Вы находитесь на странице: 1из 10

Numerical and Experimental Modal

Characterization of a 3MW Wind Turbine


Carlo E. Carcangiu
Alstom Wind
carlo-enrico.carcangiu
@power.alstom.com

Dmitri Tcherniak
Brel & Kjr SVM
dmitri.tcherniak
@bksv.com

Shashank Chauhan
Brel & Kjr SVM
Shashank.Chauhan
@bksv.com

Jon Basurko
Ikerlan-IK4
JBasurko
@ikerlan.es

Michele Rossetti
Alstom Wind
michele.rossetti
@power.alstom.com

Abstract:

1 Introduction

A thorough understanding of wind turbine


dynamics is necessary to design an efficient
and reliable wind turbine. Numerical models
are commonly used for the dynamic
characterization of wind turbines, due to the
difficulties
of
performing
full-scale
measurements. Nevertheless, such models
need to be validated against field
measurements.
Experimental modal analysis is an attractive
way to provide the necessary insight on the
wind turbine modal behaviour. However, the
application of conventional experimental
modal analysis techniques is not feasible
due to the wind turbine size. Thus,
Operational Modal Analysis (OMA) has been
selected for this study. Such modern modal
analysis technique allows estimating the
modal parameters based only on the
measured responses of the system under
operation, i.e. without artificial excitations.
An extensive experimental campaign has
been carried out on the Alstoms ECO 100
3MW wind turbine, to provide the necessary
database for the OMA. As a result, the
modal characteristics (mode frequencies,
damping and shape) of the wind turbine
under various operating conditions are made
available. The global structural modes of the
ECO 100 turbine, as well as the dynamics of
several
mechanical
and
electrical
components, have been identified, although
this paper focuses on the modal
characteristics of tower and rotor.
In
addition,
the
results
obtained
experimentally with OMA have been
correlated with the numerical models,
providing valuable indications to improve
their accuracy and the method potential is
finally demonstrated.

The design of optimized control systems


relies on a thorough knowledge of the wind
turbine dynamics, which is strongly related
to the structural characterization of the main
turbine components. Typically, numerical
models are built to simulate the wind turbine
dynamic behaviour, and experimental data
are used to increase the accuracy of such
models [1].
Therefore, the present work is aimed at
improving the numerical models, by the
correlation with an accurate experimental
characterization of the wind turbine
dynamics. More specifically, by providing a
deep insight into the complex behaviour of
the Alstoms ECO 100 (3MW, 100m rotor
diameter), this project is thought to be useful
to face design upgrades of this platform (for
instance the ECO 122, 2.7 MW, 122m rotor
span).
Ad-hoc modal analysis tools are thought to
be a valuable option in order to achieve a
better characterization of wind turbine
dynamics. A joint project has been carried
out by Alstom, Bruel & Kjaer SVM and
Ikerlan-IK4, for the dynamic characterization
of a prototype of Alstom ECO 100 3MW
wind turbine. A reliable dynamic model
(defined in terms of modal parameters i.e.
natural frequency, damping and mode
shape)
of
the
turbine
has
been
experimentally determined, in a wide range
of operating conditions. Operational Modal
Analysis (OMA), a technique to obtain the
dynamic characteristics of structures based
only on measured output responses has
been chosen for this work. It is such
capability of extracting modal characteristics
of a system without the need of artificial
excitations that makes this method
particularly suitable for large wind turbine.
Nevertheless, some drawbacks exist, due to
method assumptions which are not
completely fulfilled, and especially when

Keywords: Operational Modal Analysis,


Wind Turbine Dynamics, Experimental
Modal Analysis, FEM, Aero-elastic codes

dealing with operational wind turbines


(whereas idling conditions are more seemly
for OMA [2]).
A preliminary study on the applicability of
OMA has been therefore carried out based
on the ECO 100 aero-elastic model [3], and
possible solutions for those limitations have
been identified [4]. Among the adopted
arrangements, the use of Multi-Blade
Coordinate (MBC) transformation [5] was
successfully demonstrated in this first phase
of the project.
The design of experiment was presented in
[6], providing also details of the test planning
and execution.
Then, results of the
measurement campaign were first delivered
in [7] together with an overview of the data
postprocessing.
The
application
of
OMA
to
the
measurements data made available a large
amount of results, in terms of modal
characteristics. A selection of the most
interesting findings is summarized in this
paper, and a correlation is proposed
between the OMA results and the modal
information
obtained
using
common
numerical tools (FEM, aero-elastic models).
The paper is structured as follows. Within
section 2, an overview is given of the
numerical models used for the correlation. In
section 3 the OMA methodology is
presented, illustrating the measurement
campaign and data processing. Then, in
section 4 the results of the measured based
OMA are correlated with both finite-element
and aero-elastic models outputs. Concluding
remarks are finally drawn in Section 5.

