In this thesis, the reliability of small wind turbines is studied. Both conventional reliability analysis methods and the novel Bayesian models (Bayesian Hierarchical Modelling (BHM)) are used to analyse the reliability performance of the Gaia-Wind turbines / assemblies and components of the Gaia-Wind turbine. In Chapter 2, a simple failure mode and effect analysis (FMEA) is conducted. An approximated risk priority number (RPN) is calculated for each failure mode and assembly. The assembly that is identified to have the highest RPN is the "Rotor and Blade Assembly". As for the failure modes, "Blade Split" and "Generator Failure" failure modes are identified to have the highest RPNs. In Chapter 3, the conventional methods including the Kaplan-Meier Analysis, Weibull Plot Analysis, Homogeneous Poisson Process (HPP) Analysis, and Crow-AMSAA (Non-Homogeneous Poisson Process (NHPP)) Analysis are used to study the reliability performance of the generic turbine and the critical assemblies based on the approximated RPNs. By using these conventional methods, the L10 life can be approximated (Kaplan-Meier), the main failure modes of an assembly can be identified (Weibull Plot Analysis), the annual failure rate can be estimated (HPP), and the number of future failures can be predicted (NHPP). These methods have been implemented in a novel on-line interactive platform, named ReliaOS (Chapter 7), which effectively facilitates the process of converting the information in the warranty record to the meaningful reliability information. Three novel BHM models are proposed and implemented in WinBUGS (an open source software), namely the repair model, the environmental model, and the informative prior framework, (Chapter 5 and Chapter 6). The repair model is used to quantify the repair effectiveness of a generic repair action. The model is applied on both the turbine level as well as the component level. At the turbine level, the annual failure rate of the generic turbine is predicted to be 0:159 per turbine per year at the first year. Individual turbines can be categorised into different quality levels ("Good", "Good- Normal", "Normal", "Normal-Bad", and "Bad") based on the predicted annual failure rate values. At the component level, "Blade split", "Cracked Frame", and "Generator Failure" failure modes are studied. These are the most critical failure modes for "Rotor and Blade Assembly", "Tower, Foundation, and Nacelle", and "Generator" assemblies respectively. "Cracked Frame" failure mode is predicted to have the lowest characteristic life and a slightly increasing failure rate trend. The repair effectiveness of the "Cracked Frame" failure mode is identified to be slightly ineffective. The environmental model quantifies the influence of three environmental covariates, i.e. AverageWind Speed (AWS), Turbulence Intensity (TI), and Terrain Slope (TS). These environmental covariates are all identified to have negative impact to the reliability of the generic turbine, where TI and AWS have more pronounced impact than TS. The informative prior BHM framework offers a way of quantifying the reliability of the drivetrain frame (which corresponds to the "Cracked Frame" failure mode) in a situation where zero failure instance is recorded for the new drivetrain frame design. This is achieved by jointly considering the simulation results from SOLIDWORKS as the prior information into the BHM model. This thesis strives to understand the reliability performance of the Gaia-Wind small wind turbine from different perspectives, i.e. the generic turbine, individual turbines, and the components, by the use of conventional methods and the proposed BHM models. The novel on-line reliability platform, ReliaOS, mitigates the difficulties in converting the information in the data to the reliability information for the end users. It is believed that the proposed BHM models and the ReliaOS on-line reliability analysis platform will improve the reliability analysis of small-wind turbines.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:738978 |
| Date | January 2017 |
| Creators | Wu, JenHao |
| Contributors | Mueller, Markus ; Ingram, David |
| Publisher | University of Edinburgh |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/1842/29015 |
Page generated in 0.0157 seconds