Monitoring Of Surface Fouling Layer Characteristics And Research On Its Transformation Process In Membrane Proces

The accumulation of foulants on the membrane surface is the main incentive of membrane fouling and membrane performance degradation.The accuracy and convenience of the monitoring method for the fouling layer directly affect the judgment of the membrane fouling process.Therefore,it is particularly critical to select an appropriate method to monitor the fouling layer.In this experiment,the Ultrasonic Phased Array（UPA）method was used to monitor the formation process of the fouling layer in situ,then the membrane surface was hydrophilically modified.Monitor the deposition of different pollutants on the modified membrane surface.Add metal ions and analyze the applicability of PDA modified membranes to filtrate different pollutants with UPA methods.Finally,use the thickness and density information to train the machine learning model and predict the process of membrane fouling,which may provide a new idea for real-time and in-situ monitoring of membrane fouling.First,predict the membrane fouling process with UPA method,compare the feature difference of cake layer made by different pollutants.Kaolin is used as the signature inorganic substance,and bovine serum albumin（BSA）,sodium alginate（SA）,humic acid（HA）and their mixtures are used as the signature organic substance.Through UPA monitoring,it was found that the cake layer made of kaolin was thick and loose,and the membrane flux decreased a little after using kaolin cross-flow filtration.Among natural organic matter,the cake layer made by HA has the lowest density,so it is easier to remove by cross-flow,while the cake layer was by BSA and SA has high viscosity and high density,and will cause serious membrane fouling.Comparing the monitoring results of UPA for different pollutants,it can be found that UPA can distinguish the accumulation of foulants on the membrane surface from the beginning of filtration to the time when the filtration is stable.UPA can be used as a means of monitoring the fouling layer on the membrane surface.Second,PDA was used to hydrophilically modify the membrane surface,and UPA was used to evaluate the coating effect.It was found that PDA coating could significantly increase the hydrophilicity of the membrane surface.After coating with PDA,the anti-fouling effect of the membrane surface against BSA was significantly enhanced,while the anti-fouling performance against SA showed a downward trend.It was found that PDA coating could significantly increase the hydrophilicity of the membrane surface.This is because the polysaccharides and the PDA coating are more likely to attract each other,the foulants was trend to adhere to the membrane surface.After adding divalent metal ions,the membrane fouling caused by BSA increased under the action of different ions,indicating that the synergistic effect of metal ions and foulants has a certain influence on the performance of PDA modified membranes.Compared with Ca²⁺,Mg²⁺causes BSA to agglomerate and form larger aggregates,resulting in denser cake layer on the membrane surface and more serious membrane fouling.Finally,the density and thickness information of the cake layer is trained by two different machine learning models to predict the membrane fouling.The results show that compared with the random forest（RF）algorithm model,back propagation neural networks（BPNN）model and support vector machine（SVM）model,the convolutional neural network（CNN）model has higher prediction accuracy and is more suitable for experimental conditions.Therefore,use the CNN model to predict the membrane fouling process.At the same time,the anti-interference ability of the CNN model under different disturbances was analyzed,and it was found that the CNN model had the weakest anti-interference ability for the factor of pollutant concentration,and the strongest anti-interference ability for the filter temperature.Machine learning methods can help to improve operating conditions and reduce the membrane fouling.