficients of the one DOF magnetic

bearing system, which is written as

k eq=Kyy+Re { Kiy G (s ) };

Im { Kiy G ( s ) }

ceq= .

(8)

If the control strategy is the proportional-integral-derivative (PID) algorithm, the controller transfer function G (s ) can be

written as

Ci

G ( s )=Cp+Cd s + ,

s (9)

k eq=Kiy Cp+Kyy;

C

ceq=Kiy Cd i2 .

(10)

keq and ceq are called equivalent coefficients because these coefficients are obtained from the analysis of the theoretical

frequency-dependent feedback controller transfer function G (s ), other than from the real experimental data. This analysis

method does not consider eddy current loss, hysteresis loss, ohmic loss and time delay in the digital signal processing,

amplifiers, feedback sensors etc. More importantly, for some robustness modern control techniques such as adaptive control

strategies, it is difficult to obtain a certain controller transfer function G (s ), which makes it difficult to get the real keq and ceq

coefficients, therefore, an experimental identification is indispensible.

Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240 233

Since the theoretical equivalent stiffness and damping values are not accurate or difficult to acquire directly, estimating

these parameters through experimental data provides another reasonable solution. In the community of mechanical bearing

coefficient identification, various methods have been proposed. Here, based on the [13], we use unbalance responses to

calculate the AMBs stiffness and damping coefficients through online operation.

Here we use the updated FE rotor model to obtain experimental stiffness and damping coefficients. In Eq. (1),

Mq +( C+G) q +Kq = f ( t ), q and f are generalized displacement and generalized force vector in two radical directions, which

are written as

q= q1 qB1 qB2 qN ;

T

qi= xi yi xi yi ,

f = f1 fu fN ;

T

where qB1 and qB2 represent the displacements at the front, rear AMBs; xi and yi represent translations in the x and y

directions; xi and yi are the angular displacements about the y and x axes, respectively; fu is the unbalance excitation force

caused by unbalance mass, which can be written as

fx

cos ( t +)

real e (

i ( t + )

)

fy

fu =

=mu r2 sin ( t +) =mu r2 real( ie

i ( t + )

)

,

0 0

0

0 0

0 (12)

where mu is the unbalance mass; r is the radius between the unbalance location and the axis of shaft rotation. Note that the

periodic forced excitation with frequency caused by rotor unbalance is synchronous with rotating speed, i.e. = . In

general, f ( t )=Feit , so the unbalance response possesses the same frequency as the excitation, i.e. q = qeit . Therefore, Eq. (1)

can be written as the following form,

( KM2)+i ( C+G) q=F ,

(13)

and the transfer function between unbalance excitation and displacement (dynamic stiffness matrix) is

The stiffness K and damping C matrices include the rotor coefficients (KR , CR ) and the AMBs coefficients (KB , CB ), therefore

K and C can be described as

K=KR+KB;

where

0 0

K b1 Cb1

KB= 0 ;CB= 0 .

K b2 Cb2

0 0 (16)

To simplify the calculation, ignoring cross-coupling effect, the AMBs stiffness and damping matrices incorporate the

following (yet unknown) coefficients:

Kxx 0 Cxx 0

K bi= ;Cbi= , i=1, 2.

0 Kyy i 0 Cyy i (17)

Therefore, let the Eq. (15) into Eq. (14), the transfer function Eq. (14) can be written as

234 Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240

Here, the dynamic rotor stiffness matrix HR and dynamic AMBs stiffness matrix HB can be defined as

HB=KB+iCB. (19)

Combining the dynamic rotor stiffness matrix and dynamic AMBs stiffness matrix together, the equation of motion of the

rotor AMB system can be written as below:

In order to identify the unknown AMBs coefficients, the algebraic system of Eq. (20) is reordered by use of matrix

operations to yield

0

HR q +HB q = 0 ,

F (21)

ZB1

q = ZB2 ,

ZO (22)

where ZB1 and ZB2 are the transitional displacement vectors of AMBs supporter places, which is expressed by

T T

(23)

where ( xB yB ) are the measured synchronous rotor responses (in complex form), ZO is the transitional displacement vectors

except AMBs supporter places.

