Professional Documents
Culture Documents
Isma2012 0520
Isma2012 0520
Abstract
Practical engineering structures often exhibit nonlinear dynamic behaviour. It is essential to construct an
accurate nonlinear model from measured vibration data in order to predict the nonlinear dynamic
characteristics and response of these structures. This paper develops a novel technique to identify
structural nonlinearity and verify the nonlinear model. The method consists of two dynamic tests which
are constant-amplitude displacement test and constant-amplitude velocity test. Both tests use special
sinusoidal force excitations. The constant-amplitude displacement test is used to measure the nonlinear
stiffness behaviour and identify nonlinear stiffness terms from measured FRFs of the structure subjected to
sinusoidal force excitations which need to ensure the amplitudes of displacement is constant through
adjusting the amplitude of the force over the frequency range. The constant-amplitude velocity test is
similar to the constant-amplitude displacement and used to measure the nonlinear damping behaviour and
identify nonlinear damping terms. The identified nonlinear stiffness and damping terms are included into
the nonlinear model. The verification is done through comparison between predicted nonlinear behaviour
from the nonlinear model with measured nonlinear behaviour. This solution, taking advantage of modal
test and equivalent linearization theory, is direct and accurate. The solution is illustrated and verified
through a framed structure with bolted joints.
Introduction
In the dynamic model validation, high quality model can be obtained through establishing detailed finite
element model and further model updating using modal testing data. The updated model can be applied to
predict and analyze the dynamic response of the structure sufficiently and accurately within the interested
frequency range. Generally, most validation methods are based on linear system. The structural
nonlinearities, due to the difficulty to obtain accurate nonlinear parameters in the measurement, are often
ignored completely or considered as uncertain factors to deal with [1-4]. Worden [5] and Kershen [6]
comprehensively discussed nonlinear detection, localization and identification in structural dynamics, but
there still exist difficulties in nonlinear parameter identification suitable for engineering application with
standard vibration test because of the advantages and disadvantages of each method. Recently, sinusoidal
excitation and modal analysis techniques [7-9] have been developed for the nonlinear identification. The
method, based on the equivalent linearization theory [5], is expected to be applied widely in industry
because of its mathematical simplicity and relatively mature technique.
Most recently, Zang, Schwingshackl & Ewins [10] put forward an effective linearity method to deal with
the existed nonlinearity in an approach to the Sandia Structural Dynamics Challenge for model validation
of structural dynamic analysis. They[11] also investigated the influence of nonlinearity on uncertainty and
variability for dynamic models. Based on simulation of three degrees of freedom vibration system with a
weakly nonlinearity, two dynamic tests, namely constant-amplitude displacement test and constantamplitude velocity test, are exploited to identify the nonlinear stiffness and damping behaviour
respectively under the sinusoidal force excitations.
2611
2612
P ROCEEDINGS OF ISMA2012-USD2012
This paper will focus on the nonlinearity identification with practical modal test. A framed structure with
bolted joints is used as an example. The constant-amplitude displacement test is used to measure the
nonlinear stiffness behaviour and identify nonlinear stiffness terms from measured FRFs of the structure
subjected to sinusoidal force excitations which need to ensure the amplitudes of displacement is constant
through adjusting the amplitude of the force over the frequency range. Similarly, the nonlinear damping
features and term can be identified through the constant-amplitude velocity test. The verification is
undertaken through comparison between the predicted nonlinear behaviour from the nonlinear model with
the measurement. The results show the effectiveness of the proposed method.
2
2.1
In the case of a SDOF nonlinear system where the nonlinearity is additively separable, the equation of
motion can be written as:
mx f d x f s x F sin t
(1)
Where f d x represents the weak nonlinearity of damping behavior and f s x represents the weakly
nonlinear stiffness feature. Based on equivalent linearization theory, its equivalent linearization is written
as:
(2)
where ceq is the equivalent linear damping and k eq is the equivalent linear stiffness. Then, the equivalent
linear FRF is described as:
H eq
1
m j ceq keq
2
(3)
Simply assuming that the response to a sinusoidal excitation is a sinusoid at the same frequency, the
displacement and velocity can be expressed as
x A sin t
(4)
x A sin t
2
(5)
keq
1 2
f A sin sin d f A
A 0 s
(6)
ceq
f d A sin cos d g ( A)
(7)
According to Eq. (6) and Eq. (7), the equivalent stiffness is the function of displacement amplitude and the
equivalent damping is the function of velocity amplitude, respectively. Therefore, if the amplitudes of
displacement are kept constant through adjusting the amplitude of the sinusoidal force over the interested
frequency range, the influence of the nonlinear stiffness on the measured FRF data can be minimised.
