ABSTRACT
Several studies related to the dynamic modelling of the WT drivetrain have been reported. Some of these studies are focused on analysing the loading on bearings within the WT gearbox under
1 INTRODUCTION
The wind power industry has seen a rapid development in the last decade, especially in Europe (Smolders et al. 2010) and the United State (Daniels et al. 2004). At the end of 2003, about 35,000 MW of wind energy capacity had been installed globally (Daniels et al. 2004) and around 11 GW of capacity had been installed in Europe alone by the end of 2013. However, the wind industry faces serious challenges of premature failures of WT gearboxes occurring much earlier than the designed life. The gear and bearing failures in WT gearboxes may be attributed to the excessive loads under various operating conditions (Grujicic et al. 2014). Gearbox failures result in significant component replacements which subsequently increase the cost for wind energy (Grujicic et al. 2014). Replacing WT gearboxes needs heavy duty equipment such as cranes, which is generally expensive to hire and almost impossible to operate in adverse weather conditions for offshore wind farms (Tavner et al. 2008). A good understanding of how and why WT gearboxes are failing prematurely will improve the gearbox component design and reduce the overall maintenance cost. It is important to understand the dynamic behaviour of WT drivetrain under various operational conditions and how the flexible movement of gearbox components, such as planetary carrier and floating sun shaft, within a WT gearbox, reacts to transient loads. The multibody dynamic models developed
various operation conditions, such as normal operation and shutdown. However, these studies ignored the dynamic behaviour of some key gearbox components (Bruce et al. 2015) (Scott et al. 2012). Three types of modelling approaches have been examined by Peeters (Peeters et al. 2005). The first is the purely torsional multibody model where the gears are modelled as rigid bodies, each with a single Degree Of Freedom (DOF) in the torsional axis, connected to each other by linear springs. Such models are able to investigate torsional loads, loads on gears and bearings, and eigenfrequencies (Girsang et al. 2014) (Mandic et al. 2012) (Shi et al. 2013). The second approach is the 6-DOF rigid multibody modelling with discrete component flexibilities, which produces a more accurate model; however, the complexity of the gearbox modelling has been increased. These models take into account of individual gear stages of the gearbox and the influence of bearings stiffness. The component flexibility is represented by spring-damper systems. Such models facilitate a more detailed description of gear mesh and bearing stiffness. The third approach is the fully flexible multibody model which increases the modelling accuracy of gearbox component flexibilities from the second approach by using finite element modelling. The third approach allows the visualisation of the influence of different subcomponent flexibilities, however it is computationally expensive. Increasing flexibility does not always result in more accurate modelling results. The addition of flexible components to the WT drivetrain model increases the model complexity and affects slightly on eigenfrequencies however much more on the corresponding mode shapes (Peeters et al. 2005). LaCava et al. (LaCava et al. 2013) observed that the influence of increasing the level of gearbox component flexibilities on the bearing and gear loads using seven models of various levels of complexity. Their results concluded that the flexibility of gearbox sub-components, such as housing, carrier and the main shaft, has noticeable influences on the loading of the plant bearings. It has been reported that using a flexible coupling with reasonable stiffness can reduce the torque amplitude of the high speed shaft (Peeters et al. 2006). Increasing model complexity only has a small effect on modelling accuracy while resulting in higher computational costs (LaCava et al. 2012). The different levels of modelling complexity help gearbox designers to improve designs and assess the dynamic behaviour of chosen designs under specific loading conditions (Oyague et al. 2008). The torsional dynamic model for WT gearboxes is one of the common modelling approaches because of its fast solution time with low computational costs. The purely torsional multibody dynamic models developed in
this paper are computationally effective to capture the torsional loads and dynamic responses of key WT drivetrain components during free and forced vibrations.
