A Hierarchical System of Models for the Optimal Design of Control Strategies in Spark Ignition Automotive Engines

Abstract

The structure of a hierarchical system of models for model based optimization and rapid prototyping of control strategies in automotive spark ignition engines is presented. A distributed and hybrid modeling approach has been considered in order to reach an acceptable level of precision with conventional computational resources and limited experimental effort. Different class of models have been used, ranging from black-box (regression and/or neural networks) to physical ones (thermo-chemical); mean value dynamical model are also considered for the simulation of transient engine-vehicle manoeuvres. The advantages of the mixed approach are reviewed and the critical role played by the identification phase is discussed through a comparative analysis of the estimation procedures adopted during model development and validation process.