Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
Development of a structural metamodel in MDO for wind blades based on neural network
Submission Author:
Júlia da Silva Maschietto , SP , Brazil
Co-Authors:
Júlia da Silva Maschietto, Juan Pablo Julca Avila, Geraldo Majella Nunes Junior
Presenter: Júlia da Silva Maschietto
doi://10.26678/ABCM.COBEM2021.COB2021-0615
Abstract
The wind sector is constantly growing due to the demand for clean and renewable energy and the generation of energy by wind turbines is one of the most efficient methods of extracting wind energy from the environment. In order to capture more energy from the wind, many countries have attached great importance to the design and development of large horizontal wind turbine axes with greater power, longer blades and taller towers. With the advancement of engineering techniques and project complexity, the need for optimizations is increasingly evident, as it is necessary to minimize costs and increase the efficiency of the project. This paper deals with a problem in the wind industry, which in turn has a high relationship with aeronautics, since the wind turbine when in motion is subject to aerodynamic and inertial loads, similar to that of an aircraft. The objective of this research is to contribute to the scientific advancement in the area of structure dynamics through the implementation of a structure tracking in a Multidisciplinary Design Optimization (MDO) applied to a real problem, in the optimization of wind turbine blade parameters. The optimization methods are difficult to use directly, as the process of integrating optimization algorithms with simulation codes is time-consuming and complex. These difficulties can be overcome with the use of metamodels, which are simple approximations of expensive computational analysis, but which keep the results faithful. In addition, the Blade analysis with Multi-Body (BaBM) technique will be used to accurately describe a complex geometry with great blade deformation. Therefore, the research method will be based on a MATLAB code developed considering BaMB as a case generator, entering stiffness parameters and geometric data. The results obtained in this phase will be used to feed the neural network in order to generate a metamodel, once it is ready, the MDO integrated with several technologies in wind blades will be implemented. The expected result is a metamodel that when entering the input parameters, it will output the structural response in a short time, having good accuracy and dispensing with more complex analyzes. In conclusion, the paper brings a promising and relevant study for both academia and industry, mainly for bringing quick analyzes, saving time and resources.
Keywords
MDO, MATLAB, Metamodel, Optimization
