Computational modeling and multi-objective optimization of engine performance of biodiesel made with castor oil

In this research the engine performance of biodiesel made with castor oil through homogeneous alkali catalyzed transesterification was analyzed. The input variables for the performance analysis were biodiesel blend and engine speed while the response variables were break power (BP), basic specific fuel consumption (BSFC), break thermal efficiency (BTE), torque and unit cost. The engine performance was modeled using artificial neural network (ANN) and the ANN was subsequently used as the objective function for a non dominated sorting genetic algorithm (NSGA-II) for multi objective optimization of the engine performance. The ANN was equally coupled with a desirability function whose outputs were optimized using simulated annealing for multi objective optimization of the engine performance. Subsequent comparison of the two optimization models was done. The results show that biodiesel from castor oil could be a good replacement for biodiesels from fossil fuels. The ANN model predicted engine performance very well with the lowest value of the correlation coefficient between the experimental responses and ANN predictions being 0.9733. The multi objective optimization using desirability function performed excellently well with the optimum blend and speed being 78.7% and 1754.48 rpm respectively. The Pareto front from the NSGA-II algorithm generally has high desirability values. The Pareto front solution which is more flexible than the desirability function solution would serve as an excellent guide for engine designers. Finally, castor oil based biodiesel cost was for the first time integrated into engine performance optimization studies.