Such test is called as constant displacement test and can be used to measure the relationship between
equivalent stiffness and displacement amplitude. The nonlinear damping behaviour and terms can be also
2613
obtained in the same way through acting on the constant velocity test to the structure. Afterwards, the
identified nonlinear properties can be included into the model to predict the response of nonlinear system.
2.2
There are three dynamic test methods using the sinusoidal excitation with amplitude control: constant
force test, constant displacement test, and constant velocity test. The most widely used constant force test
needs to keep the sinusoidal excitation force amplitude constant at various interested frequency points
over the frequency range during the test. The constant displacement and constant velocity tests need to
adjust the amplitude of sinusoidal force in order to keep the displacement or velocity response amplitude
unchanged during the frequency range of excitation. The difficulty in conducting constant displacement
and constant velocity tests are how to determine the amplitude of sinusoidal excitation at each frequency
so that the response amplitudes at the steady state are maintained the same value. Generally, the feedback
control method is effective, but it is time-consuming. Here, a simple method is employed without
feedback control. As the relationship between the sinusoidal excitation and response amplitudes is affected
by nonlinearity at various frequencies, it is generally not linear but remains a monotonic increasing. This
may be true for many nonlinear structures. Therefore, the function can be measured at each interested
frequency, such as
X f (F ,i ),
i 1,2,...n
F f 1 ( X , i ),
i 1, 2,...n
(8)
(9)
So, the response amplitude can be only determined by a given excitation amplitude for a frequency in zero
initial conditions
x f (F0 ),
Freq fi ,
i 1,2,...n
(10)
If this function has been gained from test data, the excitation amplitude can be only determined by a given
response amplitude
F0 f 1 ( x),
Freq f i ,
i 1, 2,...n
(11)
In practice, the function can be created by the piecewise spline interpolation polynomial, and will be
further verified in the subsequent experimental research.
Actually, the phase difference between excitation and response is, similar to the amplitude in the zero
initial conditions, a monotonic increasing or decreasing function. Therefore, the excitation amplitude and
phase difference between excitation and response can alternatively be determined by a given response
amplitude using the data from constant force test. Therefore, the FRFs can also be determined. As a result,
the constant-amplitude displacement and constant-amplitude velocity tests can be conducted virtually
from constant force excitation.
3
3.1
In order to illustrate and verify the nonlinear parameters identification method, a series of tests has been
conducted in a framed structure with bolted joints shown in Figure 1. This structure was originally
intended to represent a 3DOF system and can represent a SDOF system after removing two mass
components. Four rubber rings are installed between the four aluminum beams and the base plate
respectively to enhance the nonlinear stiffness and damping behavior of the SDOF system.
2614
P ROCEEDINGS OF ISMA2012-USD2012
3.2
To detect the nonlinearity of the SDOF system, a series of swept sinusoidal excitation with various force
levels from 0.1N to 10N are exploited to the frame and the responses of acceleration are measured.
Obviously, the overlay of measured FRFs from Figure 2 shows the distortion characteristic of acceleration
FRFs. With the increase of excitation force levels, the resonant frequency shifts downwards to the
softening system. The zoomed changes in natural frequency and phase between 36Hz and 48Hz are
overlaid in Figure 3. The modal parameters extracted from FRFs using circle fit method from ICATS
software are shown in Figure 4.
20
0.1N
0.5N
1N
3N
5N
7N
10N
Amplitude / g/N,dB
0
-20
-40
-60
-80
-100
20
30
40
50
Frequency / Hz
60
70
10
200
0.1N
0.5N
1N
3N
5N
7N
10N
150
-10
-20
-30
-40
-50
36
38
0.1N
0.5N
1N
3N
5N
7N
10N
40
42
44
Frequency / Hz
Phase / deg
Amplitude / g/N,dB
100
50
46
48
0
36
38
40
42
44
Frequency / Hz
46
48
Figure 3: The acceleration FRFs at different force levels between 36Hz and 48Hz
2615
45
3.2
44.5
Natural frequency / Hz
44
43.5
43
42.5
2.8
2.6
2.4
2.2
2
42
0
4
6
Amplitude / N
1.8
0
10
4
6
Amplitude / N
10
(a) 10N
(b) 0.1N
3.3
The parameter setting of the sinusoidal excitation used to conduct constant force test is shown in Table 1.
