The complexity of Urban driving conditions and the human behavior introduces undesired variabilities while establishing Fuel economy for a vehicle. These variabilities pose a great challenge while trying to determine that single figure for assessment of vehicle’s fuel efficiency on an urban driving cycle. This becomes even more challenging when two or more vehicles are simultaneously evaluated with respect to a reference vehicle. The attempt to fit a generalized linear model, between Fuel Economy as predicted variable and components of a driving cycle as predictor variables produced oxymoronic and counter-institutive results. This is primarily due to existence of multi-collinearity among the predictor variables. The context of the study is to consider the event of driving on a cycle as a random sampling experiment. The outcome of a driving cycle is summarized into a list of predictor variables or components. The aim of this study is to reduce the variables which are strongly co-related using various statistical techniques, the primary and the most effective technique being Principal Component Analysis. The selected variables or principal components are then used to predict F.E using a machine learning algorithm.
Determination of Principal Variables for Prediction of Fuel Economy using Principal Component Analysis
Sae Technical Papers
Symposium on International Automotive Technology 2019 ; 2019
2019-01-09
Conference paper
English
British Library Conference Proceedings | 2019
|Backwards Principal Component Analysis and Principal Nested Relations
British Library Online Contents | 2014
|Euler Principal Component Analysis
British Library Online Contents | 2013
|Generalized Principal Component Analysis (GPCA)
British Library Conference Proceedings | 2003
|Generalized principal component analysis (GPCA)
IEEE | 2003
|