SMART SPR: identificação e análise das respostas fornecidas por sensores baseados em ressonância de plasmons de superfície

• Ano de defesa: 2020
• Autor(a) principal: Gomes, Júlio Cartier Maia
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal Rural do Semi-Árido; Brasil; Centro de Ciências Sociais Aplicadas e Humanas - CCSAH; UFERSA; Programa de Pós-Graduação em Ciência da Computação
• Palavras-chave em Português: Sensor óptico; Ressonância de plasmons de superfície (RPS); Aprendizagem de máquina; Sensor inteligente; Optical sensor; Surface plasmon resonance (SPR); Machine learning (ML); Smart sensor; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufersa.edu.br/handle/prefix/5545

Sensors based on Surface Plasmon Resonance (SPR) allow the evaluation of refractive index shifts, with the objective of detecting molecular interactions in food safety, environmental protection, detection of different chemical, biochemical and organic compounds, among several other applications. The reliability of the information transmitted by an SPR sensor must be guaranteed, given that wrong information can cause consequences, such as, misinterpretations and manipulations of molecular bonds. To improve and attest to the quality of the responses provided in the SPR-based sensor, this work investigates the use of different machine learning techniques and descriptors in acquired sensorgrams. More specifically, this work seeks to identify patterns and anomalies in their behavior in the SPR responses and to analyze the impact of these learning techniques on the quality of the sensor. The results obtained through the statistical analysis allowed to conclude that the sensorgram temporal descriptor obtained a better performance with the classification and that the use of linear regression made it possible to analyze the areas of interest in the sensorgram with satisfactory results, thus creating an SPR sensor with a intelligent response. This work also resulted in the creation of a graphical interface for the classification of sensorgrams. That way, the scientific contribution is the application of machine learning techniques that allowed to identify, analyze and classify the responses of the sensorgram