In the FE rotor model, we have 54 nodes, therefore, the dimension of q , HR and HB is 216 216. The AMBs supporting

position nodes in the FE model are located at node 14 and 46. Using qi to denote each number in the original displacement

vector q . Therefore, the specific form of original q is

q= q1 q2 q3 q4 q216

T , (24)

Therefore, the specific reordered form of q can be written as

q = q53

q54

q181

q182

q1:52

q55:180

q183:216

T , (25)

Here q53, q54 , q181 and q182 are the xB1, yB1, xB2, yB2 respectively. So ZB1=[ q53 q54 ]T ; ZB2 [ q181 q182 ]T Z O [ q1:52 q55:180 q183:216 .

T

In order to calculate conveniently, HR and HB are partitioned into sub matrices like below:

HR11 HR12 HR13 HB1 0 0

HR= HR21 HR22 HR23; HB= 0 HB2 0,

HR31 HR32 HR33 0 0 0 (26)

where HB1 and HB2 are the dynamic front, rear AMBs stiffness matrix, respectively, which are written as

Kxxi+iCxxi 0

HBi= ,i=1,2,

0 Kyyi+iCyyi (27)

HR11 HR12 HR13 ZB1 HB1 0 0 ZB1 0

HR21 HR22 HR23 ZB2 + 0 HB2 0 ZB2 = 0 .

HR31 HR32 HR33

ZO 0

0 0 ZO F (28)

In generally, for a rotor AMBs system, the sensors and actuators are not located at the same axial location, in Fig. 6.

Therefore, the measured rotor response is not the actual displacement at the AMBs. We need to transform the measured

responses to AMBs positions responses using the following relationship [14]:

ZB1 = a Z m1+AZ o;

Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240 235

B1 B2

S1 S2

input output output input

where Zm1 and Zm2 are the measured response vectors at the sensor locations near the AMBs, a and b are the known scalar

functions of the interpolation functions, and A and B are the matrix functions of the interpolation functions.

In Eq. (28), the motion of AMBs support places (ZB1,ZB2 ) and load vector F are obtained from measurements and un-

balance mass added to the rotor. So from the third row of this equation, ZO can be expressed as

{

ZO=HR331 F HR31ZB1HR32 ZB2 , } (30)

from the first and second row, we can obtain the follow equations:

HR11ZB1+HR12 ZB2+HR13 ZO= HB1ZB1;

Here, we define the AMBs transmitted forces fB1 and fB2, which can be written

fB1= (HR11ZB1+HR12 ZB2+HR13 ZO );

fB2 = ( HR21ZB1+HR22 ZB2+HR23 ZO ). (32)

Because ZO can be obtained from Eq. (30), the fB1 and fB2 values can be calculated. So the AMBs support coefficients are

determined by:

Kxx1+iCxx1 0

HB1= =f ZB11;

0 Kyy1+iCyy1 B1

Kxx2+iCxx2 0

HB2= =f ZB21.

0 Kyy2+iCyy2 B2 (33)

Note that, in the whole process of support coefficients calculation, we use the updated FE rotor model to provide an

accurate result.

Before the unbalance response experiment, the rotor has been adjusted to nearly balanced condition under the dynamic

balancing machine to avoid the residual unbalance influence from the rotor itself.

The rotor unbalance response measurements are conducted for rotor speeds from 1200 rpm to 30,000 rpm, which is

above the first bending critical speed around 480 Hz (28,800 rpm). The unbalance response less than 1200 rpm is not

recorded since the unbalance displacement amplitude is small. Table 3 shows the unbalance mass distributions adopted in

this test. Two unbalance mass are added to rotor at node 17 and node 43, respectively. Fig. 7 depicts the unbalance mass

Table 3

Unbalance distributions on test rotor.

Node 17 0.98 15 0

Node 43 0.78 15 180

236 Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240

Node 17

added hole at rotor node 17, and the unbalance mass added hole at rotor node 43 has the same structure.

