Fuel economy and exhaust emissions of a diesel vehicle under real traffic conditions

Traffic and vehicle simulations are often developed individually. However, vehicle performance is heavily affected by traffic conditions. Cosimulations of traffic and vehicle under real‐road situations can reflect the semi‐real‐world performance of vehicles, with traffic conditions being taken into considerations. This paper proposed an approach to combine the traffic and vehicle simulations that are realized by simulation of urban mobility (SUMO) and GT‐Suite software, respectively. In this paper, the sensitivities of the road grade and vehicle speed to the fuel economy and exhaust emissions were investigated; vehicle fuel consumption and regular exhaust emissions on a real‐road were analyzed; the effect of the traffic accident and congestions on fuel consumption and exhaust emissions were quantified. The results indicated that nitrogen oxides (NOx) and soot emission were consistent with fuel consumption rate, which was dominated by vehicle acceleration whose effect was aggravated by road grade. The fuel penalties caused by accident were in the range of 0.015‐0.023 kg depending on the severity of the accidents. The fuel consumption increased from 1.199 to 1.312 kg and 1.559 kg for 900 and 1800 vehicles/h traffic flow cases compared with 180 vehicles/h traffic flow.

bio-fuel, 16,17 exhaust and coolant energy recycle 18 ). Gao et al 19 adopted electrically heated catalyst (EHC) to decrease vehicle cold-start emissions, such that the emission level meets the Euro 6 limitations. However, fuel penalty is in the range of 6.49%-9.35% for different scenarios. Liu et al 20 investigated the effect of the injection strategies of dual fuels on combustion characteristics and emissions and found that a retarded pilot fuel injection slightly decreased of in-cylinder pressure and delayed the combustion phase. They found that the gas injection timing advanced by 4° crank angle (CA) resulted in the lowest soot emission, but the highest NO x emission.
Biodiesel, as a potential alternative of diesel, 21,22 is a promising route of mitigating greenhouse gas emission, 23 and it can effectively alleviate the dependency on the fossil fuel importation. Biodiesel contains more oxygen and less polycyclic aromatic hydrocarbons (PAHs) and sulfur, [24][25][26] which help improve the in-cylinder combustion and reduce PM formation. 27,28 As shown in reference, 29 biodiesel/polyoxymethylene dimethyl ethers blend fuel significantly improved the engine brake thermal efficiency and deceased PM emission. 30 The highest brake thermal efficiency reported by Oishi et al 31 is 27.79% from a dual-fuel mode of 80% biodiesel + 20% methane. It is also conducive to diesel particulate filter (DPF) regeneration 32 since the oxidation activity of biodiesel PM is much higher than the diesel counterpart. 33 Coskun et al 34 investigated the homogenous charge compression ignition (HCCI) combustion, which combined the merits of spark ignition and compression ignition engines and found higher brake thermal efficiency and lower exhaust emissions. However, the ignition of the HCCI combustion was hard to control, which led to the instability of engine operations. 34,35 To improve fuel economy on vehicle level, Mensing et al 36 optimized the vehicle speed trajectories to achieve eco-driving, and the potential improvement of fuel economy was discussed. They found that fuel economy could be improved by ~34% in theory for a free-flow urban driving, but the gain decreased by 16%~54% if safe driving conditions were considered. In the work, 37 203 drivers were monitored on a real-road of Australia and five training courses that the averaged fuel consumption dropped by 4.6% compared with pre-training.
For the current regulations, more and more attentions have been paid to real-world driving, which can reflect the real performance of vehicles. The impact of regulatory onroad test on vehicle emissions based on a Euro 6-complaint diesel vehicle was assessed by Mendoza-Villafuerte 38 who reported that NO x emission was underestimated (by up to 85%) under the current boundary conditions. Yu et al 39 improved the urban bus emissions and fuel consumption modeling by incorporating passenger load factor into real-world driving. Real-world fuel consumption and exhaust emissions were also tested by Yuan et al 40 under hot-stabilized conditions using different vehicles meeting different emission regulation. Similar work was done by Serrano et al 41  based on two identical vehicles; in this work, a new methodology using on-road tests was developed, and their approach presented excellent performance of vehicle testing.  Silva et al 42 developed a numerical model to estimate  vehicle fuel consumption and regular exhaust emissions under urban driving conditions. The model was used to simulate 14 urban trips made by two Ford vehicles. Although it predicted CO 2 emission with relatively high precision, the model did not work well with the gear-shift control strategy, which is an important factor for vehicle performance. 9 Additionally, it could not simulate the vehicle performance under real-road situation with different traffic conditions (eg, congestion, traffic light, and accident) on vehicle performance. It was difficult to develop a mathematical model to estimate the fuel consumption and exhaust emissions with a high confidence level because of the variability and nonlinearity of vehicle fuel consumption and exhaust emissions. Zhou et al 43 reviewed the vehicle fuel consumption models and factors related to fuel economy. It was shown that the factors related to road conditions, driver behavior, and traffic characteristics presented the most significant effect on fuel consumption.
Bieker et al 44 took the city of Bologna as an example to conduct the traffic simulation using SUMO software under a real-world scenario. It considered different kinds of traffic conditions, such as traffic congestion, traffic light, traffic accident; meantime, the real-road network and road characteristic (elevation, rolling resistance factor) were took into account. Dynamic traffic congestion simulation was done by Wang et al, 45 and the formation of the traffic congestion was simulated using an upgraded medium traffic model. In the work, the traffic simulation platform was constructed from the viewpoint of quantitative traffic congestion. Traffic congestion as a common problem has affected many cities around the world; in Hu's research, an actual urban traffic simulation model (AUTM) was set up for simulating traffic congestion and predicting the effect of adding overpasses and roadblocks. 46 This method could be applied to a large-scale real-world scenario in different traffic congestion situations, and the predicted accuracy of the traffic congestion reached 89%. Vissim, SUMO, and Aimsun are powerful software to simulate the dynamic traffic conditions; however, they are difficult to accurately analyze the fuel consumption and exhaust emissions to assess the effect of traffic conditions on fuel economy and pollutants.
Due to the complexity and high cost of the real-world test, especially for those heavy-duty vehicles (such as 40 t trucks) that frequent tests in a long journey and fully loaded situations are unrealistic, the real-road simulation is necessary. Additionally, vehicle performance simulation with high precision on real roads considering the traffic conditions is still a gap. SUMO has an excellent ability of traffic simulation, and GT-Suite is powerful for vehicle performance simulation. In this paper, the merits of the SUMO and GT-Suite were taken to propose a new approach combining the traffic and vehicle simulations to investigate the real-road fuel consumption and exhaust emissions. This is the first time to combine the traffic and vehicle simulations to investigate the vehicle performance under the "real-road situation" to the authors' knowledge. In the simulation, the real-road conditions were used, such as the road elevation and rolling resistance factor; hence, the simulation results were much more closer to the real-road driving. This paper was organized as the following structure: (a) the introduction of the new approach of the real-road simulation, combining the traffic and vehicle simulations; (b) the sensitivities of the road grade and vehicle speed on fuel economy and regular exhaust emissions; (c) simulations of the fuel economy and regular exhaust emissions on real road; and (d) the effects of the accident and congestions on the penalty of fuel consumption and exhaust emissions under real-road conditions.

