Академический Документы
Профессиональный Документы
Культура Документы
Renewable Energy
journal homepage: www.elsevier.com/locate/renene
a r t i c l e i n f o a b s t r a c t
Article history: In the currently booming market of wind turbines, a clear focus is put on the design of reliable and cost-
Received 18 January 2011 effective subsystems, such as the gearbox. A requirement for reliable gearbox design calculations is
Accepted 17 March 2011 sufficient insight in the dynamics of the entire wind turbine drive train. Since traditional wind turbine
Available online 15 April 2011
design codes reduce the drive train to just a few degrees of freedom, considerable research effort is spent
in advanced modelling and simulation techniques to gain more insights in the dynamics at hand. This
Keywords:
work focusses on the gearbox modal behaviour assessment by means of three more complex modelling
Gear dynamics
techniques of varying complexity: the purely torsional-, rigid six degree of freedom with discrete flex-
Drive train
Wind turbine
ibility and flexible multibody technique. Both simulation and experimental results are discussed. Typical
Flexible multibody mode categories for traditional wind turbine gearboxes are defined. Moreover the challenge of the
definition of an accurate approach to condense finite element models for representing the flexible
components in the flexible multibody models is overcome. Furthermore the interaction between the
structural modes of the planet carrier and planetary ring flexibility with the overall gearbox modes is
investigated, resulting in the definition of two new mode categories: the planet carrier modes and
planetary ring modes.
Ó 2011 Elsevier Ltd. All rights reserved.
0960-1481/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2011.03.023
J. Helsen et al. / Renewable Energy 36 (2011) 3098e3113 3099
high quality drive train component design based on traditional discusses the level of detail of information gathered from each
calculations impossible and expresses the need for more advanced modelling approach.
numerical simulation techniques.
This article focusses on wind turbine gearbox modal behaviour 2.1.1. Purely torsional multibody models
assessment and suggests the multibody technique to overcome the In this first technique, the gearbox is modelled as an assembly of
downsides of the traditional models. Multibody models with torsional DOF bodies connected by one-dimensional spring
different degree of complexity are investigated and can be classified systems. This implies that the model can only calculate torsional
according to increasing complexity as: eigenfrequencies. The models will be implemented in the software
DRESP [10].
Purely torsional multibody models
Six degree of freedom rigid multibody models with discrete 2.1.2. Six DOF rigid multibody models with discrete flexibility
flexibility In the second approach all bodies have six DOFs. Therefore more
Full flexible multibody models complicated dynamic component behaviour can be described. Six
DOF spring-damper systems introduce discrete flexibilities into the
This work indicates the improvement of the purely torsional system, facilitating a more detailed description of gear mesh and
models compared to the traditional models. However, based on the bearing stiffness. Hansen Transmissions is implementing such
discussed experimental results, the need for even more advanced models in the software Simpack [11e14], while KU Leuven is using
models is shown. Therefore the rigid six degree of freedom (DOF) Virtual.Lab Motion [15,16].
and flexible multibody modelling techniques are introduced.
Advantages as well as limitations of each modelling technique are 2.1.3. Full flexible multibody models
indicated, whereas challenges related to each modelling technique In addition to discrete flexibilities, taken into account in the
are revealed. In conclusion, the flexible multibody modelling previous model, a more detailed approach also includes the flexi-
technique is suggested as the best suited for wind turbine gearbox bility of different drive train components. This is achieved by rep-
modelling. However, given the limited literature about full flexible resenting flexibility of drive train components, such as for example
gearbox modelling quite some challenges remain in accurate flex- shafts and planet carriers by means of FE models [17]. This allows to
ible gearbox modelling. The definition of accurate coupling struc- visualize modal influence of different gearbox subcomponent
tures which facilitate the condensation of the finite element (FE) flexibilities [18]. Complex geometry and numerical convergence
models representing the flexible components and couple these requirements, however, will result in a significant number of nodal
models with the rigid multibody model. A new approach, the degrees of freedom for the FE model to accurately describe body
flexible multipoint constraints (MPCs), is suggested and compared flexibility, which makes the models computationally expensive. To
to the rigid MPC approach described in literature. Moreover the keep calculation times to a minimum, FE structures will be
influence of planet carrier and planet ring flexibility on the gearbox condensed by an appropriate model reduction technique. The Craig
modal behaviour is investigated. Bampton component modes synthesis (CMS) method will be used
as modal reduction technique. In this reduction scheme, the modal
2. Approach transformation set consists of constrained modes and normal
modes [19].
