Vous êtes sur la page 1sur 12

Chapter 37

Structural Dynamics Model Calibration and Validation


of a Rectangular Steel Plate Structure
Hasan G. Pasha, Karan Kohli, Randall J. Allemang, Allyn W. Phillips, and David L. Brown
Abstract To characterize the dynamics of a structure accurately and to minimize uncertainties, it is important to perform
modal testing in various configurations. However, performing such rigorous testing of structures can be a resource intensive
process. In addition, simulating certain conditions may not be possible in the lab. Developing a calibrated and a validated
model that can predict the dynamic response of a structure accurately can be a key to address this issue. In this paper, a model
calibration and validation case study performed on a rectangular steel plate structure is presented. The geometric and material
properties used in the model were updated to calibrate the model. The accuracy of the calibrated model was confirmed by
performing a validation process involving perturbed mass and constrained boundary condition FE modal analysis and modal
testing. The validation criteria were achieved using the calibrated model and thus proved that the model could reliably predict
the dynamic response of the structure.
Keywords Model verification Model calibration Model validation Modal correlation Model updating Perturbed
boundary conditions Constrained boundary condition testing

Notation
Symbol



b
fij
h
i
j
l
D
E
CMIF
FEA
MRIT

Description
Mass per unit area of a plate (kg=m2 )
Density of the material (kg=m3 )
Poisson ratio
Width of the plate (m)
Modal frequency (Hz)
Thickness of the plate (m)
Mode index, number of half waves in mode shape along horizontal direction
Mode index, number of half waves in mode shape along vertical direction
Length of the plate (m)
Flexural rigidity (N  m2 )
Youngs modulus (Pa)
Complex Mode Indicator Function
Finite Element Analysis
Multi-reference Impact Testing

37.1 Introduction
Testing a structure to determine its dynamic characteristics is both time consuming and expensive. This situation has made it
necessary to develop analytical/mathematical models that can predict the response of system in real time without performing
actual tests. Such models need to undergo verification, calibration and validation with actual test data before they can be
relied upon to extrapolate results for future studies. In this context, verification and validation (V&V) are defined as below:
H.G. Pasha () K. Kohli R.J. Allemang A.W. Phillips D.L. Brown
University of Cincinnati Structural Dynamics Research Lab (UC-SDRL), Cincinnati, OH, USA
e-mail: pashahg@mail.uc.edu
The Society for Experimental Mechanics, Inc. 2015
H.S. Atamturktur et al. (eds.), Model Validation and Uncertainty Quantification, Volume 3, Conference Proceedings
of the Society for Experimental Mechanics Series, DOI 10.1007/978-3-319-15224-0_37

351

352

H.G. Pasha et al.

Fig. 37.1 Verification,


calibration and validation
processes

Verification: refers to process of determining that a model implementation accurately represents the developers conceptual
description of the model and the solution to the model [1]. Frequently, the verification portion of the process answers the
question Are the model equations correctly implemented?
Validation: refers to the process of determining the degree to which a model is an accurate representation of the real world
system from the perspective of the intended uses of the model [1]. Frequently, the validation process answers the question
Are the correct model equations implemented?
In this paper, a FE model of the plate structure was developed to determine its dynamic characteristics (modal frequencies
and mode shapes). A step by step approach was taken to verify, calibrate and validate the results generated by the model in
different test configurations (Fig. 37.1).

