00 vote positif00 vote négatif

137 vues8 pagesMar 31, 2012

© Attribution Non-Commercial (BY-NC)

PDF, TXT ou lisez en ligne sur Scribd

Attribution Non-Commercial (BY-NC)

137 vues

00 vote positif00 vote négatif

Attribution Non-Commercial (BY-NC)

Vous êtes sur la page 1sur 8

Fatigue Load Computation of Wind Turbine Gearboxes by Coupled Structural, Mechanism and Aerodynamic Analysis

A. Heege, Y. Radovcic, J. Betran, SAMTECH Iberica. www.samcef.com

Introduction With a service life of about 20 years, wind turbine power trains are subjected to a very diverse spectrum of dynamic loads. Due to the high number of load cycles which occur during the life of the turbine, fatigue considerations are of particular importance in wind turbine design. Power train bearing failures occur more frequently than damage to the other components. Taking into account that these failures occur to a certain degree after a couple of years of successful operation, the importance of proper fatigue considerations becomes obvious. The respective load spectrum, in terms of load amplitudes and associated load cycles, depends on the dynamic properties of the complete mechatronical wind turbine system and cannot be calculated properly without detailed three-dimensional models. External excitations in terms of aerodynamic blade loads and electromagnetic generator torques, depend implicitly on control strategies for blade pitch and/or generator electronics, as well as on the general dynamic properties of the whole turbine. The purely dynamic character of certain gearbox loads, such as that on the planet carrier bearings, stresses the need for detailed dynamic power train models, which reproduce load frequencies and amplifications of the individual gearbox components with sufficient precision. Widely used procedures for fatigue evaluation of gearbox loads rely essentially on the load history of the rotor main shaft and account only partially for the three-dimensional character of dynamic power train loads. Experimental measurements and numerical simulations have shown that dynamic load amplifications within the gearbox can be much larger than those calculated from measurements of rotor or high speed shaft torque transients. New standards for design and specification of gearboxes for wind turbines, like the American national standard AGMA 6006-A03, requires all dynamic effects be included in the fatigue load computation. In order to cope with these requirements, the proposed fatigue procedure relies on a complete mechatronical wind turbine model, which includes a detailed gearbox model. Accordingly, the load transients are extracted from the global model for each gearbox component and fatigue cycle counting is performed individually for each power train component. This procedure has the advantage that the frequency content and the associated amplitudes of the local transients account for the non-linear character of dynamic amplifications within the power train and respect the implicit dependence of the excitations on the dynamic properties of the entire mechatronical system. The fatigue load spectra, which are obtained from the global wind turbine model in terms of component wise load transients, account for local dynamic effects within the gearbox as well. Various topics will be dealt with as follows: section 1 is a brief summary of the mathematical approach for coupling the Finite Element Method (FEM), mechanisms, control loops and finally aerodynamics in terms of Aerodynamic Blade Section Elements. In section 2 the aerodynamic-mechatronical wind turbine model is presented. In section 3 some aerodynamic results are shown for an emergency stop. Numerical results are compared to experimental and detailed results are shown for a gearbox with two planetary stages. A grid loss event with subsequent emergency stop has been chosen as an example, in order to present dynamic load amplifications separately for each bearing and gear. Component-wise fatigue diagrams are exposed in section 4 and finally in section 5 the conclusions are presented. 1. Coupling Mechanism, Structural Analysis and Aerodynamics Augmented Lagrangian approach In order to introduce the mathematical background of the coupled field problem, we start from classical non-linear Finite Element Method (FEM) equation: 61

Externer Artikel

[ M ]q

[C( q )]q

[ K ( q )]q

g (q, q, t )

external

( q, q, t )

inertia

(q, q, t )

internal

( q, q, t ) ,

(1)

In equation (1), stiffness matrix [ K ( q )] , damping matrix C q and of the right hand side g ( q , q , t ) show non-linear dependency on the solution vector q . As follows, the matrix equation (1) will be extended in order to account for mechanisms, like for example, gears, control loops, or aerodynamic follower forces. These additional equations are introduced adopting an Augmented Lagrangian /4,5/. The final expression of the Augmented Lagrangian is stated in equation (2) for a given time step n, where the non-linear FEM equations are included in the vector g . Final discretized equilibrium equation of the coupled field problem is modified by the Hilber Hughes & Taylor form (HHT) /2/ in order to improve the numerical stability in presence of constraints . The factor k is a scaling factor associated with the constraints in order to improve numerical conditioning, [ B n ], the transposed vector of the constraint gradients and is a factor introduced by the HHT modification.

