Prediction of supercritical carbon dioxide solubility in polymers based on hybrid artificial intelligence method integrated with the diffusion theory

Abstract

Solubility is one of important research hotspots of physical chemistry properties and is widely utilized in the modification, synthesis and preparation of a lot of materials. To avoid the defects of traditional thermodynamic dissolution forecasting methods, according to the mass transfer features of a two-phase system, the dissolution process is simulated. In this paper, the diffusion theory is integrated into the improvement of particle swarm optimization (PSO) so that the particles in the algorithm evolve along with the diffusion energy. In this way, the improved PSO of dual-population diffusion is obtained and used to train the parameters of the radial basis function artificial neural network. Then, a prediction model for supercritical carbon dioxide solubility in polymers is proposed. The solution experiments of 8 polymers indicate that the predicted values with the model are consistent with the experimental results. The prediction accuracy is higher and the correlation is significant. The average relative error, mean square error and square correlation coefficient are respectively 0.0043, 0.0161, and 0.9954. The prediction model has a high comprehensive performance and provides the basis for the prediction, analysis and optimization of other physical and chemical fields.