37 Structural Dynamics Model Calibration and Validation of a Rectangular Steel Plate Structure

353

37.2 Verification
37.2.1 Analytical Modeling and Formulation
Ideally, a plate is considered a two-dimensional layer of elastic material made up of a sheet that lies in a plane. A plate
as a structure possesses bending rigidity due to its thickness and elasticity. As a result, when the plate vibrates, it deforms
primarily by flexing perpendicular to its own plane.
The following assumptions were made while performing the analytical formulation of the rectangular plate structure [9]:

The plate is composed of homogeneous, linear, elastic, isotropic material


The plate is flat and has constant thickness
The plate is thin, with a thickness of less than 1=10 of the minimum lateral plate dimension
The in-plane load on the plate is zero and it deforms through flexural deformation
The deformation is small in comparison with the thickness of the plate. There is no deformation at normal to the mid
surface or nodal lines. In addition, the rotary and shear deformations are ignored.
In the case of free vibration, total transverse deformation is defined as the sum of all modal deformations such that
XX


zD
(37.1)
Aij zQ ij sin 2fij tCij
i

where zQ ij is the mode shape of the mode .i; j/, Aij is the modal amplitude and ij is the phase. The modal frequency of the
q
2
Eh3
plate is defined as fij D 2lij 2 12.1
2 / , where ij is a dimensionless parameter, which is generally a function of the mode
indices .i; j/. Flexural rigidity of plate is defined as D D

Eh3
.
12.1 2 /

37.2.2 Verification
Verification is the process of comparing different analytical models to be certain that the analysis results are correct [1, 2].
In this case, a FE model is compared to the analytical, closed form results for a rectangular plate. An analytical rectangular
plate model, with aspect ratio of 1:5 was developed, such that the first mode is a torsion mode roughly at 40 Hz and the
second mode is a bending mode roughly at 44 Hz. Closed form expressions [9] were used to develop an analytical model
(Figs. 37.2 and 37.3).
A comparable rectangular plate FE model was developed using the dimensions and material properties chosen for the
analytical closed form model of the plate. Analytical modal analysis was performed on the plate FE model. The modal
frequencies and mode shapes for the first six deformation modes were retrieved and compared with the analytical model
prediction.
The modes were in the expected sequence and the relative difference in modal frequencies was less than 2%. This verified
the FE model (Table 37.1).

37.3 Calibration
Calibration of a model involves comparing the analytical results to equivalent experimental results. Calibration in this case
involves correlating the modal parameters obtained from an analytical modal analysis of the rectangular plate FE model and
impact testing of the fabricated rectangular plate structure. The details of the various activities performed in the calibration
process are explained in the subsequent sections.

354

H.G. Pasha et al.

Fig. 37.2 2-D drawing showing


plate dimensions

4.f0.25in.(1/4-20)

2.00in.

1.00

f0.19

2.f0.25in.(1/4-20)

2.f0.25(1/4-20)

34.00in.

1.00
22.50in
Plate Thickness-0.25in

Fig. 37.3 Rectangular plate model. (a) CAD model. (b) Mesh

Table 37.1 Preliminary


verification of the FE
modelmodal frequency
comparison

Mode #
1
2
3
4
5
6

Description
First torsion
First X bending
Second torsion
First Y bending
Second X bending
Anti symmetric X bending

Modal frequency (Hz)


Analytical FE
41:55
41:34
44:59
43:95
96:3
95:12
103:8
102:99
121
119:16
139:3
137:39

Rel. diff. (%)


0:51
1:46
1:24
0:79
1:54
1:39

37 Structural Dynamics Model Calibration and Validation of a Rectangular Steel Plate Structure

355

37.3.1 Analytical Modal Analysis


r

Analytical modal analysis was performed in ANSYS Workbench 14:5. The default material model for steel in the ANSYS
material library was chosen. The modal frequencies and mode shapes were extracted in the frequency range 0250 Hz.

37.3.2 Experimental Modal Analysis


The plate structure was tested using the multi-reference impact testing (MRIT) technique with 21 accelerometers mounted on
the rectangular plate at locations identified in Fig. 37.4. For simulating the free-free boundary condition, the plate was initially
suspended using shock cords. However, the support system was dynamically coupled with the third mode of the rectangular
plate (second torsion mode) and it acted as a vibration absorber. Subsequently, the plate was supported on racquet balls,
as shown in Fig. 37.5. Though the racquet ball supports offered a lesser frequency separation ratio between the rigid body
modes and the first elastic mode (compared to the shock cords), they were well isolated from the test structure [11].

