You are on page 1of 6

30 PIERS Proceedings, Xi’an, China, March 22–26, 2010

Updating Methods for Antenna Servomechanism Structures


Hong Bao, Congsi Wang, and Jun Cheng
Key Laboratory of Electronic Equipment Structures Design of Ministry of Education
Xidian University, Xi’an, China

Abstract— A new updating method is presented based on the model condensation technique
and frequency response function (FRF) for antenna servomechanism dynamic structures As a
result of the introduction of the ratio of damping to stiffness, this method reduce the numbers
of parameters than the traditional FRF updating methods. The finite element model (FEM)
condensation technique is used to overcome the disadvantage of the traditional FRF updating
methods that the testing mode dimension is too large, it decrease the request of the testing model
dimension. Finally the numerical result shows the validity and feasibility of the method.

1. INTRODUCTION
Antenna servomechanism is a key part of the antenna, its capability determines the level of antenna’s
performance, the rationality of modeling the antenna servomechanism is very useful for the design
of the follow-up control system. However, there are a large number of uncertainties factors such
as damping, friction and gap and so on in practice works. Thus, the error exists between the
mathematical model and the physical model. It would need to update the mathematical model
when these errors beyond expectation, it means the dynamic structure updating [1, 2]. In the
past 30 years, dynamic model updating methods have great development, especially the model
updating using vibration data, which contains mode and FRF. Mode method is the algorithm
that adjusting the forecast parameters by the eigenvalue and eigenvector. At present, the mode
method developed updating process based on a few front degree of freedom (DOF) of the measured
modal frequencies and mode shapes [3], however, relying solely on a few modes to update FEM is
not entirely reasonable, for the mode set is incomplete, the updating model can only be used on
subspace of mode space, so the results just can be used in a very small range, the other shortcoming
of the mode method is that it can not be used to update damping.
The updating method of using FRF data is developed in recent years, Xu Zhang Ming [4] stud-
ied the frequency response function that get from testing and finite element model analysis, and
deduced an updating method, which based on the sensitivity analysis of FRF. Hemez [5] analyzed
the problem of using FRF to update the dynamic model, then improved this method and applied
it to the actual mode. Ting [6] presented an improved calculation method of the sensitivity of FRF
to study the technology of updating the FRF model. This approach overcome the need of corre-
sponding mode shape and it’s frequency range for updating calculation is very wide, but there are
no updating to the structure damping. RM Lin [7] improved the FRF method with combining the
structural damping characteristics, and achieved the identification updating of damping. However,
the DOF of finite element model are generally much larger than the testing’s. Because of neglecting
the problem of model reduction, Lin’s method have many limitations in practical applications. For
studying the model reduction, Xu Zhang Ming provided another improved method of reducing the
FEM using FRF [8], this method presented an independent model reduction of the structure model
unit, but in the end it needed to reduce the stiffness or mass of the unit, so it would increase the
step of reduction, and this method updated mass, stiffness and damp separately, it also increased
the number of updating parameters, it would need more frequency points, and would increase the
workload.
An improved method for damping updating using the FRF has been presented in this paper, for
FRF can not reduce the model and have low efficiency of calculating. This method combines the unit
model reduction technology, and introduces the concept of damping ratio of stiffness based on the
important of the structure damping in the project, it reduces the numbers of updating parameters
and the work of calculation. The finally numerical simulation results prove the feasibility and
reliability of the method.
2. PRINCIPLE AND PROBLEM
Usually the structural dynamics equations can be expressed as:
M θ̈ + C1 θ̇ + iC2 θ + Kθ = F (1)
Progress In Electromagnetics Research Symposium Proceedings, Xi’an, China, March 22–26, 2010 31

