Cost effective simulation-based multiobjective optimization in the performance of an internal combustion engine

Abstract

Solving real-life engineering problems can be difficult because they often have multiple conflicting objectives, the objective functions involved are highly nonlinear and they contain multiple local minima. Furthermore, function values are often produced via a time-consuming simulation process. These facts suggest the need for an automated optimization tool that is efficient (in terms of number of objective function evaluations) and capable of solving global and multiobjective optimization problems. In this article, the requirements on a general simulation-based optimization system are discussed and such a system is applied to optimize the performance of a two-stroke combustion engine. In the example of a simulation-based optimization problem, the dimensions and shape of the exhaust pipe of a two-stroke engine are altered, and values of three conflicting objective functions are optimized. These values are derived from power output characteristics of the engine. The optimization approach involves interactive multiobjective optimization and provides a convenient tool to balance between conflicting objectives and to find good solutions.

Keywords:

• multicriterion optimization
• global optimization
• interactive methods
• classification
• NIMBUS

Acknowledgements

For technical issues related to the integration of IND-NIMBUS^® software to the optimization system of this study we are indebted to Mr Vesa Ojalehto. This study was financially supported by the Academy of Finland (grant number 104641) and the Jenny and Antti Wihuri Foundation.