The accelerance FRFs obtained from constant force test and their amplitudes and phases against frequency
are plotted and shown in Figure 6 and Figure 7. The relationship of both amplitude acceleration and phase
against the force respectively at different frequencies is plotted in Figure 8. A monotonic increasing
function between amplitude and force can be seen clearly while the phase shows the decreasing function
instead.
Frequency span
Frequency interval
Force levels
Sampling Freq
Sampling time
40 Hz~47Hz
0.25Hz
0.5,1,2,3,4,5,7,9N
5120Hz
1.6s
2616
P ROCEEDINGS OF ISMA2012-USD2012
5
4
40 Hz
41 Hz
42 Hz
43 Hz
44 Hz
45 Hz
46 Hz
47 Hz
140
120
Phase / deg
Acceleration / g
160
40 Hz
41 Hz
42 Hz
43 Hz
44 Hz
45 Hz
46 Hz
47 Hz
100
80
60
40
1
20
1
5
Force / N
5
Force / N
Figure 8: The amplitude and phase of acceleration against excitation levels at different frequencies
2617
If the accelerance FRFs is H a , the mobility FRFs H v and receptance FRFs H d can be
written respectively as
H v
H d
H a
j
(12)
H a
(13)
Therefore, the displacement and velocity can be easily obtained from measured acceleration for a linear
system subjected to sinusoidal excitation without measuring directly. Based on the mobility and
receptance FRFs respectively, the amplitudes and phases of velocity vary with the excitation levels at
different frequencies are plotted in Figure 9. Similarly, the relationship between displacement and
excitation is described as Figure 10.
0.25
40 Hz
41 Hz
42 Hz
43 Hz
44 Hz
45 Hz
46 Hz
47 Hz
0.15
40
Phase / deg
Velocity / m/s
0.2
40 Hz
41 Hz
42 Hz
43 Hz
44 Hz
45 Hz
46 Hz
47 Hz
60
0.1
20
0
-20
-40
0.05
-60
1
5
Force / N
4
Force / N
Figure 9: The amplitudes & phases of velocity vary with the excitation levels at different frequencies
-4
x 10
7
6
5
-40
40 Hz
41 Hz
42 Hz
43 Hz
44 Hz
45 Hz
46 Hz
47 Hz
-60
Phase / deg
8
Displacement / m
-20
40 Hz
41 Hz
42 Hz
43 Hz
44 Hz
45 Hz
46 Hz
47 Hz
-80
-100
-120
-140
-160
1
5
Force / N
5
Force / N
Figure 10: The amplitudes & phases of displacement vary with the excitation levels at different
frequencies
As can be seen from Figure 9 and Figure 10, the relationship between the velocity/displacement and
excitation at different frequencies can be obtained. Based on those, the constant displacement and velocity
tests can be virtually conducted. For example, if the X axis (Force) in Figure 6 is replaced by the
corresponding velocity, the FRFs surface from constant velocity test can be obtained directly. And if the
surface is cut by a plane perpendicular to X axis, the intersection is the constant velocity FRFs at the
current velocity. Figure 11 reveals the relationship between constant force FRFs and constant velocity
FRFs. It also can be seen the constant force FRFs are distorted in constant velocity FRFs surface. That is
why the constant velocity/displacement test can be conducted virtually.
2618
P ROCEEDINGS OF ISMA2012-USD2012
3.4
Nonlinearity identification
3.4.1
The FRFs of the constant displacement tests under the different excitation levels are shown in Figure 12.
