Printer Friendly

Analysis of a diesel passenger car behavior on-road and over certification duty cycles.

ABSTRACT

Precise, repeatable and representative testing is a key tool for developing and demonstrating automotive fuel and lubricant products. This paper reports on the first findings of a project that aims to determine the requirements for highly repeatable test methods to measure very small differences in fuel economy and powertrain performance. This will be underpinned by identifying and quantifying the variations inherent to this specific test vehicle, both on-road and on Chassis Dynamometer (CD), that create a barrier to improved testing methods. In this initial work, a comparison was made between on-road driving, the New European Drive Cycle (NEDC) and World harmonized Light-duty Test Cycle (WLTC) cycles to understand the behavior of various vehicle systems along with the discrepancies that can arise owing to the particular conditions of the standard test cycles. The engine controller of a 2.0L diesel vehicle with active de-NOx and a particulate filter (DPF) has been monitored over 13,700km of driving. The engine speed/torque operating points showed that both the NEDC and WLTC fail to capture the complete static and dynamic usage observed on the road, and are ill-equipped to capture any driver to driver variations. A cyclic analysis of the DPF is proposed, showing up to 70% variation in soot loading at the point of regeneration. This variation can be explained by the controller waiting for favorable driving conditions for regeneration. NEDC and WLTC are poor cycles for capturing DPF effects as the former presents insufficient engine powers to trigger and the later insufficient time to complete a regeneration event. Along with understanding the general contrasts between standard CD cycles and road conditions, this is a key finding for the project overall as understanding and managing DPF loading and regeneration will enable improvements to be made in test precision.

CITATION: Chappell, E., Burke, R., Lu, P., Gee, M. et al., "Analysis of a Diesel Passenger Car Behavior On-road and over Certification Duty Cycles," SAE Int. J. Engines 9(4):2016.

INTRODUCTION

The ability to demonstrate cost/benefits is vital to the automotive sector increasing rigor demanded of the precision and accuracy of testing operations. In addition, the need to demonstrate real benefits in on-road driving is demanding that realistic, transient operating conditions be reproduced within the test cell environment. Demonstrating benefits on-road has the advantage of capturing a wide range of conditions and allows for the use of many vehicles over long distances. The major disadvantage is the need for a large quantity of data to detect small differences because of the inherent variability of on-road driving [1]. In contrast, CD testing can achieve high levels of precision over short distance, prescribed cycles [2].

This paper presents the first results from a project which aims to create new experimental methods that will improve the current limits of precision on chassis dynamometer facilities to assist with the development and demonstration of future fuels and lubricants. In going beyond the current limits of precision, it is vital to understand the sources of variability that will affect testing and demonstration work. A portion of this variability will be inherent to the vehicle whereas the remainder will be a result of the test facility and methodologies in place. The aim of this paper is understand this vehicle variability through on-road testing and to evaluate how this variability will affect chassis dynamometer testing. A crucial part of this work is understanding how real world behaviors could be reproduced in a high precision laboratory environment. In particular this work will identify and quantify areas where current and proposed standard vehicle duty cycles will fail to capture the broadness of on-road usage due to trip characteristics and driver style. This paper does not compare emissions on-road and in the test cell. The quantitative results are, of course, vehicle specific, however the analysis methodologies and lessons learnt are transferable to future testing campaigns.

The project is being undertaken within the Engineering and Physical Research Council (EPSRC) funded Centre for Low Emissions Vehicle Research (CLEVeR) at the University of Bath. The Centre boasts a unique chassis dynamometer facility combining the capability of reproducing on-road driving conditions with highly accurate state of the art emissions and fuel consumption measurement systems.

LITERATURE REVIEW

The current legislative emissions limits are enforced, as they have been for many years, by testing vehicles over the New European Drive Cycle (NEDC) in a laboratory environment. It is now widely accepted that whilst the NEDC is a useful repeatable standard to compare vehicles against, it does not represent the range of operating conditions found in real world on road driving [3]. To address these concerns the European Commission announced its intention to introduce 'Real Driving Emissions' or RDE legislation as part of Euro 6 and has now amended the standard to suit [3, 4, 5]. At the time of writing, it is planned that this will be implemented as part of Euro 6c, to be effective from 2017 onwards, and to take the form of on-road testing of vehicles utilizing Portable Emissions Measurement Systems (PEMS) equipment with a conformity factor on the emissions limit [3, 4]. In addition the United Nations working party has, for some time, been working on a replacement for the NEDC with the aim of reducing the gap between on-cycle and real world driving. This has culminated in the creation of the Worldwide harmonized Light duty Test Procedure (WLTP) that is also expected to be adopted as part of Euro 6c [3, 4, 5].

The WLTP features the Worldwide harmonized Light duty Test Cycle (WLTC) which is more dynamic and higher load cycle than the NEDC. As such it should better represent real world driving conditions compared to the NEDC, however some authors have shown that compared to the Common Artemis Drive Cycle (CADC) there are some regions of the operating range that are still not covered [6]. Other authors have shown that in terms of NOx emissions the current fleet of Euro 6 vehicles is not far off meeting the NEDC emissions limit over the WLTC, with a typical conformity factor being less than 3.0, with the inference that conforming to the WLTC will not be nearly as challenging as meeting RDE requirements [3]. There appears to be concern therefore that the real world behavior of vehicles could be significantly different to both the NEDC and WLTC.

The United States are participating in the creation of the WLTC, but their emissions regulations are already more advanced than the European counterparts. The introduction of Tier 2 emissions standards resulted from on-road studies undertaken in the 1990s [7]. This study involved 200 vehicles in two regions of the USA which were monitored over a 2 week period and the resulting data provided the basis for the US06 test for aggressive driving and the SC03 cycle to capture the impact of vehicle air conditioning systems. The WLTC has also been developed based on extensive data collection of on road vehicle use [8]. Measured speed and acceleration data are used along with traffic information to produce these cycles. These new cycles are undoubtable improvements compared to the NEDC cycle, however they each consider the vehicle as a time-invariant parameter. There are practical reasons for this, the most pertinent being the demand for testing facilities that could not cope with excessive testing times. One notable exception is the SAE J1711 standard for hybrid electric vehicle testing where recommendations for operating and evaluating these vehicles in charge-sustaining mode are proposed [9]. In this case, base legislated test cycles (US06, FTP75, NEDC, WLTC...) are repeated from a fully charged condition until cyclic battery state of charge behavior is observed - meaning that the battery has undergone a charge/discharge event, returning to a similar charge level (defined by a tolerance) over one of multiple standard test cycles. This is necessary for hybrid electric vehicles to achieve any level of performance metric because the effect of the battery on fuel consumption is so large. However when considering the limits of precision in Diesel vehicles, this cyclic nature may also become important for its dynamic components.

