Sistema de mapeamento de força com sensores de macrocurvatura em fibra ótica multiplexados

• Ano de defesa: 2022
• Autor(a) principal: Carvalho, Vinicius de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Brasil; Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial; UTFPR
• Palavras-chave em Português: Detectores ópticos; Fibras ópticas; Transdutores; Análise de regressão; Optical detectors; Optical fibers; Transducers; Regression analysis; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA; Engenharia Elétrica
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/29618

The work presents sensing systems applied in the quasi-distributed force monitoring on rigid structures. The sensing is based on the response of an array of macrocurvature sensors serially multiplexed. Systems were developed with three and five sensor elements multiplexed in a single fiber link. These devices were coupled to plate-like structures to monitor the position and magnitude of an individual force applied to these structures. The magnitudes of the forces used in the experimental tests correspond to the range between 100 gf and 2000 gf in steps of 100 gf. A one-dimensional sensing matrix was formed by a PMMA structure, 30 cm long and 5 cm wide, instrumented with three sensing elements. The monitored region had six collinear force application areas. A two-dimensional sensing matrix was also elaborated, formed by a square metallic structure with sides of 20 cm instrumented with five sensing elements. This structure had nine force application areas. Regression models aimed at processing the optical signal transmitted by the fiber to make the sensing systems operation viable. Linear and nonlinear regression models were implemented based on Elastic Net and Support Vectors with reduced dimensionality by Principal Component Analysis (PCA). The trained predictive models that presented better results were applied in a test step that simulated the actual operation of the sensor systems. The average performances detected varied according to the monitored structure, the model complexity, and the number of characteristic variables used. The minor prediction errors in the test stage were 2.41 cm and 177 gf, while the largest were 4.01 cm and 397 gf for estimating the position and magnitude of the force, respectively. The results presented show the operating capacity of the regression models with reduced dimensionality both by L1 penalty of the Elastic Net model and by PCA. From the models based on Elastic Net, an evaluation of relevant spectral bands to the operation of the systems was carried out.