Machine learning approach for classification of REE/Fe-zeolite catalysts for fenton-like reaction

Abstract

Various heterogeneous catalysts based on rare earth elements (REE) and iron supported on zeolites were selected and analyzed using machine learning approaches. REE were used in the preparation of multiple REE/Fe-zeolite catalysts with lanthanum, praseodymium or cerium obtained by ion exchange or impregnation methods, using FAU or MFI structures as supports. The efficiency of these REE/Fe-zeolite catalysts was examined in Fenton-like reaction, in the degradation of tartrazine (Tar) and indigo carmine (IC) as selected organic pollutants in the aqueous solution. The REE/Fe-zeolite catalysts demonstrated outstanding performance, with Tar being degraded by over 80% and IC 95%. Machine learning algorithms were employed for clustering and classification of the different catalysts, based on their performance. Unsupervised learning algorithms like Principal Component Analysis and K-Means were used for pattern recognition while supervised classifiers were employed to classify the heterogeneous catalysts, considering their ability to degrade dyes by Fenton reaction.