Fig. 37.4 Rectangular platereference locations

Fig. 37.5 Rectangular platesupport system

356

H.G. Pasha et al.

Fig. 37.6 Complex mode indicator function

On processing the measured data, nine deformation modes in the 0250 Hz frequency range were identified, as evident
from the CMIF plot shown in Fig. 37.6. The first torsion and the first bending mode were at 42:16 Hz and 44:16 Hz
respectively. The mode shapes are shown in Fig. 37.7.

37.3.3 Modal Correlation


The relative difference between the modal frequencies obtained from FE modal analysis and impact testing is >23 %.
The sequence of the modes was established to be in the same order for both cases by viewing the mode shape animations.

37.3.4 Model Calibration


From the correlation of the modal parameters from the preliminary analytical modal analysis and impact testing, it was
identified that there were various sources of uncertainty. On the modeling side, the sources of uncertainty were geometric
and material properties of the steel plate. In addition, the soft supports had not been modeled. On the testing side, the support
system, location of supports, sensor calibration, impact hammer tip and signal processing parameters were considered to be
the sources of uncertainty. The relative difference in modal frequencies was desired to be less than 2 % and was set as the
calibration criteria or calibration metric.

37 Structural Dynamics Model Calibration and Validation of a Rectangular Steel Plate Structure

357

Fig. 37.7 Experimental mode shapes

37.3.4.1

Model Updating

Geometric properties: It was established that the plate thickness, originally intended to be 0:25 in: was not constant. The
plate was weighed (52:5 lbm ) and the thickness of the plate model was adjusted (0:243 in:) to match the measured weight
of the plate.
Material model: The relative difference for the third mode (second torsion mode) was reasonably high compared to other
modes. A parametric study, in which the effect of varying the Youngs modulus, mass density and Poisson ratio on the
modal frequencies, was studied. For the calibrated model, the material properties were established as: Youngs modulus
E D 2:05x1011 Pa .2:9734x107 psi/, mass density  D 7; 850 kg=m3 .0:2836 lb=in3 / and Poisson ratio  D 0:29.
Modeling support system: Four racquet balls were used to support the plate during impact testing. These racquet balls act
as springs and their stiffness value was estimated to be roughly 40 lbf =in. At locations corresponding to the supports,
linear springs were added in the FE model. This reduced the relative difference for the modal frequencies further.

37.3.4.2

Testing Related Changes

Support system: Free-free boundary conditions are relatively easier to achieve compared to other boundary conditions
such as a fixed boundary condition. Support systems add stiffness and can also interact with the test structure. Certain
precautions need to be followed while selecting support systems [11].
Initially, shock chords were used to support the structure. However, it was observed that the shock chord supports
dynamically interacted with the third deformation mode (second torsion mode), and acted as a vibration absorber. As a
result, a racquet ball support system, that was reasonably isolated from the test structure, was selected.
Sensor calibration: The sensors (accelerometers and impact hammer load cells) need to be calibrated before data is
acquired. Sensors generally tend to have variability in the sensitivity in the order of 25 %, which when not accounted for
could skew the measurement magnitudes (but would have no effect on the calibration metric involving modal frequencies).

358

H.G. Pasha et al.