It has been shown previously that achieving high precision, through the control of key noise factors is highly important when conducting chassis dynamometer tests [2, 10]. But there is now an additional challenge to accuracy in the form of replication of real world driving conditions which needs further investigation to enable manufacturers to design vehicles which are compliant with RDE requirements.

Modern Diesel vehicles comprise a range of components that have complex control strategies such as filters and catalysts. These components and their controllers have inherent dynamic behavior that affect the engine performance over time periods typically greater than standard CD tests. For example, the Diesel Particulate Filter (DPF) accumulates soot over time, causing an increase in engine pumping losses and requires periodic regeneration, a process strongly affecting fuel consumption [11, 12]. There is no universal solution of how to account for these systems in vehicle testing. The simplest solution is to remove these devices from the vehicle under test [13], but the removal of such a vital system component can have significant influences beyond the emissions it is designed to reduce - in the case of a DPF this would significantly reduce engine back pressure and pumping losses. For on-road testing where large amounts of data are collected, the effects of the DPF can be captured over long driving distances and be included in overall statistical analysis. However, it would be desirable to be able to capture these effects over shorter testing campaigns, such as those conducted in CD facilities. One possible compromise solution is to test with the DPF in place, but to ignore any tests with active regeneration [5] - this is also the practice currently recommended for proposed Real Driving Emissions legislation. The practicalities of this approach depend on the frequency of regeneration events which vary from technology to technology. For the DPF system, this approach allows the effects of soot accumulation to be observed but omits the impact of regeneration events. New technologies appearing on vehicles such as lean NOx traps and SCR catalysts also have long term dynamic characteristics [14]. There is therefore a requirement for new, high precision testing approaches that capture and account for all of the long term variations of the vehicle system within short distance CD tests.

To develop and demonstrate automotive fuel and lubricant products in this regulatory framework combined with evermore complex vehicle systems, new challenges in precision and accuracy need to be addressed. In order to meet these challenges in precision and accuracy, further investigations are required to improve the understanding of the vehicle behavior on road and on CD test cycles. This paper will make a contribution to addressing these challenges by improving the understanding of imprecision inherent to the vehicle system.

METHODOLOGY

A 2014 2.0L Diesel vehicle fitted with state of the art exhaust after-treatment systems has been monitored over 13,700km of on road driving. The vehicle after-treatment consisted of a diesel oxidation catalyst, a diesel particulate filter and a Selective Catalytic Reduction (SCR) de-NOx system. In order to characterize the vehicle's on-road behavior without applying intrusive methods that might modify the vehicle's responses the vehicle was fitted with an Influx Technology Rebel CT data recorder that monitored signals on the vehicle CAN bus via the Onboard Diagnostics port. Data relating to the operating conditions of the engine and the readings from certain sensors that form part of the vehicle control system were recorded. All data has been logged at 2Hz and the key signals used in this paper are summarized in table 1. Some of these variables have physical meaning whereas others are used as indicators of vehicle behavior. It is important to note that even those variables with physical meanings do not necessarily result from a physical sensor, and can be results of models inherent to the engine management system (EMS). Attention is brought in particular to the exhaust temperature: this represents one of three measures available within the EMS which represent different locations along the exhaust system. For the purpose of this work, and for the analysis undertaken with respect to exhaust gas temperature, the location furthest from the engine, post DPF, has been chosen. However the conclusions have been verified with the other sources of exhaust gas temperature estimate.

The soot loading recording that is used as a way to characterize the behavior of the DPF. The raw data from this channel is provided as a mass estimate, expressed in grams, and will be the output of an on-board model of the DPF system such as that presented in [15]. For the purpose of this work, this value has been normalized between 0-1 with 0 representing an empty DPF (not achieved in practice) and 1 representing the highest ever seen DPF soot loading which was observed following persistent in-city driving (see figure 10). Whilst this scaling is somewhat arbitrary, it will demonstrate the variation in usage of the DPF total capacity.

The vehicle has a manual transmission and has been driven by four different drivers on the road and a different laboratory driver for NEDC and WLTC tests. Initially the vehicle was installed on a 2 wheel drive chassis dynamometer installation and driven through a number of NEDC and WLTC duty cycles as detailed in table 2. For both cycles, the time based vehicle speed profile was followed and the vehicle shift indicator was used to determine gear shifting points. The emphasis in this work was related to the overall vehicle behavior rather than emissions and fuel consumption measurements and the accuracy of these tests is deemed sufficient for comparison with on-road data, however they cannot directly be compared with standard test results that may have been established on other testing facilities.

Following chassis dynamometer testing, the vehicle was driven by four different drivers over a total distance of around 13,700km, spanning an eight month period. The drivers were not selected on a statistical basis but the intention was to represent a selection of driving patterns to stimulate the vehicle over a range of scenarios somewhat representing typical road driving. A range of road types, gradients and distances were represented. In total 314 trips were monitored and a summary of the range of data sets is given in table 3: the values described in column 1 have been calculated for each trip and the statistics detailed in subsequent columns relate to the trip to trip variations.

Table 4 gives the breakdown by driver of the trip statistics which is important to show a reasonable contribution to the overall on-road data from each participant. There are however notable differences, as driver 2 has the largest number of trips and the lowest average trip distance of all drivers. Similarly, driver 3 has tended to drive longer trips. It is important to bear this in mind when analyzing driver to driver variability which may not be solely as a result of driver behavior but could also be a result of different routes. It is beyond the scope of this paper to determine the reasons for the driver to driver variation, but this work will seek to illustrate the discrepancies that can arise compared to standard cycles.

