Evaluation of a low speed stage coupling with regard to structure-borne sound propagation in a wind turbine

Due to recent advances in the field of broadband aerodynamic noise, tonalities of wind turbines (WT) are increasingly coming into focus in the wind industry. In this case, the structure is excited inside the drivetrain and the structure-borne sound propagates through the machinery and ultimately to the surfaces of the WT, where it is radiated into the ambient air. Since any tonalities are a system characteristic, they should be considered at an early stage of product development. On the one hand, great efforts are being made to develop ever lower-toned drivetrains. On the other hand, tonalities can efficiently be neutralised by systematically decoupling the excitations in the drivetrain from the sound-emitting surfaces of the wind turbine. In addition to the well-studied behaviour regarding the decoupling of non-torque rotor loads from the drivetrain, in this paper the influence of a low speed stage (LSS) coupling on the structure-borne sound propagation inside of an integrated drivetrain is investigated. In a previous study at the Center for Wind Power Drives, it could be shown that in an integrated drivetrain, the transfer paths through the main shaft and subsequently the main bearing becomes the dominant transfer path. This is in contrast to classic bearing configurations where the torque arms of the gearbox are the dominant transfer paths of excitations from the gearbox, revealing an increased potential of LSS Couplings especially for integrated drivetrains. Detailed numerical investigations are performed in order to understand and quantify the usage of a LSS coupling for lowering sound power levels of a WT.


Motivation
As the generation of electrical energy is one of the main sources of manmade carbon emission, many countries set the goal of expanding the share of renewable energies in their production capacities. Onshore wind power is one of the driving technologies for the achievement of this goal since its high efficiency and low cost of energy compared to other renewable and also fossil sources. In order to further increase the capacity of onshore wind power, wind turbines (WT) need to satisfy legal limitations regarding sound emission when being built closer to inhabited areas [1]. Due to advancements in the lowering of broadband aerodynamic sound, tonalities from the drivetrain became an issue of concern recently. Tonalities are mainly caused by drivetrain vibrations at particular frequencies and are generally considered more irritating by human recipients than broadband sounds [2]. Surpassing legal sound limita- (2) main bearing unit; (3) bearing reinforcements; (4) supporting beams on mainframe; (5) LSSC housing tions can lead to decommissioning of a WT posing a high economic threat to the operator [1].
The acoustic behaviour of wind turbines is a topic of a large body of literature. Especially the broadband sound radiation that originates from the blade aerodynamics is investigated extensively [13][14][15][16][17]. Here, substantial advancements have been made which led to a reduced masking potential against tonal sounds. Tonal sounds are mainly caused by drivetrain vibrations [18,19]. Vibration that are excited in the drivetrain, e.g. from the gear teeth interactions, are transferred as structure-borne sound to the sound emitting surfaces like blades and tower. These oscillating surfaces radiate airborne sound into the WT surrounding [20,21]. For classical drivetrain configurations, there is already work done concerning the noise-vibration-harshness (NVH) behaviour of the WT system [20,22]. A recent review article comes to the conclusion that accurate simulation models K need to be developed in order to be able to economically develop tonality free WT [23]. The continuous development of improved WT drivetrains with ever higher rated powers recently led to increasingly integrated drivetrains. Here, main bearing unit, gearbox and generator are located in a fully connected housing. The reduced decoupling of these components leads to new questions regarding gearbox input loads and NVH behaviour of these WT, because sources of vibrations like gearbox and generator can no longer be separated regarding loads and structure-borne sound. This leads to inevitable tonalities, which need to be handled before reaching the sound radiation surfaces of tower and blades as recent studies suggest [3].
A low speed shaft coupling (LSSC) has already proven to be an effective way to lower gearbox input loads in WT drivetrains [4]. In this publication, the question will be addressed, if LSSC are additionally a solution for the limitation of structure-borne sound transfer into the sound radiation components of WT structures. For this purpose a full flexible state of the art WT model will be developed in a multibody simulation (MBS) environment. A reference WT model is compared to a WT model equipped with a LSSC between main shaft and gearbox. Transient simulation are performed and a dynamic analysis as well as a Transfer Path analysis (TPA) are carried out on the data.

