Determination of acidity in metal incorporated zeolites by infrared spectrometry using artificial neural network as chemometric approach.

The NH 3 -TPD analysis is a costly and tedious method to determine zeolites acidity. Thus, to do so, FTIR spectroscopy was quantitatively used as a fast and cost-effectively method. Back-propagation artificial neural network (BP-ANN) was used for the analysis of multivariate base on the characteristic absorbance of 11 zeolite samples after metal substitution in the ~3612 cm-1 region. The successive projection algorithm (SPA) was conducted for the uninformative variable elimination and feature selection strategies. The effect of pre-processing methods (e.g. MC and MSC) was examined. It is observed after using MSC for minimizing the light scattering effect and signal-to-noise correction, the minimum mean squared error (MSE) value of the testing set data reduced from 5.36 × 10-2 to 2.19 × 10-4 and R tot increases from 0.91 to 0.99. Also, the results of nonparametric Wilcoxon t-test and Sign test methods also confirmed that there is no clear difference between the zeolite acidity obtained by two conventional method and the proposed method. Image 1 • The zeolites FTIR vibration intensity in ~3612 cm-1 after metal loading is reduced. • Back-propagation artificial neural network (BP-ANN) as a multivariate analysis was applied for prediction of zeolites acidity. • Mean centering (MC) and Multiplicative scatter correction (MSC) were used as the preprocessing method. • After MSC, results of BP-ANN (R and MSE) were improved. • The results of nonparametric methods confirmed that there is no clear difference between the zeolite acidity obtained by two methods. [ABSTRACT FROM AUTHOR]