RESULTS

The results from on-road and on-cycle driving are presented in the following two sections. The first section describes the differences in overall operating point of the engine and the second presents a detailed analysis of the DPF behavior. The particulate trap is one of the time variant, dynamic non-linear components that is inherent to the vehicle and impacts on the precision of testing. At the same time, being part of the system, the DPF is a key component that determines the on-road behavior of the vehicle and therefore removing it to improve testing precision is not a favorable option. Therefore the aim of this second section is to understand the behavior of the DPF to improve future testing approaches.

Overview of Driving Cycles and On-Road Test

Figure 1 shows the breakdown of engine operating points on the engine speed/torque map. Individual plots have been produced for the NEDC, WLTC and on-road data and these have been split into three categories according to the vehicle speed (below 45km/h, 45-80km/h and above 80km/h). The breakdown by speed is somewhat arbitrary to show discrepancies amongst the vehicle speed regions. It is obvious from the plots that the on-road driving has covered a much broader region of the speed torque map with the most notable differences being:

1. The low speed/high torque region

2. Engine speeds above 2500rpm

These two regions are not visited during NEDC and only briefly during the WLTC cycle.

The frequency of operation in any given location of the speed/torque map is shown through the shading of the points in figure 1. This shading represents the percentage of time that is spent at a particular speed, torque and vehicle speed range with respect to the total driving time for that duty cycle. Also, the scales on the color shading has been normalized across the different duty cycles.

The NEDC data shows some significant "hot spots" at narrow speed and torque locations which correspond to the steady state cruises of the drive cycle for each of the three speed ranges. In contrast, the WLTC tends to have its operating point spread more evenly owing to the more transient nature of this cycle. The on-road data tends to have "hot spots" which correspond to narrow speed regions but covering a large range of torque. These conditions correspond to the cruising speeds of the vehicle as dictated by on-road speed limits (causing the narrow speed windows), but with a range of torques to reflect the different road loads that could occur at these speeds on the road. The NEDC and WLTC being idealized cycles, for a given vehicle speed the road load will always be the same, corresponding approximately to vehicle drag on a flat surface with no wind.

The real world data shows the highest density of points at torques below 50% maximum torque which does correspond to the areas visited by the NEDC and WLTC suggesting that whilst incomplete, the drive cycles will capture the most frequently used operating points for this speed. Crucially from this data, it would seem that the WLTC is lacking some of the stable vehicle speed driving that does appear on the road.

In the 45-80km/h region in figure 1 (c), the most frequently used points from on-road data correspond to the full torque range at speeds between 1200-1500rpm. Comparison with figure 1 (a) suggest this is perhaps the most major shortfall of the NEDC as these torque ranges are not captured. In figure 1 (b) it can be seen that the WLTC goes some way to addressing this.

At speeds above 80km/h, the majority of operating points from the on-road data are concentrated in the region of 1500-2500rpm and 40-60% maximum torque. This is a reflection of the more steady state nature of driving at these speeds which will encompass all out of town and motorway cruising. Both the NEDC and WLTC have a concentration of points in this region suggesting that they would be capable of capturing general trends in this region however both fail to cover the full range of operating points seen at these speeds which may result from a range of environmental factors including gradients.

Figure 2 breaks down the on-road usage by driver: this representation aims to show the variability in driving habits that are observed even within a small pool of drivers. It should be noted that for the purpose of this analysis, the term "driver" should be considered as both a variation in the human driver and the routes that they will have driven (i.e. this study is not a comparison of different drivers over the same routes).

There are clear differences between the drivers in terms of the range of engine speed and torque they have used. Drivers 1 and 2 rarely take the engine above 2000rpm whereas drivers 3 and 4 frequently use engine speeds in excess of 2500rpm which will be a direct result of their gear shifting strategies.

For vehicle speeds below 45km/h, the most frequently used points are simlar for all drivers, corresponding to speeds below 2000rpm and up to 50% maximum torque however drivers 3 and 4 have a tendancy to use a higher torque values and a broader speed range.

The speed range 45-80km/h presents the most significant difference between driver habits: whilst drivers 1 and 2 appear to use the full torque range at engine speeds between 1200-1500rpm, drivers 3 and 4 tend to use a broader speed range. The use of gear shift indicators in vehicles has neutralised these driver to driver variations for the current NEDC test procedure for manual transmission vehicles. In the proposed WLTP, the gear shifting logic is predetermined by the vehicle power, torque and weight characteristics [16]. However in order to capture this level of variation, CD testing will need to move away from a predetermined single trip towards a more broad approach or even an engine rather than vehicle centred test design specifying engine speed and torque rather than vehicle speed and acceleration.

Refering back to the discussion relating to figure 1, it could be concluded that the WLTC would be suitable for capturing the driving styles of drivers 3 and 4, however less so of drivers 1 and 2. Similarly, whilst the NEDC does capture the fixed speed cruise conditions, there is not the broad spread of engine torque since the vehicle load and thus the engine torque is fixed at these points on the cycle.

For vehicle speeds above 80km/h, the driver behaviour is fairly consistent between drivers with narrow speed ranges and wide torque ranges which will be caused by the steady state driving that tends to occur at these speeds. The differences in the location of the "hot spots" along the engine speed axis from driver to driver may be more related to the vehicle duty cycles and the speeds at which the vehicle was driven.

The above representations capture the static behaviour of the engine by analysing the engine speed and torque at any point in time. At a vehicle level this can be represented by considering vehicle acceleration and figure 3 presents the operating frequencies of vehicle acceleration and vehicle speed for NEDC, WLTC and on-road data (all drivers combined). Vehicle acceleration has been obtained as the numerical differential of measured vehicle speed.

