Caracterização de tamanhos de partículas sólidas utilizando emissão acústica e inteligência computacional

• Ano de defesa: 2019
• Autor(a) principal: Xavier, Gilberto Magalhães
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Elétrica; UFRJ
• Palavras-chave em Português: Tamanhos de partículas; Emissão acústica; Inteligência computacional; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://hdl.handle.net/11422/21711

Solid particles sizing plays a significant role in the quality of different products and in the performance of several manufacturing processes. The use of acoustic emission for this task is an approach that offers significant advantages such as: non-invasiveness, low cost and rapid response, resulting in relevant improvements in process control and product quality. On the other hand, the acoustic emis- sion signals are admittedly difficult to interpret and analyze, and an interesting option to overcome these difficulties is to use computational intelligence to enable the extraction of useful and relevant information from the acoustic emission signals. However, developing methods for the characterization of solid particle sizes using acoustic emission and computational intelligence is not an easy task due to several challenges such as high dimensionality of acoustic emission signals, unavailability of simulation tools that are accessible and easy to use, as well as the lack of well- characterized and reproducible databases. In this study a model for the simulation of the acoustic emission measurement chain was developed, implemented and validated; a reproducible and well-characterized database of acoustic emission signals has been produced; seven new approaches combining computational intelligence techniques and two new preprocessing methods were developed, implemented, and evaluated. The results obtained allowed to propose a new hybrid approach that combines the computational intelligence with the expert knowledge and traditional methods of acoustic emission in an unprecedented way in this context. The proposed approach has shown an average absolute deviation below 1% for the prediction of the content of particles less than 750 μm and, as an additional and innovative advantage, it does not require re-training.