| Petroleum hydrocarbon pyrolysis technology is one of the main sources of ethylene and other low carbon olefins,is also a complex chemical process.Development has a good predictive mathematical model,for the development of petroleum hydrocarbon cracking process has important guiding significance.This article chooses and the corresponding normal alkanes was cycloparaffin as model compounds,examines the reaction conditions,molecular structure and composition of raw materials on the thermal cracking conversion rate and product distribution.Combined with the free radical mechanism,focus on the molecular structure and their interaction on the influence of the thermal cracking reaction path,then build the experience model of the thermal cracking process and intelligent model.First of all,in order to explore the normal/cyclo-alkanes on the thermal cracking reaction path and product distribution,the influence of choosing hexane,heptane,cyclohexane,methyl cyclohexane as the model of reactants,analyzes its in 650-830℃the thermal cracking conversion and product distribution.The results showed that the molecular structure of alkanes determines the initial fracture and the formation of free radicals,C-C key influenced subsequent hydrogen reaction,beta fracture,combination and disproportionation reaction path.And there is interaction between different molecular structure.For example,n-hexane and cyclohexane mainly depends on the interactions between the hydrogen reaction.In n-hexane/cyclohexane mixture thermal cracking process,promoted the cyclohexane cracking n-hexane,cyclohexane inhibits n-hexane cracking.Then,in order to set up the experience of the paraffin thermal cracking model,choose is pentane,n-hexane and n-heptane,octane,nonane,decane,cyclohexane,methyl cyclohexane,n-hexane/cyclohexane,n-heptane/methyl cyclohexane as the model compounds,analyzes the reaction condition,the length of carbon chain and raw material composition on thermal cracking conversion rate and the yield of olefins.Study found,can use inverse proportion and quadratic function,linear function and exponential function,simulation ring chain length,content and reaction temperature on the decomposition of reactants and the impact of low carbon olefin formation,established by the experience of the four composite function model.The conversion,the model of ethylene yield,propylene yield showed good calculation accuracy.In addition,the model in the prediction of complex mixture of paraffin thermal cracking in the conversion and olefin yields good calculation accuracy,proves that the model has good accuracy and generality.Finally,based on paraffin thermal cracking of experimental data,the thermal cracking of alkane intelligent model is set up.The results show that Levenberg-Marquardt algorithm of BP neural network model was constructed,the training set of the R~2 value is 0.993,the texting set of the R~2 value is 0.986,the training set of MAE is 0.984,the testing set of MAE is 1.635.Using genetic algorithm to optimize the model,and greatly improve the accuracy and generalization performance of the model.In addition,and the use of Radial Basis Function(RBF)neural network to build model,the training set of the R~2 value is 0.999,the testing set of the R~2 value is 0.996.The training set of MAE is 0.914,the testing set of MAE is 1.367.The model and BP neural network model have similar properties,but compared with the BP neural network,the network structure is more simple and less network parameters and more applicability. |