Modelling approach
In order to formulate a state of the art onshore WT model with a rated power of 6 MW, 164 m rotor diameter and integrated drivetrain with 2 planetary stages, extensive modeling effort was carried out. The characteristics of this MBS model will be presented in this section. As a baseline, CAD data of a generic 6 MW WT with a 4-point suspension (shown in Fig. 1) and finite element (FE) models of tower and rotorblades were accessible. These models were developed at the Center for Wind Power Drives [5].
First, the drivetrain of the WT was modified to show the topology of the currently developed integrated drivetrain design, presented by several manufacturers in the past years. Here, the main bearing unit, the gearbox and the generator are housed in the same housing component. Gearbox and generator are not individually suspended on the supporting structure, but instead the whole drivetrain is suspended via the main bearings and the main bearing housing that sits right above the tower axis.
From the baseline drivetrain, 2 planetary stages were sufficiently designed to be used in the desired model regarding bearing location, load capacity and transmission. The parallel gearing stage, as well as the high speed shaft and the related housing were removed from the drivetrain. The generator rotor and stator including housing needed a fully new design. A housing was designed in order to be mountable directly to the gearbox housing ((1) in Fig. 1). The generator rotor diameter needed to be substantially increased in order to reach the desired power characteristics while working with much lower rotational speeds and higher torques. Also, the generator rotor is aligned with the gearbox output shaft and is also suspended via its bearings.
Due to the changed and increased requirements for the main bearings, a new set of main bearings was chosen. The main shaft is suspended via two tapered roller bearings that are mounted in opposed direction. Since gearbox and generator units are not suspended individually anymore, the loading of the main bearings is increased and it can no longer be denoted as 4-point supported due to the missing gearbox suspension. Here a static evaluation of the resulting main bearing loads was carried out and sufficiently large and robust bearings in combination with an optimal distance between rotor-sided and non-rotor-sided bearing were chosen. In order to calculate the input loads of the rotor, aeroelastic windfield simulations with the turbine model were carried out at the rated wind speed of 10 m/s, corresponding to DLC 1.1 and DLC 3.1 according to the IEC standard [6]. Next, the main shaft was scaled up to fit into the new bearing arrangement. This leads to large mass of the main shaft, resulting in a significantly heavier drivetrain (see Table 1). In addition, the introduction if the LSSC will add further 0.7 t to the mass. In further optimization steps an increasing of the inner diameter and thus a decrease in mass will be evaluated. Also a housing for the main bearing unit ((2) in Fig. 1) was developed with reinforcements ((3) in Fig. 1) and a connection to the supporting beams of the mainframe ((4) in Fig. 1).
Between main shaft and gearbox input shaft, installation space for a LSSC was provided. This space will be constant later for the reference model and the model equipped with the LSSC ((5) in Fig. 1). In the reference model, the main shaft is 650 mm longer to bypass the installation space and reach the connection to the gearbox input shaft. This design decision was made to ensure maximum comparability of the results. With an elastic coupling between main shaft and planet carrier, an additional set of bearings is necessary to suspend the planet carrier as it is no longer sufficiently supported by the main bearing. Therefore, two cylindrical roller bearings for the planet carrier are present only in the model equipped with the LSSC.
An overall comparison of the original and modified topologies can be seen in Fig. 1 and Table 1. Pitch and azimuth drives were not adopted since they will not be considered in the dynamic simulation.
Since these investigation aims to identify effects of a LSSC on the structure-borne sound distribution in the WT drivetrain, dynamic models of the respective variants need to be derived from the designed geometries. For this purpose all drivetrain components need to be considered as flexible structures during the simulation in the time domain. This includes all gear wheels as well. The geometries of all components were imported in a FE environment. Here, the structures were meshed and material parameters were associated with the mesh elements. Secondly, a modal reduction was carried out according to Craig-Bampton [7]. Selected degrees of freedom (DOF) were identified as Master-DOF in order to introduce loads at the respective locations in the dynamic simulation. The structural eigenmodes up to 200 Hz are considered since dominant excitation effects due to the gear meshing, wind and generator loads are expected up to 100 Hz. Since load distribution has a nonnegligible influence on the structure-borne sound transfer, advanced bearing models are deployed which take individual roller loads and displacements in account. A more detailed description of this method can be found in [8].
For a realistic loading situation during the dynamic simulation, the MBS model, more precisely the rotor blades, are coupled with a blade-element-momentum-theory solver that calculates local wind loads along the blade span [5]. This loads are applied to the model. The wind condition is arbitrarily selectable by the user. In this investigation a transient laminar wind profile will be considered starting from cut-in wind speed going to rated wind speed of the WT in 120 s and staying there for further 30 s. This will ensure that the structural dynamics and acoustics in all operational points can be observed.
After modification of the gearbox and generator structures, also the WT controller needed modification. The increased generator torque along with the decreased generator speed was accounted for with a scaling of the respective speed-load-characteristic. This was feasible since the windpower-characteristic, which is determined by the aerodynamics of the blades and the rated power of the WT, was held constant. Figure 2 shows the 3D visualization of the modified integrated turbine (left) and drivetrain in detail (right). From this model two variants will be derived and compared in this paper. A reference model with a slightly longer mainshaft that is directly connected to the first planet carrier of the gearbox and a LSSC model where a LSSC is integrated between mainshaft and first planet carrier. In the second variant, the planet carrier is supported via two cylindrical bearings. These models will be addressed as reference

