The effects of quasi-steady loading on a virtual spur gear model

With the projected growth of wind energy in the United States expected to account for 20% of the energy portfolio by 2030, it can be expected that wind turbines will not only increase in number, but also in size. This increase in size implies that internal components, such as the gearbox, will also increase to handle the higher loads. And given the high failure rates for gearbox components already in existence, one could expect more failure if nothing is done to improve reliability. It is well known that wind loading is not constant, rather it is random and ultimately causes fatigue loading. This thesis is concerned with studying what the effects of dynamic loading are on a smaller gear system. It is assumed the findings of this study can be scaled to a larger wind turbine system. A simple spur gear pair is first simulated at constant loading to establish a baseline and then run with a sinusoidal input with differing amplitudes and frequencies. The hypothesis is that by varying the amplitude and frequency, the responses for gear contact force and input and output shaft torques will also vary. And if these variations are noted, then conclusions may be drawn as how the frequency and amplitude influence the system. After which, it may then be correlated to a wind turbine system. Knowing what affects the frequency and amplitude have on a smaller system may help to establish guidelines. For this model, mechanical simulation software is used to build a multibody dynamics model of a spur gear system with flexible shafts. Using known wind data obtained near Amarillo, TX, a matrix of possible frequencies and amplitudes for a sinusoidal input are implemented and the solutions compared to those at constant loading. It was found that the system responded similarly regardless of input, showing RMS values for accelerations of approximately 50 m/s², gear contact forces of 520 N, input shaft torques of 42 Nm, and output shaft torques of 78 Nm. This behavior is not expected is most likely due to insufficiencies in the assumptions made to construct the model.