Case study

Precision analysis for Shell supports differentiated fuel benefits

Shell offer a range of differentiated fuels with proven benefits, but the growing complexity of vehicle systems was making more precise analysis using existing chassis dynamometers increasingly difficult. Shell needed science-based information that customers could use to inform their fuel choice. The team at Shell needed to see significant improvements in test processes that delivered robust, repeatable data at the required resolution. The challenge was brought to the IAAPS team at the University of Bath, who has an established reputation for delivery of high-precision powertrain testing of whole vehicles. A multidisciplinary team of engineers and mathematicians at Bath worked together to analyse data from both “real world” on-road driving conditions and lab-based investigations using IAAPS’ highly specified climatic chassis dynamometer. This detailed analysis increased test consistency by a factor of six, improving accuracy of fuel testing to 0.4 percent from a previous typical best of 2.5 – 3 percent.


The Challenge

Chassis dynamometers offer considerable potential for the analysis of real-world fuel economy and emissions performance, but in today’s world of multiple small incremental improvements, variability between tests can be greater than the impact of the powertrain technology change being evaluated.

Test system related variability was successfully addressed two decades ago, when the best test houses refined their processes to bring consistency up to around 0.5 percent. Since then, the growing complexity of vehicles has driven potential variability beyond 3.0 percent, even for non-hybridised powertrains. While technologies such as start-stop are relatively binary and therefore easy to manage, systems such as Selective Catalytic Reduction (SCR), Diesel Particular Filters (DPF) and smart battery charging employ more complex control strategies that can introduce substantial variability.

To provide science-based information that its customers could use to inform their fuel choice, Shell recognised that significant improvements to the test processes would be required before they could generate robust, repeatable data at the required resolution. The University of Bath had already established a considerable reputation for high-precision powertrain testing of whole-vehicles, so was a natural choice for the project.


The Approach: Isolating areas of variability

The University of Bath’s Centre for Low Emissions Vehicle Research (CLEVeR) is equipped with a highly specified climatic chassis dynamometer facility with instrumentation to research standards. CLEVeR had already developed guidance to limit imprecision resulting from test procedures,[1] so the focus of this new research programme was to identify and understand the increasingly important area of vehicle-borne imprecision.

While the solution sounds simple – identify the factors that introduce the biggest variables and develop techniques for controlling them more precisely – the complexity of modern vehicles introduces many new unknowns. The first step was therefore to identify and characterise these factors through a programme of ‘real world’ on-road driving. The investigation then moved to the university’s chassis dynamometer facility for more precise analysis and verification of the proposed mitigation techniques.

The considerable challenge of analysing more than 200 channels of time-based data was addressed through the development of a novel application of proven Partial Least Squares (PLS) regression.[2] This development by the university’s mathematicians and engineers working closely together allowed the complete dataset to be searched for sources of variability, without the need for data exclusion. To meet Shell’s objectives, fuel consumption was chosen as the dependent variable, with the PLS algorithm employed to quantify the potential importance of every other measured value.

The two most consistently significant areas of variability were found to be the way the engine management system manages the particulate filters as they fill and regenerate, causing variations of up to 2.0 percent and 1.0 percent on the Worldwide harmonised Light vehicle Test Cycle (WLTC) and New Emissions Drive Cycle (NEDC) respectively, and the increasingly sophisticated ‘smart’ management of battery charging, which introduced variability of up to 1.0 percent. Individual vehicle types, however, were found to add further areas of significant variability. One of the test vehicles was particularly sensitive to vehicle warm-up rate, which caused significant variations in catalyst light-off strategy. This one factor alone led to a 2.3 percent variability in fuel consumption.

Tailpipe NOx emissions were also seen to vary significantly between vehicles, depending on the nature of the preceding test. This was attributed to variability in management strategies for the Selective Catalytic Reduction (SCR) NOx aftertreatment systems. While the behaviour of the SCR system did not appear to significantly affect fuel consumption, this finding does highlight the need to design and implement an SCR conditioning cycle if high precision NOx measurement is required.

Most of the identified variables were greater than the effect of typical technology changes assessed on chassis dynamometer facilities. For example, an increase of 8.7 percent in fuel consumption was observed following a 90 minute battery discharge at a level equivalent to leaving the headlights on. Battery voltage was found not to be an appropriate measure of the variation in the alternator loading and simply ensuring battery charging between tests was shown to be an ineffective approach to reducing this variability.


The Outcomes: Customer benefits

The findings of the project have allowed the accuracy of fuel consumption testing using a typical commercial chassis dynamometer to be increased to 0.4 percent, from a previous typical best of 2.5 percent – 3.0 percent. This has enabled Shell to generate robust, repeatable evidence to support the fuel economy benefits of its technically differentiated fuels, fully meeting the objectives of the project.

The Bath work also helped Shell increase consistency between different facilities and engines, supporting the company’s wider research. The University conducted training at Shell offices to disseminate the revised methods and presented the results at one of the company’s Technology Forums. Through the training sessions, Bath was able to share different methods for test data analysis, which helped Shell scientists to gain new insights from their data, and provided new data analysis tools which significantly eased the burden of analysing large datasets.

The mathematical techniques developed for this research are now being applied to other vehicle test systems to identify the causes of variability. Most notably, the growing emphasis on Real Driving Emissions (RDE) is introducing new levels of randomness, particularly when combined with the growing complexity of vehicle system controls.


A view from Shell

“Understanding the interactions between our products and the increasingly sophisticated systems of modern engines is essential if we are to continue to refine our technologies and to validate their benefits. This programme has added considerable value to our work in these areas.”

Michael Gee, fuel scientist, Shell Global Solutions (UK)

Case studies

Sign up to our newsletter

Stay updated

Never miss IAAPS news, insights, events and resources.

Sign up