This content is not included in your SAE MOBILUS subscription, or you are not logged in.
A Global Sensitivity Analysis Approach for Engine Friction Modeling
ISSN: 1946-3936, e-ISSN: 1946-3944
Published August 21, 2019 by SAE International in United States
Citation: Krecker, O. and Hiltner, C., "A Global Sensitivity Analysis Approach for Engine Friction Modeling," SAE Int. J. Engines 12(5):543-566, 2019, https://doi.org/10.4271/03-12-05-0035.
Mechanical friction simulations offer a valuable tool in the development of internal combustion engines for the evaluation of optimization studies in terms of time efficiency. However, system modeling and evaluation of model performance may be highly complex. A high number of interacting submodels and parameters as well as a limited model transparency contribute to uncertainties in the modeling process. In particular, model calibration and validation are complicated by the unknown effect of parameters on the model output. This article presents an advanced and model-independent methodology for identifying sensitive parameters of engine friction. This allows the user to investigate an unlimited number of parameters of a model whose structure and properties are prior unknown. In contrast to widely used parameter studies, in which only one parameter is varied at a time, the use of the elementary effect method enables the consideration of interactions in the entire parameter space. Based on a sensitivity analysis (SA), the methodology offers a comprehensive and practicable approach to improve model performance and effectiveness of predictive optimization studies. The methodology is validated utilizing an exemplary friction model of a modern inlet valve train, and its further details are described in the form of a practicable guideline. The multifunctional application of the methodology (parameter prioritization, model calibration, and friction optimization) is also outlined. Especially the process of calibration is shown in the example of a complementary subsystem, the exhaust valve train model. Hereby, the advantages of the methodology are presented regarding the improvement of model and measurement correlation.