Proposição de novo algoritmo de controle ativo de ruído baseado no método da combinação convexa com adição de ganho dinâmico do sinal de erro

• Ano de defesa: 2021
• Autor(a) principal: Fernando Basílio Felix
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais
• Palavras-chave em Português: Engenharia de estruturas; Algoritmos adaptativos; Controle ativo de ruído; Ventiladores (Máquinas)
• Link de acesso: https://hdl.handle.net/1843/42285

Different applications of ANC (active noise control) systems are found in the literature, one of them being the application of acoustic noise control in rotating machines. Noise coming from rotating machines, such as fans, has the characteristic of having deterministic type components. As it is possible to predict the periodicity of part of the noise generated by these machines, it is possible to generate synthesized signals to be used as a reference signal of an active noise control system. This work aims to study the active noise control methodology, more specifically the application of this methodology in ventilation ducts. The aim is to reduce noise from the passage of fan blades. To reach the object, a modification of the convex combination algorithm is proposed, performing the comparison of the modified algorithm with classic ones through computational simulations and implementation in an experimental plant. The research project is being developed on the premises of the Mechanical Engineering department of the Federal University of Minas Gerais, where an experimental plant equipped with an industrial fan was built. The ANC algorithms were implemented in computer simulations and also in the experimental plant. In the simulations, parameter variations were studied, such as delay in input and output signals, number of coefficients of digital filters, gain of input signals and noise complexity. In the experimental plant, the control algorithms were implemented in an FPGA controller (Field programmable gate array) and tests were performed with and without airflow. The response data evaluated were the steady-state noise attenuation level and the convergence speed of the algorithms. According to the data obtained in different tests, the influence of parameters such as number of coefficients of digital filters, complexity of noise signals and level of input signals was clear. In general, higher levels of the error signal cause higher rates of convergence, but when applied in excess it causes the algorithm to diverge. The MC-FXLMS/F algorithm showed better results when compared to the FXLMS and C-FXLMS/F algorithm for the application of noise control in ventilation ducts. The performance improvement is mainly due to the application of dynamic gain to the error signal.