[ M ]q n 1 k (1 (1 T )[B n 1 ] n 1 k (q n ) 0 T [B n ] n (1 )g ( q n 1 , q n 1 , t n 1 ) g (q n , q n , t n )

T

) (q n 1 )

( 2)

Equation (2) represents the equilibrium state of coupled structures (FEM), mechanisms (MBS), control loops and aerodynamics for a given time instance /8/. The generalized solution vector q, yields the global solution of the coupled field problem, in terms of FEM nodal variables, MBS state variables, and aerodynamic forces. Further details on time integration procedure, error estimators, or solution strategy for equation solvers, can be found in the SAMCEF Mecano user manual /8/. Aerodynamic blade section elements for wind loads Bearing in mind that blades are represented by Super Elements, the Blade Element Momentum theory (BEM) /3,7/ can be applied very efficiently in order to introduce the wind loads by Aerodynamic Blade Section Elements. BEM theory can be considered as a two-dimensional approach that models the interaction of the incoming wind with an annular segment covered by the rotating blades. Aerodynamic loads are introduced by Finite Blade Section Elements which contribute in terms of elemental aerodynamic forces to the global equilibrium equation (2). The actual blade geometry is discretized by surface contributions AI , which correspond to the airfoil span times the chord length of section I. As shown in figure 1, aerodynamic loads are introduced at blade super-element nodes. Concerning implementation in the non-linear FEM-MBS solver SAMCEF Mecano, the induced velocities are additional degrees of freedom of the global field problem. Associated relations are included in the constraint vector of equation (2). The applied methodology permits a strong coupling, i.e. all equations associated either to aerodynamics, structures, mechanisms or control loops, are solved simultaneously. Major benefit of a strong coupling is that blade vibrations induced by aerodynamic forces affect implicitly the latter. Before computing the induced velocities at the different blade sections, some corrections are performed beforehand on the unperturbed wind field. First, the unperturbed speeds are corrected in order to account for ground effects and secondly, to account for the impact of the tower shadow.

Fig. 1:

62

Externer Artikel

Since aerodynamic torque is potentially higher than generator resistive torque and unsteady wind is considered, a pitch control is necessary to stabilize the machine at reasonably constant speed. 2. Structural and Multi-Body-System Modelling of a Wind Turbine Structural components, which are subject to elastic deformations and which have impact on the dynamic properties, are included in the complete wind turbine model in terms of FEM models. Taking into account that large time intervals - larger than 1000 [s] - have to be analysed, the total number of degrees of freedom (DOF) of the complete analysis model should stay below 20.000 DOFs. As a consequence, the FEM-structures, which are subject only to small deformations, are condensed by the Super-Element Method /1, 6/. The rotor shaft, all gearbox shafts, the generator rotor and the tower are modelled by nonlinear beam elements.

Fig. 2:

Further flexible mechanism type components like gears, bearings, drive train couplings and generator mechatronics are introduced through a Multi-Body-System (MBS) approach in terms of additional relations in the global equilibrium equation (2).

Certification

> Type Certification > Component Certification > Site Specific Certification > Certification during Development

> Type Approval > Turbulence Analysis > Periodic Inspections > Personnel Safety > Expertises for Authorities and Insurances > Damage Analyses and Damage Expertises > Risk Analyses

Services Offshore

> Assessment of Construction and Foundation > Evaluation of Design Concepts > Offshore Staff Training

DEWI-OCC Offshore and Certification Centre GmbH

Am Seedeich 9 // D - 27472 Cuxhaven Tel. ++ 49 [0] 4721 - 5088 - 0 // Fax ++ 49 [0] 4721 - 5088 - 43 info@dewi-occ.de // www.dewi-occ.de