where M and K are mass and stiffness matrices, respectively, C1 and C2 are viscous and struc-
tural damping, respectively From text [9], generally the damping force and experimental frequency
(speed) is almost irrelevant, against that, it shows that the energy consumption of damping in
a cycle is proportional to the square of the amplitude, that is We = aX 2 , This equation change
into an equivalent viscous damping can be expressed as C1 = kγ ω , substituting it into Eq. (1), the
frequency response function of the model can be written as:
1
H(ω) = (2)
−ω 2 M + iD + K
where D = ωC1 + C2 , from reference [6], the dynamic stiffness matrix of the structure can be
showed as:
Z = K + iD − ω 2 M (3)
When the product of the frequency response function HX that be tested and the dynamic
stiffness matrix ZA that be analysis in the theoretical is identity matrix, that ZA HX = I means
that the theoretical model and the actual physical model is fully consistent, when difference presence
between them, their product would not be a identity matrix, so Lin [7] update modal with this
advantage, the basic ideal present as follow:

{HX }j − {HA }j = −[HA ]∆Z{HX }j (4)

Here ∆Z is the minus between the actual dynamic stiffness matrix of the structure and the theoret-
ical dynamic stiffness matrix, {HA }j and {HX }j are the jth column of frequency response function
via FEM analysis and test, respectively. The total differential equations of the formula (3) can be
expressed as:
XN
∂Z]
∆Z = ∆pk (5)
∂pk
k=1

The pk is the parameter of the various cells, put (5) into (4) and multiply −[ZA ] in the two side
of the equation at the same time, then the equation can be gained via the nature of the FRF that
[Z][H] = [I] as:
XN
∂ZA
[I]j − ZA {HX }j = ∆p{HX }j (6)
∂p
k=1

The updating numerical value of various cells can be got by solving ∆p from the front equation.
The problem is that this method requested the same dimension of the actual test and the FEM
in the application, it is difficult in practice to meet this require, it destroy the connection of the
frame of element matrix, and make the element matrix which has been updated lose it’s physical
meaning when solving formula (6) with multiplication and division.
3. MODEL REDUCTION AND UPDATING
As the measurement dimension smaller than the dimension of the finite element model, it needs
indent the finite element model dimension to the measurement dimension, so the dynamic stiffness
matrix and the frequency response function matrix can be divided into mesurable DOF n and
inmesurable DOF s, there can get the equation via realigning the dynamic stiffness matrix and the
frequency response function matrix:
· ¸· ¸ · ¸
Znn Zns Hnn Hns Inn
= (7)
Zsn Zss Hsn Hss Iss

The equation can be got from (7)


−1 −1
Hnn = Znn − Zns Zss Zsn (8)

Let’s put the reduction matrix Z red = Znn and Znn Hnn = Inn into (8), there can be,

Z red = Hnn
−1 −1
= Znn − Zns Zss Zsn (9)
32 PIERS Proceedings, Xi’an, China, March 22–26, 2010

Formula (9) is the reduction matrix of modal dynamic stiffness matrix Z, the (n + s)-dimensional
matrix can be reduced into a n-dimensional matrix which has the same DOF with the actual testing
matrix, then use the [Z red ] matrix that has been reduced instead of ZA in (6),
S∆p = ∆I (10)
 
∂Z red ∂Z red ∂Z red
   ∂p1 {H x (ω1 )}j {Hx (ω1 )}j ··· {Hx (ω1 )}j 
 ∂p2 ∂pN 

 [I]j − Z red {Hx (ω1 )}j 
  ∂Z red 
   ∂Z red ∂Z red
[I]j − Z red {Hx (ω2 )}j  {Hx (ω2 )}j {Hx (ω2 )}j ··· {Hx (ω2 )}j 

