Predição do desempenho de hélices de pequeno porte com XGBoost: efeitos do processo de imputação da solidez por métodos de regressão

• Autor(a) principal: Rosa, Heitor Nunes
• Data de Publicação: 2022
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://hdl.handle.net/11449/217119
• topic: Machine Learning; Propeller; Surrogate model; Aeronáutica; Hélices; Modelo Substituto; Modelo substituto

This work has as main objective the elaboration of a surrogate model for the prediction of the performance of small propellers with the application of Machine Learning methods. The models were designed using the Python programming language, using the XGBoost regression algorithm, and based on the database provided by the University of Illinois at Urbana-Champaign. A data manipulation was also carried out to calculate the solidity, in order to evaluate its influence on the propeller performance. In this work, we dealt with missing data, which, despite the chosen algorithm being robust, an imputation method was applied to verify if there was an improvement in its performance. The model hyperparameters were tuned by a Bayesian optimization process. Satisfactory models were obtained for the development of preliminary designs of propellers with a relatively small prediction interval. This model will be useful to provide a better design phase, providing a faster and more efficient selection of propellers.