Suppose the equivalent stiffness is
keq a b x
D
D
Amplitude / dB,m/N
-80
D
-82
D
D
-84
D
D
-86
D
-88
40
D1
D2
-40
D3
3
4
5
6
7
8
-60
D4
-80
D5
D6
-100
D7
-120
D8
-140
D10
D9
D10
-90
(14)
-20
Phase / deg
-78
-160
41
42
43
44
Frequency / Hz
(a) Amplitude
45
46
47
40
41
42
43
44
Frequency / Hz
45
46
47
(b) Phase
2619
Displacement
Equivalent Stiffness
Natural Frequency
[10-3m]
[104m/N]
[Hz]
D1
0.0719
8.9595
43.9857
D2
0.1211
8.8508
43.7182
D3
0.1703
8.7613
43.4967
D4
0.2195
8.6883
43.3150
D5
0.2688
8.6272
43.1624
D6
0.3180
8.5749
43.0314
D7
0.3672
8.5272
42.9117
D8
0.4164
8.4831
42.8005
D9
0.4656
8.4440
42.7017
D10
0.5149
8.4106
42.6173
No.
9.271910
m
5
-3.864110
1/2
x 10
Natural frequency / Hz
44.5
44
43.5
43
42.5
42
0
0.2
0.4
0.6
0.8
1
-3
Displacement / m
x 10
Test
Fit
9.2
Test
Fit
9
8.8
8.6
8.4
8.2
0
0.2
0.4
0.6
0.8
1
-3
Displacement / m
x 10
Figure 13: Curve fitting and the test of both natural frequencies and the equivalent stiffness against the
displacement amplitude
3.4.2
Similar to the nonlinear stiffness identification, the amplitude and phase characteristics of Constant
velocity FRFs within the frequency range from 40 to 47Hz is plotted in Figure 14.
2620
P ROCEEDINGS OF ISMA2012-USD2012
V
V
-30
-32
3
4
V5
-34
V6
-36
V7
-38
V9
V8
42
43
44
Frequency / Hz
45
46
40
V3
V2
V4
20
V5
V6
V7
-20
V8
V9
V10
-60
-40
41
V1
-40
V 10
40
60
Phase / deg
Amplitude / dB,m/sN
47
40
41
42
43
44
Frequency / Hz
45
46
47
ceq p q x
(15)
The parameters p, q, and the index n can be determined with least squire method using the data from Table
4 obtained through modal analysis. The fitting results are shown in Table 5.
Velocity
Equivalent Damping
Damping ratio
[m/s]
[Ns/m]
[%]
v1
0.0199
14.0165
2.1618
v2
0.0329
14.8286
2.3007
v3
0.0458
15.2921
2.3845
v4
0.0588
15.6806
2.4553
v5
0.0718
15.9693
2.5093
v6
0.0848
16.2260
2.5574
v7
0.0977
16.5065
2.6089
v8
0.1107
16.7664
2.6569
v9
0.1237
16.9640
2.6946
v10
0.1367
17.0932
2.7205
No.
12.3924
13.0957
1/2
3
Test
Fit
2.8
2.6
2.4
2.2
2
0
0.05
0.1
0.15
0.2
Velocity / m/s
2621
19
18
Test
Fit
17
16
15
14
13
0.25
0.05
0.1
0.15
0.2
Velocity / m/s
0.25
Figure 15: Curve fitting and the test of both the viscous damping ratio and the equivalent viscous damping
against the velocity amplitude
3.5
Based on the nonlinearity identification of stiffness and damping property, the equation of motion of the
bolted frame can be written as
1
mx c0 x cn x x 2 k0 x kn x x 2 F sin t
(16)
keq k0 0.9153kn x 2
(17)
ceq c0 0.9153cn x 2
(18)
k0 a
kn b / 0.9153
(19)
c0 p
cn q / 0.9153
The results of parameter identification are shown in Table 6.