The x axis has been limited to the range of -3m/[s.sup.2] to +3m/[s.sup.2] in order to provide the details of common acceleration events. In particular, this study is not interested in high deceleration events under braking. The hot spots on this graph clearly show the steady state point from the NEDC. The WLTC has a broader coverage of positive and negative accelerations in the range of -1.5 to 1.5m/[s.sup.2] and less distinct steady state hot spots. This is consistent with the composiiton of the duty cycles where the WLTC is more transient with no significnat steady state periods. The on road data shows two main features:

* There is significant operation with largely steady vehicle speeds

* There is a much broader range of accelerations than both the WLTC and NEDC, but the frequency of operation in these regions is low.

A second transient analysis is presented below in form of a spectral analysis of the different duty cycles. The spectral analysis is computed using Welch's approach [17] to estimating the fourier transform of vehicle speed and engine torque. For vehicle transients, the time constants for the physical phenomena of interest are of the order of 0.1-20s, corresponding to frequencies 10-0.05Hz. However due to the limitations of logging frequency to 2Hz, the analysis in this work is restricted in upper frequency to 1Hz. To match these time periods of interest, a Hanning window function was used during the fourier transform calculation with the intervals described in Table 5. In all cases the window interval is larger than the period of interest to avoid power losses in the calculation.

Figure 4 (a) presents the frequency spectra of vehicle speed for the NEDC, WLTC and on-road data. The spectrum is presented over the frequency region from 0.01 to 1Hz. The vehicle speed spectra show that the on-road data has a higher frequency content in the frequencies approaching 1Hz than both NEDC and WLTC. This suggests that the on-road data is more dynamic in vehicle speed than the other cycles. This agrees with the representation in figure 3 and the engine speed/torque plots presented in figure 1.

Figure 4 (b) presents similar frequency spectra for engine torque. This is important because at different points in time, the engine may be producing a given torque at a given speed, however if in one case the engine is experiencing a torque transient, the air fuel ratio in the cylinders may be much closer to the smoke limit than if the same torque is encountered in a steady state situation.

Generally speaking, the amplitude of torque changes is higher for the on-road data than both the NEDC and the WLTC over the frequency ranges 0.1 to 1Hz. There are exceptions in specific frequency ranges where the amplitudes of the chassis dyno cycles are similar to or even greater than the on-road data, for example:

* 0.3-0.45Hz and 0.7-1Hz for WLTC

* 0.7-0.8Hz for NEDC

Analysis of this rate of change in torque is limited somewhat by the logging frequency of the data sets which will be unable to capture the rapid changes in torque purely induced as a result of fuelling changes, however load changes limited by the turbocharger response would be reasonably expected to appear. Further analysis would be recommended here using higher logging frequencies such as up to 20Hz to understand the trend of these spectra into the higher frequencies.

The static and dynamic analysis of the on-road data and chassis dynamometer test cycles have shown the differences and similarities between them. Considering only the speed and torque usage, there are aspects of both the NEDC and the WLTC that can be found in on-road driving meaning both could remain relevant in this respect for developing and demonstrating automotive fuel and lubricant products. However, individually both cycles have limitations in their ability to represent the full breadth of on-road driving. It is worth emphasizing that no practical drive cycle for emissions homologation could be expected to contain all possible driving scenarios as the cycle has to fit within the legislative framework and be applied in a finite number of test facilities for a large number of vehicles. The WLTC has been developed with a predetermined duration of 1800s which represents a compromise of statistical representativeness and practicality [8].

Diesel Particulate Filter Analysis

DPF Cyclic Breakdown

The test vehicle was equipped with a particulate filter system which collects soot emissions exhausted from the cylinders. Over time, the soot will build up; increasing the back pressure on the engine and eventually the soot will need to be purged through a regeneration event [11, 12]. A typical regeneration is illustrated in figure 5. In order to increase the temperature in the exhaust to burn off the collected soot particles, the engine management system was observed to:

1. Activate an engine throttle to reduce the air fow

2. Inject additional fuel into the cylinder very late in the cycle in the form of three post-injections.

Over the 13,700km of on-road testing, 30 particulate filter regenerations were observed: the modelled soot loading at the point of regeneration is plotted against the distance between regenerations in figure 6. The soot loading at regeneration tends to correlate with the distance between regenerations; this is logical as the longer distance travelled, the more soot will accumulate in the filter, though it is recognized that this will be affected by operating conditions such as EGR use and exhaust gas temperature. The measurements have shown that the distance between regenerations varied between 260 and 710km. The soot loading at regeneration, as estimated by the engine management system, was never seen to be lower that around 60% of maximum observed loading. These values are of course specific to this vehicle and DPF sizing, however they illustrate the concept of minimum and maximum soot loading and the soot loading window for activating particulate filter regeneration.

A cyclic analysis concept to quantify the behavior of the DPF system is presented in figure 7. In this figure, a complete DPF cycle is presented from the end of a previous regeneration until the current regeneration being analyzed. In the upper graph is shown the evolution of vehicle speed and in the lower graph is an indicator within the engine management system to indicate a state of active DPF regeneration. The plot is composed of a number of different trips concatenated in the time domain to give a completed record of vehicle behavior between the two regenerations. Over the period, six key points are identified as follows:

1. Start of the regeneration cycle - this is the point immediately after the end of the previous regeneration and the point at which the filter soot loading is at its lowest point.

2. Activation of current regeneration - this is the point at which the engine management decides to activate the current regeneration and corresponds to the highest point of soot loading in the DPF

3. End of current regeneration cycle - this point corresponds to the end of the regeneration being analyzed and also the beginning of the subsequent regeneration cycle (point 1 for next cycle)

4. Two minutes prior to regeneration activation - this point allows a two minute period just prior to regeneration activation to be determined between points 4 and 2

5. Earliest expected regeneration activation - this corresponds to the point at which the DPF loading reaches 60% of maximum loading, corresponding to the earliest observed regeneration activation and is an interesting point as it corresponds to the point after which we might expect a regeneration

6. Two minutes prior to 60% soot loading - similar to point 4, this point determines a two minute period just prior to the earliest expected regeneration activation time.

