FLUID PROPERTY PREDICTIONS WITH THE AID OF NEURAL NETWORKS

Two group-contribution-based artificial neural networks were developed to predict a fluid's normal boiling point, critical properties, and acentric factor. Similar to the conventional group-contribution methods, the trained networks are capable of estimating those characteristic properties upon a fluid's molecular structure being known. Generally, promising results have been obtained by using the neural network as an alternative tool for predicting the thermodynamic properties.