Research article

ANN IN NATURAL GAS TREATMENT PROCESS: CARBON DIOXIDE (CO2) - ETHANE (C2H6) AZEOTROPE SEPARATION

Daniel Chuquin-Vasco 1, *, Wendy Dávila3, Nelson Chuquin-Vasco2, Juan Chuquin-Vasco2 and Salvador C. Cardona 4

Online First: March 30, 2023


The aim in this study was to develop an artificial neural network (ANN) to forecast the mole fractions of the CO2- C2H6 azeotropic separation during the natural gas treatment process. The ANN was designed with experimental data (150 data pairs) obtained in DWSIM from a pre-viously process described in bibliography. The sample used to train the ANN is structured by three inputs: pressure, temperature and solvent / feed ratio, and six outputs: the mole fractions of distilled CO2 and residual ethane in the extractive column, the mole fractions of distilled ethane and residual propane in the solvent recovery column and the mole fractions of distilled ethane and residual ethane in the concentrator column. The neural network was designed using 80 hidden neurons in its architecture and the Bayesian regularization algorithm for training (MSE=0.0036 and R=0.9554). The ANN’s prediction was validated using statistical parameters (ANOVA and Kruskal Wallis) which indicate that the designed ANN is statistically valid and can be used to predict the mole fractions of the CO2- C2H6 azeotropic separation and can be used in the improvement of the processes of sweetening of natural gas after its demethanization.

Keywords

Azeotrope; Extractive Distillation; Natural Gas; DWSIM; Carbon Dioxide; MATLAB; Artificial Neural Networks (ANN)