During the whole unbalance test, the rotor displacements and rotating speed information are saved every 5 Hz

(300 rpm). Since the experimental data from the unbalance responses contains lots of noise, we need band pass filter to

process these signals. After storing the test data, the measured displacement values go through unbalance signal filter using

zero phase digital filters to help us extract the steady state amplitude and phase information. Then use one order Fourier

series based the least square method to fit the filtered signals, which could be written in the follow expression

y=a 0+a1 cos ( x)+b1 sin ( x), (34)

from the Eq. (34), the amplitude A and phase values under each rotating speed can be acquired though the parameters

a0, a1, b1, , which is written as Eq. (35)

2 2

A= a1 + b1

a1 .

=arctan

b1 (35)

Fig. 8 details the experimental unbalance responses data before and after filter under 100 Hz (6000 rpm) at sensor 1 in x

direction. Fig. 9 shows the unbalanced response test data process procedure.

Fig. 10 shows the measured unbalance responses at the sensor locations resulting from the unbalance masses under

rotating condition. Note that the measured rotor displacements are at sensor location radial planes and we need use Eq. (29)

to transform them from the measurements to AMBs actuators locations.

Fig. 11 shows the theoretical equivalent stiffness and damping coefficients calculated using the system transfer function

from Eq. (10) and the identified actual AMBs supporting stiffness and damping coefficients, obtained from the measured

rotor unbalance responses, using the method presented in this paper in Section 3. The theoretical equivalent coefficients

calculated from Eq. (10) of two AMBs are drawn using green line. Note that the control parameters are the same for both x

and y directions of each AMB, therefore only one green line indicates the theoretical values.

It is seen that before the first bending critical speed (around 480 Hz), the equivalent stiffness and damping values in-

creases steady with the running speed. Since the control parameters for both x and y direction of each AMB are the same,

the identified coefficients possess the same trend for the two orthogonal directions. However, there are still some small

unclose agreement, which may be attributed to the small discrepancy in mechanical and electrical performance. The

x 10

2

1

Amplitude (m)

-1

-2

1 1.01 1.02 1.03 1.04 1.05 1.06 1.07 1.08

Time (s)

Before filtering

After filtering

Fig. 8. Unbalance responses data under 100 Hz ( 6000 rpm) at sensor 1 in x direction.

Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240 237

Experimental unbalance

displacement data

design

Zero phase digital filter

fitting fitting

amplitude data phase data

Fig. 9. Flowchart of test data process procedure.

-5 -5

x 10 x 10

2 2

xm1

1.5 ym1 1.5

Amplitude (m)

Amplitude (m)

1 1

xm2

0 0

100 200 300 400 500 100 200 300 400 500

Rotate speed (Hz) Rotate speed (Hz)

200 200

100 100

Phase (degree)

xm1

Phase (degree)

ym1

0 0

xm2

-100 -100

ym2

-200 -200

100 200 300 400 500 100 200 300 400 500

Rotate speed (Hz) Rotate speed (Hz)

Fig. 10. Measured rotor unbalance responses (amplitude and phase) at sensor locations.

identified results fluctuate greater around the critical speed since the vibration near the circuital speed becomes greater. The

equivalent and test actual damping coefficients have the same trend although the test values fluctuate, however the dis-

crepancy for equivalent and test stiffness coefficients is great. The reason of these differences may attribute to that the

equivalent coefficients are obtained from the analysis of the theoretical frequency-dependent feedback controller transfer

function is simplified, ignoring the non-linearity in electromagnets, the eddy current loss, hysteresis loss, ohmic loss and

time delay in the digital signal processing, amplifiers, feedback sensors etc. Therefore the simplified theoretical calculation

cannot precisely stand the real AMBs stuffiness and damping and the experiment identification is indispensible.

5. Results verification

In order to make sure the identified values are accuracy, the verification is indispensible. We use these identified

coefficients as supporting parameters of the FE rotor model to calculate the unbalance responses under various operating

condition from 1200 rpm to 30,000 rpm. And then compare estimated the magnitude and phase with the measured during

the course of the experiment. Figs. 12 and 13 shows the comparisons between simulated and test amplitude and phase data

238 Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240

7.5 7.5

Kxx1

Stiffness (N/m) (log10)

7 Kyy1 7

Kxx2

Keq1

6.5 6.5

Kyy2

6 6 Keq2

5.5 5.5

5 5

100 200 300 400 500 100 200 300 400 500

Rotate speed (Hz) Rotate speed (Hz)

6000 6000

Cxx1

5000 5000 Cxx2

Cyy1

Damping (N.s/m)

Damping (N.s/m)

4000 4000 Cyy2

Ceq1

Ceq2

3000 3000

2000 2000

1000 1000

0 0

100 200 300 400 500 100 200 300 400 500

Rotate speed (Hz) Rotate speed (Hz)

Fig. 11. Comparisons between effective and actual identified stiffness and damping coefficients. (For interpretation of the references to color in this figure,

the reader is referred to the web version of this article).