Externer Artikel

Gearbox Modelling: Coupled MBS & FEM Approach Gearbox is included in the global analysis model combining FEM and MBS approach /9,10,11/. As written, all gearbox shafts are represented by non-linear beam elements. Gearbox housing is modelled by a solid FEM model, which is condensed to a Super Element in order to reduce the number of degrees of freedom. The frictional contact problems between flexible gears are reduced to geometrically variable, and point wise flexible contacts. Gear geometry is defined by helix-, cone- and pressure angles, normal modulus, respective teeth number and, if needed, further correction factors for the gear teeth. Gear teeth flexibility is defined either according to ISO 6336, or by non-linear gap-functions. It is emphasized that the proper modelling of gear and bearing clearances is of crucial importance when evaluating gearbox loads during backlashes. Every bearing of the wind turbine, including the rotor main shaft, the entire gearbox and the generator, is modelled by non-linear stiffness functions, which account for the coupling of radial and axial bearing properties. All bearing clearances in radial and axial directions are accounted for. If detailed stress computation is required, frictional contact between bearing components, or between gears is modelled by fully non-linear FEM models. Generator and Controllers The electro-mechanical generator characteristics are defined in terms of non-linear torque versus speed functions. This non-linear generator function may be modified during certain operation modes due to control actions, or in order to simulate grid failures. It is emphasized that the resulting transient generator torque is an analysis result and not a boundary condition. Note that general control loops are defined either directly using the controllers implemented in SAMCEF, or can be imported from Matlab/Simulink controller models. 3. Application Examples Emergency stop: aerodynamic results Our first numerical example is an emergency stop simulation. Emergency stop or E-stop, is the process that bring the wind turbine to idle as fast as possible. The E-stop is commanded by the control system in determined cases: grid loss, excessive vibrations,... Figure 2 shows a typical Numerical Wind Turbine Model including on one hand structural FEM Components like blades, rotor- and gearbox shafts, tower structure, gearbox housing, planet-carrier, Fig. 3: Lift, drag and moment coefficients for some blade sections bedplate etc. and on the other hand MBS type components like gears, bearings, elastic couplings or bushings, overload clutch, and finally the generator model & control loops. As depicted schematically in figure 2, the gearbox model is based on two planetary stages and one parallel helical stage where the numerical model accounts for every relevant gearbox components. Aerodynamic loads are entered through the finite aerodynamic blade elements. Figure 3 shows lift, drag and moment coefficients for some blade sections. In this simulation, wind turbine runs at constant wind of 15 m/s. The generator disconnects at 44 [s]. The disk brake located at gearbox exit starts acting immediately but reaches full power after some tenths of a second. Blades are pitched in the wind at the same time to bring turbine to rest (figure 7, pink curve referring to right ordinate). As shown in figure 7, blade pitch is nearly constant up to time 45 [s] and turned in the following seconds 90 [degrees] into the wind in order to invert the rotor torque. Figures 4 and 5 show the lift and drag forces acting on one blade at different radii. Figure 6 shows the resulting total aerodynamic torque and total thrust during the whole process. Note that the effect of wind shear and tower shadow are considered and the latter is visible in figures 4, 5 & 6.

64

Externer Artikel

Fig. 4:

Fig. 5:

Fig. 6:

Fig. 7:

Dynamic load transients in gearbox during emergency stop A grid loss event is chosen in order to demonstrate the non-linear character of the dynamic load amplifications within the power train. A grid loss event is characterized by the instantaneous dropping of generator torque due to an electrical grid failure. As a consequence, the pretension of the power train is lost and large dynamic oscillations occur. These frequently produce backlashes. In order to prevent the wind turbine running into over speed, immediate control actions like blade pitching are necessary. Figure 8 presents the rotor shaft and high speed shaft (HSS) torques during the grid loss event. Figure 9 & 10 present the axial & radial bearing forces of each planetary gearbox bearing and figure 11 depicts the gear forces. It is interesting to note, that the largest transient forces do not occur immediately after the grid loss, but some seconds later due to the fact that the wind turbine runs into over speed without any pretension of the power train. Note that the torque oscillations at the rotor or HSS shaft do not reveal important dynamic load amplifications (see figure 8). However, important load amplifications are detected numerically within the gearbox in terms of gear forces, radial and axial bearing forces (see figures 9, 10, 11). Load amplifications during emergency stop Taking into account that visible load amplifications at rotor or HSS shaft do not provide enough information on load amplification within the gearbox, derivation of reliable component specific load spectrums is absolutely crucial in order to size properly each bearing & gear. Transient curves depicted in figures 8, 9, 10 & 11 are processed to allow an intuitive interpretation relative to reference loads at nominal power. Figures 12, 13, 14 & 15 are derived from figures 8, 9, 10 & 11 by extracting the extreme values and by relating them to the loads at nominal power. Abscissas in figure 13, 14 & 15 represent the respective gearbox component. The right ordinates of figures 13,14 & 15 represent the maximum and nominal loads [N]. The left ordinate of figures 13, 14 & 15 present the load amplification with respect to the nominal loads. A load amplification of 1 means: loads 100% above nominal load: i.e. the maximum load is 2 times the nominal load. For the load components without reference values like for example, axial loads of planet and planet car-

65

Externer Artikel

Fig. 8:

Fig. 9:

Fig. 10:

Fig. 11:

rier bearings, there is no reference value and loads are of purely dynamic nature. In that case, we show average dynamic loads at nominal power and maximum dynamic loads occurring during and after the event. It is crucial to recognize that the load amplifications within the gearbox are much larger than the amplifications which would be detected by experimental measurement or numerical simulation at rotor shaft and/or HSS shaft. Comparison of numerical results to experiment Figure 16 shows the comparison of the numerical rotor shaft torque results to the corresponding experimental data during an emergency stop at low wind conditions. The sudden augmentation of rotor shaft torque at about time =6[s] is due to the activation of the disc brake at the gearbox exit. It can be seen that the numerical model reproduces with a very satisfactory precision the system change which occurs at the transition from braking to turbine at idle. During the time interval [6s 10 s] first drive train mode is visible at about 1.1 [Hz]. At about time [10 s], the turbine starts idling and the dominating frequency drops to about 0.5 [Hz]. Figure 16 also includes a zoom on one single oscillation at turbine idle. The comparison to experimental data shows that the numerical model reproduces properly the zero torque instances due to clearances. 4. Fatigue Evaluations In widely used design practices of gearboxes, fatigue evaluations are based essentially on the time history of the rotor shaft torque. Looking at the transient curves depicted in figures 8-11, it can be seen that this design practice is only of very limited precision. Better load spectra for the different gearbox components are obtained, if the non-linear dynamic load amplifications are introduced component wise to correct the input rotor shaft time histories. However, this approach only corrects load amplitudes seen by component, but not the associated load frequencies. Further improvement to fatigue load spectra of power train components is obtained by extracting Rain Flow Counting (RFC) and Load Duration Distributions (LDD) separately for each gearbox component. In this approach, transient loads are extracted for each power train component from the global mechatroni66

Externer Artikel

Fig. 12:

Fig. 13:

Fig. 14:

Gear loads

Fig. 15:

cal wind turbine model and, RFC and/or LDD are performed separately for each bearing and gear. Thus, fatigue load spectra account as well for local dynamic effects within the gearbox. As a consequence, load cycles of higher frequency content are included in the fatigue spectra. Furthermore, dynamic amplifications of operation states producing local resonance are taken into account. In the case of structural components, FEM models might be used in order to compute the stress state associated with each load cycle and corresponding Fig. 16: Comparison to experiment, Main shaft torque [Nm]. load case. Structural components like blades or bedplate are included in the global model, but condensed by the Super Element Method to reduce CPU time. Stress within these components can be recovered at any instance of the transient analysis, by a backtransformation from the condensed Super Element. Damage can then be computed from this stress history. Bearings and gears are modelled by a MBS approach, thus reducing the analysis results to threedimensional load transients. Theses load transients might be used as boundary conditions for detailed FEM models. However it might be more convenient to use instead analytical methods in order to deduce respective component damage from RFC results. Figure 17 presents individually for each bearing the radial bearing load cycles in terms of RFCs obtained by gathering relevant load cases over 20 years of operation. Analogously, figure 18 presents the cumulated radial bearing load duration distribution (LDD) for 20 years of operation. 5. Conclusions The implicit dependence of power train loads on the dynamic characteristics of the assembled wind turbine, excludes a decoupling of analysis techniques, in order to reduce the complexity of the numerical models. If a gearbox is analyzed without accounting for the other properties of the wind turbine, there is 67

