Influence analysis of driving pattern factors on bus fuel economy by the SEM

This study is aimed at proposing a method to find principal driving pattern factors to describe the dimensions of city bus driving cycle, sorted by the effect of factors on fuel economy. 27 driving pattern factors were calculated for 22353 driving patterns collected in actual traffic. These factors included traditional driving pattern factors, bus route geographic factors, driving habits factors and environmental factors. The structure equation model (SEM) was proposed to analyze the direct and indirect relations between 27 factors and fuel economy, and the stepwise iteration was applied to reduce meaningless relations. 13 driving patter factors from 27 factors were finally abstracted to describe the dimensions of city bus driving cycle, illustrating the influence coefficients of fuel economy. Compared with the traditional multiple linear regression analysis, the SEM not only reflects the direct and indirect influence between variables, but also gives the minimum value of the whole estimated residual. The proposed structure model can effectively illustrate the complex influence between 13 variables of driving pattern factors and fuel economy. The results of traditional multiple linear regression were only in consistent with the direct influence of the SEM. The proposed analysis method will support the simplification of principal variables measuring driving cycle and bus fuel economy.