OF VEHICLE FUEL ECONOMY AND EXHAUST EMISSIONS
The current vehicle simulations are mostly based on the driving cycles suggested in emission regulations, such as New European Driving Cycle (NEDC), 47 Federal Test Procedure (FTP), 15 and Worldwide harmonized Light vehicles Test Cycles (WLTC). 48 Solomon 49 researched the influence of road geometry on vehicle emissions and fuel consumption, with the conclusion that exhaust emissions and fuel consumption have a direct relation with the road grade, which indicated the importance of considering the real-road conditions. The simulations based on these driving cycles contained few information of the real-road network and the traffic conditions. Figure 1 shows the new method of vehicle simulation on a real-road network. The real-road driving simulation could be achieved by the combination of traffic and vehicle simulations. The traffic simulation is conducted by SUMO software 50 ; meanwhile, the vehicle simulation is done using GT-Suite software. The process of the real-road vehicle simulation is as following: (a) obtain the 2D real-road network; (b) integrate the road elevation into the 2D real-road network; (c) extract the targeted route for further simulation; (d) load the real-road network and traffic information into SUMO software; and (e) input the traffic simulation results into GT-Suite software to finish the vehicle simulation. It should be noted that in the 3rd step, the crossways of other roads and the target one remained and there was traffic flow as well. 2D OpenStreetMap contained the information of the real-road network (eg, number of lanes, road geometries, and speed limitations), except for road elevations. The road elevations (from Nasa SRTM) were integrated into the real-road network using the method in osmosis-srtm-plugin instructions. [47][48][49][50][51] The traffic demands were imported using the commands of SUMO, where the traffic flow, vehicle types, and vehicle routes were generated automatically. TraCI4Matlab was used to integrate the SUMO traffic simulator and Matlab, 52 in which target vehicle could be monitored. Madireddy et al 53 also combined the microscopic traffic simulation model with the emission model; however, the emission model was set up based on the vehicle speed and acceleration, and this method was with a low precision. Table 1 shows the specifications of the diesel vehicle, which meets the Euro 6 emission regulation. The engine is a fourcylinder, turbocharged, direct-injection engine. The maximum power output and torque are 103 kW and 325 N·m, which are corresponding to 4000 and 1500 rpm, respectively. Figure 2 shows the tested brake-specific fuel consumption (BSFC) as the functions of engine speed and torque. The optimal BSFC zone is in the range of 1500-3100 rpm and 90-310 N m, where the BSFC is below 230 g/(kW h). Figure S1 shows the exhaust emission maps, which were obtained using engine test bench. BSFC map and exhaust emission maps are basic inputs of the vehicle model in order to investigate the fuel economy and exhaust emissions. Figure 3 is the validation of vehicle model under WLTC, which indicates a high precision of the vehicle model.