Analysis methods
In this section, the methods will be described that will be used to compile the raw data from the dynamic simulations into the data described in Sect. 3.
The raw data consists of time series data of all model states, e.g. forces, displacements and modal parameters. The simulation is sampled with 2000 Hz. In Sect. 3.1 data will be plotted either raw in the time domain or processed via Fast-Fourier-transformation into the frequency domain. In Sect. 3.2 the data will be analyzed further using a Transfer Path analysis (TPA). This method is described below.
The Aim of a TPA is to study the transmission of vibrations from an active source to a passive structure. To describe the transfer characteristics, the excitation and transfer characteristics are described separately. In the literature, different types of transfer path analysis are distinguished. There is the classical TPA, the component-based TPA and the operational TPA [9]. In classical TPA, which is applied in this paper, the FRF is determined for the assembled system, and the excitation is measured with the interface forces at the transfer path (Fi) from the dynamic simulation. The transfer characteristics are described by separately measuring the Frequency Response Function (FRF) [9].
The individual path contributions are calculated by multiplying the FRF and the Interface Force (1). The sum of the path contributions can be compared here with the measurements at the reference point (yk).
The graphical representation of the path contributions is done using a Partial Path Contribution Plot (PPCP). This allows the amplitude of the path contributions for each path to be plotted on top of each other depending on the speed, frequency at an operating point or order [10].

Dynamic analysis
To give an overall picture about the effect a LSSC has on the described drivetrain, in this section, force spectra from 10 Hz up to 200 Hz at all interfaces between rotating drivetrain and drivetrain housing will be presented and discussed. Finally, the accelerations at the blade root and tower top will be compared in order to come to preliminary conclusions regarding tonality of the systems since structure borne sound is typically transferred to the large surface structures and from there, radiated as airborne sound. Loads are exchanged between the drivetrain and the housing at the following positions which are also represented in Fig. 3     Partial path contribution plots of the Reference system (a) and the LSSC system (b) for lateral tower top acceleration in comparison since they are only present for the LSSC system. Anyhow such spectra might be of interest because they are contributors which are exciting within the reference system. Such sources might ad critical excitations.
From the spectra below it can be derived that the gear meshing of both stages are the main excitations in both systems. The gear meshing excitation of the 1st stage has an overall smaller effect on all transfer paths between drivetrain and housing. The radial loads of the main bearing are significantly excited at 14.6 Hz and its second to fourth harmonics in the reference system. The introduction of a LSSC successfully almost eliminates the excitation of the main shaft with vibrations from the first gearbox stage. This is achieved by smaller non-torque loads entering the gearbox, because of the LSSC and the additional planet carrier bearings, which ultimately lead to smaller force amplitudes at the tooth flanks of the first stage. Additionally, the excitation from the second stage is suppressed to a large extent for the generator-side main bearing, indicating that with the additional elasticity of a coupling, a significant transfer path for structure borne sound through the main shaft was cut. When observing the radial gearing loads from the second stage, it can be seen that the LSSC reduces the radial contact force amplitude at 95.7 Hz by 56.3%. The circumferential amplitude component is reduced by 30.3%.
When further inspecting the loads of the gearbox bearings (Fig. 5), differences are noticeable, but not clearly in favor of one system variant. Differences in the acoustic system behavior are therefore most likely due to the cutting of the transfer path between gearbox and main shaft via LSSC and the reduced gearing force amplitudes. It can firstly be stated that the gearbox excitations due to the 1st can successfully be decoupled from the low speed part of the drivetrain. Secondarily, the excitation due to the 2nd stage can successfully be reduced at its root and thus its effect in almost all parts of the system. At this point it is K worth mentioning that the transfer path housing-mainframetower is also significant and cannot be addressed directly by the use of a LSSC.
In a subsequent step, the accelerations at the interface points to the sound emitting structures tower and rotor blades are compared. Differences in the acceleration at this points are strong indicators for the transfer of structure borne sound into those structures. When being excited, vibrations are radiated into the ambient air as airborne sound which poses a risk for tonalities.
First the tower top acceleration in axial (x) und lateral (y) direction are shown in Fig. 6. In order to ensure comparable global deformation of the tower, the comparison of the total deflection as time series is shown. The spectrum of the tower top acceleration shows significantly reduced amplitudes at all major excitation frequencies. The amplitude at 95.7 Hz is reduced by 39.7%, the second harmonic at 191.4 Hz is reduced by 70.8%. Even though the direct transfer path between the tooth contacts and the tower via the housing-mainframe-tower path could not be isolated, a significant reduction of vibration could be achieved here due to the reduction of the excitation of the housing.
For the rotor blades similar observations can be made. Figure 7 on the right shows the angle of the rotor as time series to ensure concurring loading due to local wind conditions and gravity direction. On the left, the spectra of the blade tip acceleration and the blade root acceleration are shown. Both are measured relative to the pitch bearing ring on the hub side. Again, the 2nd gear meshing frequency and its second harmonic are the dominant frequencies. At the blade root, the reduction of vibration amplitudes due to a LSSC amounts to 85.3% at 95.7 Hz and 71% at 191.4 Hz.
Summarizing it can be stated that, although both tower and blades are still coupled to the gearbox via the drivetrain housing, the vibrations of the blade root is significantly reduced with a LSSC. This indicates that the main transfer path of structure-borne sound from the gearbox into the blades is directly through the main shaft. Much less is transferred from the housing back into the main shaft via the main bearing and then transferred to the blades. Similarly for the tower, one of the main transfer path of structure-borne sound appears to be via the main shaft over the bearings into the main frame.