2.1. Modelling techniques
2.2. Experimental validation
This section gives a general overview of the discussed modelling
techniques. Table 1 gives an overview of the level of complexity Multibody models can only add value to the design process if
contained in each modelling approach. The table is divided in two simulation results prove to be representative and reliable, which
parts. The first part, einput parameterse, illustrates the differences requires sufficient experimental validation. Test rig based valida-
in the model components. The second part of the table, eoutputse, tion is performed on a 13 MW test rig, shown in Fig. 1, on which two
Table 1
Overview of the modelling techniques used.
Fig. 1. The concept of the back-to-back gearbox test rig is based on the vision that it is possible to transform wind turbine behaviour into test rig conditions. The gearboxes in this
photo are gearboxes for the REpower 6M turbine. This photo is only used for clarification of the test rig concept. Measurement results from other gearboxes were used for the
research in this paper. (*) represents an optional speed reducer (3:1 gearbox).
gearboxes can be placed in a back-to-back set-up and be subjected 3. Purely torsional multibody modelling
to dynamic load cases representing wind turbine conditions [20].
Analogous to the operation of a wind turbine, the ‘wind’ side of the 3.1. Modelling approach
test rig is speed controlled, whereas the ‘generator’ side is torque
controlled [7]. The National Renewable Energy Laboratory (NREL) A model consisting of the two gearboxes with both planetary
applies a similar approach for investigating dynamic loads and and helical gear stages, as well as the test rig set-up was devel-
improving wind turbine gearbox reliability. Testing is performed on oped. Main model inputs are inertias, derived from CAD-geometry
a dynamometer test rig and in a wind turbine in the field [21,12]. and component stiffness values, calculated based on detailed FE
This work experimentally proves the limitations of the purely analyses. Shaft torsional and bending deformation, gear mesh
torsional models in describing gearbox modal behaviour. Further- stiffness and bearing flexibility were calculated analytically.
more a verification of a full flexible multibody model with focus on Planet, planet carrier and planet ring deformations were accoun-
eigenfrequencies is discussed. ted for by means of torsional stiffnesses determined on the basis of
Table 3
Results from verification of natural frequencies from modes with experimental results.
Fig. 3. Outline of helical gear stage multibody model with discrete flexibility: (a)Helical gear stage, (b)Multibody implementation including bearings and gear mesh.
Fig. 4. Outline of planetary gear stage multibody model with discrete flexibility: (a)Planetary gear stage, (b)Multibody implementation including bearings and gear mesh.
J. Helsen et al. / Renewable Energy 36 (2011) 3098e3113 3103
describing variable tooth stiffness and accounting for backlash, 6. Model implementation
dynamic separation distance, multiple tooth contact and material
properties of both gears [11,39]. 6.1. Full test rig model
Fig. 7. Planetary mode shapes: (a)Rotational-axial mode, (b)No modal deformation (reference), (c)Translational-tilting mode.
Fig. 8. Two exemplary Helical mode shapes: (a)Axial translation mode corresponding to eigenfrequency at 562Hz, (b)No modal deformation (reference), (c)Mode of ISS about one of
its’ radial axes corresponding to eigenfrequency at 485Hz.
6.1.2. Model verification with focus on eigenfrequencies by means of intermediate shaft (ISS) (4). On the ISS a high speed wheel is
measurements on dynamic test facility mounted, which makes contact with the teeth on the high speed
The experimental eigenfrequencies determined from test rig shaft (HSS) (5).
measurements can be used to validate the flexible multibody model
describing the test facility including the two back-to-back gear- 6.2. Model results analysis: full gearbox model
boxes. A similar approach as discussed in [35] is used to realize
a systematic and objective comparison. In this approach the In general, the modes of a full gearbox system differ from the
measured and simulated modes corresponding to the different ones of the stand alone gearbox components [40]. Therefore the full
eigenfrequencies are correlated at the measurement sensor loca- gearbox should be modelled. The modes of the full gearbox con-
tions. The results of this eigenfrequency matching are shown in sisting of both planetary and helical stages, can be classified in three
Table 3. For confidentiality reasons the exact eigenfrequencies categories: planet modes, helical modes and global modes.
cannot be presented. Instead, frequency ranges that contain the
found eigenfrequencies (numbered f1 to f22) are indicated. More-
over, the difference percentage in frequency between model and
measurements is indicated. For 17 of the eigenfrequencies the
difference between model and measurements is well below 10%.
