Numerical Analysis Of Direct Reduction Process Of Carbon-containing Pellet And Optimization Based On Neural Network

The steel industry is a high-consumption,heavy-polluting industry,and the large amount of metallurgical dust generated during the production of steel is one of its main sources of pollution.The metallurgical dust contains a large amount of iron,which has a high recycling value.However,in practice,a large proportion of the dust is not treated effectively and is disposed of in piles or landfills.Not only does this take up a lot of land,but it also causes a lot of environmental problems.With the increasing pressure of our national environmental protection policy,it has become urgent to find a green and efficient process for treating metallurgical dust in the steel industry.At present,the main method of recovering metallurgical dust is to collect the dust and mix it with coal power or toner to make carbon-containing pellet,which are then subjected to a direct reduction process in a rotary heat furnace and then used as a high quality raw material for blast furnace iron-making to recover the iron in the dust.The reduction of carbon-containing pellet in a rotary heat furnace is a complex process involving physical and chemical changes within the pellet as well as heat and mass transfer.Some research on this process has been carried out by different scholars for different research objects and purposes,but it is not yet adequate and there is less research on the rapid prediction of target results to guide the regulation of various process parameters in the actual production process.Therefore,COMSOL multiphysics software is used to simulate the complex reduction process of pellet and to analyse the influence of various parameters on the pellet reduction process in this paper.Then,based on the calculation results of the direct reduction process model of carbon-containing pellet,a BP neural network is used to establish a prediction model for the metallization rate of the direct reduction process of carbon-containing pellet and to carry out optimization of the process parameters of the reduction process on the basis of the established model.Firstly,a model for the direct reduction process of the carbon-containing pellet was developed using a metallurgical dust-containing pellet with specific components as the research object.The reduction reaction of iron oxides,the gasification reaction of carbon,the heat and mass transfer as well as the variation of physical parameters and pellet porosity during the reduction process were fully considered in the model.The calculation results show that the consumption rate of C during the initial 50s and temperature ramp-up of the pellet in the first 50s of the initial reaction are small,Fe₃O₄will first be produced from Fe₂O₃ in large quantities during this process.The Fe₃O₄concentration reaches its peak at about 200s and then Fe O is produced,and at this time the consumption rate of C and Fe₂O₃ increases and Fe starts to be generated,indicating that during the reduction of the carbon-containing pellet,the sequence of iron chemical forms in reduction is from Fe₂O₃ to Fe₃O₄ to Fe O to Fe.During the reduction process,the increase in reduction temperature,the increase in carbon to oxygen ratio and the decrease in pellet radius all contribute to the increase in the metallization rate of the carbon-containing pellet.The higher the furnace chamber temperature,the larger the pellet radius and the greater the temperature difference between the surface and the centre of the pellet.The porosity changes over time in a decreasing and then increasing manner.Comparing the sphere and ellipsoid pellet models,it is found that the ellipsoid pellet model is more favourable to the reduction process and therefore is more advantageous.Comparing the variable and constant furnace temperature shows that suitable temperature control not only ensures the reduction quality of the pellets,but also reduces the energy consumption in the production process,which is of great importance to the economic green production of enterprise.Based on the calculated results of the direct reduction model for carbon-containing pellet,a prediction model was developed using BP neural networks for the response of the direct reduction process of carbon-containing pellet,which provides a relatively fast and accurate prediction of the metallization rate of the direct reduction process.In addition,a genetic algorithm was used to optimize the prediction model to improve its prediction accuracy and generalization capability.In the modelling process,comparative analysis and study of the prediction model structure and model parameter configurations,including the number of hidden layer neurons,the neural network model training algorithm,the learning rate and the population size of the genetic algorithm,are highlighted.A sensitivity analysis of the parameters of the direct reduction process was also carried out by the developed prediction model,and it was obtained that reduction reaction time and temperature had greater influence on the reduction,while pellet carbon to oxygen ratio and size had less influence,which fits basically with the actual state.Finally,the NSGA-II optimization algorithm was used to optimize the process parameters of the direct reduction process by combining the mathematical model and the BP neural network prediction model,the Pareto optimal solution set was obtained,which could provide valuable guidance and reference for the actual producers in the process of regulating process parameters to achieve production targets.