Impact hammer tip: The impact hammer tip should be selected such that the modes in the required frequency range are
adequately excited without exciting higher frequency modes. When the focus is to acquire data in lower frequency range,
using a harder tip would possibly cause overloads on the data acquisition channels associated with sensors near the impact
location. As a result, an appropriate hammer tip should be selected. A soft hammer tip was selected as the plate structure
was lightly damped.
Signal processing parameters: The rectangular plate is a very lightly damped structure. It was noted that after every impact,
it took over 16 s for the response to die out. After selecting the frequency bandwidth, the frequency resolution should
be chosen such that the response at all the reference locations is completely observed in the captured time segment.
If this condition is not met, it would result in leakage errors. If desired, the force-exponential windows can be applied to
condition the signal to avoid leakage and measurement noise. The number of ensembles acquired per average also affects
the ability to obtain good quality data and minimize measurement noise. In this test, data was acquired in the 0100 Hz
and 0250 Hz frequency ranges. The frequency resolution (F) was chosen for each case such that each ensemble was
observed for 16 s. The force-exponential window was chosen, with a cutoff value of 10 % for the force window and an
end value for exponential window as 20 %.

37.3.4.3

Modal Correlation after Calibrating the FE Model

Modal correlation was performed after updating the FE model and incorporating the changes identified for the test setup.
The relative difference for modal frequencies reduced to a value less than 2 %.
FRFs from the FE model were synthesized at a few locations corresponding to the driving-points in the impact test.
The measured FRFs were compared with the corresponding synthesized FRFs as a further calibration metric that involves
damping and sensor calibration issues. A plot comparing the measured FRF with the synthesized FRF for a driving-point
is shown in Fig. 37.8. It is apparent from the comparison that the peaks of the synthesized FRF are sharp as damping was
not prescribed in the FE model. However, the peaks of the measured FRF are damped. In reality the modes do not have the
amount of damping that is evident in the FRF comparison plot as the rectangular plate is a lightly damped structure. The
damped peaks in the measured FRF are a result of the artificial damping introduced by the force-exponential window that
was used to eliminate leakage while performing the impact testing.

Fig. 37.8 Comparison of a


driving-point FRF

37 Structural Dynamics Model Calibration and Validation of a Rectangular Steel Plate Structure

359

Table 37.2 Correlation of modal frequencies for free-free case with no perturbed mass and
updated FE models
h D 6:17 mm .0:243), E D 2:05x1011 Pa .2:9734x107 psi/
FE model I:  D 7; 810 kg=m3 .0:2821 lb=in3 /, m D 23:8 kg .52:44 lb/
FE model II:  D 7; 850 kg=m3 .0:2836 lb=in3 /, m D 23:9 kg .52:69 lb/

Mode
First torsion
First X bending
Second torsion
First Y bending
Second X bending
Anti sym. X bending
Third torsion
X and Y bending
Third X bending

Modal frequency (Hz)


Experimental FE model I
42:16
41:33
44:16
43:71
95:58
95:04
104:6
103:1
119:06
118:19
137:98
137:38
176:38
175:52
203:72
202:86
244:98
243:65

FE model II
41:34
43:65
95:12
102:99
118:16
137:39
175:81
203:25
244:1

Rel. diff. (%)


FE model I FE model II
2
1.98
1.02
1.16
0.57
0.48
1.45
1.56
0.74
0.76
0.44
0.43
0.49
0.32
0.42
0.23
0.54
0.36

The results of modal correlation are presented in Table 37.2. The calibration criteria, a relative difference for modal
frequencies less than 2 % was met and a calibrated rectangular plate FE model was obtained.

37.4 Validation
Validation is the process of determining the correctness of a model in its description of the reference system under a set of
test conditions [1, 2]. For validating the calibrated rectangular plate FE model and thereby establishing the robustness of the
model, it was decided to analyze and to perform testing for the following two cases:
1. Perturbed mass analytical modal analysis and testing
2. Constrained boundary analytical modal analysis and testing
The relative difference in modal frequencies was chosen as the validation criteria or the validation metric. For the perturbed
mass case, the validation criteria was set to less than 2% relative difference in modal frequencies.
Constraining specific points is possible theoretically in a FE model; however, it is impractical to achieve while testing.
In addition, modeling real boundary conditions is involved. As a result it was decided to set the validation criteria less than
5% relative difference for modal frequencies for the constrained boundary condition case.

