Developing Stabilized Multi-Layer Perceptron for Prediction of Hydrate Formation Conditions in the presence of Tetra n-butylammonium Fluoride

Recent studies reveal that presence of semi-clathrate formers such as tetra n-butylammoniumfluoride (TBAF) moderates the thermodynamics of gas hydrate formation dramatically.Although a great number of studies accomplished on the thermodynamics of semi-clathratehydrates, while, few attempts have been made to investigate the kinetics of this type of hydrates.In the present study, an in-house stabilized network algorithm is developed to estimate thehydrate dissociation conditions for the nitrogen+water+tetra-n-butylammonium fluoridesystems. In order to train efficient algorithm, large set of experimental data containing TBAFpromoter with different mass fraction of 0.018 - 0.4328 have been gathered from open literaturealong with our experimental set-up to develop various models. The reliability of stabilizationtechniques was checked by using 80% of experimental data for training and remains forvalidation purposes. The obtained results show that stabilized MLP network has superiorperformance over seen data with correlation coefficient of 0.9989 and root mean square errorof 0.1423. Comparison of stabilized and un-stabilized networks clearly revel that the optimalstabilization level is essential for filtering the noise and can improve the prediction accuracy. Itcan be concluded that stabilized MLP network is a reliable tool for anticipating hydrateformation pressure with TBAF promoter additive.

Developing Stabilized Multi-Layer Perceptron for Prediction of Hydrate Formation Conditions in the presence of Tetra n-butylammonium Fluoride Keywords:

TBAF , Hydrate , MLP , Stabilization , Nitrogen