Exploración y aplicación de "Diagnostic Feature Designer" para desarrollar modelos de diagnóstico de fallos en plantas eólicas

This degree work explores the application "Diagnostic Feature Designer" (DFD) of MATLAB in order to understand its functionalities and, consequenrly, create models for diagnosing failures in wind turbines. For this purpose, the MathWorks documentation was used as a guide in the development of the scripts, covering the preprocessing, data processing, and feature extraction in the time, time-frequency, and frequency domains. To find these diagnostic models, the temperatura data of the converter provided by the Electric Corporation of Ecuador (CELEC) was selected. These data were chosen because they represent 33% of failures in wind turbines. With these temperaturas, processed in all domains, 5 important characteristics were identified using the supervised two-class test method (T-test): mean, peak value, root mean square (RMS), peak value with intrinsic mode functions, and band power. These features were used according to the order of importance determined by the T-test method. These characteristics or condition indicators were trained in the Classification Learner application, with the tree model delivering the highest accuracy during training, achieving an accuracy of 97.97%. This result from the DFD application underscores the importance of using this tool to obtain diagnostic models and condition indicators, particularly in the case of wind turbines. Keywords: DFD, MathWorks, T-test, supervised method, Classification Learner, condition indicators.

• Autor principal: Santos Ordoñez, Wilmer Andrés (author)
• Formato: bachelorThesis
• Idioma: spa
• Publicado em: 2024
• Assuntos: DFD; MATHWORKS; T-TEST; MÉTODO SUPERVISADO; CLASIFICATION LEARNER; INDICADORES DE CONDICIÓN
• Acesso em linha: https://dspace.unl.edu.ec/jspui/handle/123456789/30387