2 Numerical Models
2.1

Aero-Elastic Models

Aero-elastic models are commonly used for


the performance and loads assessment of
wind turbine generators, as well as for the
control design and validation [8].
The GL-Garrad Hassans aero-elastic code
Bladed, versions 3.82 and 4.1 have been
considered for this study.
The structural model of Bladed 4.1 is based
on a multibody method combined with the
modal representation of the flexible
components [9]. The components, that can
be either flexible or rigid, are connected at
the nodes using the finite element
technique. Modal properties are computed
according to the Craig and Bampton's
method, which allows identifying the coupled
eigenvalues and eigenvectors of the whole
wind turbine.

Respect to older Bladed versions (e.g. 3.82),


based on the component mode synthesis
approach [9], the multi-body approach is
expected to improve the accuracy of the
numerical modeling, and so to reduce the
need of corrections from measured data.

2.2

Finite Element Models (FEM)

Finite Element Models (FEM) are widely


used to estimate the modal parameters of a
system during the design phase. Especially
in the case of complex systems, they can
give an overview of the main mode shapes
and frequencies. This information can be
used to select the sensors, decide their
location for an experimental measurement
and facilitate the interpretation of the
experimental data, especially when the
entire system is not being measured or
when only incomplete data is available. In
this project, two different FE models have
been used: one simplified and one detailed
model, as extensively described in [11]. Both
models have been built with the ANSYS
software.
The simplified model was built in order to
give an overview of the main structural
modes. The other model, being more
detailed allows identifying a larger number of
modes. The comparison of the two FEMs
mode shapes (using the Modal Assurance
Criterion or MAC) permitted to identify local
structural modes: a mode shape from the
detailed model that is not paired with the
simplified one indicates that it is probably a
local structural mode shape (e.g. related
with the nacelle or with the drive train).

3 Experiments based OMA


3.1

Innovative Approach

This technique allows extracting the modal


characteristics of the wind turbine, without
knowing the external perturbation, i.e. the
wind, but just observing the outputs of the
system. Due to the unpredictable nature of
the wind, and to the difficulties to provide
another external excitation to the system,
such output-only approach has significant
advantages for the modal characterization of
a wind turbine.
In a previous work by Hansen et al.[12], the
OMA approach was utilized to extract the
modal characteristics of an operating wind
turbine, based on the records of several
strain gauges. Nevertheless, to the authors
knowledge, Operational Modal Analysis has

Figure 1: Wind turbine regimes. Each point is a 15-minutes dataset.


Red: pitch-regulated regime; green: RPM-regulated regime; blue: idling.
been comprehensively performed for the first
time with this project on a full-scale MultiMW wind turbine, by using dedicated field
measurements.

3.2

OMA fundamentals

A dynamic response of a linear mechanical


system is related to the excitation forces by
a following expression:

x() H ()f ()

(1)

where x() and f() are the vectors of the


response and excitation spectra respectively
and H() is the Frequency Response
Functions (FRF) matrix. From modal
analysis theory, it is known that FRF matrix
contains all necessary information to extract
modal parameters [14]. Multiplying (1) by its
Hermetian and averaging in time yields:

G xx ( ) H( ) G ff ( )H ( ) H

(2)

where Gxx() and Gff() are the output and


input power spectra matrices respectively.
The main assumptions OMA is based upon
are that the excitation forces are
uncorrelated, have a uniform spectra and
distributed over the entire structure thus the
input
cross-spectra
matrix
becomes
diagonal:

G ff ( ) I

(3)

G xx ( ) H( )H ( )
H

(4)

The last two expressions show that, if the


excitation
assumptions
fulfilled,
the
response cross-spectrum matrix contains

the full information required to obtain an unscaled modal model of the system.

