Study On Low Eutectic Properties Of Nitrile Resin Based On GA-BP Neural Network

Resin matrix composite materials are widely used in various fields due to their high specific strength,high specific modulus,and strong designability.Their heat resistance and mechanical properties are mainly determined by the resin matrix,so does the processability.Therefore,a high-performance resin with high temperature resistance,excellent mechanical properties and low temperature liquid phase processability is the key to the development of advanced resin matrix composite materials.In the preliminary work of our research group,the eutectic technology was used to reduce the melting point and viscosity of the resin,and the low-temperature liquid-phase processability of the resin was achieved.The self-made phthalonitrile monomers were used to prepare multiple groups eutectic of phthalonitrile resin.Obvious hydrogen bonds can be detected in common eutectic organic mixtures,but there is no obvious hydrogen bond found in this new eutectic nitrile-based resin system.Further research found that the introduction of different groups on the nitrile-based resin monomer can affect the eutectic melting point of the eutectic resin.The introduction of different groups actually changes various molecular structure parameters of the compound,and thus changes the interaction force between different molecules.First of all,from the perspective of molecular structure,this article obtained 6 kinds of molecular structure parameters of 9 self-made nitrile-based resin monomers through Gaussian,including energy,polarizability,dipole moment,LUMO energy,HOMO energy,and LUMO energy minus HOMO energy,and carry out the mathematical processing of the binary system.We use the factor analysis module in SPSS to perform principal component analysis,the results show that the first two principal components after dimensionality reduction can represent 90%of the information of the previous six molecular structure parameters.The DOE module in Isight was used to analyze the sensitivity of its eutectic degree,and the results showed that the molecular structure parameters most related to eutectic degree were LUMO energy,HOMO energy,and LUMO energy minus HOMO energy.Then,with key molecular structure parameters as input and eutectic degree as output,GA-BP neural network is established to obtain a quantitative relationship model between the two.The model results show that when the principal component after subtraction is selected as the input parameter,when the number of hidden layer neurons is 3?12,the coefficient of determination R~2 is greater than 0.9,the best model R~2 is0.990,and the relative error of eutectic degree prediction is within 5%.At the same time,an optimized eutectic degree of 86.7?is given,which is nearly 40?higher than the experimental maximum of 48.5?.When the value of the first principal component is 0.23?0.38 and the value of the second principal component is between 0.25 and 0.97,the optimized eutectic degree of the binary eutectic can reach 85.2?86.7?,realizing the simulation for the purpose of optimization.Finally,from the perspective of the intermolecular interaction force,Gaussian is used to calculate the interaction energy between the binary eutectics to characterize the interaction force between different molecules.Then use Matlab to establish a GA-BP neural network between interaction energy and eutectic degree.The model results show that when the number of hidden layer neurons is 3?12,the coefficient of determination R~2 is greater than 0.9,the max R~2 is 0.990,and the relative error of eutectic degree prediction is within 5%.At the same time,an optimized eutectic degree of 76.9?is given,which is nearly 30?higher than the experimental maximum of 48.5?.When the value of the interaction energy is between-5.84 and-5.01kcal/mol,the eutectic degree can reach 75.5 to 76.9°C,which achieves the purpose of fitting optimization.