A grid environment for diesel engine chamber optimization

The goal of this paper is to show that computer modelling techniques can be used to solve the real life problem of Diesel Engine performance improvement. The purpose of our work is to achieve the lowest emissions level and improved fuel efficiency with respect to European Norm for emissions. Particularly, we are interested to reduce NO + HC and soot emissions and to maximize the PMI, a pressure proportional to engine power. These parameters depend on combustion chamber geometry too, so we propose its optimization turning to the new micro Genetic Algorithms (micro-GA) technique. The idea consists in the automated random generation of a lot of meshes. each of these representing a different chamber geometry, with respect to some common geometric constraints: then in the use of the micro-GA to optimize, at each iteration, the results obtained in the previous steps. The innovative feature of our work is the multiobjective nature of the optimization process. This is the main reason to chose micro-GA rather than simple Genetic Agorithms. Emissions level and fuel efficiency can be evaluated using a modified version of KIVA3 code that outputs three values, each of these related to one of the three specific fitness functions to be maximized. The optimization process involves the execution of a lot of KIVA3 simulations to calculate fitness values of the chamber geometries taken in consideration during all of the optimization steps. We propose the use of Grid Computing technologies to increase the performance of the KIVA-micro-GA, showing how a distributed environment allows to reduce the computational time needed by the optimization process, taking advantage of intrinsic parallelism of micro-GA. In fact, their structure allows executing simultaneously KIVA3 simulations over the random meshes and over the geometries that populate the micro-population at each iteration. The services offered by the system are the micro-GA parameters definition. the submission of the optimization process and the monitoring of the process status. A trusted user can access the implemented services using a grid portal, called DESGrid (Grid for Diesel Engine Simulation). The analysis of the results, achieved by execution of KIMA-micro-GA on three ES40 Compaq nodes, each one equipped by four processors, shows a good reduction in both emissions and fuel consumption. In the paper we show numerical values and related geometries representation obtained after the first steps of the global optimization process execution.