m ( kg )
k0 ( N/m )
kn ( N/m1.5 )
c0 ( Ns/m )
cn ( Ns1.5/m1.5 )
1.1730
9.2719104
-4.2216105
12.3924
14.3074
2622
P ROCEEDINGS OF ISMA2012-USD2012
Test - 0.5N
Test - 3N
Test - 7N
Prediction - 0.5N
Prediction - 3N
Prediction - 7N
Acceleration /g
5
4
3
2
1
40
41
42
43
44
Frequency /Hz
45
46
47
160
-2
140
-4
120
Phase /deg
Amplitude /dB,g/N
Figure 16: The overlay of acceleration responses between the model prediction and the measurement
-6
Test - 0.5N
Test - 3N
Test - 7N
Prediction - 0.5N
Prediction - 3N
Prediction - 7N
-8
-10
-12
-14
40
41
42
43
44
45
Frequency /Hz
Test - 0.5N
Test - 3N
Test - 7N
Prediction - 0.5N
Prediction - 3N
Prediction - 7N
100
80
60
40
46
47
20
40
41
42
43
44
45
Frequency /Hz
46
47
Figure 17: The overlay of acceleration FRFs between the model prediction and the measurement
Conclusions
Nonlinearities are very common in real structures. In most cases, the functional types of nonlinearities are
unknown; therefore it becomes crucial to decide the nonlinear functional forms. The constant
displacement test and the constant velocity test under the ssinusoidal force excitations can be used to
identify the nonlinear stiffness and the nonlinear damping behaviours respectively from measured FRFs of
the structure. Both methods are mainly based on equivalent linearization theory and only considered the
first harmonic in the nonlinear identification. The experimental research based on a SDOF system with a
weakly nonlinear system shows the satisfactory results. Future work will extend to a MDOF system and a
FEM model, for example, the bolted joint beam etc.
Acknowledgements
The financial support of the National Natural Science Foundation of China (Project No. 51175244) and
Research Fund for the Doctoral Program of Higher Education of China (Project No. 20093218110008) are
gratefully acknowledged. Zang also acknowledges the project sponsored by the Scientific Research
Foundation for the Returned Overseas Chinese Scholars, State Education Ministry and the Fundamental
Research Funds for the Central Universities of China (Project No. kfjj20110204).
2623
References
[1] R.G. Ghanem, A. Doostan and J. Red-Horse, A probabilistic construction of model validation,
Computer Methods in Applied Mechanics and Engineering, Vol.197 No.29-32 (2008), pp. 2585-2595.
[2] T. Hasselman and G. Lloyd, A top-down approach to calibration, validation, uncertainty
quantification and predictive accuracy assessment, Computer Methods in Applied Mechanics and
Engineering, Vol.197 No.29-32 (2008), pp. 2596-2606.
[3] L.G. Horta, et al., NASA Langley's approach to the Sandia's structural dynamics challenge problem,
Computer Methods in Applied Mechanics and Engineering, Vol.197 No.29-32 (2008), pp. 2607-2620.
[4] J. McFarland and S. Mahadevan, Error and variability characterization in structural dynamics
modeling, Computer Methods in Applied Mechanics and Engineering, 2008. 197(29-32), pp. 26212631.
[5] K. Worden and G.R. Tomlinson, Nonlinearity in structural dynamics: detection, identification, and
modeling, 2001: Taylor & Francis.
[6] G. Kerschen et al., Past, present and future of nonlinear system identification in structural dynamics,
Mechanical Systems and Signal Processing, Vol.20 No.3 (2006), pp. 505-592.
[7] O. Arslan, M. Aykan and H. N Ozguven, Parametric identification of structural nonlinearities from
measured frequency response data, Mechanical Systems and Signal Processing, Vol.25 No.4 (2011),
pp. 1112-1125.
[8] O. Arslan, M. Aykan and H. N Ozguven, Modal identication of non-linear structures and the use of
modal model in structural dynamic analysis, in: 26th International Modal Analysis Conference
IMAC, Orlando, USA, 2008, Paper no.95.
[9] A. Carrella and D.J. Ewins, Identifying and quantifying structural nonlinearities in engineering
applications from measured frequency response functions, Mechanical Systems and Signal
Processing, Vol.25 No.3 (2011), pp. 1011-1027.
[10] C. Zang, C.W. Schwingshackl and D.J. Ewins, Model validation for structural dynamic analysis: An
approach to the Sandia Structural Dynamics Challenge, Computer Methods in Applied Mechanics
and Engineering, Vol.197,No 29-32,(2008), pp. 2645-2659.
[11] C.W. Schwingshackl, C. Zang and D. J. Ewins. The influence of nonlinearity on uncertainty and
variability for dynamic models, Proceedings of the 1st International Conference on Uncertainty in
Structural Dynamics. Sheffield, UK, May 2007, pp. 311-319.
2624
P ROCEEDINGS OF ISMA2012-USD2012