3.3

Applicability to operational
Wind Turbines

Turbulent wind excitation fulfils the


abovementioned assumptions, which makes
the application of OMA to standstill wind
turbines rather straightforward. However,
when dealing with operational turbines the
method assumptions are no more strictly
valid.
First, due to the rotor rotation, the analysed
system structure is no more time invariant.
The effect of the rotor rotation manifests
itself in the equation of motion: the mass,
stiffness and gyroscopic force matrices
become time-dependent. Formulating and
solving the eigenvalue problem for this case
would thus lead to time-dependent
eigenvalues and eigenvectors, which are
meaningless as modal parameters. The socalled Coleman coordinate transformation
(also known as Multi-Blade Coordinate
transformation, MBC [5]) has been utilized to
eliminate time dependency of the system
matrices, thus converting the original timevarying eigenvalue problem to a timeinvariant one. The modal parameters: modal
frequencies, damping and mode shapes are
hence
obtained
by
solving
the
corresponding eigenvalue problem. This
approach
has
been
extended
to
experimental modal analysis, as illustrated in
previous work [3]. Forward Coleman
transformation is applied to the data
measured on the wind turbine blades, which
is then combined with responses measured
on the tower. The OMA methods are then
applied to the transformed data, resulting in
modal frequencies, damping and mode
shapes. Backward Coleman transformation
is finally employed for the mode shapes
visualization.
In addition to the structure time invariance,
the aerodynamic forces do not fulfil OMA
assumptions on external excitations. In fact,
the forces acting on different parts of the
blades
are
correlated
around
the
fundamental frequency (1P) and its
harmonics. Moreover, the forces have a
periodic nature, which manifests itself by
peaks on force frequency spectra. The
peaks are again located at the fundamental
frequency and its harmonics and have thick
tails, which narrows the regions where OMA
assumptions are valid.
In order to avoid the frequency regions
where the OMA assumptions are not

completely fulfilled, a careful design of


experiment was carried out [3] and a long
term measurement campaign was finally set
up.

3.4

Measurement campaign

Typically, dynamic characteristics of the


wind turbine are expected to depend on the
different wind regimes. Moreover, being the
ECO 100 variable speed variable pitch
turbine, its dynamics also varies with those
characteristics. To ensure that the measured
data cover the whole operational conditions
of ECO 100 turbine, a long term
measurement campaign had run over a
period of 3 months from August, 2010 to
October, 2010 [6].
A total of 80 channels were recorded,
including
68
acceleration
channels
measuring
the
output
response
accelerations at various points on the tower,
nacelle, gearbox and generator. The
position of the accelerometers was selected
in order to capture the required number of
structural modes and also for the ease of
estimating the mode shapes (animated
mode shape have been extensively used to
get a clear understanding of the mode
vibration patterns). In addition to the
accelerometer channels, 10 control signals
and 2 tachometer signals were recorded
during this campaign. These control signals,
such as wind speed, rotor rpm, yaw angle
etc.,
are
extremely
important
for
characterizing the operating conditions and
classifying the vast amount of gathered
data.
With sensors available on the blades, the
turbine modal characteristics would be more
easily observable, and especially damping.
However, a number of practical issues make
a challenging task to get measurement on
the blades. These include obtaining data
from the rotating part and synchronizing it
with the static part. Also, instrumenting
turbine blades is difficult and costly,
requiring specially trained people and
equipment. Thus, in this project, the blades
were not instrumented. Moreover, it was
proven with simulations that rotor related
modes can be estimated from tower and
nacelle mainframe measurements [3],
although with some difficulties. In this
context, prior information available from
finite element models was helpful in
determining good locations for the
accelerometers.

3.5

Data processing

During the long term measurement, more


than 4000 datasets of 15 minutes each were
recorded successively using a sampling rate
of 256 Hz. Metadata provided by control
signals were then used to create an indexed
database. Querying this database on
various metadata parameters such as wind
speed etc., relevant datasets were obtained
for performing OMA.
Three main operational regimes have been
identified (Figure 1): idling, speed-regulated
and pitch-regulated. In Idling only all OMA
assumptions are fulfilled, however, this is not
an operational case. In the pitch regulated
regime, the rotor modes are strongly
influenced by the pitch changes and it is
difficult to find a dataset with a nearly
constant pitch value. The RPM-regulated
regime is the most suitable for modal
analysis as in this regime the pitch does not
change in time. In addition, the natural
frequencies are not as sensitive to the
changes in RPM as they are to the changes
in pitch angle.
Within the RPM-regulated regime, 65
datasets were selected for the further
analysis based on two criteria: (i) minimal
variation of the rotor speed within each
dataset, and (ii) selected datasets covering
the entire RPM range. The latest should
guarantee enough points to perform
statistical estimation of the uncertainties and
plot a Campbell diagram for the RPM range.

