ANFIS models for prediction of biodiesel fuels cetane number using desirability function

Cetane number (CN) is one of the most important properties of a fuel and indicates its ignition delay in combustion process. The aim of this study is to model and predict the cetane number of biodiesel from its fatty acid methyl esters composition using ANFIS technique. The input variables were the number of biodiesel fuels' double bonds (dn), and their molar weight (Mw). For designing ANFIS models, three fuzzy inference systems (FIS) structures were generated: grid partition, subtractive clustering and fuzzy c-means (FCM). Comparison of the developed models was performed by statistical criteria, such as coefficient of determination (R2), root mean squared error (RMSE), standard deviation of error (STD) and mean absolute percent error (MAPE) coupled with desirability function. The obtained results showed that the maximum coefficient of determination is related to grid partition FIS (pimf) and ranges from 0.939 to 0.951 for various data. The minimum values of RMSE, STD and MAPE criteria varied in 3.62-3.91, 2.85-3.92 and 0.53-5.19 ranges, respectively. According to the obtained ANFIS models, it can be concluded that all models have a good potential to determine the cetane number of biodiesel fuel. Consequently, the results showed that the ANFIS models developed by grid partition FIS (pimf) and fuzzy c-means (FCM) technique have a higher final desirability of 0.857 and 0.718, respectively.