Investigation of the effect of zeolite on emission and performance in a diesel engine by artificial neural networks (ANN)

The negative effect of toxic substances emitted from the internal combustion engine exhaust to human health and natural environment is an issue that is desired to be eliminated. In this study, it is aimed to design a zeolite filter working with the Pressure Swing Adsorption principle, to purify the air to oxygen and to improve the exhaust emission and performance parameters of the engine by feeding this pure oxygen to the suction line of a diesel engine. The zeolite was modified with 2N HCl and 4N NaOH in order to expand the surface area and improve ion exchange, adsorption and molecular sieve properties. Natural zeolite, synthetic zeolite, base modified zeolite and acid modified zeolite are used in the filter system. The oxygen enriched by ZF was fed into the combustion chamber through the oxygen inlet record mounted on the intake manifold of the engine. In the engine experiments, CO, CO2, HC, NOx, smoke density, vibration, noise, EGS, ÖYT, BSFC values were recorded in the standard diesel engine and under 4 different zeolite types. As a result of the experimental measurements, the data obtained were entered into ANN as training and test data and modeled in MATLAB R2016a program. The reliability of the program was checked by comparing the predicted values and experimental results of ANN. For this purpose, graphs, regression graphs and mean square error data are taken into consideration for the estimated values and experimental values for each exhaust emission and performance parameter. As a result of the measurements, it was determined that ZF absorbs the nitrogen in the air with the principle of PSA and provides oxygen enrichment as large to small, SZ, BMZ, DZ, AMZ. As a result of the process, CO, HC, smoke density and ÖYT values were decreased. On the other hand, CO2, NOx, EGS, vibration, noise, values have increased. The performance of ANN, which was formed as a result of experimental measurements, was examined and it was seen that the ANN's predicted results were in acceptable error rates. Keywords: Zeolite, Pressure Swing Adsorption, Diesel engine, Artificial Neural Networks