∆I = , S =  ∂p1 ∂p2 ∂pN 

 ··· 
  
   ··· ··· ··· ··· 
[I]j − Z red {Hx (ωn )}j  ∂Z red red red 
∂Z ∂Z
{Hx (ωn )}j {Hx (ωn )}j ··· {Hx (ωn )}j
∂p1 ∂p2 ∂pN
{Hx (ωi )}j is the frequency response function measured when frequency is ωi , the equation can be
got from (9),
∂Z red ∂Znn ∂Zns −1 ∂Z −1 −1 ∂Zsn
= − Zss Zsn − Zns ss Zsn − Zns Zss (11)
∂pi ∂pi ∂pi ∂pi ∂pi
Put the theory unit parameters as the initial value into (10) and (11), ∆p can be got in (10), but
S, ∆I are complex number can be known from (2) and (3), so the number of mass sell and stiffness
sell matrix in ∆p also may be complex number, this is inconsistent with the physical concept, so
divide the real and imaginary part can get,
· ¸½ ¾ ½ ¾
Re(S) −Im(S) Re(∆p) Re(∆I)
= (12)
Im(S) Re(S) Im(∆p) Im(∆I)
In order to facilitate the solution of equation, introduction the error rate of variable Xi and Yi of
mass and stiffness unit. So consider there is the connection between the initial value and the actual
value of finite element model,
N
X
M = (1 + Xi )Mie (13)
i=1
XN
K = (1 + Yi )Kie (14)
i=1

There, N is the sum of the frame cell, Mie and Kie means the ith unit’s mass and stiffness matrix.
In the finite element model because the cell structure’s damping matrix and stiffness matrix has
the same form, so use the stiffness damping coefficient γ, so the ith unit’s damp matrix can be
expressed as,
Die = γi Kie (15)
According formula (13)–(15) the error of frame dynamic stiffness matrix can be expressed as,
N
X N
X
∆Z = ∆K + i∆D − ω 2 ∆M = (Xi + iγi )[K e ]i − ω 2 Yi Mie (16)
i=1 i=1

Via using the above variable, it can change the parameter problem about unit stiffness matrix,
mass matrix and damping matrix into the problem about solving the error rate variable and stiffness
damping ratio coefficient, put (13) and (14) into (11), it can gain that,
∂Z red
= −ω 2 Mie + ω 2 Mie Zss
−1
Zsn − ω 2 Zns Zss
−1
Mie Zss
−1
Zsn + ω 2 Zns Zss
−1
Mie (17)
∂Xi
∂Z red
= Kie − Kie Zss
−1 −1 e −1
Zsn + Zns Zss Ki Zss Zsn − Zns Zss −1 e
Ki (18)
∂Yi
∂Z red ∂Z red
= i (19)
∂γi ∂Yi
Progress In Electromagnetics Research Symposium Proceedings, Xi’an, China, March 22–26, 2010 33

Combine (16) and (12), can get that,


½ ¾
{Xi }
Re({∆p}) = (20)
{Yi }
½ ¾
{γi }
Im({∆p}) = (21)
{0}

Put (20), (21) into (12), and eliminate {0} vector, the finally result can be that:
· ¸( {X}
) ½ ¾
Re([S]) (−Im([S]))K Re({∆I})
{Y } = (22)
Im([S]) (Re([S]))K {γ}
Im({∆I})

There (([S]))K means the part of updating parameter that connect with siffness matrix. Put the
{X}, {Y } and {γ} solved from (22) into formula (13)–(15), can gain the [K], [M ] and [D] updated.
According to the description above, the model updating method steps summarized as follows:

1. According the definition of frequency response function obtaine [Z] using the numerical pa-
rameters finite element model as the initial value. Then according (8) and the measurable
node, gain the [Z red ]0 that updated the modal;
2. Via (17), (18) and (19) and unite the first step [Z red ]0 obtaine the sensitivity matrix S that
after the reduction of dynamic stiffness matrix, and finally express the equation as (12);
3. Translate (12) into (22), and gain the unknown {X}, {Y } and {γ}, then obtaine updated
matrix [K], [M ] and [D];
4. Repeat the above steps until kZ red HX − Ik∞ ≤ ε (ε is the given iteration stop condition).

