Molecular Spectroscopy Analysis And Prediction Of Natural Organic Matter Membrane Fouling In Low-pressure Membrane Filtration

Low-pressure membrane filtration technology represented by microfiltration and ultrafiltration has developed rapidly in recent years due to characteristics of low operating pressure and good filtration quality,and has been developed rapidly and favored by industrial wastewater,urban water supply,and urban sewage treatment fields.Membrane fouling has always been an unavoidable problem in low-pressure membrane filtration technology,which largely hinders the promotion and development of membrane technology.Natural organic matter（NOM）,as a complex heterogeneous system composed of different organic matter,has a significant impact on membrane fouling.The large functional groups such as hydroxyl,carboxyl,amino,carbonyl,and phenol in NOM are generally chromophores or auxochromes,and when light interacts with them,they could show a specific molecular spectral front.Therefore,molecular spectroscopy techniques have become a powerful tool for studying NOM and its membrane fouling behavior.Therefore,exploring the correlation between NOM molecular spectral information and membrane fouling,and establishing a quantitative analysis of spectral information and membrane fouling is the key to realizing intelligent early warning,rapid diagnosis,and targeted regulation of membrane fouling.First,the correlation between NOM spectral characteristics and membrane fouling potential was established.In this study,the effects of filtrate particle size and hydrophilic and hydrophobic characteristics on membrane fouling behavior of microfiltration and ultrafiltration were investigated,and the unified membrane fouling index（UMFI）was used to characterize membrane fouling potential.The results showed that organic particles with larger particle sizes could aggravate membrane fouling,while inorganic particles with small particle sizes were more likely to enter and block membrane pores and caused serious membrane fouling.Hydrophilic neutral and hydrophobic neutral had the fastest fouling rate and the largest values of UMFI among all fractions.Combined with the ultraviolet-visible,fluorescence,and Raman spectral characteristics of the filtrate and the infrared spectral characteristics of the fouled membrane surface of the hydrophilic and hydrophobic components,the correlation between the spectral indices and the membrane fouling potential were established,respectively.The results showed that ultraviolet-visible and fluorescence spectra had a significant correlation with UMFI and could be used as sensing spectra to characterize membrane fouling in both microfiltration and ultrafiltration systems.Secondly,the multi-spectral fusion of ultraviolet-visible and fluorescence spectrum was realized,and a model for predicting UMFI was established.This study focused on the process of membrane fouling induced by coagulants and investigated the correlation between the solution spectral parameters and UMFI after adding different doses of coagulants.It was found that with the addition of coagulants,the particle size and Zeta potential of aggregates gradually increased,and the formed filter cake layer was loose and more permeable.The multi-spectral fusion for predicting UMFI was first established via multiple linear regressions（MLR）by using the parameters with both slope ratio（S_R）and specific fluorescence intensities（SFI）in various conditions with different solutions and coagulants.The backpropagation neural network（BPNN）was used to establish the suitable model under various conditions with good accuracy,robustness,and adaptability,indicating machine learning had a great application potential in membrane fouling.In addition,in order to achieve accurate predictions of membrane fouling potential in a long-term running reactor,the continuous membrane filtration-submersed reaction was built and four different types of surface water were simulated to change in influent to facilitate validation of the organic matter-spectra-fouling relation.The results of variance partitioning analysis showed that the components of polysaccharides（PS）,proteins（PN）,and humic acids（HA）acted on different stages of membrane fouling respectively.For the initial membrane fouling,it was mainly contributed by the individual PS content and the interaction of PS-PN-HA,while for UMFI was mainly contributed by the individual HA and PS-PN-HA.In addition,K-Nearest Neighbor（KNN）,BPNN,Support Vector Machine（SVM）,and Random Forest（RF）machine learning models were compared,and the results showed BPNN and SVM were more suitable to predict the membrane fouling potential.Finally,the quantitative characterization of PN and HA in NOM on the membrane surface at the initial stage of fouling was realized,and the structural characteristics of the complex foulant layer were analyzed.In this study,an in-situ online fluorescence optical fiber monitoring platform was successfully established to realize real-time monitoring of the accumulation behavior of foulant during the filtration process.The results showed that at the initial stage of filtration,PN substances were preferentially adsorbed on the membrane surface.With the progress of filtration,the loading of HA components gradually increased,and the PN-HA aggregate was gradually formed.In addition,the effects of different valence ions on the membrane fouling process and membrane cleaning were also discussed.The results showed that the structures of the foulant layers formed by the two solutions are different in the electrostatic and hydration interactions of Na⁺ions and Ca²⁺ions.The combined effect of electrostatic shielding and bridging caused by Ca²⁺ions resulted in more serious fouling of the gel layer,but the nanopore size of the gel layer had a higher retention rate of organics and ions.The Na⁺ions could cause the molecules of the polymer in the NOM solution to compress to a certain extent,making it easier for organic substances to block the membrane pores,and the foulant layer formed while aggravating the membrane fouling is more difficult to be removed by physical cleaning.