The impact of real driving on vehicle CO2 emissions

On-road (real) driving of a conventional vehicle having free gear-shift strategy and driving style, non-zero positive altitude gain, and uncontrolled ambient conditions is expected to produce different emissions values, including CO2, in respect to those produced in a controlled laboratory experiment following a standardized velocity profile. This paper aims to investigate the impact of driving style characterized by driving dynamics on CO2 emissions of a passenger car equipped with an internal combustion engine. The study is based on recorded on-road trips of a passenger car that are postprocessed using the EU real driving calculation procedures. Selected segment trips of urban, rural and motorway driving are used as input for a detailed vehicle model developed in LMS Imagine. Lab AMESim in order to assess the impact of real driving on vehicle CO2 emissions.


Introduction
The transport sector and in particular the emissions from passenger car internal combustion engines represent a large fraction of pollutants and CO2 emissions in the European Union (EU).The latest EU regulations are imposing limits for fleets average CO2 emissions and define a Real Driving Emission (RDE) test procedure in order to limit tailpipe emissions under all normal conditions of use.For now, the focus of the RDE regulation is on air pollutant emissions but RDE testing is also measuring the CO2 emissions.
It became obvious that with more available data regarding the real-world performance of new vehicles in terms of CO2 emissions, the vehicles manufacturers must focus on reducing them under all conditions.On-road (real) driving is different from roller bench testing due to varying rolling resistance (different road surfaces and cornering) and aerodynamic drag (wind), free gear-shift strategy and driving style, non-zero positive altitude gain, and uncontrolled ambient conditions.
The driving style is practically defined by gear-shift strategy and driving dynamics.The driving dynamics during the trip can be evaluated using specific global parameters such as: VSF (Vehicle Speed Fluctuation), average positive acceleration, RMS acceleration, RPA (Relative Positive Acceleration), a specific percentile of the product of vehicle speed per positive acceleration etc.In [1] the influence of different standardize cycles (US 06, Japanese 10-15, HWFET, NEDC and WLTC) was considered based on VSF parameter.Average positive acceleration and RMS acceleration are considered in [2] for correlation of real-world vehicle emissions.The RDE procedure uses a different method to assess the overall excess or absence of dynamics during trips [3].This paper aims to investigate the driving dynamics influence on CO2 emissions based on recorded trips on a passenger car based on RDE driving dynamics indicators.The recordings were done on a passenger car for 22 days and a total of 920 km of combined urban, rural and motorway driving were recorded.The data are postprocessed and analyzed with the aim of selecting the most relevant trips segments to be evaluated using a detailed model developed in LMS Imagine.Lab AMESim and previously validated [4].

Trip recordings and postprocessing
The recordings were made using a DAWN OBD Mini Logger on a C-class passenger car.There were recorded 920 km of trips during 22 days of combine urban, rural and motorway driving.The postprocessing of the trips in terms of driving dynamics and classification was done using RDE procedure [3].
As shown in [3], the dynamic parameters are determined from the speed signal with an accuracy of 0.1 % above 3 km/h and a sampling frequency of 1 Hz.This accuracy requirement can be usually attained by wheel rotational speed signals.
The following calculations were performed over the whole time-based speed trace: distance increment per data sample, vehicle acceleration, the product of vehicle speed per acceleration.
=   /3.6 (1) where: di is the distance covered in time step i; vi is the actual vehicle speed in time step i [km/h]; Nt is the total number of samples.
The values vi , di , ai and (v·a)i are ranked in ascending order of the vehicle speed.To show the high variability of the recorded data, the (v·apos) values for each point of the trips with minimum and maximum average (v·apos) are plotted against vehicle speed and shown in figure 1.Moreover, the RPA values for each point of the trips with minimum and maximum average RPA are plotted against vehicle speed in figure 2.  where, for urban (k=u), rural (k=r) and motorway (k=m) shares: Nk is the total sample number.
A number of 22 trips are recorded, but it is useful to separate into trips segments that are classified as urban, rural and motorway.Only the trips segments with a number of datasets with acceleration values ai > 0.1 m/s 2 and bigger or equal to 150 seconds are considered.Therefore, after applying those conditions, a number of 15 urban trips segment, 6 rural trips segment and 7 motorway trips segments were left.
For each trip the 95 th percentile of v·apos is calculated by ranking the (v·a)i per trip (for ai ≥ 0.1 m/s 2 ) and the total number of these Mk is determined.Afterwards, percentile values are allocated to the (v · apos)j,k values by assigning for the lowest value the percentile 1/Mk, for the second lowest 2/Mk, and so on until the highest value for which the Mk/Mk =100 % value is assigned.(v · apos)k _ [95] is the (v · apos)j,k value, with j/Mk = 95 %, or is calculated by linear interpolation between consecutive samples j (with j/Mk < 95 %) and j+1 (with (j+1)/Mk > 95 %).
The relative positive acceleration is calculated for each speed bin.
for j=1 to Mk and i=1 to Nk (5) where, for urban (k=u), rural (k=r) and motorway (k=m) shares: RPAk is the relative positive acceleration; Mk is the sample number with positive acceleration; Nk is the total sample number and Δt is the time difference (1 s).
The average velocity, 95 th percentile of the v·apos and RPA of each trip segment are calculated and shown in figure 3 and 4 together with an average of each trip segment class.
The limits considered in [3] for high dynamic and shown with red line in figures 1 and 3 are: The limits considered in [3] for low dynamic and shown with red line in figures 2 and 4 are:

Results
The speed profiles for the relevant trip segments are used as test cycle input in a complex C-class SUV model developed using LMS Imagine.Lab AMESim.The model presented in [4] is based on the submodels from the IFP Drive library and can compute the CO2 emissions for a variable speed test cycle.The CO2 emissions were determined for low and high dynamics cycles in every category (urban, road and motorway) in order to assess the influence of using the proposed classification criteria.The results are shown in table 1 and figure 5.  To compare the influence on fuel consumption of different test cycles we define a parameter called CO2 emission ratio (CO2_ER) in a similar way with FCV (Fuel Consumption Ratio) defined in [5]: where: CO2_cycle is the CO2 emission obtained in the test cycle; CO2_ct.speed is the CO2 emission obtained at constant speed movement equal to the average speed of cycle.
From the Table 1 data, we can observe that there is not a simple corelation between chosen cycle dynamics parameters and CO2 emission.In urban driving, the high dynamic corresponds to a cycle with low average velocity.The engine will be operated in a disadvantageous point and the fuel consumption (thus the CO2 emissions will be high), in this particular case 75% higher than low dynamic driving.For motorway driving, the low dynamics corresponds to a cycle with high average velocity and grater aerodynamic drag that will increase the CO2 emissions.For the simulated examples is 27% higher when the driving dynamic is low.In rural driving the two tendencies are compensating and the differences between high and low dynamic driving are under 0.2%.

Conclusions
It was shown that in real-world the driving dynamics can have a big influence especially in urban and motorway driving.The test cycles were obtained as trip segments from 920 km of combine urban, rural and motorway driving recorded for 22 days using RDE procedure.
One limitation of the resulted trip segments is the low specific power of the test vehicle that can influence the overall driving dynamics.Therefore, the recordings must be used for studies on the vehicles with similar specific power.However, for every driving segment type substantial difference between the driving dynamics characterisation parameters was obtained: from 67% to 145% for RPA and from 8% to 47% for the 95 th percentile of the v·apos.

Figure 1 .
Figure 1.(v·apos) values for each point of the extreme trips.

Figure 2 .
Figure 2. RPA values for each point of the extreme trips.The datasets are sorted, those with vi ≤ 60 km/h belong to the 'urban' speed bin, for 60 km/h < vi ≤ 90 km/h belong to the 'rural' speed bin and if vi > 90 km/h belong to the 'motorway' speed bin.The average vehicle speed is calculated for each speed bin.̅  = ∑  ,

Figure 3 .
Figure 3.The 95 th percentile of the v·apos and average velocity of each trip segment.

Figure. 4 .
Figure. 4. The RPA and average velocity of each trip segment.

Figure 5 .
Figure 5. CO2 emissions for low and high dynamics in every category (urban, road and motorway).

Table 1 .
CO2 emissions for low and high dynamics cycles.