Analysis and evaluation of direct injection diesel engines to optimize performance and emissions

In this paper, optimization of fuel consumption and NOx and soot emissions in a direct diesel engine is done using neural network and ant algorithm by applying the parameters of inlet air temperature, rate of fuel injection mass and engine speed. Complexity of the behavior of internal combustion engines was first determined by using experimental experiments to establish the relationship between the input and output parameters by the neural network. The artificial neural network with the Levenberg-Marguerite training algorithm is used to model and train the existing relationship between the above parameters and is used as a subroutine for predicting optimal values in the ant colony algorithm. Results show the engine optimized parameters are drawn to lower temperatures due to lower NOx and soot emissions by lowering the inlet air temperature. Also the results of modeling and prediction performed by neural network show 98% and 94% concordance with the experimental data in emissions and fuel consumption, respectively. On the other hand, improving the quality of NOx values, because of its high weight in the objective function, affects the overall optimization result and the behavior of the objective function in convergence is very similar to NOx behavior. Also, combination of neural network-Ant algorithm approach due to its fast convergence and consequently short response time can be used as an effective method of diesel engine intelligent control systems to reduce emissions and fuel consumption.