Light hydrocarbons solvents solubility modeling in bitumen using learning approaches

In this paper, two types of machine learning, namely neuro-fuzzy inference system (ANFIS) and multilayer perceptron (MLP), have been studied to model light hydrocarbons' solubility solvent in bitumen. The 268 number of experimental data is used in this work from different articles. The input parameters are Temperature (T), pressure (P), and molecular weight (MW) of hydrocarbons. The result shows the high performance of the MLP model with a two layers to predict the experimental values. The estimated values were investigated by statistical parameters such as R (2), MSE, and MARD%. According to, statistical parameters, the values of 0.99, 0.00081, and 0.68 for MLP, and 0.96, 0.0029, and 0.78 for ANFIS indicate the high performance of the MLP model. Comparison between established models and previous work indicates that the developed model can be a suitable technique for solubility modeling

Automated summary

This study investigated the use of neuro-fuzzy inference system (ANFIS) and multilayer perceptron (MLP) for modeling the solubility of light hydrocarbons in bitumen. The MLP model showed high performance in predicting experimental values, with statistical parameters indicating its superiority over the ANFIS model.