FTIR/PLS and SVM multivariate calibrations to determination of the coke amount into the deactivated catalysts and the product of the methanol to gasoline conversion
2020; Elsevier BV; Volume: 105; Linguagem: Inglês
10.1016/j.infrared.2020.103229
ISSN1879-0275
AutoresMaryam Juybar, Mohammadreza Khanmohammadi, Amir Bagheri Garmarudi,
Tópico(s)Analytical Chemistry and Chromatography
ResumoThe rapid deactivation of zeolite catalysts in the methanol-to-gasoline conversion due to the accumulation of carbonaceous species into the pores is a technological and economic problem that has always been considered. In this work, after zeolites framework modification with various metals, the coke formation into zeolites pores and the bulky by-products (C10, C10+), as a precursor of the coke, were determined using the Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFT) in the wavenumber range of 1000–2000 and 1000–4000 cm−1, respectively. The results of this method were modeled using linear PLS and non-linear SVM approaches. The TGA and GC–MS were used as a reference method. Before implementing PLS and SVM regression algorithms, preprocessing methods like the baseline correction and the multiplicative scatter correction (MSC) were performed on the spectroscopic data. RMSEP and R2 for coke amount by PLS and SVM algorithm were (0.33, 0.25), (0.96, 0.98), and for organic fraction (bulky by-products) were (0.38, 0.33), (0.95, 0.98) respectively. Owing to the non-linearity of chemical parameters, the nonlinear SVM regression algorithm presents better results with less error in predicting of the coke and bulky hydrocarbons amounts.
Referência(s)