Application of Unsupervised Learning for the Evaluation of Aerogels’ Efficiency towards Dye Removal—A Principal Component Analysis (PCA) Approach

Abstract

Water scarcity is a growing global issue, particularly in areas with limited freshwater sources, urging for sustainable water management practices to insure equitable access for all people. One way to address this problem is to implement advanced methods for treating existing contaminated water to offer more clean water. Adsorption through membranes technology is an important water treatment technique, and nanocellulose (NC)-, chitosan (CS)-, and graphene (G)- based aerogels are considered good adsorbents. To estimate the efficiency of dye removal for the mentioned aerogels, we intend to use an unsupervised machine learning approach known as “Principal Component Analysis”. PCA showed that the chitosan-based ones have the lowest regeneration efficiencies, along with a moderate number of regenerations. NC2, NC9, and G5 are preferred where there is high adsorption energy to the membrane, and high porosities could be tolerated, but this allows lower removal efficiencies of dye contaminants. NC3, NC5, NC6, and NC11 have high removal efficiencies even with low porosities and surface area. In brief, PCA presents a powerful tool to unravel the efficiency of aerogels towards dye removal. Hence, several conditions need to be considered when employing or even manufacturing the investigated aerogels.

Towards Understanding Aerogels’ Efficiency for Oil Removal—A Principal Component Analysis Approach

“…The first two PCs exhibited 64.02% of the total variance (42.18% for PC1 and 21.84% for PC2; Figure 1 a). Interestingly, higher variance in the case of oil removal was obtained in comparison to the dataset investigated for the case of ion and dye removal [ 12 , 13 ]. This indicates a higher scope of applicability of the adopted method for the sake of revealing the hidden patterns and the certain correlation between physico-chemical properties from one side and adsorption parameters from another side, in the case of oil removal.…”

“…This indicates a higher scope of applicability of the adopted method for the sake of revealing the hidden patterns and the certain correlation between physico-chemical properties from one side and adsorption parameters from another side, in the case of oil removal. The aforementioned findings may indicate the presence of some other properties influencing the activity of the investigated aerogels towards oil removal, yet to a lower extent than the case of ion and dye removal [ 12 , 13 ]. For organic molecules, the stipulation of molecular polarity and capacity to produce hydrogen bonds should be taken into consideration, for the sake of better seeking its influence, in regard to the molecular interaction of pollutants with the investigated type of matrices [ 37 ].…”

“…In this study, we attempted to apply the so-called “Principal Component Analysis” (PCA) method for the sake of enhancing our understanding towards the efficiency of mostly nanocellulose, along with chitosan and graphene oxide-based aerogels, for oil and organic pollutants removal. The application of PCA has given a higher representativeness for oil removal compared with the previous findings of dye and ion removal [ 12 , 13 ]. This has been shown by a higher total variance yielded by the PCA of this study.…”

“…However, further research is needed to explore their potential applications and optimize their performance in different water treatment scenarios. One way to seek potential uses for these types of membranes is by applying data analysis on the dataset encompassing physico-chemical properties, adsorption parameters, and even manufacturing conditions and trade-offs [ 12 , 13 ]. One of the most suited data analysis techniques is “Principal Component Analysis” (PCA).…”

“…PCA is widely used in various fields, such as geology [ 15 , 16 ], biomass characterization, and valorization [ 17 ]. For aerogels, our previous studies applied PCA for the sake of revealing hidden patterns between the physical and chemical properties and adsorption parameters [ 12 , 13 ]. The machine learning approach found its importance in evaluating several water treatment processes, including the efficiency of nanofiltration membranes and the fabrication of sustainable materials [ 18 , 19 , 20 ].…”