Contact
    Start Publications Novel Surrogates for Membrane Fouling and the Application of ...
    ←  Publication Finder
    QSense

    Novel Surrogates for Membrane Fouling and the Application of Support Vector Machine in Analyzing Fouling Mechanism

    Year: 2021

    Journal: Membranes, Volume 11, DEC

    Authors: Meng, Xianghao; Wang, Fukuan; Meng, Shujuan; Wang, Rui; Mao, Zhongyuan; Li, Yue; Yu, Meifeng; Wang, Xuye; Zhao, Qian; Yang, Linyan

    Organizations: National Natural Science Foundation of China [51808019]

    Keywords: membrane fouling; extracellular polymeric substances; surrogates of foulants; support vector machine

    It is difficult to recognize specific fouling mechanisms due to the complexity of practical feed water, thus the current studies usually employ foulant surrogates to carry out research, such as alginate and xanthan gum. However, the representativeness of these surrogates is questionable. In this work, the classical surrogates (i.e., alginate and xanthan gum) were systematically studied, and results showed that they behaved differently during filtration. For the mixture of alginate and xanthan gum, both filtration behaviors and adsorption tests performed by quartz-crystal microbalance with dissipation monitoring (QCM-D) indicated that alginate plays a leading role in fouling development. Furthermore, by examining the filtration behaviors of extracellular polymeric substances (EPS) extracted from practical source water, it turns out that the gel layer formation is responsible for EPS fouling, and the properties of gel layer formed by EPS share more similarities with that formed from pectin instead of alginate. In addition, with the use of experimental data sets extracted from this study and our previous studies, a modeling method was established and tested by the support vector machine (SVM) to predict complex filtration behaviors. Results showed that the small differences of fouling mechanisms lying between alginate and pectin cannot be recognized by Hermia's models, and SVM can show a discrimination as high as 76.92%. As such, SVM may be a powerful tool to predict complex filtration behaviors.
    View publication