1. Repositório Institucional - Universidade Federal de Uberlândia
2. Faculdade de Engenharia Mecânica (FEMEC)
3. TCC - Engenharia Aeronáutica

ORCID:	http://orcid.org/0009-0001-6279-581X
Tipo do documento:	Trabalho de Conclusão de Curso
Tipo de acesso:	Acesso Aberto
Título:	Artificial neural networks for critical flutter speed prediction in semi-span flat-plate wings with geometric variations
Autor(es):	Amaral, Vinícius Nunes
Primeiro orientador:	Almeida, Odenir de
Primeiro coorientador:	Morais, Tobias Souza
Primeiro membro da banca:	Venson, Giuliano Gardolinski
Segundo membro da banca:	Morais, Tobias Souza
Resumo:	This thesis presents a data‐driven method to predict modal damping and the critical flutter speed of semi‐span flat‐plate wings. A linear FE model (FEMAP/MSC Nastran SOL 145, PKNL with Doublet–Lattice aerodynamics) was calibrated with impact‐hammer modal tests and used to generate mode-by-mode g(V) and f(V) curves for spans 0.35- 0.50 m. From these runs we built a supervised dataset with inputs {S,V, f1,..., f4} and targets {g1,...,g4} and trained a regularized MLP that predicts damping across modes; flutter speed is inferred as the }rst zero‐crossing of the predicted g(V). On held-out simulation cases the network achieved low errors and reproduced the span-dependence of flutter speed; comparisons with wind-tunnel data (limited to ∼25 m/s) showed agreement within the experimental resolution. The approach provides near-instant estimates suitable for preliminary design, while remaining limited by the scope of the parametric sweep and the tunnel envelope. Future work includes broadening geometry variation (e.g., dihedral, sweep), expanding numerical and experimental coverage, and exploring physics-guided or unsupervised models that learn stability directly from raw time-series.
Palavras-chave:	Aeroelasticity Flutter Semi-span wing Modal damping Multi-Layer Perceptron (MLP) Finite Element Method (FEM) Machine Learning MSC Nastran SOL 145 PK/PKNL flutter methods Doublet–Lattice Method (DLM)
Área(s) do CNPq:	CNPQ::ENGENHARIAS::ENGENHARIA AEROESPACIAL::ESTRUTURAS AEROESPACIAIS::AEROELASTICIDADE
Idioma:	eng
País:	Brasil
Editora:	Universidade Federal de Uberlândia
Referência:	AMARAL, Vinícius Nunes. Artificial neural networks for critical flutter speed prediction in semi-span flat-plate wings with geometric variations. 2025. 56 f. Trabalho de Conclusão de Curso (Graduação em Engenharia Aeronáutica) – Universidade Federal de Uberlândia, Uberlândia, 2025.
URI:	https://repositorio.ufu.br/handle/123456789/49039
Data de defesa:	30-Set-2025
Aparece nas coleções:	TCC - Engenharia Aeronáutica

Os itens no repositório estão protegidos por copyright, com todos os direitos reservados, salvo quando é indicado o contrário.