-5 -5

x 10 x 10

2 2

1.5 1.5

simulated data in xm1 simulated data in ym1

Amplitude (m)

Amplitude (m)

1 1

0.5 0.5

0 0

0 100 200 300 400 500 0 100 200 300 400 500

Rotate speed (Hz) Rotate speed (Hz)

50 200

test data in xm1

0 100

Phase (degree)

Phase (degree)

test data in ym1

-50 0

simulated data in ym1

-100 -100

-150 -200

0 100 200 300 400 500 0 100 200 300 400 500

Rotate speed (Hz) Rotate speed (Hz)

Fig. 12. Comparisons between simulated and test amplitude and phase for front AMB.

for both AMBs. It can be observed that the unbalance response in the x and y axes from the finite element model agree with

the experimental measurements very well even around the critical speed (28,800 rpm, 480 Hz). Note that, we obtain the

same conclusion using the commercial finite element software to calculate the rotor unbalance responses under various

operating condition from 1200 rpm to 30,000 rpm. Here we just give the results calculated from FE rotor model created in

MATLAB in part 2.2, which has been updated and verified by our previous work. The verification results demonstrate that

the identification method proposed in this paper is effective and accurate.

Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240 239

-5 -5

x 10 x 10

2 2

1.5

Amplitude (m)

Amplitude (m)

1.5

1

test data in xm2 test data in ym2

1

0.5 simulated data in xm2 simulated data in xm2

0 0.5

0 100 200 300 400 500 0 100 200 300 400 500

Rotate speed (Hz) Rotate speed (Hz)

200 150

100 100

Phase (degree)

Phase (degree)

simulated data in xm2

-100 0

-200 -50

0 100 200 300 400 500 0 100 200 300 400 500

Rotate speed (Hz) Rotate speed (Hz)

Fig. 13. Comparisons between simulated and test amplitude and phase for rear AMB.

6. Conclusion

This paper presented a supporting parameters identification strategy of AMBs under rotating condition. In order to

improve the identification accuracy, a Timoshenko beam FE rotor model is created. The proposed unbalance response based

identification method considers the differential location between sensors and AMBs. The identified supporting parameters

are verified by comparing the simulation and experimental rotor unbalance responses. The good consistency between es-

timated and experimental unbalance responses shows the proposed method is accurate and effective.

Compared with the theoretical analysis of control laws to obtain the stiffness and damping coefficients, unbalance mass

responses method is simply but reliable. The identified results show that the stiffness and damping coefficients on both x

and y axes vary along with the rotating speed, indicating that the magnetic force is affected by the operating speed. The

differences from the comparison between theoretical equivalent coefficients and experiential actual coefficients show that

the simplified theoretical calculation cannot accurately describe the real AMBs stuffiness and damping since various factors

are ignored in the idea condition.

Although the unbalance excitation is an easy and convenient excitation form for running rotor, some drawbacks cannot

be neglected. For example unbalances beyond certain level may generate excessive unbalance force, which may break or

damage the rotor system, especially for AMBs rotor system. In the experiment adopted in this paper, above 30,000 rpm

(500 Hz) rotating speed is not performed due to the safety reason. Hence, acquiring the dynamic parameters by impulse

response measurement or from the non-simplified system model may be a good choice. Current efforts are directed toward

the identification method mentioned above.

Acknowledgment

This research has been supported by Research and Innovation Program of Jiangsu Postgraduates (KYLX15_0296), National

Natural Science Foundation of China (51275240 and 51075200).

Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.ymssp.

2016.06.009.

240 Y. Xu et al. / Mechanical Systems and Signal Processing 83 (2017) 228240

References

[1] G. Schweitzer, E.H. Maslen, Magnetic Bearings: Theory, Design, and Application to Rotating Machinery, Springer, Berlin, Germany, 2009.

