Author : Shriniwas D. Chivate and Y. P. Reddy

The invention of common rail fuel injection system in diesel engine has become a prominent solution to many emission problems owing to provision of liberty in multiple injections. The complex combustion mechanism in diesel engines produces combustion noise which adds up to mechanical noise to give total radiated noise. The presented work involves the assessment of effect of injection strategy modification on combustion noise. The combustion noise of baseline injection strategy is evaluated using a software tool. The injection parameters such as pilot injection fuel quantity and separation, main injection timing, rail pressure are modified to arrive at minimum noise producing combination at various speed and load conditions. The combustion noise is evaluated again with modified injection strategy. The combustion noise is found to be reduced by maximum of 4 dB. The total noise is also reduced by 1 dB at lower speed and loads. Structural attenuation is found to be same as expected. This approach will provide an alternative to time consuming and costly approach of structural modification to reduce the radiated noise. As a novel activity, a neural network and Graphical User Interface is developed for prediction of radiated noise and sound quality parameters when cylinder pressure is provided as an input. The trained neural network is found to give good co-relation between target values and predicted values. This network can be effectively used for noise and sound quality prediction using only cylinder pressure and speed as an input.

Keywords: Combustion noise, common rail diesel engine, function fitting, feed forward neural network, injection strategy, structural attenuation.