Two measured eigenfrequencies could not be matched with the
model. Based on this comparison it can be concluded that the
multibody model shows big potential to describe the dynamic
behaviour of a gearbox. For further model verification the reader is
referred to [35].
Fig. 10. Low Speed Shaft (LSS): (a)LSS, (b) Multipoint constraint locations in LSS.
Fig. 11. Displacements for the constraint mode unit translation in Y direction: (a)Rigid multibody constraint set-up, (b)Flexible multibody constraint set-up.
Table 4
Categorized eigenfrequencies for rigid generic gearbox model.
7. Full flexible multibody models differences in modelling is given in Table 1. Schlecht indicated the
need for the introduction of flexibility of wind turbine structural
As shown in the previous section, six DOF models can already components: i.e.bedplate within multibody turbine models [42,43].
capture complex dynamic behaviour. Nevertheless, it remains From correlation of their flexible multibody model of a wind
impossible to describe local component flexibilities and the cor- turbine to experimental data, Heege et al. show the need for fully
responding effect of component modes on the global gearbox coupled three dimensional models and take flexibility within
behaviour. One possible solution is the use of FE models to repre- turbine and gearbox into account [44]. Moreover, Rigaud indicates
sent structural gearbox components. A detailed description of the the need for shaft and housing flexibility within gearbox multibody
Table 5
Categorized eigenfrequencies of the full gearbox model with flexible planet carrier.
Fig. 15. Planet Carrier auto-FRF for torsional DOF of condensation node of turbine main shaft planet carrier connection MPC.
modelling [45]. This work will therefore further elaborate the effect structure, the flexible MPC, the coupling between the interface node
of flexibility within multibody gearbox modelling. The assessment and the FE nodes is governed by a weighted average:
of the effect of local flexibilities is divided into two parts. Very little
has been published about the accurate coupling between multi- M$ui $ri
Fi ¼ (8)
body and FE models with regard to force and moment transfer. u1 $r12 þ u2 $r22 þ . þ un $rn2
Therefore in the first part a detailed comparison is made between
where:
the available coupling structures for interconnecting the compo-
nents’ FE models to the global multibody gearbox model. Secondly,
M ¼ moment at interface node
modal analyses of both the six DOF full gearbox model and of a full
Fi ¼ force at DOF i
gearbox model with respectively flexible planet carrier and flexible
planetary ring are compared to assess the influence of planet
ui ¼ weighting factor at DOF i
ri ¼ radius from interface node to DOF i
carrier flexibility and planetary ring flexibility.
n ¼ number of interface nodes
7.1. Finite element multibody coupling The main advantage of this approach is that the flexible MPC
does not add any extra stiffness to the FE modelled component.
In order to drastically reduce calculation times, the components’ However, since a weighted average is used, certain displacements
FE models are reduced to just a few degrees of freedom using the of the FE nodes will not result in displacement of the interface node,
Craig Bampton component mode synthesis technique. Main chal-
lenge in doing so, is in finding the right balance between accurately
Table 7
transferring forces and moments onto the component’s FE model
Rigid components of gearbox planet carrier related modes.
and minimizing the kept number of degrees of freedom. Consider
the case of forces and moments transfer between a bearing Rigid components of planetary modes of multibody model with flexible planet
carrier
-modelled as a single spring-damper system in the multibody model
and the shaft’s FE model. In this example, a single force or moment Rotational-Axial Modes Translational-Tilting Modes Global Modes
introduced at the interface node should be distributed over the 643 502
entire contact area between shaft and bearing. Two possibilities are 659 672
759 1055
discussed. In the first, referred to as rigid MPC, the interface node is
772
rigidly connected to the FE nodes. This results in a rigid shell of 819
nodes in which the distance between the different nodes is constant 835
at all times. This drastically stiffens the flexible gearbox component 842
reducing the effect of the flexibility. In the second coupling 847
958
962
Table 6
986
Eigenfrequencies of gearbox structural components.