[2] G. Li, Z. Lin, P.E. Allaire, J. Luo, Modeling of a high speed rotor test rig with active magnetic bearings, J. Vib. Acoust. 128 (3) (2006) 269281.

[3] T.M. Lim, S.B. Cheng, Parameter estimation and statistical analysis on frequency-dependent active control forces, Mech. Syst. Signal Process. 21 (5)

(2007) 21122124.

[4] R.R. Humphris, R.D. Kelm, D.W. Lewis, P.E. Allaire, Effect of control algorithms on magnetic journal bearing properties, J. Eng. Gas Turbines Power 108

(4) (1986) 624632.

[5] R.D. Williams, F.J. Keith, P.E. Allaire, Digital control of active magnetic bearings, IEEE Trans. Ind. Electron. 37 (1) (1990) 1927.

[6] T.M. Lim, D. Zhang, J. Yang, S. Cheng, S.H. Low, L.P. Chua, X. Wu, Design and parameter estimation of hybrid magnetic bearings for blood pump

applications, Mech. Syst. Signal Process. 23 (7) (2009) 23522382.

[7] N.C. Tsai, H.Y. Li, C.C. Lin, C.W. Chiang, P.L. Wang, Identification of rod dynamics under influence of active magnetic bearing, Mechatronics 21 (6) (2011)

10131024.

[8] D. Kozanecka, Z. Kozanecki, T. Lech, Experimental identification of dynamic parameters for active magnetic bearings, J. Theor. Appl. Mech. 46 (1)

(2008) 4150.

[9] O.C. De Santiago, L. San Andrs, Field methods for identification of bearng support parameterspart I: identification from transient rotor dynamic

response due to impacts, J. Eng. Gas Turbines Power 129 (1) (2007) 205212.

[10] O.C. De Santiago, L. San Andrs, Field methods for identification of bearing support parameterspart II: identification from rotor dynamic response

due to imbalances, J. Eng. Gas Turbines Power 129 (1) (2007) 213219.

[11] Y. Xu, J. Zhou, L. Di, C. Zhao, Q. Guo, Active magnetic bearing rotor model updating using resonance and MAC error, Shock Vib. 2015 (2015).

[12] H.D. Nelson, A finite rotating shaft element using Timoshenko beam theory, J. Mech. Des. 102 (4) (1980) 793803.

[13] L. San Andres, O.C. De Santiago, Identification of bearing force coefficients from measurements of imbalance response of a flexible rotor, in: Pro-

ceedings of ASME Paper GT2004-54160.

[14] O.C. De Santiago, L. San Andrs, Experimental identification of bearing dynamic force coefficients in a flexible rotorfurther developments, Tribol.

Trans. 50 (1) (2007) 114126.

- MEMB343 Mechanical Vibrations Presentation SlidesTransféré parRanjani Vasu
- UT Dallas Syllabus for opre7346.001 06s taught by Alain Bensoussan (axb046100)Transféré parUT Dallas Provost's Technology Group
- Ratio Control TechnologyTransféré parBastian Elias Cardona
- Ali et al (2007)Transféré parJulio Dutra
- Design of Type-III Control Loops Using Explicit Analytical PID Tuning TechniqueTransféré parIOSRjournal
- MEEG_5113b.pptTransféré parShyam Suryawanshi
- Dugdale 1988Transféré parjihmaine
- ThesisTransféré parKarthikeyan Krishnan
- Andersen_2002_Post-Critical Behavior of Beck's Column With a Tip MassTransféré parJuan Carlos Mosquera
- Centac Vibration Procedure%2EpdfTransféré parrashm006ranjan
- Study Guide 2013Transféré parPortia Shilenge
- vibrationTransféré parghubade_anil
- Midya ThesisTransféré parDipanjan Das
- w25-29Transféré parAtakan OzturK
- SkopjeSE40EEE_GoerguelueTransféré parAsia Ward
- Brinkgreve&Kappert&Bonnier2007Transféré parAnonymous D5s00DdU
- Dynamic Analysis of Steel Frame Using Manufactured Viscous DamperTransféré parInnovative Research Publications
- MV Motor Publication Jan 2012(2).pdfTransféré parManjunath Aithal
- A Frequency Domain TechniqueTransféré parGibs Rivera
- Lab ManualTransféré parGökhan Yüksek
- Block Contorl System FlyerTransféré parARMFX
- Fundamentals of Automatic Control for Building SystemsTransféré partrung
- Guranteed Cost Nolinear Tracking Control of Bioler Turbine Unit Using LMITransféré parAliAlMisbah
- Modelling, Identification and Experimental Validation of a Hydraulic Manipulator Joint for ControlTransféré parSeyed Ali Tabatabaee
- Experiment 1Transféré parVikram Saini
- Control System _ Closed Loop Open Loop Control System.pdfTransféré parSiva B Positive
- Gheorghe_ICAE Design and Implementation of Membrane Controllers for Trajectory Tracking of Nonholonomic Wheeled Mobile RobotsTransféré parmont21
- process control glossary.pdfTransféré parddcguy3190
- Distillation DecouplingTransféré parLim Xiu Xian
- 2018년Transféré parAngela James

