Chapter 16: Computational Approaches to the Biodiesel Production Process and Optimization: Neem Oil Methyl Esters as an Example

Data-based machine learning techniques or computational intelligence techniques have been gaining research interest recently because of the complexity, ambivalence, and non-linear nature of biodiesel production systems. In this chapter, a minireview of the various modeling and optimization techniques regarding biodiesel processes is presented. To demonstrate the application of some of these tools, two learning machine methods, viz. adaptive neuro-fuzzy inference system (ANFIS) and the artificial neural network model (ANN), were utilized in modeling the production process of neem oil methyl esters (NOME) via transesterification of esterified neem oil (ENO). The results showed that the ANFIS model [correlation coefficient (R) = 0.9993 and mean relative percentage deviation (MRPD) = 0.2624] was marginally superior to ANN (R = 0.9989, MRPD = 0.3038). ANN combined with a genetic algorithm (ANN–GA) and ANFIS–GA were used to optimize the biodiesel system to obtain the most favorable operating conditions. ANFIS–GA gave a maximum NOME yield (99.45 wt%) using a methanol/ENO molar ratio of 9 : 1, solid catalyst loading 0.98 wt%, and reaction time 75 min compared with ANN–GA with a maximum NOME yield (98.85 wt%) using a methanol/ENO molar ratio of 13 : 1, solid catalyst loading 0.56 wt% and reaction time 60 min. The results showed that both machine learning tools could accurately predict the NOME yield and represent the complex system investigated.