A methodology for engine performance optimization

Design and optimization of intake and exhaust systems and valve timing is crucial in development of a naturally aspirated engine. Nowadays numerical simulation is a fundamental tool for this area. Unfortunately to perform an optimization of engine performance by setting even only a few parameters needs great effort in terms of time and engineering resources even with simple architecture engines.To overcome this problem the authors have developed an optimization methodology: the use of a 1_D simulation code allows one to build a neural network (NN) that characterizes engine working conditions for several input data variations (such as intake/exhaust systems and valve timing). A genetic algorithm (GA) coupled with the neural network is used to carry-out the multi-parameter optimization of engine performance. As an initial application, this methodology has been used for a 1-cylinder four stroke engine for off-road motorcycle application: inlet and exhaust valve phase angles were the input parameters and the maximum shaft power was the fitness function of the optimization process.In this paper the optimization methodology is described and the results of the above-mentioned initial application presented.