- 5c572dd5f6f90493c8d1a6e73f392445 Mvsd Objective BitsTransféré parsandip Thorat
- cccccccc.pdfTransféré parsandip Thorat
- aaa.txtTransféré parsandip Thorat
- 1-s2.0-S0094576512000458-main.pdfTransféré parsandip Thorat
- Mill is 1999Transféré parsandip Thorat
- 10.1016@j.jnucmat.2017.01.033Transféré parsandip Thorat
- Venner i 2015Transféré parsandip Thorat
- Process Compressors en 2013Transféré parsandip Thorat
- GEOG_Turbocompressor_Availability.pdfTransféré parsandip Thorat
- IEEE Compressor FuelCellSystemTransféré parsandip Thorat

- Harvard Government 2000 SyllabusTransféré parJ
- UG- EC303 DSP Part-3 Fixed Point DSP Addressing Modes -PrintTransféré parShreyas Tater
- Crypto DictTransféré parDianna Ion
- FAQ QA InterviewTransféré parbalspars
- Electric Power SystemsTransféré parFebrian Dwi Cahyono
- 1992 E Applied Automatic Control Technology ShortTransféré parDany Syahputro
- Test 8. ProbabilityTransféré par台大人家教中心
- Compiler Design -- Compilers Principles and Practice - A.hosking - Compiler Course SlidesTransféré parAjay Chacko
- M5 PS1Q2Transféré parJoseph Simon Madrinan
- LT3763Transféré parjinxy06
- Sotoudeh Zahra 201108 PhdTransféré parnoneofyourbusinees
- Applied Longitudinal Data Analysis Slides PT1 August-2005Transféré parBenjamin Koh
- Vector Analysis TutorialTransféré parMohamed Mostafa
- Worded Problems With Solution in PhysicsTransféré parMa Denden
- Engg Physics NotesTransféré parravithimmegowda
- Circular Motion Lesson 2Transféré parrebbieg
- 284910991-A-Solution-for-Cross-Docking-Operations-Planning-for-Cross-Docking-Operations-Planning.pdfTransféré parLuca Marius Cristian
- 03-Hydrostatics Hydrokinematics HydrodynamicsTransféré parSarmad Khan
- NT - Rationals datingTransféré parIngrida Griksaite
- 91-025Transféré parArdie Carderas
- Harmonics and Wind POwerTransféré parcphcricri
- Excel VBA Good BookTransféré parpinku_malhotra007
- TYMCM Mass PointsTransféré parAshutosh Kumar
- The Secret of the ElectronPositron PairdocTransféré pargnostication
- solving systems of equationsTransféré parapi-234448837
- kanderson techlesson plan 4Transféré parapi-306013615
- Writing Task 1Transféré parAndrea Paola Niño
- BA201 Engineering Mathematic UNIT8 - Cramer's Rule and Inverse Matrix MethodTransféré parAh Tiang
- Finite string pattern recogniserTransféré parAditi Kanjolia
- Improved PrefixSpan Algorithm for Efficient Processing of Large DataTransféré parEditor IJRITCC

## Bien plus que des documents.

Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.

Annulez à tout moment.