The main objective of this research is to study factors that effect on the vehicles emissions on Egyptian roads. Vehicle emission models were investigated using the application of (SPSS) computer program Version (26). The models were calibrated using vehicles emission records collected during the study for the period (November 2017). Data recorded for eight vehicles, emission data were classified according the fuel type to three categories (Diesel, Natural Gas and Petrol Vehicles), to conduct a comparative analysis of various statistical modeling techniques such as "Linear Regression with Link Function of Identity, Linear Regression with Link Function of Log, Gamma Regression with Link Function of Log and Tweedie Regression with Link Function of Log" which classified to generalized linear regression models to predict vehicle emission rates as a function of the independent variables.
The study based on collecting data of the travel-related factors, highway characteristics and vehicle characteristics in addition to the effect of climate for the three different vehicles categories, also vehicles emission measurements (CO2 [g/s], CO [mg/s], HC [mg/s], and NOX [mg/s]) used in this study were obtained from Egyptian Environmental Affairs Agency (EEAA) recorded for the period (November 2017), Six independent variables were selected in this research (vehicle speed, profile grade, ambient temperature, ambient pressure, ambient relative humidity and numbers of rotation per minute for vehicle engine) which affect directly on vehicle emissions from transportation on the different vehicles categories then a comparison of these results obtained from the (SPSS) mathematical model.
Finally, it was found that linear regression model with Link Function of Identity (LRMLFI) was the best generalized regression model to represent the correlation between Co2 emissions for Diesel vehicles, while linear regression model with link function of log (LRMLFL) was the best generalized regression model for CO, HC and NOX emission for Diesel vehicles.
Linear Regression Model with Link Function of Identity (LRMLFI) was the best generalized regression model to represent the correlation between Co2 and NOX emissions for Natural Gas vehicles, Co and HC emissions for Natural Gas vehicles provide the best models using linear regression model with Link Function of Log (LRMLFL).
Petrol vehicles emission measurements (CO2 [g/s] and HC [mg/s]) were well presented with Linear Regression Model with Link Function of Identity (LRMLFI), while linear regression model with link function of log (LRMLFL) was the best generalized regression model to represent the correlation between Petrol vehicles emission measurements (CO [mg/s] and NOX [mg/s]).
|