1047
Structural eigenfrequencies 1059
1098
Planet Carrier Ring wheel with three gear contact multipoint 1107
constraint 1218
800 (m ¼ 2) 70 442 937 1240 1239
950 (m ¼ 2) 70 447 1017 1242 1241
961 143 621 1096 1275 1342
1045 144 632 1107 1302 1386
1140 197 820 1110 1512 1392
1164 200 833 1209 1531 1487
1392 (m ¼ 2) 381 918 1237 1630 1490
J. Helsen et al. / Renewable Energy 36 (2011) 3098e3113 3109
as shown in Fig. 9. This implies underestimation of stiffness of the model. Fig. 12 shows auto-FRF’s, where the interface node most on
structure and can cause penetration of the respective interface the left in Fig. 10 is exited and all other interface nodes are free.
surfaces of two coupled structures. Mathematical details of both Dynamic behaviour differs significantly. Up to 2000 Hz, the differ-
MPCs are discussed in [46]. To illustrate the influence of coupling ence is approximately 18%. The first eigenfrequency in case of rigid
structures, a FE model of a shaft will be discussed. Multipoint MPCs is 26% higher than for the model with flexible MPCs. There
constraints are used to introduce bearing, gearwheel and spline can be concluded, that the choice of a coupling structure, optimal
forces and are shown in Fig. 10. for the problem at hand, is of prime importance as it both influ-
A first comparison consists of comparing one of the constraint ences the static as well as the dynamic behaviour of the full gearbox
modes for the rigid and flexible multipoint coupling, characterizing model. In this respect, paying special attention to the used MPCs is
shaft behaviour under static deformation, is shown in Fig. 11. Both suggested, when building up gearbox models.
differ significantly. However, dynamic behaviour can best be
compared by means of frequency response functions (FRF’s), rep- 8. Modal behaviour comparison between full gearbox six DOF
resenting the absolute value of amplitude of displacement over rigid multibody model with discrete flexibility and full
harmonic excitation. Rigid body dynamics are not accounted for, as gearbox model with flexible planet carrier
in the floating frame of reference technique the reduced model of
the body flexibility should only describe the elastic deformation of 8.1. Model implementation
the flexible body [47]. The rigid body motion of the flexible body
should not be taken into account a second time by the modal The rigid model with discrete flexibility of stand alone gearbox,
representation, as it is already accounted for in the multibody discussed in Section 4, forms the base for the flexible model. To get
Fig. 16. Planetary ring MPCs for connecting the planetary ring mesh to the gearbox housing and torque arm of the rigid multibody model.
3110 J. Helsen et al. / Renewable Energy 36 (2011) 3098e3113
the flexible model, rigid planet carrier was replaced by a CMS motion to describe large translations and rotations of the reference
reduced FE model. Meshes of different mesh element size were frame superposed by FE model deformation to represent small
compared to get stabilization of first 15 eigenfrequencies within structural component deformations. Both can be approached sepa-
a range of 1%. The CMS reduction set contained static constraint rately to get an impression of the form of the mode shape found in
modes and flexible modes up to 3000 Hz. Steel was used as planet the rigid multibody model which couples with the structural one of
carier material. Multipoint constraints facilitated coupling between the planet carrier. Table 6 lists descriptions of the rigid parts of the
FE and multibody models as shown in Fig. 13. Both at planet carrier different mode shapes corresponding to the eigenfrequencies of the
bearing interfaces (1) as well as at planet bearing interfaces (2), second range. Helical modes are not listed, as these were hardly
flexible MPCs were preferred. However, to represent the stiffening influenced by planet carrier flexibility. Global modes however are
effect of the turbine’s main shaft at the planet carrier turbine main influenced by component flexibility. In Table 4, a planet carrier mode
shaft coupling (3), the use of rigid MPCs was deemed better to is found at 672 Hz. From Table 3 it is clear, that in the rigid body model
represent this interface. All modelling was performed under free no planet carrier related eigenfrequency existed in the range of
boundary conditions. 672 Hz, whereas Table 6 shows the rigid component of the modes of
the flexible models. It is expected that the first planet carrier struc-
8.2. Simulation results tural mode at 800 Hz has coupled with the global mode at 468 Hz to
result in the found planetary mode at 672 Hz. For the extra found
For the full gearbox model with flexible planet carrier, the same planetary modes, a similar mechanism is expected. The structural
mode categories are found as for the rigid model. However, one planet carrier modes interacted with the translational-tilting modes
