Prediction efficiency of artificial neural network for CRDI engine output parameters

Abstract

Abstract Considering the environmental and energy concerns, the need for finding an alternative fuel is increasing greatly. But also, the toil associated with finding the suitable fuel for the engine under study is very high. As a measure to reduce this work, theoretical analysis can be used in the search of suitable fuel blends. The present study deals with the creation of an Artificial Neural network (ANN), training, testing the network and finally comparing the predicted values with the experimental values. The test engine was operated with lemon peel oil, with different ratios of Di tertiary butyl peroxide (DTBP), rice husk nanoparticles and water injection. Six engine output parameters (BTE, BSFC, HC, NOx, CO & Smoke) were predicted for different input values of engine load, DTBP proportions, RH nanoparticle concentration and water injection percentage. A single cylinder constant speed direct injection diesel engine is used for obtaining all the experimental values. From the obtained results a data set of 70 experimental values were used to train and test the neural network formed. The predicted values were correlated with the experimental values and the R2 was found for each correlation. The R2 values, thus obtained is in the range of 0.98209 to 0.99744 (which is very close to unity) thus proving that the network created were able to predict the performance and emissions of engine accurately. This work shows the prediction of multi component fuel mixture using ANN, which is new and also helps to reduce the experimental works needed to find the engine output parameters.