Vehicle emissions from internal combustion engines are the result of the simultaneous influence of the engine operating point, the combustion process, and the exhaust aftertreatment systems. Consequently, predicting these emissions from a limited set of operating parameters is challenging, particularly under time-varying driving cycles. This master's thesis presents a data-driven approach for predicting pollutant emissions of a compressed natural gas bus based on real-world driving emissions measurements. Utilizing the measured concentrations of CO$_{2}$, CO, NO$_{x}$, and THC, a predictive model was developed that assigns characteristic emission concentrations to each operating point – defined by engine speed and torque – in the form of discrete tables. By incorporating a simplified thermal model of the catalytic converter, the model was extended to include separate tables for cold and warm operating states, and was subsequently evaluated on a separate validation section of the cycle. The best agreement was achieved for CO$_{2}$, which is directly linked to fuel consumption and engine load. For CO, NO$_{x}$, and THC, the instantaneous values are less accurate, and the additional consideration of the monolith temperature did not result in a statistically significant improvement in predicted accuracy. Nevertheless, the model provides a good estimation of total emissions over the driving cycle, making it suitable for a rapid assessment of cumulative emissions and driving-cycle analysis, rather than for predicting instantaneous peaks.
|