The proposed cyclic breakdown of the DPF cycle can be used to illustrate and analyze data relating to this behavior. In figure 8 the average exhaust gas temperature during different periods are compared for all 30 regeneration cases recorded during on-road driving. Each of these four average temperatures yield particular information:

* The average exhaust temperature during period 1-2 lies close to 200[degrees]C for all regeneration cases. This represents long periods of driving when the vehicle is not in an active state of regeneration and could be interpreted as the natural average exhaust temperature for this vehicle when driven by the four drivers in this study.

* The average exhaust temperature during period 2-3 is between 400-600[degrees]C. This represents the periods when the engine is in active regeneration mode and the exhaust temperature is higher because of the measures taken to increase this temperature illustrated in figure 5.

* Period 4-2 represents the period just prior to each regeneration activation. For the majority of cycles, the average temperature during this period is 25-50[degrees]C hotter than average temperature in period 1-2. This is an indicator the conditions prior to a regeneration tend to favor hotter exhaust temperatures and could be interpreted as higher power driving conditions. There are some notable exceptions to this in particular regenerations number 19 and 25: these correspond to regenerations that occurred very soon after vehicle start-up and with very high soot loadings close to 1. This would suggest these two cases correspond to extreme regenerations that have been triggered by an upper soot loading level.

* The period 5-4 corresponds to the period just prior to the earliest expected regeneration. The average exhaust temperature during this period varies from cycle to cycle and can be similar to the average temperature during period 1-2 or similar to period 4-2. This is a result of the random nature of driving and that in some cases, driving conditions will require more engine power, and cause higher exhaust temperatures than others. The period 5-4 shows the snapshot of the conditions at the point the DPF soot loading reaches 60% maximum loading. It would be expected that for cases where the average temperature in this period 5-4 is similar to regular driving (period 1-2), then the engine management would delay activation of regeneration mode. In other cases, where the average temperature in period 5-4 is high and similar to that seen in period 4-2, then the engine would activate regeneration mode at that time. In this latter case it should be noted that points 2 and 5 become very close in time and for this reason some points do not exist for period 5-4 as point 4 may occur before point 5.

As exhaust gas temperatures are a result of the engine operating condition and depend on the amount of fuel being burnt and the quantity of fresh air being ingested by the engine, the driver to driver effects observed in figure 4 will also affect when these favorable conditions for regeneration occur. The results for average vehicle speed yield similar if somewhat less clear results. Average speed will be more strongly affected by the driving routes rather than the driving style and therefore may have less variability driver to driver.

* The average speed in period 1-2 is predominantly lower than for period 2-3 showing that DPF regenerations tend to occur at higher vehicle speeds.

* The average vehicle speed during period 4-2 is higher in most cases than during period 1-2 and similar to the average speed encountered during regeneration (period 2-3). This illustrates that the engine management system waits for a period of high vehicle speed/power before activating the regeneration state. Cycles 15, 19 and 25 have very low average speeds during this period: this is because the regenerations occur soon after vehicle start up.

* The average vehicle speed during period 5-4 fluctuates from being similar to that of period 1-2 or that of 4-2. This is because in some cases this period 5-4 will correspond to general driving and in others this will correspond to the period just prior to regeneration. Figure 9 summarizes this discrepancy by plotting the soot loading at the start of regeneration against this average speed during period 5-4. Cycles with higher average speed during this period 5-4 tend to trigger the DPF regeneration with lower soot loadings closer to the 60% loading trigger, whereas a lower average vehicle speed results in a wide range of soot loadings at regeneration up to the maximum observed in this study.

Figure 10 shows the case of an exceptional regeneration cycle where the vehicle has been stopped during an active regeneration on multiple occasions. This circumstance can arise for prolonged city running and in the case of this study was observed during a prolonged series of short trips. It can be seen that the regeneration is attempted over three periods (evidenced by the three drops in DPF soot loading). These scenarios are difficult to analyze according to the cyclic approach described above, however it is a scenario that can easily occur both in real driving and over low power cycles such as NEDC.

Correlations of Vehicle Operation and DPF Cycles

Figure 11 shows the correlation between regeneration duration (period 2-3), initial soot loading and average engine power during the regeneration. To improve visualization of the observed trends, the raw data has been fitted with a quadratic surface with fit correlation [R.sup.2] of 0.70. Two observations are made:

1. The duration of the regeneration is longer with higher initial soot loading, i.e. when a greater quantity of soot needs to burnt from the filter, the regeneration takes more time.

2. When the average engine power during regeneration is higher, the regeneration tends to be shorter, i.e. higher engine power seems to assist with the regeneration and reduce the total time required.

In the same way, figure 12 shows the correlation between the fuel consumed by post injections during the DPF regeneration (period 2-3), the initial soot loading and the average engine power during the regeneration. It can be seen, as expected, that the total fuel consumed through post injections increased when the soot loading was higher.

Figure 12 shows that there is a general trend of the total quantity of extra fuel injected by the post injections during period 2-3 being proportional to the normalized soot loading on the DPF. In other words the higher the soot mass that is held in the DPF at the start of the regeneration the higher the mass of post injection fuel that is needed. Figure 12 also shows that the extra mass of fuel injected has no trend with the average engine power, so all the additional fuel mass is used to burn off soot in the DPF and not to generate additional shaft power from the engine.

It should be noted that an exceptional regeneration cycle produced an extremely high post-injection fuel consumption of approximately 700g to burn only 60% of maximum soot loading with highest average engine power. This point was removed from the 3D plot but is included in the 2D projections. The reasons for this point are unclear but cannot be explained by any errors in data acquisition.

Figure 13 shows the fuel consumption of the post injections that are active only during the regeneration as a percentage of the total fuel injected over:

a. The full regeneration cycle (period 1-3)

b. The active regeneration (period 2-3)

The fraction of fuel injected as post injections represents between 1-4% of the total fuel consumed over the full cycle (period 1-3). This is plotted against the duration between regeneration in figure 13 (a). There is a general trend for the post injection fraction to reduce as the duration increases. This is explained by the longer period 1-2 representing a larger quantity of fuel. The fraction of fuel injected as post injections represents between 40-60% of the total fuel injected during the regeneration itself (period 2-3). Figure 13 (b) shows that this is not linked to initial soot loading.

