Artificial Intelligences Tools for Prediction of Hydrate Formation Conditions for the System of Methane + Tetra n-butylammonium fluoride + Water

Tetra n-butylammonium fluoride (TBAF) regarded as a hydrate former that promote the gas hydrate formation conditions, considerably. There are few thermodynamic models available to predict the systems containing TBAF. The in-house RN, ANFIS and conventional MLP networks were utilized to predict the methane + TBAF hydrate formation conditions, in the present study. Large experimental data set conainting TBAF with mass fraction of 0.02 - 0.4482 have been gathered from open literature along with our experimental set-up to develop various models. All the networks have been checked using 80% of experimental data for training and kept remained 20% to examine estimation performances. It was found that RN has the best performences over seen data with correlation coefficient of 0.9996, while conventional MLP shows the superior estimation for unseen data with R2 0.9951. Error analysis reveal that RN, ANFIS and MLP networks are reliable tools for anticipating hydrate formation pressure with TBAF promoter additive