Externer Artikel

1000000 900000 800000 700000 600000 Load 500000 400000 300000 200000 100000 0 1,00E+00

Cumulated count Rad_main_shaft Load unit: N Cumulated count Rad_PLC_rot Load unit: N Cumulated count Rad_PLC_gen Load unit: N Cumulated count Rad_planet_1 Load unit: N

1,00E-02

1,00E-01

1,00E+01

1,00E+03

1,00E+04

1,00E+05

1,00E+06

Fig. 17:

Fig. 18:

some risk that cycle count, as well as load amplitudes are underestimated. Further on, possible operation deflection modes might affect the alignment of the power train and should be taken into account in fatigue evaluations. In that context it is crucial to recognize that the load amplifications within the gearbox are generally much larger than the amplifications which would be detected by experimental measurement or numerical simulation at rotor shaft and/or HSS shaft. The need for complete three-dimensional models is further emphasized by the purely dynamic character of certain gearbox load components like for example the axial planet bearing loads, or the axial and radial planet carrier bearing loads. Simplified methods like purely torsional models, can not properly reproduce the loads in the planetary stage. In the case of fatigue considerations being based only on rotor shaft time history, the introduction of dynamic load amplitude & load cycle correction factors for different gearbox components and for different load directions might permit to improve the fatigue calculations. However, due to the non-linear and three-dimensional character of wind turbine dynamics, it is recommended that the respective Load Duration Distribution and/or Rain Flow Counts be extracted from a global dynamic model, individually for each power train component. This requirement leads to the use of implicitly coupled analysis techniques like the Finite Element Method, Multi-Body-System approaches and aerodynamic load calculations. The presented examples demonstrate the feasibility of such an implicit coupling approach. Detailed fatigue analysis can be performed on a PC. It is expected that the availability of more precise fatigue load spectra will contribute to improve the design of wind turbine power trains. 6. References [1] R. Craig and M. Bampton. Coupling of substructures for dynamic analysis, AIAA Jnl.,6 no. 7. pages:1313-1319,1968. [2] H.M. Hilber, T.J.R. Hughes and R.L.Taylor. Improved numerical dissipation for time integration algorithms in structural dynamics, Earthquake Engng. Struct. Dyn., Vol. 5, pages: 283-292, 1977. [3] Anderson, John D., Fundamentals of Aerodynamics, McGraw-Hill, Inc. 1984. [4] A.Cardona and M. Geradin. Time integration of the equations of motion in mechanism analysis, Computers and Structures, 33, No. 3, pages: 801-82, 1989. [5] A. Cardona, M. Geradin, and D.B. Doan. Rigid and flexible joint modelling in multibody dynamics using finite elements. Comp. Meth. Appl. Mech. Engng., 89, pages: 395-418,1991. [6] A. Cardona and M. Geradin. Modelling of Super Elements in mechanism analysis, International Journal for Numerical Methods in Engineering, 32 No. 8, pages: 1565-1594,1991. [7] Manwell, J.F., McGowan, J.G., Rogers, A.L., Wind Energy Explained, John Wiley and Sons, Ltd. 2002. [8] Samcef/Mecano . User Manual version 11.0.5. Samtech SA, http://www.samcef.com. [9] A. Heege. Computation of dynamic loads in wind turbine power trains. DEWI Magazin Nr. 23, pages: 59-64, August 2003. [10] A. Heege. Quantification of Wind Turbine Gearbox Loads by Coupled Structural and Mechanism Analysis, Proceedings 7th German Wind Energy Conference DEWEK 2004, 20-21 October 2004, Germany. [11] A. Heege. Computation of Wind Turbine Gearbox Loads by Coupled Structural and Mechanism Analysis, Proceedings NAFEMS World Congress 2005, Malta, 17th-20th May 2005. 68

## Bien plus que des documents.

Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.

Annulez à tout moment.