Comparison of DPF Behavior during NEDC, WLTC and On-Road Driving

Figures 14 and 15 show the DPF soot loading and exhaust gas temperatures for the NEDC and WLTC cycles respectively. In both cases, a combination of cold and hot start cycles have been presented together, illustrated by the differences in exhaust gas temperature over the first 800s of each cycle.

For both cycles, it can be seen that there are a number of tests where the DPF soot loading was above the 60% threshold, which based on the on-road data is believed to be the minimum loading to trigger a DPF regeneration. However, during each cycle, no complete regenerations have been observed. In the case of the NEDC, the exhaust gas temperature remains below 300[degrees]C throughout the test, and it can be seen that no attempt is made by the engine management system to regenerate the filter. In the case of the WLTC cycle, a regeneration was systematically activated during the last seconds of the test, evidenced by the reduction in soot loading, however the regeneration is unable to complete and is interrupted by the end of the cycle, similar to the behavior presented in figure 10.

Analysis of the final 2 minutes of the NEDC shows that the average exhaust gas temperature was 210[degrees]C and the average vehicle speed 92km/h, whereas for the WLTC cycle these were 286[degrees]C and 113km/h respectively. Clearly this difference is sufficient to trigger a regeneration in the case of the WLTC but not for the NEDC.

The comparison between NEDC, WLTC and on-road driving shows that starting with a regenerated DPF, 45 NEDC tests and 14 WLTC tests would need to be run in order to complete a full regeneration cycle. The large difference in these estimates is a consequence that the WLTC does in fact cause driving conditions to trigger a DPF regeneration, although this would be expected to influence multiple sequential tests.

Current, high precision CD and vehicle tests for fuel economy tend to ignore the effects of DPF by excluding tests where regenerations occur. These are easy to detect as figure 13b shows that a DPF regeneration can more than double the instantaneous fuel consumption. However figure 13a shows that by ignoring DPF regenerations, fuel consumption could be underestimated by up to 4% over long term driving. It is therefore interesting to consider how a high precision test procedure could be designed that includes the DPF behavior.

With regards to the precision of CD testing, it could be proposed that the SAE J1711 [9] standard for hybrid vehicle testing presents an interesting solution to this DPF cycle problem. In the same way that the battery is charged and discharged as the vehicle is driven, the DPF is loaded and unloaded by action from the engine. This would involve cycling the vehicle through repeated standard test cycles until a repeated level of soot loading was achieved. The complete DPF cycle would then be considered as the test. The disadvantages of this is that as has been shown from the on road data, this would be expected to take a large number of NEDC or WLTC cycles and this is impractical from an emissions perspective. A second option would be to consider the expected behavior of repeated NEDC and WLTC cycles:

* Repeated NEDC cycles: As there are no favorable conditions for DPF regeneration, the soot loading would increase up to a maximum level at which point a regeneration would need to be triggered in non-favorable conditions. Owing to the low load nature of the NEDC, that may only result in a partial regeneration of the filter after which the soot loading would again increase up to the maximum level

* Repeated WLTC cycles: As there are favorable condition for regeneration during the extra high period, a regeneration would be triggered once the vehicle reaches the minimum threshold of 60% soot loading. However, only a partial regeneration would be completed which may be continued in subsequent cycles.

In both cases, this is very different to the observation of on-road behavior where the majority of DPF regenerations are completed. A direct adoption of SAE J1711 [9] would therefore not be sufficient and care should be taken to ensure that a cycle offering favorable regeneration conditions is included within the test procedure.

Examining the on-road data in isolation showed that for individual drivers carrying out repeated routes, the DPF regenerations tended to occur in similar locations during their trips. Additionally there was a significant difference in DPF regeneration behavior when these routes were driven in different directions. For example figure 16 and 17 show that for one driver carrying out a repeated trip, there were six regeneration events out of seventeen trips in the outbound direction compared with only one regeneration out of nineteen trips completed in the inbound/return direction. For the outbound trips, where six regeneration events occurred, the soot loading was always above the 60% minimum threshold and yet from the time of engine start, the regeneration was not triggered until the average vehicle speed was consistently high, needing to be approximately 100km/h. The region for regeneration also corresponds to the region where the vehicle does not stop. It appears that, unless the soot loading is at the highest threshold, the vehicle needs to be in a high speed cruise condition before a regeneration is triggered. This explains why the regenerations are normally seen in the same portion of the trip; since these regions are higher speed roads with less traffic. These conditions are in broad agreement with the trends seen in figure 8 where the average vehicle speed in the period 4-2 was always in the region of 80-100 km/h. The notable exceptions (regenerations #15, 19, 23 and 25 in figure 8) are all cases where the soot loading at the start of the regeneration (point 2 in figure 7) is high. In these cases the regeneration is considered to be triggered by an upper threshold on soot loading rather than by favorable driving conditions - this would correspond to behavior for prolonged and persistence low speed and low load driving such as in cities.

Figure 17 shows a less clear trend for the return trips, there appear to be multiple trips where the conditions are favorable for regeneration, i.e. a higher vehicle speed cruising condition with no vehicle stoppages and soot loading above 60%, yet only on one trip does the DPF regenerate. Clearly the algorithms modelling soot loading and regeneration are complex and have a strong time-dependency; the variability of events is therefore high, and the consequences in terms of vehicle operation (e.g. fuelling & other settings) significant. So there are dual challenges of transferring "representative" behavior to a CD, and ensuring that testing minimizes the effects of vehicle operating variables versus target differences from fuels and lubricants. These challenges will be addressed in future parts of the program.

CONCLUSIONS

In this paper the behavior of a single diesel passenger car has been compared on NEDC, WLTC and over 13,700km of on-road driving. Various parameter data were recorded from the vehicle's engine management system to provide indicators of vehicle behavior. These have given vital insights into the variability that is inherent within this test vehicle and that will affect the precision and accuracy of chassis dynamometer testing. In particular, the on-road testing has demonstrated the time varying aspects of vehicle behavior. The comparison with standard vehicle test cycles has shown how these vehicle variations may be expected to manifest themselves during development or demonstration testing campaigns.