37.4.1 Perturbed Mass Modal Analysis


Two separate cases were studied by attaching two cylindrical masses, 1 lb 1:5 oz and 2 lb 1:5 oz each, to the calibrated plate
FE model. The perturbation masses were also attached to the fabricated plate structure and impact testing was conducted.
The results obtained from the modal analysis and impact testing were correlated. The correlated results for 1 lb 1:5 oz and
2 lb 1:5 oz perturbed mass cases are listed in Tables 37.3 and 37.4 respectively. It is evident that the validation criteria,
relative difference for modal frequencies less than 2 %, was met.

37.4.2 Constrained Boundary Condition Modal Analysis


The plate was clamped to ground at two locations near one of the edges using steel spacers, which in turn were grounded to
a huge isolated mass. The test setup is shown in Fig. 37.9. Impact testing was performed on the clamped plate.

360

H.G. Pasha et al.


Table 37.3 Correlation of modal frequencies for free-free case with perturbed mass of
0:496 kg (1 lb 1:5 ozand updated FE models)
h D 6:17 mm .0:243), E D 2:05x1011 Pa .2:9734x107 psi/
FE model I:  D 7; 810 kg=m3 .0:2821 lb=in3 /, m D 23:8 kg .52:44 lb/
FE model II:  D 7; 850 kg=m3 .0:2836 lb=in3 /, m D 23:9 kg .52:69 lb/

Mode
First torsion
First X bending
Second torsion
First Y bending
Second X bending
Anti sym. X bending
Third torsion
X and Y bending
Third X bending

Modal frequency (Hz)


Experimental FE model I
38:95
38:4
42:42
42:15
89:53
89:73
99:96
98:93
117:39
117:09
128:95
128:94
167:56
168:34
197:19
198:36
235:5
237:82

FE model II
38
41:84
88:86
97:79
116:11
127:56
166:33
196:76
233:8

Rel. diff. (%)


FE model I FE model II
1.43
1.06
0.63
0.76
0.23
0.98
1.04
1.17
0.26
0.84
0.31
1.08
0.46
1.21
0.59
0.81
0.98
0.71

Table 37.4 Correlation of modal frequencies for free-free case with perturbed mass of
0:95 kg(2 lb 1:5 oz) and updated FE models,
h D 6:17 mm .0:243), E D 2:05x1011 Pa .2:9734x107 psi/
FE model I:  D 7; 810 kg=m3 .0:2821 lb=in3 /, m D 23:8 kg .52:44 lb/
FE model II:  D 7; 850 kg=m3 .0:2836 lb=in3 /, m D 23:9 kg .52:69 lb/

Mode
First torsion
First X bending
Second torsion
First Y bending
Second X bending
Anti sym. X bending
Third torsion
X and Y bending
Third X bending

Modal frequency (Hz)


Experimental FE model I
36:86
36:47
41:22
41:09
86:62
86:92
96:26
95:49
115:98
115:85
125:37
125:29
163:3
164:29
195:1
196:22
229:84
231:37

FE model II
36:17
40:82
86:28
94:5
114:99
124:39
162:74
195:05
230:8

Rel. diff. (%)


FE model I FE model II
1.07
1:92
0.33
0:97
0.34
0:39
0.8
1:86
0.11
0:86
0.26
0:79
0.6
0:34
0.57
0:23
0.67
0:42

Subsequently, the FE model was analyzed after applying constraints in vertical direction at the location corresponding to
the clamping points. The modal frequencies for the constrained boundary analysis and modal testing are listed in Table 37.5.
The larger relative difference for this case was expected, and can be attributed to the uncertainties involved in modeling
the boundary constraints used in the test. While it is aimed to constrain only the vertical direction translation of the plate
at two points, it is practically not possible to achieve this boundary condition. The support system constrains the motion of
the plate at a small but finite patch near the clamping location. The boundary condition of the constrained plate test setup is
difficult to incorporate in a FE model. As the FE model is not a true representation of the test setup, the relative difference
for this case is higher. However, the validation criteria was met for the constrained boundary condition case as well.

