A Reduced Order Model based on Artificial Neural Networks for nonlinear aeroelastic phenomena and application to composite material beams

[EN] Applications of composite materials in industry have increased due to their high stiffness-to-weight ratio. In the particular case of unidirectional fibers or perpendicular fabrics, the materials behavior is orthotropic, so that an extra degree of freedom, related to the orientation of the fibe...

Descripción completa

Detalles Bibliográficos
Autores: Torregrosa, A. J.|||0000-0003-0933-1626, Gil, A.|||0000-0001-7192-6992, Quintero-Igeño, Pedro-Manuel|||0000-0003-4373-2079, Cremades-Botella, Andrés|||0000-0002-7052-4913
Tipo de recurso: artículo
Fecha de publicación:2022
País:España
Institución:Universitat Politècnica de València (UPV)
Repositorio:RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Idioma:inglés
OAI Identifier:oai:riunet.upv.es:10251/196137
Acceso en línea:https://riunet.upv.es/handle/10251/196137
Access Level:acceso abierto
Palabra clave:Aeroelasticity
Reduced Order Model
Artificial Neural Networks
Structural coupling
Flutter
INGENIERIA AEROESPACIAL
MAQUINAS Y MOTORES TERMICOS
Descripción
Sumario:[EN] Applications of composite materials in industry have increased due to their high stiffness-to-weight ratio. In the particular case of unidirectional fibers or perpendicular fabrics, the materials behavior is orthotropic, so that an extra degree of freedom, related to the orientation of the fibers, must be included in the structural optimization. Composite material thin walled beam models have been developed for reducing the computational cost of the simulations. Traditionally, these models have been coupled with potential aerodynamics to calculate the aeroelastic response, and thus, the viscous nonlinear effects have been omitted. In order to capture these effects, this manuscript focus on the development of a Reduced Order Model enhanced by an Artificial Neural Network for the analysis of composite structures under aerodynamic loads. The presented methodology shows the training process of the neural network, the comparison with high fidelity simulations and the design optimization of a carbon fiber laminated foam beam. It is demonstrated that the model reduces the computational cost by orders of magnitude, while still capturing structural couplings and being capable of increasing the flutter velocity by more than 10% with respect to the longitudinal orientation.