Flash point prediction of tailor-made green diesel blends containing B5 palm oil biodiesel and alcohol

Abstract

The flash point prediction accuracy of the Liaw model through UNIFAC-type models was evaluated and improved for B5 palm oil biodiesel–alcohol blends. The UNIFAC group interaction parameters of alcohol–alkyl chains (OH–CH2), alcohol–double bonded alkyl chains (OH–CC) and alcohol–esters group (OH–CCOO) were revised comprehensively to attend this improvement. The prediction accuracies (calculated as average absolute relative deviation (AARD)) were improved by solely revising the interaction parameters of OH–CH2 with similar AARD for all UNIFAC type models (from 7.0% and 5.8% to 1.1% for Original and NIST UNIFAC; from 6.6% and 6.2% to 1.3% for Modified UNIFAC (Dortmund) and NIST-Modified UNIFAC). The revised parameters were further validated using the test set data of B5-alcohol blends and B5-ethyl levulinate (EL)–butanol (BU) blends. Satisfactory improvements were obtained; but for B5–EL–BU blends, only the Original UNIFAC and NIST-UNIFAC model showed improvement. The presented study showed that by solely revising the interaction parameters of OH–CH3, a better flash point prediction is gained for the green diesel blends containing palm oil biodiesel and alcohol.