Automated Limit Space Characterisation & Mapping 2004-01-1616

More stringent legislative guidelines on emissions and fuel economy have led to an increased number of engine parameters to be optimised. Steady state engine mapping to produce an empirical engine model remains a fundamental step in this optimisation process. Recently statistical techniques such as design of experiments have been introduced to improve the efficiency of this modeling phase. Before undertaking an experimental design it is first necessary to determine the permissible envelope of the various design parameters. The importance of this limit space is two fold, firstly to ensure that the engine is not operated in regions, which may cause damage to it, and secondly so that subsequent experimentation yields test data wholly valid for the subsequent engine model.

Currently this limit space is defined in a largely manual process, requiring expert input many times in the test and characterization process. This is a constraint on the activity since progress is inhibited by the availability of the engineer. This paper introduces a fully automated parameter limit search procedure prototyped in MATLAB and run on a turbocharged gasoline engine test-bed.

The aim of the software is to automatically move through multiple engine operating points and to carry out a limit search at each. The software supervises the test-bed and engine, dealing intelligently with limit violations to ensure the search is robust. Two different search methods are assessed as to their respective performances. In addition a prior knowledge mechanism has been developed to feed-forward limit space information from one operating point to the next, in order to enhance search efficiency.