Application Research Of GM And Gas-liquid Two-phase Flow Field In MBR

Membrane bioreactor(Membrane Bio-Reactor)is a new type of water treatment technology used in the field of wastewater treatment and water reuse.The technology is mainly composed of two parts:membrane separation unit and biological treatment unit,which is significant for the study of membrane fouling in membrane separation unit.A lot of research has focused on how to slow down the membrane fouling of MBR.In the various measures to mitigate MBR membrane fouling,the conditions and parameters of the regulation of the system are highly operable,flexible and effective.The important parameters of MBR aeration system in operation in order to mitigate the impact of MBR membrane fouling,distribution of gas-liquid two-phase flow field simulation of the first aeration when using the Fluent simulation software,to explore how to achieve better effect of aeration mitigation membrane fouling from the micro perspective.Then,we use GM algorithm and Linear algorithm in deep learning to predict the optimal aeration intensity of MBR system at different time.We build a model from the perspective of data analysis to predict the optimal aeration intensity corresponding to different time.And then the function of reference data that can be used to provide the best aeration intensity at different time is achieved.From microcosmic angle of two-phase flow simulation when aeration experiment has two main difficulties,the first difficulty is to understand the theory of fluid mechanics,especially the fluid of gas and liquid in two states in the mass conservation equation and momentum conservation integration when needed to establish constant equation;second is to difficult to operate analysis of fluid mechanics software Fluent,main control mesh to construct the container space gas-liquid fluid exists,and master Fluent software built on the use of different methods of fluid mechanics model.Prediction of the data analysis view,the best aeration intensity values corresponding to different time is the focus of this paper,firstly,according to the different time has the best aeration intensity values corresponding to the structure of a sequence,the sequence as the grey system GM algorithm input,GM algorithm can be used to meet the conditions,the best aeration time corresponding the output of the next value;and then use the best aeration intensity sequence with the same GM algorithm,split the sequence and data normalization form Linear algorithm in deep learning required for input and output,using mean square loss function as the loss function of Linear linear neural network model training,using stochastic gradient descent algorithm as the optimization algorithm to optimize the convergence loss function parameters,and rapidly obtained the corresponding Linear model.The final implementation of the Linear linear neural network algorithm mainly uses the recently popular Pytorch depth learning framework.The program implementation of the Linear network model in this paper uses the framework and gets more accurate prediction results.