37.5 Conclusions
Testing is expensive and resource intensive. Therefore, it is necessary to develop a well calibrated and validated FE model
to assist in predicting system response.
This paper presents an example of model calibration and validation of the FE model of a rectangular steel plate based on
correlation with experimental results.

37 Structural Dynamics Model Calibration and Validation of a Rectangular Steel Plate Structure

361

Fig. 37.9 Rectangular plate with


constraints. (a) Test setup;
(b) supports and clamping
location

Table 37.5 Comparison of modal frequencies for constrained boundary condition


case
Mode #

Description

1
2
3
4
5
6
7
8
9

Psuedo-pitch
First X bending
Second torsion
First Y bending
Second X bending
Anti symmetric X bending
Third torsion
X and Y bending
Third X bending

Modal frequency (Hz)


FE
Experimental
10:76
11:02
33:34
34:1
53:99
52:71
82:14
83:56
94:92
91:75
114:7
118:59
124:4
128:43
151:9
154:53
205:4
208:28

Rel. diff. (%)


2:42
2:28
2:37
1:75
3:34
3:39
3:24
1:73
1:4

A plate structure was fabricated and tested to estimate its dynamic properties. A FE model of the plate was developed to
perform modal correlation and validation.
The results obtained from both the model and the test showed some differences in modal frequencies due to uncertainties
in geometric parameters, the material model and boundary conditions.
Once corrections to material model and geometric parameters were applied, the FE model was calibrated. An improved
support system was used for testing. Subsequently, a good agreement between FE and modal testing results was achieved.
The relative error was reduced to 12 %, which met the validation criteria.
Both FE model and modal testing results had similar modal frequencies obtained in the frequency range of 0250 Hz.
The first nine elastic modes showed good agreement in determining torsion, bending or combination of both torsion and
bending modes.
The robustness of the validated model was checked by comparing the results predicted by the FE model with the results
of perturbed mass and constrained boundary testing.

362

H.G. Pasha et al.

References
1. AIAA (1998) Guide for the verification and validation of computational fluid dynamics simulations. AIAA G-077-1-1998
2. ASME (2006) Guide for verification and validation in computational solid mechanics. ASME V&V 10-2006
3. Oberkampf WL, Roy CJ (2010) Verification and validation in scientific computing. Cambridge University Press, Cambridge
4. Allemang RJ (2010) Validation of the digital twin of a reusable, hot structure, hypersonic vehicle. In: US Air Force Summer Faculty
Fellowship Program
5. Paez TL (2009) Introduction to model validation. In: Proceedings of the 27th IMAC, SEM, Orlando
6. Hirsch C (2002) Verification and validation for modeling and simulation in computational science and engineering applications. In: Foundations
for verification and validation in the 21st century workshop. Johns Hopkins University/Applied Physics Laboratory, 2223 October 2002
7. Trucano TG, Swiler LP, Igusa T, Oberkampf WL, Pilch M (2006), Calibration, validation, and sensitivity analysis: Whats what, reliability
engineering and system safety 91:13311357
8. Mariappan D (2009) Simulation and experimental validation of automotive components. In: Proceedings of the 27th IMAC, SEM, Orlando
9. Blevins RD (1979) Formulas for natural frequency and mode shape, Butterworth-Heinemann, USA, pp 252253
10. Schwarz BJ, Richardson MH (1999) Experimental modal analysis. In: CSI reliability week, Orlando, October 1999
11. Pasha HG, Allemang RJ, Brown DL, Phillips AW, Kohli K (2014) Support systems for developing system models. In: Proceedings of the
international modal analysis conference (IMAC) XXXII a conference and exposition on structural dynamics

Vous aimerez peut-être aussi