4. SIMULATION RESULTS
In order to verify the accuracy of methods, in this paper, the multi-axis gear transmission system
(Figure 1) gave in the paper of Wu [8] using the above analysis of the theory of model updating.
From the reference literature can get the model’s finite element dynamic equation and its various
units’ mass and stiffness matrix:
Table 1: The mass matrix of the unit.
M1 M2 M3 M4 M5 M6
(1, 1) 9721.9 (2, 2) 198.925 (3, 3) 0.0501 (4, 4) 0.0257 (2, 2) 104.7260 (6, 6) 26.8135
(2, 1) 1.0 (3, 2) −1.1775 (4, 3) 0.0250 (5, 4) 0.0129 (6, 2) −2.1950 (7, 6) −0.0805
(1, 2) 1.0 (2, 3) −1.1775 (3, 4) 0.0250 (4, 5) 0.0129 (2, 6) −2.1950 (6, 7) −0.0805
(2, 2) 988.3 (3, 3) 403.0158 (4, 4) 234.05 (5, 5) 234.0257 (6, 6) 300.911 (7, 7) 81.6191

Table 2: The stiffness matrix of the unit: 1.0e+006∗ .


K1 K2 K3 K4 K5 K6
(1, 1) 5.4600 (2, 2) 9.3508 (3, 3) 35.2800 (4, 4) 68.7000 (2, 2) 45.0372 (6, 6) 152.0510
(2, 1) −5.4600 (3, 2) 4.0337 (4, 3) −35.2800 (5, 4) −68.7000 (6, 2) 9.3473 (7, 6) 18.0806
(1, 2) −5.4600 (2, 3) 4.0337 (3, 4) −35.2800 (4, 5) −68.7000 (2, 6) 9.3473 (6, 7) 18.0806
(2, 2) 5.4600 (3, 3) 1.7400 (4, 4) 35.2800 (5, 5) 68.7000 (6, 6) 1.9400 (7, 7) 2.1500
Note: The (x, y) in table means the (x, y) position in matrix, there be 0 where had no data.

Assume the unit damping stiffness rate γ expressed in Table 3.


Assume the test frequency ω = [10 20 50 100], so there can get the full FRF combining the
above data. In this paper, they use node 2 as the actuator, choosing the FRF on node 1, 2, 3, 4 go
along the iteration. At the same time, in order to verify the effectiveness of the algorithm, assuming
the unit model error rates that assigned the finite element initial number shown in Tables 4 and 5.
And assume the damping stiffness ratio of the every finite element unit no updating are all 0.
Table 6 is the damping element stiffness of each unit after the 5th iteration.
34 PIERS Proceedings, Xi’an, China, March 22–26, 2010

Figure 1: The picture of multi-axis gear transmission system.

The stiffness matrix and mass matrix after 5th iteration,


 
9722.1 1.000 0 0 0 0 0
 1.000 1291.9 −1.1775 0 0 −2.194 0 
 
 0 −1.1775 403.07 0.025 0 0 0 
 
M =  0 0 0.025 234.08 0.0129 0 0 
 0 0 0 0.0129 234.03 0 0 
 
 0 −2.194 0 0 0 327.64 −0.0805 
0 0 0 0 0 −0.0805 81.620
 
5.46 −5.46 0 0 0 0 0
 −5.46 59.85 4.034 0 0 9.348 0 
 
 0 4.034 37.02 −35.28 0 0 0 
 
K = 1.0e + 006 ×  0 0 −35.28 103.98 −68.70 0 0 
 0 0 0 −68.70 68.70 0 0 
 
 0 9.348 0 0 0 153.51 18.02 
0 0 0 0 0 18.02 2.143

After comparative the stiffness matrix, mass matrix given in literature [8] and a given initial
value in this article, as well as updating the various units of the stiffness, mass, it can show that
the relative error of cell parameters that adding the man-made unit stiffness (the biggest difference
between the 70 times) and damping matrix (the biggest difference between the 30 times) before
updating at the maximum of 0.031% (occurred in the stiffness matrix in the first 6 units) others
are all less than 0.03% after 5 iterations with the using of the method of this paper It should
be noted that ignores the existence of damping before updating, while after updating (compare
Table 6 and Table 3), they are very close.Analysed the FRF using the system equation before and
after updating, the frequency response function curves after updating and the frequency response
function curves using the analysis method in paper [8] can be showed.
Therefore, combining the above data, the thing can be believed that this method in this paper

Table 3: Unit damping stiffness rate.