category is added: the Planet Carrier modes. These modes are of the rigid planetary system. Table 6 clearly shows that most of the
dominated by local flexible planet carrier displacements. An example planet carrier modes have an underlying translational-tilting rigid
of such a Planet Carrier Mode is shown in Fig. 14. For the gearbox mode. This translational-tilting mode couples with the planet carrier
under investigation, eigenfrequencies for the model with flexible structural modes to result in the observed modal behaviour. This
planet carrier are categorized in Table 4. A first conclusion from significantly increases the number of found eigenfrequencies for the
comparison of the eigenfrequencies listed in Tables 3 and 4 is that system. Previous findings imply that planet carrier flexibility influ-
planet carrier flexibilities have relatively low influence on modes ences a large frequency range and that it should only be accounted for
from the helical modes category. A second finding is the increase in when these frequency range is of interest in the specific gearbox
the number of eigenfrequencies for the model with flexible planet
carrier. This is due to the structural eigenfrequencies of the flexible
planet carrier, listed in Table 5. Moreover, interaction exists between
overall gearbox modes and structural planet carrier modes. Therefore
the frequency range of interest can be divided in two ranges based on
the structural frequencies of the planet carrier: the first range
contains all frequencies well below the lowest eigenfrequency of the
planet carrier; whereas the second range stretches from the end of
the first one to the highest frequency of interest.
In the first range, flexibility of the planet carrier results in
a small decrease of the eigenfrequencies for the Planetary modes.
Given the use of the floating reference frame approach [47] to
create the flexible multibody model, structural component defor-
mation consists of rigid motion to describe large translations and
rotations of the reference frame superposed by FE model defor-
mation to represent small structural component deformations.
Modal behaviour of the FE model can therefore best be described by
means of frequency response functions (FRF’s), representing the
absolute value of amplitude of displacement over harmonic exci-
tation. Rigid body dynamics are not accounted for, as the reduced
model of the body flexibility should only describe the elastic
deformation of the flexible body. Fig. 15 shows the auto-FRF for the
planet carrier, where the interface node at the turbine main shaft
planet carrier connection is excited and all other interface nodes are
free. The first range, indicated in grey in Fig. 15, stretches approx-
imately from 0 Hz to 600 Hz.
The main effect of the flexibility in the first range is a very small
decrease of the eigenfrequencies. Based on these findings it can be
concluded that planet carrier flexibility as such is of less influence in
the lower frequency range. Therefore, if the goal of the gearbox model
is to investigate dynamic behaviour only for this first frequency
range, discrete structural component flexibility representation could
be considered. Such a representation could for example be realized
by discrete springs to represent static bending, torsion and tilting
stiffnesses of the planet carrier. In the second frequency range,
coupling occurs between the structural planet carrier modal behav-
iour and the overall gearbox modal behaviour found in the rigid six
DOF multibody model. Given the use of the floating reference frame Fig. 17. Gear mesh force introduction multipoint constraint that introduces gear
approach, structural component deformation consists of rigid meshing forces in a 120 section of the planetary ring FE model.
J. Helsen et al. / Renewable Energy 36 (2011) 3098e3113 3111
Table 8
Categorized eigenfrequencies for generic gearbox with flexible ring wheel with three gear mesh for introduction MPCs.
Multibody Model With Flexible Planet Ring With Three Multipoint Constraints
operating field. Since the Helical modes, are not influenced by the However, rim thickness reduction should be implemented with
introduced flexibility of the planet carrier, this results in a categori- care as it increases hoop stresses, which can cause catastrophic
zation in planet-, planet carrier-, helical and global modes (Table 7). failures [49]. Parker et al. extensively investigated the beahaviour of
planetary systems with flexible planetary ring by means of 2-
dimensional models [50e52]. Parker et al. extend their classifica-
9. Modal behaviour comparison between full gearbox six DOF
tion for two dimensional planetary system modal behaviour to the
rigid multibody model with discrete flexibility and full
following categories:
gearbox model with flexible planet ring
Fig. 18. Planetary ring auto-FRF in x translational direction for the condensation node of one of the three gear mesh MPCs.