| Sensitivities of the road grade and vehicle velocity to the fuel economy and exhaust emissions
Due to the significant changes of the road elevations for the mountain motorways, the road grade being higher than 3% is in a large proportion. Vehicle speed and road grade are two of the most important factors which affect vehicle fuel economy and exhaust emissions. Figures 4-6 show the sensitivities of vehicle speed and road grade to the fuel economy and exhaust emission factors. The fuel consumption was much low if the road grade was negative, except for extrahigh and extralow vehicle speed zones. The fuel consumption was almost doubled when the road grade increased from 0% to 4% that was in the range of the mountain road grade. NO x and soot emissions are still challenges in meeting stricter emission regulations that it should combine the advanced vehicle technologies and driver training to pursue better performance. The tendency of NO x emission factor was similar to the fuel consumption because high fuel consumption caused high in-cylinder combustion temperature and, further, more NO x emission. NO x and soot emissions were at a low level when the road grade was negative due to low in-cylinder combustion temperature, which was caused by low engine load. The highest soot emission zone was located in the ranges of 2%-4% road grade and 100-130 km/h vehicle speed, which was the same for carbon monoxide (CO) emission factor, since the formations of CO and soot were under the condition of oxygen shortage. It should be noted that CO and hydrocarbon (HC) emission factors were still high for low vehicle speed and negative road grade situations. In reference, 7 the fuel sensitivities of the road grade and averaged vehicle speed were also investigated. Costagliola et al 54 researched the impact of road grade on real-driving emissions from Euro 5-complaint diesel vehicles on urban, rural, and motorway roads with different grade (−4% to 5%). The road grade had a significant influence on fuel consumption and NO x emission. CO 2 emission was almost linearly related to the variations of road slope. As indicated by Solomon, 49

| Vehicle fuel consumption and exhaust emissions on real-road
Hooftman et al 1 indicated that Europe's emission regulations of passenger car were proved to fail when it came to NO x by diesel engines, which made it necessary for the real-road investigation. Figure 7 shows the fuel consumption, NO x , and soot emissions on real-road situations. The elevation of this road was in the range of 510-590 m. The maximum and minimum vehicle speeds were ~110 and ~75 km/h, respectively. There were many sections where the vehicle speed changed suddenly, which simulated the brake cases caused by traffic perturbations. Fuel consumption of the vehicle changed significantly, being from 0-21 kg/h, due to uphill, downhill, acceleration, and deceleration. Yu et al 39 indicated that the fuel consumption showed closely related to vehicle acceleration, which was similar to the authors' opinion. Most of the peak positions of NO x and soot emissions were corresponding to the fuel consumption's, and these peaks were almost three times higher than regular values. The uphill aggravated the effect of acceleration on NO x and soot emissions. Gallus's 55 results indicated that the penalty of fuel consumption and NO x emission caused by a 5% road grade was in the range of 85%~115% based on two diesel vehicle road testing. In order to decrease the fuel consumption and exhaust emissions, an acceleration advisory tool was applied to a vehicle, 56 which was avoided to run under aggressive acceleration. It was achieved by the resistance in the acceleration pedal when drivers tried to accelerate rapidly. As can be seen from Figure 8, HC and CO emissions present less dependent on vehicle acceleration and road grade than NO x and soot emissions. Because the tendency of CO and HC emission rates are not consistent with the elevation changes, meantime, the peak of CO and HC emission rates are not only in the acceleration process or on uphill roads.

