Algoritmo genetico na seleção de variaveis em calibração multivariada de dados espectroscopicos

AUTOR(ES)
DATA DE PUBLICAÇÃO

1998

RESUMO

This work describes the development and application of genetic algorithm in variable selection of spectroscopic data in multivariate calibration. The genetic algorithm is based on Darwin evolution theory, and in last years has increase the interest in the scientific community .A program for calculations was developed in MATLAB language. For program evaluation were realize three experiments: i - Simultaneous sugars determination in solution with midi infrared spectroscopy using attenuated total reflection ( ATR ); ii -Simultaneous determination of Metyl- terc-butyl-ether and Ethanol in gasoline based on near infrared spectroscopy; iii - Simultaneous optimization in more than one variable dimension using genetic algorithm in generalized rank annihilation method, applied in flow system for determination of copper and zinc. In the first two calibration methods was used partial least squares ( PLS ) and multiple linear regression ( MLR ) and the results were compared with obtained, using variable selection. The third experiment consisted in the variable selection in bilinear method, determing the best sets of time ( pH ) and wavelengths, in order to optimize the best condition for the analytes determination.

ASSUNTO(S)

calibração algoritmos espectroscopia de infravermelho

ACESSO AO ARTIGO

Documentos Relacionados