Transfer path analysis
In this section the excitation of the tower will be studied in more detail using the TPA described in Sect. 2.2. The contribution of the various paths from the rotating drivetrain to the drivetrain housing are compared for the reference system and the LSSC system. The direction of acceleration of the tower top is distinguished between axial direction (parallel to the drivetrain rotational axis, Fig. 8) and lateral (horizontally, Fig. 9).
In the reference system no contribution of the 1st planet carrier bearings can be detected, since they are only present in the LSSC system. Furthermore these are cylindrical roller bearings leading to no contribution in the axial direction.
For the axial direction, the main contributors to the tower acceleration are both main bearings for both systems, specifically the loads in axial direction. In the reference system, the axial bearing loads of the generator shaft additionally become significant contributors with higher rotational speeds.
For the lateral direction, the lateral loads of the main bearing and the 1st planet carrier bearings are the main contributors in the reference system. In the LSSC system vibration of the tower top are mainly driven by the loads transferred through 1st and 2nd planet carrier bearings.
These results are in accordance with the results from Sect. 3.1. The PPCP of the lateral tower top acceleration shows the cutting of the load transfer between gearbox and main shaft to a large extent, preventing the main bearing from being a major transfer path here. However, in axial direction this effect is less distinct. But since the amplitudes are already lowered here due to the LSSC, this nevertheless leads to a reduction of structure-borne sound transfer into the tower.

Conclusion
In this paper, the reduction of structure-borne sound introduced into sound-radiating structures of a 6 MW onshore wind turbine by a Low Speed Shaft Coupling was investigated. It was found that, in the integrated drivetrain design that was investigated, the structure-borne sound that is excited in the gearbox is transferred into the main shaft and housing and distributed further in the system from there on. This is a disadvantage to classical separated drivetrains since the housing of main bearing, gearbox and generator are naturally decoupled via design. However, significant reductions in accelerations at interface points to the large surface structures indicate the potential to decrease radiated sound from the gearbox. This can be done without redesigning the complete drivetrain but instead connecting the main shaft and the gearbox with a LSSC. It is strongly expected that similar technologies at other critical interface points could lead to further positive effects. Especially the gear meshing frequency of the second planetary stage significantly excited vibrations in the whole system. Although the LSSC clearly reduced this effect, an additional damping of the 2nd stage, for example mounted to the planet carrier, could further decrease the excitation of the structure [11]. Another option can be the introduction of reduced stiffness between the 2nd stage and the generator shaft by a semi torsional elastic coupling element [11]. In future research, the actual sound radiation of the sound emitting structure will be included in the simulation via recovering the surface displacements of the respective structures using the modal parameters of the dynamic simulation as presented in [12]. This will help to quantify the findings of this study. Also the effect of sound excitation from the electromagnetic airgap forces becomes of interest for the new integrated drivetrain concept with a midspeed generator. The authors see room for further research, especially in the interaction between generator and gearbox in a rigidly mounted housing.
Funding Open Access funding enabled and organized by Projekt DEAL.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4. 0/.