The rise in the number of automobiles on the road increases the air pollution caused by automotive exhaust. Therefore it becomes important to continually analyze and find new and improved methods to reduce engine emissions. This paper augments the studies carried on engine emissions by establishing a method of prediction of concentration various species in the CI engine exhaust based upon the instantaneous pressure rise in the combustion chamber.
The rate of rise of cylinder pressure depends upon the combustion process, which in turn is controlled by various parameters such as injection timing, compression ratio, inlet air properties and most importantly the quality of the fuel used. This rate of pressure rise is assumed to control the rate at which the various species are formed as it depends upon the combustion process itself. In this experiment the fuel alone is changed maintaining all other parameters constant. The combustion chamber pressure at every crank angle is measured using a pressure transducer and the emissions of CO, CO2, NOx, HC and soot are measured using an Exhaust Gas Analyzer. This paper correlates the matrix containing the pressure for every crank angle in the entire cycle with the measured emissions. Artificial Neural Networks are employed for providing the necessary correlation as they are well known to perform well when correlations are non-linear and extensive train data is available.
A feed forward back propagation neural network is used. The network is trained and then tested with a series of experimental data produced from by testing a CI engine run on petroleum diesel as well as on renewable fuels such as vegetable oils and esters, which are most likely to replace petroleum derived diesel as a CI engine fuel in future. The network is optimized by varying the number of hidden layers, number of hidden layer neurons, activation functions and training algorithms.