Parameter estimation
in one dimensional models
for non-stationary flow
in fuel injections systems

Brief English version (recommended - pdf, 13 pages, 180 KByte)

Full German version (pdf, 77 pages, 800 KByte)

Abstract

By means of experiments and parameter estimation based on one dimensional models a deeper understanding of flow phenomena in fuel injection equipment for diesel engines (also operated on dimethyl ether) is obtained and the regarding physical models are improved.

One dimensional models are used as they reflect the system's behaviour very well. Phenomena with a certain cross-sectional profile are represented by special parameters. Further attention is drawn to the demand that all parameters should have a physical meaning. So the results can be used for layout and analysis of equipment and also direct conclusions about physical phenomena (e.g. cavitation) can be drawn.

Parameter estimation methods are used to exhibit physical information which by experimental techniques could hardly be assessed. The optimization procedure is carried out by a Levenberg-Marquardt algorithm and the partial differential equations are solved by a new, efficient algorithm approximating the analytical solution.

Some of the results are: injected fuel quantity per cycle can be calculated from the mere injection line pressures, the distance between the last pressure measurement along the injection line and the nozzle can be determined, losses in orifices can be identified as well as flow phenomena and information regarding geometry. Even information about flow velocities within the spray holes is attained and suggestions are made how to improve the regarding results.

There is also a little longer description in German available.