The dynamic loads on the rollers inside the bearings of large wind turbine gearboxes operating under transient conditions are presented with a focus on identifying conditions leading to slippage of rollers. The methodology was developed using a multi-body model of the drivetrain coupled with aeroelastic simulations of the wind turbine system. A 5 MW reference wind turbine is considered for which a three-stage planetary gearbox is designed on the basis of upscaling of an actual 750 kW gearbox unit. Multi-body dynamic simulations are run using the ADAMS software using a detailed model of the gearbox planetary bearings to investigate transient loads inside the planet bearing. It was found that assembly and pre-loading conditions have significant influence on the bearing's operation. Also, the load distribution in the gearbox bearings strongly depends on wind turbine operation. Wind turbine start-up and shut-down under normal conditions are shown to induce roller slippage, as characterized by loss of contacts and impacts between rollers and raceways. The roller impacts occur under reduced initial pre-load on opposite sides of the load zone followed by stress variation, which can be one of the potential reasons leading to wear and premature bearing failures.While the focus herein is on the gearbox, important WT aspects like aeroelastic interactions and control are included. Aeroelastic models allow interactions between inertial, elastic and aerodynamic forces and their servo mechanisms. Examples of aeroelastic software are FAST 10 developed by NREL, HAWC2 11 by DTU Wind Energy, Bladed 12 developed by DNV-GL and AdWiMo 13 by MSC Software. These calculate WT loads and response in the time domain on the basis of input environmental conditions, but usually, in these models, the gearbox dynamics are neglected. 14 The detailed models for drivetrain components, usually for gearboxes, can be built by lumped parameter models, multi-body system or finite element (FE) method. These models allow simulation of the internal loading resulting from dynamical interactions in the drivetrain. It is important to determine the level of complexity required to balance between accuracy and computational efficiency. According to Guo et al., 15 for drivetrain modeling, the gearbox alone is insufficient to capture component loads and motions; coupling with the main bearing and generator is important. For simulation of loads in a gearbox, it is important to include the clearance in bearings and pre-load, which can influence the non-torque load transfer path, as well as deformations of housing and carrier that can change bearing loads. From the point of view of excitation sources, besides the torque input, the non-torque loads and gravity are relevant. 15 Oyague 14 has presented the state of the art on simulation codes for WT's gearboxes. The author has shown four stages of drivetrain modeling due to model complexity. A purely torsional model of WT drivetrain based on the lumped parameter system with 5 degrees of freedom (DOFs) was used by Mandic...

No abstract

In the paper an application of laser measurements for validation of the gears dynamic model have been presented. The model of the gears was compared with an experimental data in case of diagnostics of manufacturing failures. The comparison was done on the basis of vibration signals generated by the model and the real object. In the experiment correct work as well as incorrect one of the gears was studied. Purpose of the paper is to compare the model with the experimental data; to this end, the Multibody model of the gears was build in MSC ADMS software. The model consists of two gears and two shafts, with both gears and shafts treated as rigid bodies, but contact surfaces between them are exible. To conclude, in the study comparison of acceleration signals estimates behavior due to other gears faults in the experiment and in the model has been presented. For optimal model validation, laser measurements on the shaft in the experiment were conducted.

In the paper, two-stage signal pre-processing algorithm based on the ltration is presented. The developed algorithm is dedicated for the diagnostic programmable device PUD-2. The PUD-2 is the real-time analyzer based on programmable logic devices FPGA, as well as on ARM processor. Application of FPGA programmable devices and ARM processors allows to merge advantages of hardware and software implementations. Further, analysis of digital lters parameters in case of its ecient realization on the FPGA is presented. The aim of the study is to select digital-lter parameters in such way that the available resources of FPGA are used eciently and lter characteristics meet established criteria. In the study, low pass nite impulse response and innite impulse response lters are compared. For the rst stage of the signal pre-processing algorithm, hardware implementation of the innite impulse response lter is proposed, contrary to the second stage, where software realization of the nite impulse response lter is suggested. Combination of hardware and software ltration algorithms allows for fast and ecient realization of signal pre-processing algorithm used in analysis carried out on the PUD-2.

No abstract