Unit number 1 2 3 4 5 6
Unit damping
0 0.4 0.6 0 0 0
stiffness rate γ

Table 4: Stiffness error rate Y . Table 5: Mass error rate X.


Unit number 1 2 3 4 5 6 Unit number 1 2 3 4 5 6
Error rate/% 0 −70 −50 0 −10 15 Error rate/% −20 0 20 0 30 0
Progress In Electromagnetics Research Symposium Proceedings, Xi’an, China, March 22–26, 2010 35

Table 6: Damping and stiffness ratio of each unit after updating.

Unit number 1 2 3 4 5 6
Damping and
7.82e − 018 0.4000 0.6000 4.11e − 013 4.03e − 015 3.58e − 013
stiffness ratio γ

not only can reduce the stiffness of each element of error significantly, but also can identify all of
the damping unit correctly that it has be assumed at the beginning.
5. CONCLUSION
In this paper, a new updating method be showed based on the traditional FRF model updating
method, its characteristic showed as follow:
1. It can update the damping model, and overcome the shortage of modal updating method.
2. It reduce the test number using the model reduction, and introduce into the unit damping stiff-
ness ratio, compare to the traditional FRF method it reduce the unknown value of updating,
improve computational efficiency, and the results of updating are more accurate.
3. Consider the effect of the nonlinear factors, it is the focus of future research.
ACKNOWLEDGMENT
This work was supported by National Natural Science Foundation of China (50775170, 50805111)
and Natural Science Basic Research Plan of Shaanxi Provincial (SJ08E203) and Basic Research
Foundation of Universities, Ministry of Education of China.
REFERENCES
1. Zhu, H., B. Xu, and Y. Huang, “Comparison and evaluation of analytical approaches to struc-
tural dynamic model correction,” Advances in Mechanics, Vol. 32, No. 4, 513–525, 2002.
2. Li, H. and H. Ding, “Progress in model updating for structural dynamics,” Advances in Me-
chanics, Vol. 35, No. 2, 170–180, 2005.
3. Bao, H., D.-Z. Zhao, and Y.-Q. Qiao, “Methods updating for the dynamic model with orthog-
onal vector basis,” Journal of Xidian University, Vol. 36, No. 1, 151–155, 2009.
4. Xu, Z.-M., R.-Y. Shen, and H.-X. Hua, “Updating finite element midel by the sensitivity
analysis of FRF correlation function,” Journal of Mechanical Strength, Vol. 25, No. 1, 5–8,
2003.
5. Hemez, F. M. and G. W. Brown, “Improving structural dynamics models by correlating
simulated to measuredfrequency response functions,” 39th AIAA/ASME/ASCE/AHS/ASCS
Structures, Structural Dynamics and Materials Conference, AIAA-98-1789, 772–782,
Apr. 1998.
6. Ting, T., “Design sensitivity analysis of structural frequency response,” 4th
AIAA/USAF/NASA/OAI Symposium on Multidisciplinary Analysis and Optimization,
AIAA-92-4799-CP, 878–881, Sep. 1992.
7. Lin, R. M. and J. Zhu, “Model updating of damped structures using FRF data,” Mechanical
Systems and Signal Processing, Vol. 20, 2200–2218, 2006.
8. Xu, Z.-M., T.-M. Gao, and R.-Y. Shen, “Improved finite element model updating method
based on frequency response functions,” Journal of Vibration and Shock, Vol. 21, No. 3, 43–46,
2002.
9. Fu, Z.-F., Vibration Modal Analysis and Parameter Identification, Machine Press, Beijing,
China, 1990.
10. Wu, J. S. and C.-H. Chen, “Trisonal vibration analysis of gear-branched systems by finite
element method,” Institute of Naval Architecture and Marine Engineering, 2000.

You might also like