Factors Affecting Car Fuel Consumption

Abstract

This paper describes a programme of statistical analyses on test data conducted to investigate the effects of a number of factors on car fuel consumption. The data used for the analyses were obtained from a comparative study of the fuel economy of petrol and diesel cars, conducted at the Department of Transport Technology, Loughborough University, on behalf of the Transport and Road Research Laboratory (TRRL). Three Vauxhall Cavalier cars (1300 cc, 1600 cc petrol and 1600 cc diesel models) and six drivers, categorized by sex and age, participated in the tests. Each test was completed over an extensive test route on public roads encompassing all types of driving environments. Factor analysis was used to determine the underlying relationships between a number of measured variables and fuel consumption. Under the ‘free-flowing’ environment of the motorway route section, fuel consumption was primarily influenced by ambient pressure, vehicle speed and frequency of gear change. In suburban conditions, only throttle velocity made any significant contribution to fuel consumption. In the tests over urban sections, the significant variables were found to be vehicle speed and deceleration, frequency of gear change and ambient pressure. Of the ambient factors examined, only pressure had any significant independent effect. The associations derived from factor analyses point the way to realistic correction techniques for vehicle test data. Regression models were constructed to fit the measured fuel consumption data to an expression derived to give fuel consumption as a function of engine, vehicle and environmental parameters. For purely urban, suburban and motorway travel and an overall journey based on a 50/40/10 per cent split of urban, suburban and motorway road types, 93.8, 92.4, 88.2 and 95 per cent of fuel consumption variance was explained by the respective regression models. Regression variables based on engine power correction factors for ambient temperature and humidity showed a degree of statistical significance in all models.