Predicting thermodiffusion in an arbitrary binary liquid hydrocarbon mixtures using artificial neural networks

A previously presented neural network-based thermodiffusion model that was valid for n-alkane type components has been extended to predict the thermo-solutal diffusion in an arbitrary binary hydrocarbon mixture. The enhanced model uses additional input information about the binary system and is based on a significantly large database of thermodiffusion data. Apart from the development and validation with respect to an extensive set of experimental data on the binary mixtures from the literature, the ability of the model to predict the known thermodiffusion trends has been demonstrated. The model can be potentially extended to multi-component mixtures and for any type of mixture, viz., polymers, molten metals, water-alcohol, colloidal mixtures etc.