| The effect of accident on fuel consumption
Large amounts of fuel are consumed if an accident happens on the road as vehicles start to stop and go. The fuel penalties caused by the accident differ significantly from vehicle to vehicle, which depends on how far the target vehicle is from the accident, and how seriously the vehicle speed is affected. However, the fuel penalty of the whole traffic is almost the same, as long as the severity of the accident is known (the accident-affected area, accident-affected time). In this section, the fuel penalty caused by the accident was analyzed from the point of individual vehicle and the group. The scenarios of the real-road accident are shown in Figure 9. As soon as the accident happened, a sign of slowing down was set up 200 m before the accident place. Then, the other vehicles had to change the lanes and decelerated, and the accident-affected region was enlarged gradually. The vehicle speed was set as 10 km/h from the deceleration sign to the accident vehicle in order to avoid the secondary accident. The individual vehicle speed under the effect of real-road accident is shown in Figure 10. In the accident-affected area, the vehicle speed decelerated gradually to a rather low value and accelerated to a The fuel penalty was in the range of 0.015-0.023 kg for the given scenarios, as shown in Table 2. It should be noted that NO x emission changed slightly even though the traveling time increased; however, HC, CO, and soot emissions increased significantly. Based on the data, the relation of individual vehicle fuel penalty and the distance between the target vehicle and accident vehicle can be obtained, as Equation 1, where x i is the distance between the accident vehicle and the affected vehicle i (the distance started from the affected point, which differed for individual vehicle). Further, the total fuel penalty could be calculated if the traffic flow and accident evolution could be gotten, as Equation 2.
F, the total fuel penalty caused by the accident; f i , the fuel penalty of individual vehicle i. Due to much fuel consumption and exhaust emission, penalty will be caused by (1)   traffic accident, and accident charging of wreckers can be considered to be a warn to further enhance the driving safety.

| The effect of congestion on fuel consumption
The effect of vehicle speed caused by congestion (heavy traffic flow) is different from that of accident situations, where the vehicle would recover from low speed situation once the target vehicle passed the accident; however, it had several low speed regions for regular congestions, as shown in Figure 11. The accumulated fuel consumption as the series of distance is shown in Figure 12. In the simulation, three different traffic flows were set to simulate traffic congestions. The simulation was also based on the real-road network. As indicated in the reported work 57 that the commonly used traffic simulation model (average speed models) exclude the effect of traffic congestions on fuel consumption and exhaust emissions. The vehicle was almost under the free-flow condition for 180 vehicles/h traffic flow situation that the vehicle was almost kept at a constant speed, which greatly decreased the times of acceleration and deceleration. It significantly improved the fuel economy because the acceleration dominated the fuel consumption rate, as discussed above. The fluctuations of the vehicle speed increased greatly for 900 and 1800 vehicles/h traffic flow, and vehicle speed fluctuations seriously worsen the fuel economy. Greenwood et al 58 estimated the penalty of vehicle fuel consumption and emissions caused by traffic congestion using the acceleration noise. This approach provided a better predictive ability than those of traditional speed-flow methods, whose errors reached 200% for passenger cars,  (Table 3); meanwhile, NO x emission increased by 67.3% and 170.1% for 900 and 1800 vehicles/h traffic flow, respectively. However, HC, CO, and soot emissions were the least for 900 vehicles/h traffic flow in the three cases. As for HC and CO, the emission factors decreased to a rather low level due to the use of advanced after-treatment technologies; however, NO x emission is still an issue. The congestion made it even worse to achieve the ultralow emission vehicles. Figliozzi 59 analyzed CO 2 emission for different levels of congestions that congestion and speed limits had a significant impact on vehicle CO 2 emission, which was caused by longer travel time. Bharadwaj et al 60 investigated the impact of congestion on fuel consumption for road transport in Mumbai that ~51% of traveling time caused by congestion would led to ~53% more fuel consumption. In order to decrease the traffic congestion, the congestion price was recommend, and a mathematical approach for dynamic congestion price was proposed. 61

| CONCLUSIONS
The real-world test of vehicle performance is time-consuming and expensive, even unrealistic, and the test is enslaved to the real traffic conditions that many factors are uncontrollable for investigators. The real-road simulation can reflect the vehicle performance with relative high precision to some extent. In this paper, a novel approach was proposed to jointly conduct the traffic and vehicle simulations on real-road network in order to investigate the effect of traffic conditions on vehicle performance, further to investigate the penalty of fuel consumption and regular exhaust emissions caused by traffic accident and congestion. The main conclusions are as the following: 1. The approach of the combination simulations of traffic and vehicle system contains the following: (a) 3D OpenStreetMap extraction (including road elevation); (b) traffic condition loading and simulation; and (c) vehicle simulations based on the traffic simulation results. 2. The tendency of the fuel consumption and NO x emissions as the function of road grade and vehicle speed was similar; high soot emission region was located at high vehicle speed and road grade that the maximum soot emission factor reached 3.2 g/100 km.
3. In the real-road simulation, vehicle acceleration dominated the fuel consumption, which was aggravated by road grade. The tendency of the NO x and soot emission rates was consistent with that of fuel consumption rate whose peaks were mainly happened in the acceleration process. 4. The fuel penalties caused by accidents were in the range of 0.015-0.023 kg under the given scenarios for individual vehicle. The fuel consumption increased from 1.199 to 1.312 kg and 1.559 kg for 900 and 1800 vehicles/h traffic flow, compared with 180 vehicles/h traffic flow situation; also, NO x emission increased by 67.3% and 170.1%, respectively.