Application of Artificial Neural Networks to Modelling Vapour–Liquid Equilibrium in the Fischer–Tropsch Reaction

Modelling vapour–liquid equilibrium (VLE) requires a choice of the equation of state (EOS), the choice of pure component parameters, and the choice of mixing rules. The more constants the EOS has, the more mixing rules are required, and the more data is required for evaluating pure-component parameters. The Fischer–Tropsch (FT) reaction mechanism has been studied for over 90 years yet an adequate description of its kinetics still seems to escape us. Including VLE in modelling an FT reactor only compounds the intricacy.

The field of computational intelligence is revolutionizing many areas of science and engineering. The ability to extract useful insights from vast amounts of weak and incomplete information is not only fuelling the current interest in “big data”, but has also enabled rapid progress in more traditional disciplines such as computer vision, estimation, and robotics, where data are available but difficult to interpret.

The FT synthesis face similar challenges, as many molecular and reaction pathways are still poorly characterized and available data are incoherent. Therefore, machine learning techniques such as artificial neural networks may be instrumental in overcoming these challenges by modelling the dependencies between random variables and using them to extract and accumulate the small amounts of information each random event provides. Hence, this study applies artificial neural networks to the complex problem of modelling VLE in the FT reactor. The data used to train, validate, and test the model is obtained from literature.