Caracterização potenciométrica simultânea em fluxo, de espécies aniônicas, empregando análise das componentes principais e redes neurais artificiais

• Ano de defesa: 2006
• Autor(a) principal: Penoni, Nayara
• Orientador(a): Reis, Efraim Lázaro
• Banca de defesa: Reis, César , Queiroz, Maria Eliana Lopes Ribeiro de , Fidencio, Paulo Henrique , Milagres, Benjamin Gonçalves
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Agroquímica
• Departamento: Agroquímica analítica; Agroquímica inorgânica e Físico-química; Agroquímica orgânica
• País: BR
• Palavras-chave em Português: Potenciometria; Eletrodos seletivos a íons; Redes neurais; Análise de componentes principais; Análise por injeção de fluxo
• Palavras-chave em Inglês: Potentiometry; Anions selective eletrodes; Neural networks; Principal components analysis; Flow injection analysis
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::QUIMICA::QUIMICA ANALITICA
• Link de acesso: http://locus.ufv.br/handle/123456789/2057

A potentiometric system was mounted. It was composed of 6 electrodes, ion selective (ISE) of bromide, iodide, nitrate, nitrite, sulfate and cyanide. These electrodes had been placed in line, allowing it to do potentiometric determination in continuous flow. The variables, flow control, and the stabilization of readouts had been verified through a developed computational program in Delphi language. With the use of this system, instrumental techniques of calibration had been developed, through chemometric, capable to separate answers of different ions. The techniques of multivariate analysis used had been the principal components analysis (PCA) and the artificial neural networks. With the use of the cubical experimental planning of net simplex lattice , it was possible to get 63 mixtures with the six described anions with concentrations that had varied between 10-2 mol L-1 to 1,33x10-3 mol L-1, others 3 mixtures had been used for validation. Also, in different environments, 8 samples of water had been collected. These samples had been separated in a group in the graph of PCA, which seems to indicate the presence of different ions from the six that had been analyzed. The other mixtures had been separate by the PCA according to the concentration and the interferings. The PC1 explained 79.09% of the variance and the PC2 explained 10.28% of the variance. The architecture of the artificial neural networks was optimized, and the minors mid square-errors of forecast (RMSEP) had been obtained. The used architecture has three layers. The entrance layer has 6 neurons, the intermediate layer has 13 neurons and the exit layer has 6 neurons. As a post of transferences, the sigmoidal tangent was used for the intermediate layer, and the linear tangent was used for the exit layer. The estimate error by the neural networks was in a order of 10-3, which shows a considerable interference of anions, causing an inadequate forecast by the net. The samples of collected water could not have been submitted to the forecast by the artificial neural networks because the concentrations of anions may have values that surpass the interval of training of the artificial neural networks.