Intelligent Design of Experiment for an Efficient Gauging of Combustion Engines
Combustion engines are traditional gauged with lattice like arranged input data. I.e. all relevant input data like engine speed, injection mass, etc. will be adjusted in all combinations (e.g. 8 engine speeds x 8 injection masses x 8 ...) and the corresponding outlet values like engine torque, consumption, exhaust components are gauged. To be in accordance with modern exhaust emission standards (Euro 4 and 5) and for further consumption reduction engines are designed with more degrees of freedom. These additional degrees of freedom such as exhaust gas recirculation, variable adjustable turbocharger, multiple injections, etc. make a lattice like measurement in the future impossible, as the number of measuring points increases exponentially according to the number of inlets.
Subject of this study is the development of new strategies to prevent a lattice like measurement. The basic idea is to gauge (gathering information) solely where this is relevant. To make this estimation a model idea is necessary. Among other things local linear and polynomial model approaches shall be evaluated, which can also be interpreted as neural networks and fuzzy-systems.
This study ought to be accomplished in cooperation with a manufacturer or electronic supplier of the automotive industry.
Figure. An intelligent strategy of measurement shall detect automatically, that in the upper right corner denser measuring is necessary, because there is stronger nonlinear behaviour. Therewith many measurements for comparable quality of the model can be conserved.