3.6

Performing OMA

The selected datasets were analysed by


Stochastic Subspace Identification (SSI)
f

6P
3P
1P

algorithm [13][15] using Bruel & Kjaers


OMA software (Type 7760). The SSI is a
time domain method, which has been
selected for being the most suitable for wind
turbine systems during the preliminary
phase of the project. In particular, it was
found that frequency domain methods fail
when dealing with wind turbine harmonics.
Moreover, frequency domain techniques
require more manual intervention (e.g. peak
picking), which makes them less suitable
when large amount of data has to be
treated. In fact, the analysis procedure was
automated, taking care that all datasets
were processed using same algorithmic
parameters.
The typical result of the automated
identification is shown in Figure 2. Each
column of the points (e.g. like the ones in
the dashed-line box) represents the
automatically selected poles resulting from
modal analysis performed on a dataset. The
inset shows the example of a stabilization
diagram obtained for one dataset with
automatically selected poles. Abscissa of
each column corresponds to datasets
median rotor RPM. The frequencies of the
poles become the ordinates of the points in
Figure 2.
Identifying true structural modes of the
turbine amongst the many automatically
chosen poles is not trivial. Hierarchical
clustering based approach was utilized to
help in this task. It is expected in fact that
structural modes are consistent in terms of
mode shape along the RPM axis (the global
modes shape should not depend on rotor
speed). Therefore, hierarchical clustering
uses mode shape similarity as a criterion for
clustering OMA identified poles, and also
Stabilization Diagram

Figure 2: Results of SSI performed on 65 selected datasets in the RPM-regulated regime. The
blue dots are the OMA identified poles of the system, vertically aligned for being corresponding
to one dataset each at a given average RPM. A close-up shows the stabilization diagram for
one dataset, with frequency in the abscissa and the red lines indicating the identified poles.

provides a mean to filter out extra


computational poles identified by OMA.
The next step is providing the identified
modes with the correct nomenclature. This
task is made difficult by the unavailability of
mode shape information from the blades.
Nevertheless, with a-priory knowledge of the
wind turbine dynamics and its expected
mode shapes (e.g. resulting from FE
modelling) the identified modes can be
finally classified.

3.7

4 Correlating numerical and


experimental results

OMA results

Once identified the mode characteristics and


associated nomenclature, the results can be
summarized into a Campbell diagram
(Figure 3). The global turbine mode
frequencies against the RPM regime are
thereby
plotted,
along
with
their
corresponding
confidence
intervals
(confidence level 95%, = 0.05). The main
modes of interest are summarized in Table
1, along with the adopted nomenclature.
Table 1: Global modes nomenclature
Sys.

Tower

Rotor

Mode Name

Acron.

st

T1FA
T1SS
T2FA
T2SS
T1T
O1BW
O1C
O1FW
I1BW
I1FW
I1C
O2BW
O2C
O2FW
I2BW
I2FW
I2C

Tower Fore-Aft 1
st
Tower Side-Side 1
nd
Tower Fore-Aft 2
nd
Tower Side-Side 2
st
Tower Torsional 1
st
Out-of-plane backward whirling 1
st
Out-of-plane collective 1
st
Out-of-plane backward whirling 1
st
In-plane backward whirling 1
st
In-plane forward whirling 1
st
In-plane collective 1
nd
Out-of-plane backward whirling 2
nd
Out-of-plane collective 2
nd
Out-of-plane forward whirling 2
nd
In-plane backward whirling 2
nd
In-plane forward whirling 2
nd
In-plane Collective 2

all. Further, the confidence intervals of the


2nd and 3rd out-of-plane modes are much
wider
compared
to
their
in-plane
counterparts. Possible reasons for difficult
identification of out-of-plane modes can be
the fact that they are heavily damped due to
considerable aero-elastic effects and
perhaps also have less influence on the
tower and the nacelle than the in-plane
modes.

All in-plane modes present in the frequency


range of interest are identified, and the
confidence interval is quite narrow. The
same is observed for the tower modes, but
for the 2nd tower for-aft mode the confidence
interval is slightly wider. The dependence of
the 1st and 2nd in-plane backward and
forward whirling modes on the rotor speed is
also clearly observed. In contrast to in-plane
modes, the out-of-plane modes are much
more difficult to identify. The first out-ofplane modes, which are expected to be in
the range between the first and third
harmonics (1P and 3P), were not found at

4.1

OMA vs. FEM

A correlation between the mode shapes


obtained from the measured based OMA
and the FEM computed eigenvectors has
been carried out. FEMTOOLS software has
been used for this purpose. The results for
the RPM-regulated regime are only
presented in this paper for brevity.
Modal Assurance Criterion (MAC) is the
technique employed to quantify the
differences between two mode shapes.
Eight modes have automatically been
paired, by using a criterion of MAC > 50%
and frequency difference < 10%. The first 6
identified mode shapes for the RPM-regime
are finally shown in Figure 4.

