Construction of a catalyst screening method for furfural hydrogenation to furfuryl alcohol based on machine learning

As global focus on environmental protection and sustainable development intensifies, the chemical industry is shifting towards greening and decarbonization. Efficient conversion of furfural (FFR) to furfuryl alcohol (FOL) has thus become a key research area. Traditional catalyst screening for the hy...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of cleaner production Jg. 523; S. 146357
Hauptverfasser: Yang, Xiaohui, Liu, Dongyu, Pan, Xinran, Shen, Lu, Liu, Wenman, Pang, Ruixin, Yu, Shitao, Liu, Shiwei, Geng, Sai, Li, Lu, Yu, Longzhen, Liu, Yue, Liu, Xiao
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 10.09.2025
Schlagworte:
Catalyst
Clustering algorithm
Furfural
Furfural hydrogenation
Machine learning
Furfural
Clustering algorithm
Furfural hydrogenation
Catalyst
Machine learning
ISSN:0959-6526
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:As global focus on environmental protection and sustainable development intensifies, the chemical industry is shifting towards greening and decarbonization. Efficient conversion of furfural (FFR) to furfuryl alcohol (FOL) has thus become a key research area. Traditional catalyst screening for the hydrogenation of FFR to FOL is time-intensive and labor-intensive, relying on trial-and-error methods that consume many resources with low success rates. Machine learning offers advantages, rapidly processing large datasets, conducting multi-dimensional analyses, identifying optimal catalysts, enhancing screening efficiency, and cutting costs. In this study, we established and improved a dataset for the FFR-to-FOL hydrogenation process. We refined K-Means cluster analysis to better understand data distribution for model comparison and optimization. To address dataset imbalance, we used an oversampling technique, significantly improving model performance. After a comprehensive evaluation, we selected Support Vector Machine (SVM) and Neural Network (NN) as core models. Using genetic algorithms and multi-objective optimization, we identified NiZr/CoOx as the optimal catalyst for the SVM model, and PdCu/MgO for the NN model. This work is expected to promote the industrial use of FFR derivatives and advance catalytic sciences with new catalyst screening and optimization ideas. To screen the optimal catalyst for furfural hydrogenation to furfuryl alcohol, we used machine learning techniques to innovate the catalyst screening method. The multi-objective optimization method based on genetic algorithm was applied to ultimately determine the best catalysts under the model: NiZr/CoOX (SVM), PdCu/MgO (NN). [Display omitted]
ISSN:0959-6526
DOI:10.1016/j.jclepro.2025.146357