60097 - A Comprehensive Model for Cetane Number Prediction Using Machine Learning 

Machine learning based predictive models are being extensively applied for predicting combustion properties like cetane number (CN), which is the measure of a fuels ignition quality. In the present work, a comprehensive model was developed using artificial neural networks (ANN) that can predict the ignition quality of fuels containing a large number of chemical classes, namely parraffins, iso-paraffins, olefins, naphthenes, aromatics, alcohols, ethers, ketones, esters and carboxylic acid. Experimental CN's of 555 fuels was used as a dataset for devolping the CN model. The compositional data of the fuels in the dataset, expressed in the form of eleven functional groups along with branching index (BI) and molecular weight were used as the input parameters of the model. These input parameters for the transportation fuels in the dataset were measured from 1H nuclear magnetic resonance (NMR) spectra, whereas the input parameters for pure components (and their blends) were calculated from their molecular structure. A shallow neural network with a single hidden layer was developed using Bayesian regularization. The developed model was validated with 15 % of the data points that were randomly generated and kept aside for model validation. A regression coefficient (R²) of 0.98 was observed between the predicted  and the experimental values along with an average absolute error of 1.1. The results showed that the developed model was succesful in predicting the CN of fuels and can be applied to pure compounds, blends and real fuels containing varied chemical functionalities.