Research On Data Driven Modeling And Multi-objective Optimization Method For Sugar Crystallization Process Based On CFD

The boiled sugar crystallizing tank is a relatively vacuum-tight convection heat exchange equipment.The good convection circulation condition and convective heat transfer efficiency in the tank are the key to improving the efficiency of boiling sugar and reducing the energy consumption of production.At present,the geometric parameters of the crystallizing tank and the operating parameters of the boiled sugar process are basically set according to experience,the coupling relationship between the parameters and the parameters and the effect of each parameter on the scouring performance are not clear.In order to fully study and optimize the physical structure parameters and operating parameters of the crystallization tank to improve the convective circulation and convective heat transfer efficiency in the tank,and improve the performance of the tank to reduce energy loss,this paper is based on the CFD model study of the boiled sugar.process,combined with parametric simulation.The experimental method and the PSO-SVR data-driven modeling method are used to obtain the two objective functions of entropy production and pressure drop.The multi-objective optimization problem is solved by NSGA-II and the parameter optimization problem is solved.Specific research content includes:(1)According to the internal fluid characteristics and experimental data of the boiled sugar crystallizing tank,based on the inter-phase drag coefficient,the gas-liquid two-phase flow model of the boiled sugar crystallizing tank is constructed based on the mass conservation,energy conservation and momentum conservation equations to achieve reliable prediction of two performance indicators of entropy production and pressure drop.(2)Study the CFD parametric simulation experiment method,with the air drum heating tube height,the air drum heating tube diameter,the air drum lower baffle height,the auxiliary steam inlet diameter and the sugar paste inlet and outlet diameter,a total of 5 geometric parameters and the initial absolute pressure in the tank,the tank The internal operating temperature,the air drum temperature and the auxiliary steam inlet speed have a total of four operating parameters as parameterized variables.The data sample set of parameterized variables and entropy production and pressure drop are obtained by random sampling simulation experiment,based on PCA.Principal Component Analysis implements the dimensionality reduction of the input sample set.(3)Based on the sample set after dimension reduction,the construction method of PSO-SVR data-driven regression model is studied to realize the regression prediction of entropy production and pressure drop of boiled sugar crystallizing tank.With RMSE,MAE,MRE,SSE and R2 as performance indicators,the effectiveness of PSO-SVR algorithm is verified by comparison with regression models constructed by generalized linear model,deep learning model,decision tree,random forest and gradient lifting tree under the same conditions.Sex and superiority.(4)Using the physical structure parameters and operating parameters as decision variables,construct a multi-objective optimization model with minimum entropy production and pressure drop,obtain Pareto frontier by NSGA-II solution,and decide multi-objective optimization problem of boiling sugar process based on TOPSIS method.The ultimate optimal solution.The optimized design variables are under the condition of CFD numerical analysis under the same conditions.Compared with before optimization,the entropy production is reduced by 9.73%,the pressure drop loss is reduced by 11.36%,and the velocity field,pressure field and temperature field are significantly improved,and the convection and transmission are improved.Heat optimizes the design goal of reducing unnecessary energy losses.