3112 J. Helsen et al. / Renewable Energy 36 (2011) 3098e3113
Planet Modes: These are not important for the gearsystem Acknowledgments
under investigation, as these modes only occur for systems
with more than three planets This research was performed in the framework of a research
Purely Ring Modes: The displacements of the rigid elements is project supported by the Institute for the Promotion of Innovation
zero and there is only elastic ring deformation. by Science and Technology in Flanders (IWT Vlaanderen).
[33] Parker RG, Agashe V, Vijayakar SM. Dynamic response of a planetary gear [43] Schlecht B, Rosenlcher T, Schulze T. Strategy for user orientated simulation of
system using a fe/contact mechanics model. Journal of Mechanical Design large drive trains to calculate realistic load conditions. Warsaw: Proceedings
2000;122:304e10. of ECCOMAS Multibody Dynamics; 2009.
[34] Parker RG. Nonlinear gear dynamics in real world applications. In: 10th [44] Heege A, Betran J, Radovcic Y. Fatigue load computation of wind turbine
international conference on recent advances in structural dynamics. South- gearboxes by coupled finite element, multibody system and aerodynamic
ampton; 2010. loads. Wind Energy 2007;10:395e413.
[35] Helsen J, Vanhollebeke F, De Coninck F, Vandepitte D, Desmet W. Insights in [45] Rigaud E, Sabot J. Effect of elasticity of shafts, bearings, casing and couplings
wind turbine drive train dynamics by validating advanced models on a newly on the critical rotational speeds of a gearbox. VDI Berichte 2007;1230:
developed 13.2mw dynamically controlled test-rig. Mechatronics, in press, 833e45.
doi: 10.1016/j.mechatronics 2010.11.005. [46] Heirman GHK, Desmet W. Interface reduction of flexible bodies for efficient
[36] Ambarisha VK, Parker RG. Nonlinear dynamics of planetary gears using analytical modeling of body flexibility in multibody dynamics. Multibody Syst Dyn
and finite element models. Journal of Sound and Vibration 2007;302:577e95. 2010;24(2):219e34.
[37] Cai Y, Hayashi T. The linear approximated equation of vibration of a pair of spur [47] Wasfy T, Noor A. Computational strategies for flexible multibody systems.
gears (theory and experiment). Journal of Mechanical Design 1994;116:558e64. Applied Mechanics Reviews 2003;56:553e613.
[38] Cai Y. Simulation on the rotational vibration of helical gears in consideration [48] Kahraman A, Ligata H, Singh A. Influence of ring gear rim thickness on
of the tooth separation phenomena (a new stiffness function of helical invo- planetary gear set behaviour. Journal of Mechanical Design 2010;132:021002-
lute tooth pair). Journal of Mechanical Design 1995;117:460e9. 1e021002-8.
[39] Eberahimi S, Eberhard P. Rigid-elastic modelling of meshing gear wheels in [49] Seager DL. Load sharing among planet gears, SAE paper No.700178.
multibody systems. Multibody Sytem Dynamics 2007;16:55e71. [50] Wu X, Parker RG. Modal properties of planetary gears with an elastic
[40] Qin D, Wang J, Lim T. Flexible multibody dynamic modelling of a horizontal continuum ring gear. Journal of Applied Mechanics 2008;75:031014-
wind turbine drive train system. Journal of Mechanical Design 2009;131(11): 1e031014-12.
114501-1e114501-8. [51] Parker RG. Vibration of planetary gears with elastically deformable ring gears
[41] Eritenel T, Parker RG. Modal properties of three-dimensional helical gears. parametrically excited by mesh stiffness fluctuations. In: Proceedings of the
Journal of Sound and Vibration 2009;325:397e420. ASME 2009 IDETC/CIE. San Diego; 2009.
[42] Schlecht B, Schulze T. Analysis of dynamic loads in multi-megawatt-drive- [52] Parker RG, Wu X. Vibration modes of planetary gears with unequally spaced
trains. In: Proceedings of the international conference on noise and vibration planets and an elastic ring gear. Journal of Sound and Vibration 2010;329:
engineering ISMA. Leuven; 2006. 2265e75.