ON THE DETERMINATION OF NORMAL BOILING POINT OF PURE COMPONENTS: A SOFT COMPUTING APPROACH

In this study, a new generalized non-group contribution method is developed to predict normal boiling point of pure chemical components. A robust and fast estimation method based on feed forward artificial neural networks with back propagation algorithm is presented. Molecular weight and specific gravity was selected as the input parameters for the proposed model. In order to develop the model, the experimental data of 563 pure components are gathered from literature sources. The collected data includes experimental data points from 13 different groups including: paraffins; cycloparaffins; monooleffins and dioleffins; cyclooleffins and actylens; benzene derivatives; condensed ring aromatics and derivatives; acids, alcohols, and phenols and aldehydes; amines and nitrogen containing components; esters; ethers, ketones; halogenated hydrocarbons; sulfur containing hydrocarbons. The prediction results using the proposed method were compared to two of the most conventional and accurate previously published methods in estimating normal boiling point using statistical and graphical error analyses. Comparisons showed that the proposed model is more reliable and accurate than the available methods. The average absolute percent relative error of the obtained model is only 3.41%, much lower than the pre-existing correlations.

کلیدواژه ها:

Normal boiling point ، non-group contribution ، artificial neural network ،