4.2

OMA vs. Aero-elastic models

The OMA outputs are hereby correlated with


the modal characteristics computed with
those output from aero-elastic models.
The GL-Garrad Hassan aero-elastic code
Bladed version 3.82 and version 4.1 have
been considered for this study, with the new
version generally matching better with the
experimental results.
So, in Figure 5 the Campbell diagram is
shown, correlating OMA and Bladed 4.1
results. As expected, 1st tower modes
perfectly
match,
whereas
some
discrepancies are observed for the
corresponding 2nd modes. The agreement
between in-plane modes is also good,
whereas out-of-plane modes, as already
mentioned, are less observable with OMA. It
should be noted that for control design and
tuning, tower and in-plane modes are
generally more important.

Frequency

I1FW
I1BW
O2C
O2BW
T2FA
T2SS
O2FW
I1C
O3C
O3BW
I2FW
I2BW
O3FW
T1FA
T1SS

Out-of-Plane-1?

Rotor speed

I2FW
I2BW

Frequency

I1C

T2FA
T2SS

I1FW
I1BW

T1FA
T1SS
Rotor speed

O3FW

O3C

Frequency

O3BW

O2FW
O2C
O2BW

O1FW
O1C
O1BW

Rotor speed

Figure 3: Modal frequencies as functions of rotor speed. From top to bottom: a) All lowest
global modes; b) In-plane and tower modes; c) Out-of-plane modes. Confidence intervals are
shown as grey bands.

Tower Side-Side 1st

Tower Fore-Aft 1st

Out-of-Plane 2nd Yawing

Out-of-Plane 2nd Tiliting

Tower side-Side 2nd

Tower Fore-Aft 2nd

Figure 4: Mode shapes correlation in RPM-regulated regime: OMA (red) vs FEM (blue).
With OMA, the total damping of the system
has also been estimated, i.e. including both
structural and aero-elastic components.
As expected, higher damping ratios are
found for out-of-plane modes as well as for
tower fore-aft modes. On the other hand,
confidence intervals are also wider for the
rotor out-of-plane modes, as well as for
tower fore-aft modes.
Indeed, OMA algorithms are known for being
not precise in absolute damping estimates,
but relative comparison of different modes
damping is still feasible. The values of the
mode damping ratio at the nominal speed
have also been correlated with the aeroelastic code prediction (Table 2), showing a
reasonable matching, although OMA values

are generally higher for tower modes and


smaller for the rotor modes.
Table 2: Damping ratio, and confidence
intervals, at rated speed.
System

Tower

Mode
T1FA
T1SS
T2FA
T2SS
I1BW
I1FW
I1C
O2BW
O2C
O2FW
I2BW
I2FW

Bladed
8.0%
1.0%
2.3%
1.1%
1.1%
1.3%
1.2%
14.0%
14.0%
14.0%
1.5%
1.2%

OMA
12.5%
1.0%
3.5%
0.5%
1.5%
1.7%
0.7%
7.8%
5.9%
6.2%
1.2%
0.8%

OMA
6.0%
1.3%
4.1%
0.3%
2.1%
1.9%
1.0%
8.1%
7.7%
10.9%
0.5%
0.7%

nd

In-Plane 2
modes (bw,fw)

Frequency

In-Plane 1
Collective

Tower 2
modes

st

nd

st

In-Plane 1
modes (bw,fw)

Tower 1
modes

st

RPM

Figure 5: Modal frequencies as functions of rotor speed. Aero-elastic model (solid lines) vs.
measurements based OMA (marks).

