Effect of fuel on gasoline particulates emissions

Report Nº 2/24: Concawe conducted a research programme in cooperation with VTT examining the relationship between gasoline physical-chemical properties and particulate number (PN) emissions. The complete programme was executed in two distinct phases, during which altogether 4 vehicles using 23 fuels were tested in VTT’s emission laboratories equipped with a chassis dynamometer. All the combinations of vehicles and fuels tested showed tailpipe PN emissions compliant with the latest Euro 6d standards. For each phase, mathematical models were developed to examine the link between fuel properties and experimentally measured PN emissions.

During the first phase, 13 formulated fuels (surrogates) were tested on a single vehicle equipped with a gasoline direct injection (GDI) engine and a gasoline particulate filter (GPF). The fuel matrix was designed to intentionally and independently vary different fuel properties suspected to impact PN emissions: volume evaporated at 150°C (E150) as a proxy of the heavy fraction of gasoline, total aromatics content, heavy aromatics content (more than 9 carbons) and ethanol content. The vehicle was tested using an “ambient start” (23°C) WLTC (Worldwide harmonized Light vehicle Test Cycle), a “hot start” WLTC and a test cycle simulating RDE (Real Driving Emissions) conditions. During the laboratory tests, both gaseous and particulate engine-out (EO) and tailpipe (TP) emissions were sampled. The particulate sampling included continuous PN10 and PN23 (PN having a diameter respectively bigger than 10 nm and 23 nm). In this first phase, it was concluded that it was possible to establish a fairly good and simple model between TP PN emissions and the fuel properties targeted in the fuel matrix, and more particularly E150 and total aromatics content. The experimental data was also used to check the correlation to other PN models referenced in the literature: “Honda PM Index”, “Yield Sooting Index” (YSI), simplified PM index (based on E130 and E170) or a simple correlation with E150. It was found that none of these models correlate with the experimental data collected, showing the incapability of these literature models to actually predict PN emissions from the test vehicle on which they were not calibrated.

The second phase of the study was conducted on three vehicles. Two of the studied vehicles were equipped with GDI technology (vehicles A and B), while the third one (vehicle C) was equipped with a port fuel injection (PFI) engine. All of them were equipped with GPFs. Eight market fuels, sampled from European refineries, were tested on each of the vehicles. The fuel matrix was designed to vary different fuel properties such as E150, total aromatics and olefins content or ethanol content by targeting specific samples in the refineries, but without any specific intervention in the fuel design. Additionally, two fuels had to be specifically formulated to complete the fuel matrix, reaching a total of ten fuels.  This second phase followed a similar structure as the first one: It consisted of an experimental part for the purpose of vehicles testing with an experimental setup similar to the first phase (using a different RDE cycle and with a cold start at 12°C to be representative of average real-world conditions in Europe) and a modelling part, focused mainly on relationships between fuel properties and PN emissions with a specific part on vehicles cross-comparisons regarding their fuel response.

In the second phase, a relatively great variation in how the different vehicles responded to the fuel properties was observed. E.g. the magnitude of PN10 EO emissions for the two GDI vehicles were found relatively similar, and the PFI equipped vehicle, vehicle C, produced lower PN10 EO emissions. However, the magnitude of PN23 EO emissions was found similar between vehicles A and C and the PN23 EO emissions were significantly higher for the second GDI vehicle (vehicle B). Furthermore, the filtration efficiency of vehicle A’s GPF was found fairly high, resulting in extremely low PN TP emissions overall.

In the second phase vehicle emissions varied significantly case by case, both in terms of gaseous and particulate emissions and PN TP emissions were strongly affected by GPF properties. These experimental results demonstrated that vehicle technology is much more impactful on PN emissions than fuel composition. Highest fuel effect and correlation between fuel characteristics and PN emissions was found for vehicle B but no absolute, general or common variables were found connecting all three different vehicles. These observations have important consequences: they imply that it is not possible for the fuel producer to design a fuel that simultaneously reduces the PN emissions in all vehicles.