An innovative methodology to improve the design and the performance of direct injection diesel engines

Abstract

The present investigation aims at defining a new methodology to analyse the effect of combustion chamber shape and injection strategy on diesel engine emissions and to find an optimized configuration to reduce NO_x, soot and HC without significantly decreasing engine performance. This was achieved by using a multistep optimization process based on genetic algorithms. The first step was related to the optimization of combustion chamber shape for a fixed injection strategy. The optimization was firstly performed with respect to a single-pulse injection strategy and a six-hole injector design. Then the process was repeated for an innovative injection strategy characterized by the injection of a relevant quantity of fuel far from top dead centre (large early injection) and a seven-hole injector. The combustion chamber shape was defined according to five geometric parameters, whose ranges of variation were very wide. For all the investigated configurations, the bowl volume and squish-bowl volume ratio were kept constant so that the compression ratio was the same for all the investigated chambers. The spray injection angle was also considered as a variable parameter. The optimization was simultaneously performed for different engine operating conditions, i.e. load and speed values, weighted according to their occurrence in the European Driving Test. From the results of this step, an optimized chamber configuration was selected and, at the same time, the effect of each geometric characteristic on NO_x and soot emissions was underlined. In the second step, the combustion chamber shape was kept constant and the optimization process was aimed at identifying an optimized injection strategy for a fixed engine operating mode. Injection pressure, pilot and main advances and pilot energizing time were considered as input for the second step of the optimization process.

The numerical tools used in the investigation include a modified version of the KIVA-3V code and a multiobjective genetic algorithm. The code capability to simulate engine performance was assessed by comparing numerical results with experimental data available for the baseline engine configuration. Three fitness functions were defined according to engine emission levels (soot, NO_x and HC) and a penalty function was used to account for engine performance and fuel consumption.