Research On Non-uniform Solid Stress Model Based On Artificial Intelligenc

Gas-solid fluidization is widely used in industrial production and everyday life,and has been involved in many fields such as energy,chemical industry and environmental protection.As the main simulation method for fluidization,MP-PIC(Multiphase particle-in-cell)method has the advantage of packing particles with the same properties into a numerical particle for analysis and calculation,thus reducing the time needed for calculation.Among then,the solid stress model which describes particle-particle interaction is one of the most important models for MP-PIC method.Recent research work has tried to introduce EMMS theory into the MP-PIC method to establish the heterogeneous EMMS solid stress model to account for the effect of non-uniform solid distribution.However,the calculation process is very complex and also very time consuming for this heterogeneous solid stress model.The expression of the heterogeneous EMMS solid stress can be obtained by manual fitting method.However,the fitting variable describes heterogeneous solid distribution as well as the fitting function describe the shape of solid stress are required to be determined manually to fitting.Since the heterogeneous solid stress function is highly nonlinearity in nature,the fitting precision is not high enough for the manual fitting model.And there is an obvious deviation between the fitting correlation and the original result,because it is hard to find out an appropriate parameter to characterize the heterogeneous solid concentration distribution as well as to find out an appropriate fitting function.To resolve those problems for heterogeneous solid stress model,machine learning method has been introduced in the correlative modeling of the heterogeneous solid stress model in the current work.At first,the solid volume fraction and the parameter representing particle distribution are taken as independent variables,to establish an artificial neutral network(ANN)model.After training of the artificial neutral network model,the ANN solid stress model of solid is obtained.The error caused by the artificial definition of fitting function can be avoided by applying the artificial neural network based machine learning method.Subsequently,the gas-solid fluidization reactor was simulated with the newly developed ANN solid stress model under different grid resolution and coarse-grained ratio to verify the accuracy of the model.Generally,the simulation results agree with the experimental data.However,the same problem in manual fitting method can also be found in the double variables based ANN model,that the fitting precision is not high enough due the question that,it is difficult to find out an optimum parameter to characterize the heterogeneous particle concentration distribution.Thus,as an improvement of the model it is proposed to use the solid volume fraction distribution as the variable of solid stress function.Specifically for the two dimensional condition with regular grid,the solid volume fraction of the central cell together with the volume fractions of the eight neighboring cells are adopted to approximate the solid distribution inside the central cell,and they are used as the input parameter and solid stress is designed as the output to build an artificial neural network model.The network is trained repeatedly to correlate the solid stress and solid volume fraction distribution.Finally,the solid distribution based ANN solid stress model can be obtained.Compared to the ANN solid stress model with two markers,fitting precision is obviously elevated for the improved ANN model.In the simulation of the gas-solid fluidization reactor,the simulation accuracy is also improved in comparison with the ANN solid stress model with two markers.And the simulation dependence on the grid resolution and coarse-grained ratio is also reduced by applying the improved ANN solid stress model.In summary,the current work has successfully introduced machine learning in the correlative modeling of heterogeneous solid stress model and the ANN based heterogeneous solid stress model was obtained.Finally,the results and conclusions are summarized and the further research work is proposed.