Methods of sensitivity analysis in model-based calibration

Abstract

The effort and time demand for the calibration of electronic control systems for internal combustion engines on test benches is rising constantly for a number of years. This is mainly driven by new engines and powertrain technologies as well as by the rising quantity of series and vehicle variations. In the engine calibration process with the objective for optimization of fuel consumption and emission values, the number of parameters is large and the evaluation on a test bench is expensive. Since a certain target quantity is usually dependent on a range of various parameters, the sensitivity of system inputs on outputs should be identified. The goal of this approach is a reduction of the dimension in the design of experiments to the most important factors. In this study, the approaches by linear models, nonlinear models and mutual information are introduced and are compared with measurement data.