A hybrid ANN-Fuzzy approach for optimization of engine operating parameters of a CI engine fueled with diesel-palm biodiesel-ethanol blend

Abstract

This paper investigates use of artificial neural network (ANN) model in prediction of brake specific energy consumption (BSEC), nitrogen oxides (NOx), unburnt hydrocarbon (UHC), and carbon dioxide (CO2) emissions of a single cylinder diesel engine operates with diesel-palm biodiesel-ethanol blends. The engine is run at different load form 20–100% and 1500 rpm constant speed. The fuel used in this present study are diesel and six different diesel-palm biodiesel-ethanol blends. The Levenberg-Marquardt back propagation training algorithm with logistic-sigmoid activation function results best prediction of performance and emission characteristics with accurate overall correlation coefficient (R) (0.99329–0.99875) and minimum mean square error (MSE) (0.000179082–0.000465809). The mean absolute percentage errors (MAPE) are observed to be in range of 2.32–4.54% with the acceptable margin of mean square relative error (MSRE). Furthermore, experimental and ANN predicted data are compared in fuzzy interface system (FIS) to find optimum engine operating parameters. Compared to other blends, at 20% load, D85BD10E5 blend exhibits the highest MPCI (multi performance characteristics index) values of 0.718 and 0.705 for experimental and ANN predicted data respectively. Robustness and reliability of the proposed techniques clearly explain the application of ANN and fuzzy logic system in the prediction and optimization of engine parameters.