|
چکیده
|
The UV-spectrophotometric data analysis of a binary and ternary mixtures containing vanillin, vanillic acid, syringaldehyde by using multivariate calibration methods, such as partial least-squares regression (PLS), and artificial neural networks (ANN) was described [1]. The simultaneous determination of binary and ternary mixtures of these compounds by using spectrophotometric methods is a difficult problem, due to spectral overlapping. For each compound a linear calibration was obtained in the concentration range of 0.61-20.99μ gml-1, 0.67-23.19 μgml-1 and 0.73-25.12 μgml-1 for vanillin, vanillic acid and syringaldehyde, respectively. The parameters controlling behavior of the systems were investigated and optimum conditions were selected. The calibration sets were based on 49 and 27 reference samples, consisting of binary and ternary mixtures, respectively. Feed-forward neural networks with backward propagation algorithm have been trained to quantity determination in ternary mixtures under optimum conditions [2]. The ANN and PLS models were compared and their predictive performance was tested with the use of synthetic samples for simultaneous determination of the three compounds, and in general. Satisfactory results were obtained.
|