ABSTRACT
Nowadays, due to the high demand for petroleum product, people are looking for another alternative fuel, in which biodiesel is on top. Researchers perform the experiment with various combinations of fuels. To have maximum information in less number of experiments, various predication models are present. In this chapter, response surface methodology (RSM), artificial neural network (ANN), factorial design of experiment, and fuzzy inference system (FIS) method have been compared for modeling and prediction of engine parameter and emission characteristics. For initial experimental data, the experiment is performed on single-cylinder, four-stroke compression ignition engine with biodiesel–ethanol–diesel as fuel. For modeling, four factors are considered, that is, biodiesel (0–20%), ethanol (0–20%), speed (1500 rpm), and load (0–5 kg). RSM and factorial method is performed 416by using Design Expert and Minitab statistical software, respectively, and MATLAB is used in developing ANN and FLN model. In ANN, three factors are taken as input and output response and it is generated by using transfer function. In current work, feed forward with back propagation neural network model is used. In FIS, multi-input multi-output fuzzy models are developed, trained, and validated with experimental data. The developed RSM and FIS describe the process with high accuracy. Optimal condition obtained by RSM is better than factorial design. However, ANN shows advantage over RSM on basis of absolute deviation and coefficient of determination (R2).
