Automatic Model Calibration for Engine-Process Simulation with Heat-Release Prediction 2006-01-0655

The development of powertrains with DI diesel Engines as primary drive demands accurate analytic specification of the cylinder process, with sufficient consideration taken of gas-exchange, turbo charging, and - especially - combustion. Until now, engine-process simulation has proved to be a powerful tool for satisfaction of these requirements. In addition to simple modeling of the heat-release rate with Vibe functions, complex heat-release models based on injection-rate profiles have been developed and implemented in existing simulation tools such as GT Power. To calibrate the parameters to achieve satisfactory agreement between experimental data and three-dimensional Computational Reactive Fluid Dynamics Models (CRFD) simulations, an appreciable amount of expertise and time is required.

This paper deals with a universal method that automatically identifies and calibrates the relevant model parameters to the experimental data. This technique simplifies and accelerates the application of engine-process simulation with more accurate heat-release models, which, however, allow heat-release prediction only with certain constraints. This approach allows intensification of the interface between engine-process simulation and combustion-rate development at the test bench or in CRFD simulation.

A comprehensive cylinder process analysis represents the core of this method. It can additionally analyze transient operating points, and offers an algorithm that optimizes the model parameters. Comprehensive cylinder-process analysis provides the following from the plotted cylinder-pressure curve: the reference-input variable, the heat-release rate, as well as required model specific-input variables such as gas composition, cylinder temperature, and thermodynamic gas properties. The optimizer must be able, in stable operation, to optimize a system with many degrees of freedom.