In the European Union the legislation governing fleet CO2 emissions is already in place with a fleet average limit of 130g/km currently being imposed on all vehicle manufacturers. With the target for this legislation falling to 95g/km by 2020 and hefty fines for noncompliance automotive engineers are working a pace to develop new technologies that lower the CO2 emissions and hence fuel consumption of new to market vehicles. As average new vehicle CO2 emissions continue to decline the task of measuring these emissions with high precision becomes increasingly challenging. With the introduction of real world emissions legislation planned for 2017 there is a development driven need to precisely assess the vehicle CO2 emissions on chassis dynamometers over a wide operating range. Furthermore since all type approval and certification testing is completed on chassis dynamometers, any new technology must be proven against these test techniques. Typical technology improvements nowadays require repeatability limits which were unprecedented 5-10 years ago and the challenge now is how to deliver this level of precision. Detailed studies are conducted into the four key areas that cause significant noise to the CO2 emissions results from chassis dynamometer tests. These are the vehicle electrical system, driver behaviour, procedural factors and the chassis dynamometer itself. In each of these areas, the existing contribution of imprecision is quantified, methods are proposed then demonstrated for improving the precision and the improved case is quantified. It was found that the electrical system can be controlled by charging the vehicle battery, not using auxiliary devices and installing current measurement devices on the vehicle. Simply charging the vehicle battery prior to each test was found to cause a change to the CO2 emissions of 2.2% at 95% confidence. Whilst auxiliary devices were found to cause changes to the CO2 emissions of up to 43% for even a relatively basic vehicle. The driver behaviour can be controlled by firstly removing the tolerances from the driver’s aid which it was found improved the precision of the CO2 emissions by 43.5% and secondly by recording the throttle pedal movements to enable the validation of test results. Procedural factors, such as tyre pressures can be easily controlled by resisting the temptation to over check and by installing pressure sensing equipment. Using a modern chassis dynamometer with low parasitic losses will make the job of controlling the dynamometer easier, but all dynamometers can be controlled by following the industry standard quality assurance procedures and implementing statistical process control tools to check the key results. The implementation of statistical process control alone improved the precision of unloaded dynamometer coastdown checks by reducing the coefficient of variation from 6.6 to 4.0%. Using the dynamometer to accelerate the vehicle before coastdown checks was found to approximately halve the variability in coastdown times. It was also demonstrated that verification of the dynamometer inertia simulation and response time are both critically important, as the industry standard coastdown test is insufficient, in isolation, to validate the loading on a vehicle. Six sigma and statistical process control techniques have shown that for complex multiple input single output systems, such as chassis dynamometer fuel economy tests, it is insufficient to improve only one input to the system to achieve a change to the output. As a result, suggested improvements in each noise factor often have to be validated against an input metric rather than the output CO2 emissions. Despite this, the overall level of precision of the CO2 emissions and fuel consumption seen at the start of the research, measured by the coefficient of variation of approximately 2.6%, has been improved by over six times through the simultaneous implementation of the findings from this research with the demonstration of coefficient of variation as low as 0.4%. Through this research three major contributions have been made to the state of the art. Firstly, from the work on driver behaviour an extension is proposed to the Society of Automotive Engineers J2951 drive quality metric standard to include the a newly developed Cumulative Absolute Speed Error metric and to suggest that metrics are reviewed across the duration of a test to identify differences in driving behaviours during a test that do not cause a change to the end of test result. Secondly, the need to instrument the vehicle and test cell to record variability in the key noise factors has been demonstrated. Thirdly, a universal method has been developed and published from this research, to use response modelling techniques for the validation of test repeatability and the correction of CO2 emissions. The impact of these contributions is that the precision of chassis dynamometer emissions tests can be improved by a factor of 6.5 and this is of critical importance as the new real world driving and world light-duty harmonised emissions legislation comes into force over the next two to five years. This legislation will require an unprecedented level of precision for the effective testing of full vehicle system interactions over a larger operating range but within a controlled laboratory environment. If this level of precision is not met then opportunities to reduce vehicle fuel consumption through technology that only has a small improvement on fuel consumption, which is likely given the large advances that have be achieved over the last few decades, will be missed.

No description available.