A comparison of the engine speed and torque operating characteristics illustrated the small operating regions covered by these certification cycles compared with on-road driving. These differences were shown to be both static (absolute speed and torque levels) and dynamic through differences in frequency spectra. Whilst this analysis does not relate to the vehicle variability, it does highlight how operation over the standard test cycles will fail to capture certain characteristics that would be seen in real world operation. Furthermore the analysis of on-road data from individual drivers in the 45-80km/h speed range, showed that two drivers tended to cruise at constant speeds using a broad spread of engine torque; a behaviour that is not represented by either the NEDC or WLTC, since the former has fixed engine load cruise conditions and the latter does not have longer duration fixed speed cruise conditions.

The breakdown of on-road data by driver showed that at speeds below 45km/h and above 80km/h, there are similar characteristics in terms of the principal engine speed/torque operating points. For vehicle speeds between 45-80km/h, differences were observed which could be attributed to different driving styles and gear shifting strategies. It is clear that predetermined gear shifting such as that used in the NEDC and the WLTP will not capture these variations. Crucially, as the engine operation determines whether there are favorable conditions for DPF regeneration, the driving style could affect this behavior as well as route to route characteristics. These findings suggest that for CD testing to be fully representative of on-road conditions, an engine-centered rather than vehicle-centered testing approach could be adopted.

The cyclic nature of DPF regenerations was captured from on-road driving through 30 independent regenerations. These were typically separated by 200-500km with soot levels varying between 60-100% of maximum soot loading. This means that a full DPF cycle would require 45 NEDC or 14 WLTC cycles which should be considered when planning test campaigns. This is particularly important as the DPF cycle directly affects fuel economy with a regeneration accounting for 40-60% of fuel consumption during regeneration and 1-4% over a full cycle. Simple correlations were observed between the average driving conditions and the regeneration cycles. In particular it was observed how driving conditions can affect the triggering of the DPF regeneration with lower engine powers and vehicle speeds delaying this trigger. Crucially these regeneration conditions are not achieved during the NEDC and only partially triggered on the WLTC. Due to both the first order impacts (post injection) and second order impacts (soot loading affecting exhaust backpressure) of DPF operation it is concluded that DPF duty must be managed during test programs to achieve good precision of fuel economy measurements.

The DPF behaviour aspects of this program has laid the foundation to understanding some of the requirements for highly repeatable test methods to measure very small differences in powertrain performance from fuels and lubricants. This has been underpinned by identifying and quantifying the variations inherent to this specific test vehicle, both on-road and on-CD, that create a barrier to improved testing methods.

REFERENCES

[1.] Nattrass, S. and Jones, W., "The Application of Telematics to the High-Precision Assessment of Fuel-Borne Fuel Economy Additives," SAE Technical Paper 2012-01-1738, 2012, doi:10.4271/2012-01-1738.

[2.] Brace, C. J., Burke, R. and Moffa, J., Increasing accuracy and repeatability of fuel consumption measurement in chassis dynamometer testing. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, 2009. 223 (9): p. 1163-1177.

[3.] Yang, L., Franco, V. , Campestrini, A., German, J. and Mock, P., [NO.sub.x] Control Technologies for Euro 6 Diesel Passenger Cars. 2015, ICCT. Available from: http://www.theicct.org/[NO.sub.x]-control-technologies-euro-6-diesel-passenger-cars [Accessed 01/07/2016]

[4.] May, J., Bosteels, D., and Favre, C., "An Assessment of Emissions from Light-Duty Vehicles using PEMS and Chassis Dynamometer Testing," SAE Int. J. Engines 7(3):1326-1335, 2014, doi:10.4271/2014-01-1581.

[5.] Andersson, J., May, J., Favre, C., Bosteels, D. et al., "On-Road and Chassis Dynamometer Evaluations of Emissions from Two Euro 6 Diesel Vehicles," SAE Int. J. Fuels Lubr. 7(3):919-934, 2014, doi:10.4271/2014-01-2826.

[6.] Sileghem, L., Bosteels, D., May, J., Favre, C. and Verhelst, S., Analysis of vehicle emission measurements on the new WLTC, the NEDC and the CADC. Transportation Research Part D: Transport and Environment,

[2014.] 32: p. 70-85.

[7.] n.d, 1993, Federal Test Procedure Review Project: Preliminary technical report, EPA 420-R-93-007, U.S. Environmental Protection Agency, [online] Available from: https://www3.epa.gov/otaq/regs/ld-hwy/ftp-rev/ftp-tech.pdf [Accessed 01/07/2016]

[8.] Tutuianu M., Bonnel P., Ciuffo B., Haniu T., Ichikawa N., Marotta A., Pavlovic, J., Steven, H., Development of the Worldwide harmonized Light duty Test Cycle (WLTC) and a possible pathway for its introduction in the European legislation, Transportation Research Part D: Transport and Environment. 2015; 40:61-75.

[9.] n.d., 2010, Recommended Practice for Measuring the Exhaust Emissions and Fuel Economy of Hybrid-Electric Vehicles, Including Plug-in Hybrid Vehicles, SAE Standard J1711_201006, SAE International, Warrendale, USA

[10.] Chappell, E.C., Brace, C.J. and Ritchie, C., The control of chassis dynamometer fuel consumption testing noise factors and the use of response modelling for validation of test repeatability. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, 2013. 227 (6): p. 853-865.

[11.] Konstandopoulos, A., Kostoglou, M., Skaperdas, E., Papaioannou, E. et al., "Fundamental Studies of Diesel Particulate Filters: Transient Loading, Regeneration and Aging," SAE Technical Paper 2000-01-1016, 2000, doi:10.4271/2000-01-1016.

[12.] Stenning, L., "Strategies for Achieving pre DPF Regeneration Temperatures using in Cylinder Post Injection on a Common Rail Diesel Engine with EGR, DOC and Intake Throttle," SAE Technical Paper 2010-36-0306, 2010, doi:10.4271/2010-36-0306.

