Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Mechanical and Aerospace Engineering

Committee Chair

Gregory J. Thompson.


When developing regulations or preparing emissions inventory models for on-road mobile sources, it is of critical importance to obtain representative values of fuel consumption and emissions rates that accurately represent the vehicle behavior for varying vehicle operation. Unfortunately, fuel consumption and emissions are often only known for a few test cycles which generally represent limited types of vehicle activity. It is known that distance-specific emissions and vehicle fuel consumption strongly depend on the vehicle activity and this dependency causes difficulties when trying to estimate fuel consumption and emissions over different vehicle activity than the original test cycle(s). The central hypothesis of this research is that linear relationships exist between metrics that quantify vehicle activity and the resulting emissions and fuel consumption. These metrics are calculated on the basis of vehicle speed-time traces and are termed driving cycle properties. A methodology of linear interpolation in the properties' space is used to calculate appropriate linear combination weights for varying vehicle activities in order to predict fuel consumption and emissions over "unseen" driving cycles or in-use routes. The proposed modeling approach and method evaluation provides accurate oxides of nitrogen and fuel consumption inventory data estimations for diverse vehicle operation using either chassis dynamometer or in-use data collected with portable emission measurement systems. The predictive accuracy of the in-use model was improved when using grade-related metrics. Fuel consumption average prediction errors of 5.7% and 3.6% were obtained for chassis dynamometer data and in-use data, respectively. Oxides of nitrogen average prediction errors of 10.2% and 7.2% were obtained for chassis dynamometer data and in-use data, respectively. The prediction errors were slightly higher than the run-to-run variations found in chassis or in-use testing.