5 Conclusions

References

A more profound knowledge of the modal


behaviour of large wind turbines has been
achieved
with
this
project.
With
experimental-based OMA, the main global
modes of an operational wind turbine have
been identified for three main operational
regimes (idling, RPM-regulated and pitch
regulated), which are key-important for the
design of the controller. Moreover, valuable
indications have been gathered regarding
the modes of the turbine structural
components and auxiliaries.
The OMA results have been correlated vs.
the numerical models, revealing analogies
and discrepancies. Since different numerical
models have been tested, indications have
also been stemmed for the use of multi-body
based codes instead of simpler modal based
models, being the first matching better with
the OMA outputs.
Finally, the data obtained during the project,
the gained experience and the detailed
knowledge of the advantages
and
drawbacks of the proposed methodology,
become an important base for further
improvements in this field. Among all the
indications derived from this project, it must
be highlighted that making available
measurements on the blades would be
necessary in order to better identify rotor
modes.

[1]. Rossetti, M. 2010. Validation of Wind


Turbine Dynamics. Proc. 28th Int.
Modal Analysis Conference (IMACXXVIII).
[2]. Chauhan,
S.,
Hansen,
M.
H.,
Tcherniak, D. 2009. Application of
Operational Modal Analysis and Blind
Source Separation / Independent
Component Analysis Techniques to
Wind Turbines. Proc. 27th Int. Modal
Analysis Conference (IMAC-XXVII).
[3]. Tcherniak, D., Chauhan, S., Rossetti,
M., Font, I., Basurko, J., Salgado, O.
2010. Output-only Modal Analysis on
Operating Wind Turbines: Application to
Simulated Data. Proc. of European
Wind Energy Conference (EWEC2010).
[4]. Tcherniak, D., Chauhan, S., Hansen,
M.H. 2010a. Applicability Limits of
Operational
Modal
Analysis
to
Operational Wind Turbines. Proc. 28th
Int. Modal Analysis Conference (IMACXXVIII).
[5]. Skjoldan PF, Hansen MH. On the
Similarity of the Coleman and
Ljapunov-Floquet Transformations for
Modal Analysis of Bladed Rotor
Structures, Journal of Sound and
Vibration, 2009; 327:424-439
[6]. Chauhan, S., Tcherniak, D., Basurko,
J., Salgado, O., Urresti, I., Carcangiu,

C.E., Rossetti, M. 2011. Operational


Modal Analysis of Operating Wind
Turbines: Application to Measured
Data. Proc. 29th Int. Modal Analysis
Conference (IMAC-XXIX).
[7]. Tcherniak, D., Basurko, J., Salgado, O.,
Urresti, I., Chauhan, S., Carcangiu,
C.E., Rossetti, M. Application of OMA
to
operational
wind
turbine,
Proceedings of 4th International
Operational Modal Analysis Conference
(IOMAC-2011), Istanbul, Turkey, 2011.
[8]. Moriarty, P.J. and Butterfield, S.B.,
2009.
Wind
Turbine
Modelling
Overview for Control Engineers. Proc.
American Control Conference, St.LouisMO, USA, June 10-12, 2009
[9]. Bossanyi, E.A. 2010. Bladed v4
Multibody dynamics Theory manual. GL
Garrad Hassan and Partners.
[10]. Bossanyi, E.A. 2008. GH Bladed 3.82
Theory manual. Garrad Hassan and
Partners.
[11]. Basurko, J., Salgado, O., Urresti, I.,
Tcherniak, D., Chauhan, S., RodrguezTsouroukdissian, A., Carcangiu, C.E.,
2011. Test / Model correlation in the
ALSTOM 3 Megawatt wind turbine,
Proc. 29th Int. Modal Analysis
Conference (IMAC-XXIX).

[12]. M. H. Hansen, K. Thomsen, P.


Fuglsang, T. Knudsen, Two Methods
for Estimating Aeroelastic Damping of
Operational Wind Turbine Modes from
Experiments, Wind Energy, Volume 9,
Issue 1-2, pages 179191, January April 2006
[13]. Brincker, R., Andersen, P. 2006.
Understanding Stochastic Subspace
Identification, Proceedings of the 24th
International
Modal
Analysis
Conference, St. Louis, Missouri.
[14]. Ewins, D.J., Modal Testing: Theory,
Practice and Application; Research
Studies Press Ltd.: Baldock, England,
2000.
[15]. Zhang L, Brincker R, Andersen P.
2005. An Overview of Operational
Modal Analysis: Major Development
and Issues, Proc. of 1st Int. Operational
Modal Analysis Conference (IOMAC2005).

Acknowledgements
This work has been performed within the
InVent project, and was co-funded by
Generalitat de Catalunya, through ACC1
(CIDEM | COPCA) and by FEDER, Fons
Europeu de Desenvolupament Regional.

Вам также может понравиться