Aplicacao das transformadas wavelets em dados de espectroscopia no infravermelho para a determinacao de biodiesel e oleo vegetal em blendas biodiesel/oleo vegetal/diesel

Partial least squares (PLS) calibration models were compared for the determination of biodiesel and vegetable oil in different blends using infrared spectra data (FTIR/ATR). Eighty binary and ternary blends containing biodiesel, vegetable oil and diesel were made, 48 were employed to compose the calibration set and 32 for prediction set. Initially the spectral signals of samples containing biodiesel and vegetable oil between 0 and 8%(v/v) were compressed using Discrete Wavelet Transform. The best models with compressed or uncompressed signals were compared, using the interval partial least squares algorithm (iPLS), also identifying which region of the infrared showed better correlation. The results using the compressed data showed similar errors in the determination of the vegetable oil and of the biodiesel in blends. The combination of the regression algorithm iPLS applied to data obtained by FTIR/ATR has shown considerable promise in the development of a simple, rapid and non-destructive method for the determination of adulteration with vegetable oil in biodiesel or biodiesel/diesel blends Keywords: wavelets; infrared; multivariate calibration; biodiesel; adulteration