[13.] Vojtisek-Lom, M., Fenkl, M., Dufek, M., and Mares, J., "Off-cycle, Real-World Emissions of Modern Light Duty Diesel Vehicles," SAE Technical Paper 2009-24-0148, 2009, doi:10.4271/2009-24-0148.

[14.] Majewski, W.A., Selective Catalytic Reduction, Ecopoint Inc., 2015, [online], Available From: https://www.dieselnet.com/tech/cat_scr.php, [Accessed 05/07/2015]

[15.] Kladopoulou, E., Yang, S., Johnson, J., Parker, G. et al., "A Study Describing the Performance of Diesel Particulate Filters During Loading and Regeneration - A Lumped Parameter Model for Control Applications," SAE Technical Paper 2003-01-0842, 2003, doi:10.4271/2003-01-0842.

[16.] n.d., Report of the Working Party on Pollution and Energy (GRPE) on its seventy-second session. Addendum 1: Adopted proposal for Amendment 1 to global technical regulation (gtr) No. 15 on Worldwide harmonized Light vehicles Test Procedures (WLTP), UNECE report ECE/TRANS/WP.29/GRPE/72/Add.1, 17 March 2016, [online] Available from: http://www.unece.org/fileadmin/DAM/trans/doc/2016/wp29grpe/ECE-TRANS-WP29-GRPE-72a1e.pdf [Accessed 06/07/2016]

[17.] Welch P. The use of fast Fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms, IEEE Transactions on Audio and Electroacoustics. 1967;15(2):70-3.

CONTACT INFORMATION

Contact Author: Dr Richard Burke Address: Powertrain and Vehicle Research Centre, Dept. of Mechanical Engineering, University of Bath, Bath, BA2 7AY, UK R.D.Burke@bath.ac.uk

Phone: +44 (0) 1225 38 3481

ACKNOWLEDGMENTS

This research was carried out using the equipment and facilities of the Centre for Low Emissions Vehicle Research (CLEVeR) which is part of the Powertrain and Vehicle Research Centre at the University of Bath. The equipment in this facility was funded by an EPSRC equipment grant and their funding is acknowledged.

The assistance of Influx Technology in logging data from the vehicle's engine controller is gratefully acknowledged (http://www. influxtechnology.com/).

DEFINITIONS/ABBREVIATIONS

CADC - Common Artemis Driving Cycles

CD - Chassis Dynamometer

DPF - Diesel Particulate Filter

ECT - Engine Coolant Temperature

EGR - Exhaust Gas Recirculation

EMS - Engine Management System

LNT - Lean NOx Trap

NEDC - New European Driving Cycle

NOx - Oxides of Nitrogen

RDE - Real Driving Emissions

RW - Real World or On-road Conditions

SCR - Selective Catalytic Reduction

[T.sub.ex] - Exhaust Temperature

VSS - Vehicle Speed Sensor

WLTC - Worldwide harmonized Light duty Test Cycle

Edward Chappell, Richard Burke, and Pin Lu University of Bath

Michael Gee and Rod Williams Shell Global Solutions (UK)

Table 1. Common data readings including descriptions and recorded
ranges

Data Channel   Definition

Distance       Odometer within the EMS to
between        measure distance between DPF
regenerations  regenerations
               Engine coolant temperature (sensed
ECT            by EMS)
Engine         Measured engine speed from the
Speed          EMS.
               Estimated engine torque from the
Engine         EMS. (Non-dimensionalised relative
Torque         to maximum torque)
Post           Estimated quantity of fuel injected
injection      into the cylinder through post
quantities     injections
DPF Regen.     Timer within the EMS to indicate
indicator      period of active DPF regeneration
               Model built into the EMS to estimate
               the state of the DPF (Non-
               dimensionalised relative to maximum observed loading)
               Model or Measured exhaust gas
Tex            temperature downstream of DPF
               Indicator of engine throttle closing
Throttle       (note this refers to a physical throttle
activity       valve in the engine air path, not the
               accelerator pedal of the vehicle)
               Vehicle speed from vehicle speed
               sensor

Data Channel   Range

Distance
between        0-1000km
regenerations

ECT            -10-100[degrees]C
Engine
Speed

Engine         0-100%
Torque
Post
injection      0-20mg/st
quantities
DPF Regen.
indicator

               0-100%


Tex            0-800[degrees]C

Throttle
activity

Table 2. Summary of chassis dynamometer test campaign

                                        NEDC  WLTC

Duration (s)                            1180  1800
Distance (km)                             11    23.3
Average Vehicle Speed (km/h)              31    47
No. of Hot start tests                     7     1
No. of cold start tests (25[degrees]C)     5     1

Table 3. Data Statistics from the 314 trips recorded from on-road
driving

                          Mean  St. dev.  Min  Max

Trip Duration (s)         3680  3460      60    21150
Trip Distance (km)          44    45       0.1    208
Trip Average Vehicle        39    17       2       88
Speed (km/h)
Trip Average Ambient
Temperature ([degrees]C)

Table 4. Driver by driver breakdown of on-road data collection

                     All          Driver
                     Data     1      2     3     4

No. of Trips           314    80    115    70    49
Total Distance (km)  13700  3330   3680  4430  2260
Average Speed           39    44     30    47    37
(km/h)
Average trip
distance (km)           44    42     32    63    47

Table 5. Parameters for calculation of power spectrums using Welch's
approach, note: RW denotes Real World on-road driving

                                        Windowing
Variable  Cycle    FFT Method /  FFT    interval
                   Windowing     Order  (data
                                        points)

          NEDC                          256
          WLTC                          256
          On road                      1024
          NEDC                          128
          WLTC                          128
          On road                      1024
COPYRIGHT 2016 SAE International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Chappell, Edward; Burke, Richard; Lu, Pin; Gee, Michael; Williams, Rod
Publication:SAE International Journal of Engines
Article Type:Report
Date:Dec 1, 2016
Words:9388
Previous Article:Development of TWC and PGM Free catalyst combination as gasoline exhaust after treatment.
Next Article:Extension of analytical methods for detailed characterization of advanced combustion engine emissions.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters