Emissions Measurement
A report for the
National ITS Implementation Center
Authors:
Hesham Rakha
Charles Via Jr. Department of Civil and Environmental Engineering, Virginia Tech
3500 Transportation Research Plaza (0536)
Blacksburg, VA 24061
Phone: (540)
231-1505
Fax: (540)
231-1555
E-mail: hrakha@vt.edu
Sangjun Park
Charles Via Jr. Department of Civil and Environmental Engineering, Virginia Tech
3500 Transportation Research Plaza (0536)
Blacksburg, VA 24061
Phone: (540)
231-1505
Fax: (540)
231-1555
E-mail: sangjun@vt.edu
Linsey C. Marr
Charles Via Jr. Department of Civil and Environmental
Engineering, Virginia Tech
411 Durham Hall
Blacksburg, VA 24061
Phone: (540) 231-6071
E-mail: lmarr@vt.edu
Richard Olin
Virginia Department of
Environmental Quality, Richmond, VA 23240
Phone: (804) 698-4425
E-mail: rdolin@deq.virginia.gov
summary
Remote Sensing Devices (RSDs) are used as supplementary tools to identify high emitting vehicles (HEVs) in order to help achieve the U.S. National Ambient Air Quality Standards (NAAQS). Accordingly, we conduct two studies to enhance the procedures of HEVs detection using RSDs.
First, the
conversion of volumetric concentration to mass emissions is studied. Tailpipe emissions in grams cannot be directly measured using remote
sensing (RS) systems because they utilize a concentration-based technique. Consequently,
converting emission measurements from concentration to mass emissions is
needed. The study combines the carbon balance equation with fuel consumption estimates
to make this conversion. In estimating vehicle fuel consumption rates, the
VT-Micro model and a Vehicle Specific Power (VSP)-based model (the PERE model)
are considered and compared. The results of the comparison demonstrate that
both of the VT-Micro and PERE models provide reliable fuel consumption estimates
(R2 of 90% and higher for a 1993 Honda Accord with a 2.4L engine). The study clearly demonstrates
that the proposed procedure works well in converting concentration measurements
to mass emissions and can be applicable in the screening of HEVs and normal
emitting vehicles for several vehicle types such as sedans, station
wagons, full-size vans, mini vans, pickup trucks, and SUVs.
Second, remote sensing cut points that are sensitive to engine load conditions is proposed to enhance the HEV screening procedures using RSDs. Emission compliance is determined by comparing the concentration of pollutants measured by an RSD to on-road remote sensing (RS) emission standards. The literature shows that the current RS emission standards are insensitive to vehicle speed and acceleration levels even though vehicle emissions are highly affected by the engine load. Consequently, the study firstly demonstrates variability of vehicle exhaust emissions to motivate this study and proposes a procedure for constructing on-road remote sensing emission standards that are sensitive to vehicle speed and acceleration levels. Given the proposed remote sensing cut points, a comparison to the existing cut points is presented as well as sample tests using the proposed cut points. The results of the sample tests demonstrate that the approach is effective for HE (high-emitting) 3 and 4 vehicles of the VT-Micro emission models and require further enhancements for HE1 and 2 vehicles.
This report consists of two separate papers which follow.
SOLUTIONS FOR ENHANCING REMOTE SENSING HIGH
EMITTER VEHICLE SCREENING PROCEDURES
INTRODUCTION
To reduce air pollutant emissions to meet the National Ambient Air Quality Standards (NAAQS), many state environmental agencies are focusing their efforts on identifying high emitting vehicles (HEVs). HEVs are vehicles whose emissions of hydrocarbons (HCs), nitrogen oxides (NOx) are two and/or carbon monoxide (CO) are three times higher than the certification emissions level for the vehicle (EPA, 1999). Although HEVs comprise only a small fraction of the vehicle fleet, they contribute to a large fraction of total emissions. For example, one study found that 7.8 percent of the fleet is responsible for 50 percent of the total emissions based on a gram of CO per gallon of fuel burned (Lawson et al. 1990). Another study found that 5 percent of the vehicles emitted 80 percent of the emissions (Wolf et al. 1998).
Many of the states in the U.S. operate their own Inspection and Maintenance (I/M) Program, in order to identify and repair HEVs. In addition, other supplementary devices, such as RSDs (remote sensing devices), are used to identify HEVs. Several states are now using RSDs because they can collect on-road emission data from the in-use vehicle fleet. In contrast to some I/M tests that quantify emissions on a mass per time basis over a driving cycle that can last up to four minutes, RSDs report mole fractions, or concentrations, of pollutants in exhaust at a single point in time. The advantage of RSDs is that they are able to capture site-specific measurements under real-world conditions as vehicles are driven on-road. However, several issues remain in screening HEVs and normal emitting vehicles using RSDs, including converting from concentrations to mass emission rates and setting RSD-based standards to identify HEVs.
The objectives of this paper are to validate the use of RSD measurements to predict mass emission rates, to compare and contrast different methods for estimating fuel consumption rates, and to evaluate the accuracy with which RSDs can be used to screen HEVs using the proposed methods.
In terms of the paper layout, the paper first presents the validation of the procedure developed to estimate mass emissions. Secondly, the Physical Emission Rate Estimator (PERE) model that is based on vehicle specific power (VSP) and the VT-Micro model are compared, because these models can be used to estimate fuel consumption rates. The following section presents the mass emission estimations and a comparison of the emission estimates against field measurements. Finally, the conclusions of the study and recommendations for further research are presented.
VALIDATION
OF MASS EMISSION PROCEDURE
Conversion of Concentration
Measurements to Mass Emissions
Measurements of vehicle exhaust emissions are very important because they are used in many air-quality improvement activities such as I/M programs and the development of emission models and inventories. In practice, two test methods are widely used in quantifying vehicle exhaust emissions: mass emission tests and concentration tests. Mass emission tests directly measure the mass of several pollutants emitted from a vehicle running a simulated driving cycle. In these tests, exhaust emissions are measured in units of grams per unit time or grams per unit distance. A group of tests that are named based on the underlying drive cycle fall into this category. The Federal Test Procedure (FTP) is used to certify new vehicle emissions. Other tests used by state I/M programs include the IM240, BAR31, IM93 (CT93), and IM147 (NRC 2001).
Concentration tests measure the pollutants in vehicle exhaust emissions and report results in units of percentage or parts per million (PPM) of total exhaust volume. Idle and ASM tests fall into this category and are used in I/M programs in several states. Additionally, RSDs measure the concentrations of emissions from on-road vehicles. RSDs are considered a supplemental tool for I/M programs, due to their ability to capture on-road emissions. Consequently, several states in the U.S. are trying to improve their I/M programs using RSDs. However, in order to estimate the mass emissions per unit of time, a relationship between concentrations and mass emission rates needs to be developed.
The literature describes two approaches for developing conversion equations. The first approach is based on regression models. One thing should be addressed is that the regression models require to use of both concentration and mass emission measurements of a sample of vehicles to develop coefficients. For instance, Austin et al. (1989) proposed a new emission test procedure, the Acceleration Simulation Mode (ASM) test, that can correctly and economically identify 90% of vehicles that emit excessive nitrogen oxide (NOx) emissions for I/M programs. In the study, they concluded that the ASM 5015 test is best for identifying high NOX emitting vehicles and the 2500 rpm test could most correctly identify high CO and/or HC emitting vehicles. In addition, formulae were developed for the estimation of carbon monoxide (CO), hydrocarbon (HC), and nitrogen dioxide (NO2) emissions using regression methods. In estimating CO and HC mass emissions, the concentration of CO and HC emissions are measured from the 2500 rpm test based on the engine size and used as the regressors for CO and HC mass emissions. Engine displacement is also used as a regressor for CO and HC mass emissions. On the other hand, the NOX mass emissions are regressed from the concentration of NOX emissions measured by the ASM 5015 test and the emission test weight (vehicle weight plus 300 lbs for light duty vehicles) rather than the engine size.
DeFries et al. (2002) constructed models for simulating
Virginia IM240 emissions from concentration measurements taken from ASM 5015
and ASM 2525 test procedures, because Virginia must report emission reductions
in terms of mass emissions to the EPA. In this study, a dataset of 1702 paired
ASM and IM240 emissions were utilized for the modeling purpose. The models for the conversion
were constructed by utilizing full ASM tests, not “fast pass” ASM tests. First,
raw emission concentration measurements are corrected for dilution and humidity
effects. Using the corrected measurements, the intermediate predictor
variables, HC, CO, and NOX terms, are computed for the input
variables. Finally, the IM240 mass emissions are regressed from HC, CO, and NOX
terms, vehicle engine displacement, vehicle age, vehicle type, and a carbureted-or-fuel
injected flag. Specifically, the HC term, NOX term, engine
displacement, and vehicle age are used as regressors for IM240 HC emissions.
The model for IM240 CO emissions includes the CO term, engine displacement, and
vehicle age as the input variables. Lastly, the model for IM240 NOX
emissions utilizes the HC term, CO term, NOX term, engine
displacement, vehicle age, vehicle type, and carbureted-or-fuel injected engine.
The second approach for developing conversion equations is to use carbon balance for converting concentrations to mass emission rates per unit of fuel burned (NRC 2001). For example, Stedman, developer of the FEAT system (an RSD for on-road vehicle emissions), and his colleagues derived the equations for the conversions. Initially, they developed only one equation for CO emissions. This equation was then extended to HC and NOx emissions when the RSD system was updated to measure these pollutants (Bishop et al. 2003). In addition, Singer and Harley (1996) proposed a fuel-based methodology for computing motor vehicle emission inventories. In this study, the inventory was estimated as the product of mass-based emission factors with fuel consumption rates. In the process of calculating emission factors, the concentrations of on-road vehicle emissions are converted into mass emissions in units of grams of emissions per fuel consumed. Since the equation that they used is also based on carbon balance, it has the same structure as the equations that Stedman used. Specifically, mass emissions per fuel burned are computed by multiplying the number of moles of HC, CO, NOX emissions per fuel burned and the molecular weight of HC, CO, and NOX. In order to compute the number of moles for pollutant, the ratio of pollutant to the sum of CO2, CO, and HC is multiplied by the number of moles of carbon per fuel burned.
Data Description
The study utilizes a dataset of second-by-second IM240 emission measurements that were taken by TESTCOM since a comparison between measured emission rates and estimated emission rates can be done easily for validating a proposed procedure and for testing its effectiveness. The measurements were taken between September 2001 and April 2002. The vehicle model years ranged from 1981 to 2001, and body types included sedans, station wagons, full size vans, mini vans, pickup trucks, and sport utility vehicles.
A second-by-second IM240 emission test reports the vehicle’s
speed profile, HC, CO, and NOx emission rates as a function of time,
as illustrated in Figure 1.
The tested vehicle in Figure 1
is a 1993 Honda Accord with a 2.4L engine.
Validation Procedure
The mass emission equations that are presented in the literature were validated by first applying them to calculate pollutant concentrations from mass emission rates measured during a sample IM240 test run. The calculated concentrations were then used together with fuel properties and the rate of fuel consumption to predict mass emission rates. The fuel consumption rate was computed using the carbon balance equation, and exhaust concentrations were estimated from the mass emissions using the combustion equation. Finally, predicted mass emission rates were compared to the original mass emission rates.
All the carbon enters the engine as fuel leaves in the form of HC (g/s), CO (g/s), CO2 (g/s), and a typically negligible amount of particulate matter that will be ignored here. Given that the molecular weight of carbon and oxygen are 12 and 16 g/mole, respectively, the molecular weight of CO2 can be calculated to be 44 g/mole (12+16x2). Therefore, CO2 contains 27.3 percent (12/44) carbon. Similarly, the molecular weight of CO is 28 g/mole (12+16) yielding 42.9 percent carbon in CO. Also, according to the Code of Federal Regulations Title 40 Part 86 (40 CFR 86), HC emissions from a gasoline powered vehicle contain 86.6 percent carbon by weight. Consequently, the instantaneous carbon emission rate in units of g/s can computed as
.
Recognizing that average gasoline sold in the US contains 86.4 percent of carbon, and has a density of 738.8 g/L (or 2800 g/gallon), there are 638.31 (0.864×738.8) grams of carbon in a liter of gasoline. Consequently, the fuel consumption rate (L/s) can be computed as
.
Using the mass emissions of HC, CO, NOx, and CO2 available from IM240 test runs, the emission concentrations were computed by first estimating the mass emissions of N2 through the use of the combustion equation, which can be cast as
,
where CH1.9 represents gasoline; O2 + 3.76 N2 represents air composed of 21% O2 and 79% N2 (with argon and other non-oxygen components lumped with N2); combustion is assumed to be complete with an equivalence ratio of one; and formation of minor species such as NO and CO can be neglected relative to the amount of major species such as N2 and CO2 emitted in the exhaust.
Consequently, the mass ratio of N2 to CO2 can be computed as
.
The N2 emissions in g/s are then computed as
.
The volumetric concentrations of HC, CO, NOx, and CO2 can be computed as
,
,
, and
.
where NOx is reported as NO2.
The estimated mass emissions of HC, CO, NOx, and CO2 (HC’, CO’, NOx’, and CO2’) are then computed as
The mass emission estimates HC’, CO’, NOx’, and CO2’ for a sample vehicle (1993 Honda Accord equipped with a 2.4L engine) were found to be consistent with the field measurements, as clearly demonstrated in Figure 2. Specifically, the slope of the line ranges from 1.0011 to 1.0012 with an R2 of 1.0 for all model estimates. This exercise demonstrates that the mass emission equations that are proposed are valid and thus can be used to estimate mass emissions.
ESTIMATION OF MASS EMISSIONS
Comparison of VSP and the VT-Micro
Model Fuel Consumption Estimates
As demonstrated in the previous section, it is clear that the
accuracy of the mass emission estimates hinges on the accuracy of the fuel
consumption rates that are used to compute the mass emissions. For purposes of
this study we investigated two approaches for estimating a vehicle’s
instantaneous fuel consumption rate, namely: an approach based on the vehicle
specific power (VSP) and the use of the VT-Micro model. Each of these
approaches is described in some detail in this section.
VSP is a measure of engine load that has been proposed as a primary causal variable in emissions formation for modeling purposes and has been implemented in the Environmental Protection Agency’s Physical Emission Rate Estimator (PERE). One study also suggested an approach using VSP to estimate fuel consumption rates of on-road vehicles (Environ International Corporation 2003). However, PERE is only used to estimate fuel consumption rates in this study. PERE is meant to supplement the data driven portion of a Multi-scale mOtor Vehicle and equipment Emission System (MOVES) and fill in gaps where necessary. The model is essentially an effort to simplify, improve, and implement the Comprehensive Modal Emissions Model (CMEM) developed at the University of California, Riverside. PERE is based on the premise that for a given vehicle, (engine out) running emissions formation is dependent on the amount of fuel consumed. As such, it models the vehicle fuel rate as well as CO2 generation with some degree of accuracy. Being a physically based model, it has the potential (with some modification) to model new technologies (vehicles meeting new emissions standards), deterioration, off-road sources, I/M programs, as well as being able to easily extrapolate to areas where data are sparse.
The VSP approach to emissions characterization was flourished by Jimenez-Palacios (1999). VSP is a measure of the road load on a vehicle; it is defined as the power per unit mass to overcome road grade, rolling, and aerodynamic resistance in addition to the inertial acceleration. VSP is computed as
where v is vehicle speed (assuming no headwind) in m/s, a is the vehicle acceleration in m/s2, Є is a mass factor accounting for the rotational masses (~4%), g is the acceleration due to gravity, G is the roadway grade, Cr is rolling resistance coefficient (~0.0135), CD is aerodynamic drag coefficient, A is the frontal area, and m is vehicle mass in metric tonnes.
The equation can also have an added vehicle accessory loading term (air conditioner being the most significant) added to it. Moreover, higher order terms in rolling resistance can be added to increase the accuracy of the model (Gillespie 1992). Using typical values for coefficients, in SI units the equation and assuming CDA/m ~ 0.0005, the equation can be written as
The introduction of future technologies such as low rolling resistance tires and more aerodynamic forms can be reflected by adjusting the coefficients in the equation. It should be noted that the while it may be reasonable to assume typical values for rolling and aerodynamic resistance constants, it may pose a problem to assume a single mass for all cars (or vehicle types). There is approximately a factor of 2 difference in CDA/m between an empty compact car and a full large passenger car (J.L. Jimenez 1999). Using a single value for all LDVs (for example) can result in a significant error (in VSP) at high speeds when the aerodynamic resistance term dominates and when feed gas emissions are relatively high.
The fuel rate in L/s can be computed as
where φ is the fuel air equivalence ratio (mostly = 1), K(N) is the power independent portion of engine friction (dependent on engine speed), N(v) is the vehicle engine speed, Vd is the engine displacement volume, η is a measure of the engine efficiency (~0.4), Pacc(T,N) is the power drag of accessories such as air conditioning (AC), which is a function of the ambient temperature and the humidity level. Without an AC it is some nominal value (~1 kW), and LHV is the factor lower heating value of fuel (~11.6 kJ/L) (EPA 2003; EPA 2005).
The fuel rate is relatively insensitive to K; consequently VSP remains the primary driver of vehicle fuel consumption. The model of represents the Physical Emission Rate Estimator (PERE), which is implemented within EPA’s MOVES (EPA 2003; EPA 2005). This model was used to estimate the fuel consumption for the same sample vehicle that was described earlier. Currently, the PERE models are only implemented in spreadsheets. “PEREld.xls” was utilized since it is for light duty conventional vehicles. The parameters of a 1993 Honda Accord were input to the model and the fuel consumption estimates were compared to the in-laboratory measurements over the entire IM240 drive cycle, as illustrated in Figure 3. The figure clearly demonstrates that the PERE model tends to under-estimate the fuel consumption rate (slope of line 0.8418) with a R2 of 0.8043.
Using the fuel consumption rates that were estimated by the PERE model, the vehicle emissions were computed and compared to in-laboratory measurements, as demonstrated in Figure 4. As was the case with the fuel consumption estimates, the figure clearly demonstrates that the model tends to under-estimate vehicle emissions but has a small amount of prediction error (R2 ranges from a minimum of 90% to a maximum of 97%).
In addition to the PERE model, the VT-Micro model was tested as an alternative tool for predicting the vehicle fuel consumption rate. The VT-Micro model, unlike the PERE model, is a statistical as opposed to a physical model. The model estimates vehicle fuel consumption and emission rates using a combination of speed and acceleration levels by means of a dual-regime model as
,
where Lei,j and Mei,j represent model regression coefficients for MOE e (HC, CO, NOX, CO2, fuel) at speed exponent i and acceleration exponent j (Ahn et al. 2002; Rakha and Ahn 2004; Rakha et al. 2004). The model was developed using a sample of 101 light duty vehicles (LDVs). The data were gathered by EPA on a chassis dynamometer at the Automotive Testing Laboratories, Inc. (ATL) in Ohio, and EPA's National Vehicle and Fuels Emission Laboratory (NVREL) in Ann Arbor, Michigan in the spring of 1997. All vehicles at ATL were drafted as a stratified random sample at Inspection and Maintenance lanes utilized by the State of Ohio. The vehicles tested at the EPA laboratory were recruited randomly. All vehicles were tested under as-received condition (without repairs). Of the total 101 vehicles 62 vehicles were tested at ATL and 39 vehicles were tested at NVREL. Of the 101 vehicles, 96 vehicles had complete datasets. Furthermore, of these 96 vehicles, 60 vehicles were classified as normal vehicles. These 60 normal vehicles were grouped into homogenous groups using a Classification and Regression Tree (CART) algorithm. The CART algorithm is a data-mining technique that uses a regression tree method that automatically searches for important patterns and relationships and quickly finds hidden structures in highly complex data. Tree structured classifiers or binary tree structured classifiers are built by repeating splits at active nodes. An active node is divided into two sub-nodes based on a split criterion and a split value. The splitting process is generally continued until (a) the number of observations in a child node has met a minimum population criteria or (b) a minimum deviance criteria at a node is met, where the deviance criteria D is defined as the Sum of Squared Error (SSE) (Breiman 1984; Insightful 2001; Wolf et al. 1998).
The dependent variable (Y) was a 60-by-4 matrix that included 4 dependent variables for 60 normal vehicles. The dependent variables included HC, CO, CO2, and NOx emissions averaged over 15 drive cycles. Similarly, the independent variable (X) was a 60-by-n matrix that included a number of vehicle attributes, including the vehicle model year, engine technology, engine size, and vehicle mileage. Alternatively, the X matrix can be thought of as a set of vectors Xk, each composed of 60 elements, where k is the vehicle attribute index under consideration in the CART algorithm.
The vehicles were classified into 5 LDV and 2 LDT categories,
as demonstrated in Table
1. The Honda Accord vehicle would fit in category LDV2
because its mileage was 81,360. However LDV5 was the closest to the sample vehicle in terms of fuel consumption.
Thus the VT-Micro LDV2 and LDV5 models were utilized to estimate fuel
consumption rates.
The results of the analysis demonstrate a high degree of correlation between the estimated instantaneous fuel consumption rate and the measured rate. In addition, the LDV5 model produced closer fuel consumption rates than the LDV2 did, as demonstrated in Figure 5. The impact of the use of a moving average (MA) size 5 on the degree of correlation and errors is also demonstrated in Figure 5. The MA is used to smooth some of the peaks in the model estimates and to account for the historical effects on vehicle fuel consumption and emission rates. The estimated and smoothed VT-Micro LDV5 model fuel consumption rates in conjunction with the emission concentrations were then utilized to estimate the vehicle emissions of HC, CO, NOx, and CO2. The results clearly demonstrate a minimum systematic error (slope of line close to 1) and a high degree of correlation (R2 in excess of 90%), when using the VT-Micro LDV5 model.
Comparing Figure 3 to Figure 5, both of the VT-Micro and PERE models appear to provide reliable estimates of vehicle fuel consumption since all the R2 values are greater than 0.80. In terms of errors, the VT-Micro LDV5 model produced closer fuel consumption estimates to the measured fuel consumption rates, while the VT-Micro LDV2 and PERE models under-estimated fuel consumption rates.
Different Vehicle Type Analysis
As was demonstrated in the previous section, both of the VT-Micro and PERE approaches provided reliable estimates of vehicle fuel consumption and mass emission estimates for the sample Honda vehicle. This section expands the analysis by considering different vehicle types including station wagons, full size vans, mini vans, pickup trucks, and sport utility vehicles, as summarized in Table 2.
The classification of the sample vehicles was achieved using two methods. The first method involved selecting the vehicle category based on the vehicle parameters and matching these parameters with the CART classifications that were demonstrated earlier in Table 1. The second approach categorized vehicles based on their fuel consumption rates by comparing each sample vehicle to the VT-Micro model vehicle classifications in terms of fuel consumption rates using the IM240 test cycle. The second approach was utilized because it provided better results in terms of systematic errors and degree of correlation.
Using the
second-by-second IM240 emission measurements for the five sample vehicles a
comparison of the VT-Micro and PERE estimates was conducted, as summarized in Table 3. The results summarize the slope
and R2 of the regression line for each of the vehicle types. The
results of the analysis demonstrate that both of the
VT-Micro and PERE models provide reliable estimates of vehicle fuel consumption and emission rates, with
low systematic errors and high degrees of correlation. It is hard to determine which model is superior
than the other based on the results, although the VT-Micro models have slightly
higher R2 values. As can be seen in Table 3, the slope of the regression line
ranges from 0.75 to 1.35 and 0.73 to 1.13 for the VT-Micro and PERE models, respectively. Alternatively, the R2
ranges from 0.60 to 1.00 and 0.43 to 0.99 for the VT-Micro and PERE models, respectively.
CONCLUSIONS
The study presents a new approach for estimating vehicle mass emissions from concentration emission measurements using the carbon balance equation in conjunction with the either the VT-Micro or PERE model fuel consumption rates. The study demonstrates that the proposed approach produces reliable mass emission estimates for different vehicle types including sedans, station wagons, full size vans, mini vans, pickup trucks, and SUVs. Finally, the study demonstrates that the proposed procedure can be used to enhance current state-of-the-art HEV screening procedures using RSD technology.
As is the case with any research effort, this study demonstrates the need for further research to identify the engine load conditions that provide optimum HEV screening and to develop instantaneous load-specific cut points for identifying HEVs. Since this study provides a procedure to convert concentration measurements into mass emissions per unit time at specific engine loads, HEV screening could be enhanced by identifying the engine loads that result in largest differences between normal vehicles and HEVs. Any screening procedure can produce erroneous vehicle screening depending on the vehicle speed and acceleration levels. Consequently, further research is required to identify the engine loads that are required to minimize false alarms (erroneous identification of normal vehicles as HEVs) and detection errors (erroneous identification of an HEV as a normal vehicle).
ACKNOWLEDGEMENTS
The authors are greatly indebted to the financial support provided by the Virginia Department of Environmental Quality and the ITS Implementation Center to conduct the research presented in this paper.
REFERENCES
Ahn, K., Rakha, H., Trani, A., and Van Aerde,
M. (2002). "Estimating vehicle fuel consumption and emissions based on
instantaneous speed and acceleration levels." Journal of Transportation Engineering, 128(2), 182-190.
Austin, T. C., and Sherwood, L. (1989).
"Development of Improved Loaded-Mode Test Procedures for Inspection and
Maintenance Programs." SAE Technical Paper 891120.
Bishop, G. A., Burgard, D. A., and Stedman, D.
H. (2003). "On-Road Remote Sensing of Automobile Emissions in the La Brea
Area: Year 3, October 2003." Prepared for: Coordinating Research Council,
Inc. Contract No. E-23-4.
Breiman, L. (1984). Classification and regression trees, Wadsworth International Group,
Belmont, Calif.
DeFries, T. H., Limsakul, B., and Kishan, S.
(2002). "Models for Estimating Virginia IM240 Emissions from ASM
Measurements." Report prepared for the Virginia Department of
Environmental Quality.
Environ International Corporation. (2003).
"Analysis of EPA's Draft Plan for Emissions Modeling in MOVES and MOVES
GHG. Final report to the Coordinating Research Council, project E-68."
EPA. (2003). "Proof of Concept
Investigation for the Physical Emission Rate Estimator (PERE) to be Used in
MOVES (EPA420-R-03-005)." EPA420-R-03-005,
Ann Arbor, Michigan.
EPA. (2005). Fuel Consumption Modeling of Conventional and Advanced Technology
Vehicles in the Physical Emission Rate Estimator (PERE), Draft, EPA420-P-05-001.
Gillespie, T. (1992). "Fundamentals of
Vehicle Dynamics." Society of
Automotive Engineers.
Insightful. (2001). S-Plus 6 guide to statistics volume 1 and volume2, Insightful
Corporation, Seattle, WA.
J.L. Jimenez, P. M., G.J. McRae, D.D. Nelson
and M.S. Zahniser (1999). "Vehicle Specific Power: A Useful Parameter for
Remote Sensing and Emission Studies." Proceedings
of the 9th CRC On-Road Vehicle Emissions Workshop, San Diego CA.
Lawson, D., Groblichki, P., Stedman, D.,
Bishop, G., and Guenther, P. (1990). "Emissions from In-Use Motor Vehicles
in Los Angeles: A Pilot Study of Remote Sensing and the Inspection and
Maintenance Program." Journal of Air
and Waste Management Vol. 40: 1096-1105.
NRC. (2001). Evaluating Vehicle Emissions Inspection and Maintenance Programs,
National Academy Press, Washington DC.
Rakha, H., and Ahn, K. (2004).
"Integration modeling framework for estimating mobile source emissions."
Journal of Transportation Engineering,
130(2), 183-193.
Rakha, H., Ahn, K., and Trani, A. (2004).
"Development of VT-Micro model for estimating hot stabilized light duty
vehicle and truck emissions." Transportation
Research, Part D: Transport & Environment, 9(1), 49-74.
Singer, B. C., and Harley, R. A. (1996).
"Fuel-based motor vehicle emission inventory." Journal of the Air & Waste Management Association, 46(6),
581-593.
Wolf, J., Guensler, R., Washington, S., and
Bachman, W. (1998). "High-emitting vehicle characterization using
regression tree analysis." Transportation
Research Record(1641), 58-65.
List of Tables
Table 1 CART Algorithm Vehicle
Classification
Table 2 Specification of Tested Vehicles
Table 3 Slope and R2 of Trend Line
List
of Figures
Figure 1 Second-by-Second IM240 emission test
Figure 2
Model Validation Results
Figure 3
PERE Estimated vs. In-laboratory Measured Fuel Consumption Rates
Figure 4
PERE Estimated vs. In-laboratory Measured Emission Rates
Figure 5
VT-Micro Estimated vs. In-laboratory Measured Fuel Consumption Rates
Figure 6
VT-Micro Estimated vs. In-laboratory Measured Emission Rates
Table
1
CART Algorithm Vehicle Classification
|
Vehicle Category |
Number of Vehicles |
|
Category for Light Duty Vehicles |
|
|
LDV1: Model Year < 1990 |
6 |
|
LDV2: 1990<=Model Year<1995, Engine Size < 3.2 liters, Mileage < 83653, |
15 |
|
LDV3: Model Year >= 1995, Engine Size < 3.2 liters, Mileage < 83653, |
8 |
|
LDV4: Model Year >=1990, Engine Size < 3.2 liters, Mileage >= 83653 |
8 |
|
LDV5: Model Year >=1990, Engine Size >= 3.2 liters |
6 |
|
LDV High Emitters |
24 |
|
Category for Light Duty Trucks |
|
|
LDT1: Model Year >= 1993 |
11 |
|
LDT2: Model Year < 1993 |
6 |
|
LDT High Emitters |
12 |
|
Total Vehicles |
96 |
Table 2 Specification of Tested Vehicles
|
Vehicle |
Engine |
ETW (lb) |
Make |
Model |
Model |
Number of
Cylinders |
Odometer (mi) |
Tr. |
TRLHP |
|
Sedan |
2.4 |
3500 |
Honda |
Accord |
1993 |
4 |
81,360 |
A |
11.3 |
|
Station wagon |
1.9 |
2750 |
Ford |
Escort |
1993 |
4 |
111,471 |
A |
11.4 |
|
Full size |
5 |
4000 |
Ford |
E150 Econoline |
1988 |
8 |
169,231 |
A |
20 |
|
Minivan |
3 |
4000 |
Mazda |
MPV |
1991 |
6 |
124,733 |
A |
14.8 |
|
Pickup |
1.6 |
2750 |
Geo |
Tracker |
1991 |
4 |
5,014 |
M |
16 |
|
SUV |
4 |
4500 |
Ford |
Explorer 2-DR. |
1993 |
6 |
127,928 |
A |
16.5 |
*ETW:
Equivalent Test Weight
*Tr.:
Transmission Type (A:Auto, M: Manual)
*TRLHP:
Track Road Load Horse Power
Table 3 Slope and R2 of Trend Line
|
|
Vehicle Type |
VT-Micro Model1) |
PERE2) |
||||||||
|
HC |
CO |
NOX |
CO2 |
Fuel |
HC |
CO |
NOX |
CO2 |
Fuel |
||
|
Slope |
Sedan |
0.98 |
0.97 |
1.02 |
0.97 |
0.97 |
0.86 |
0.86 |
0.84 |
0.83 |
0.83 |
|
Station Wagon |
1.00 |
0.99 |
0.98 |
0.97 |
0.97 |
0.95 |
1.01 |
0.91 |
0.96 |
0.96 |
|
|
Full Size Van |
0.82 |
0.81 |
0.91 |
0.83 |
0.83 |
1.09 |
1.05 |
1.13 |
1.12 |
1.12 |
|
|
Mini Van |
0.85 |
0.85 |
0.84 |
0.85 |
0.85 |
0.89 |
0.86 |
0.85 |
0.87 |
0.87 |
|
|
Pickup Truck |
1.28 |
1.24 |
1.35 |
1.14 |
1.14 |
0.94 |
0.87 |
0.73 |
0.87 |
0.87 |
|
|
SUV |
0.78 |
0.81 |
0.75 |
0.76 |
0.76 |
1.03 |
0.96 |
0.91 |
0.91 |
0.91 |
|
|
R2 |
Sedan |
0.92 |
0.95 |
0.96 |
0.91 |
0.91 |
0.90 |
0.93 |
0.97 |
0.91 |
0.91 |
|
Station Wagon |
0.60 |
0.90 |
0.94 |
0.92 |
0.92 |
0.43 |
0.87 |
0.93 |
0.87 |
0.87 |
|
|
Full Size Van |
0.98 |
1.00 |
0.96 |
0.95 |
0.95 |
0.96 |
0.99 |
0.90 |
0.87 |
0.89 |
|
|
Mini Van |
0.99 |
0.99 |
0.99 |
0.97 |
0.97 |
0.93 |
0.93 |
0.94 |
0.89 |
0.89 |
|
|
Pickup Truck |
0.96 |
0.97 |
0.99 |
0.89 |
0.90 |
0.97 |
0.98 |
0.89 |
0.93 |
0.93 |
|
|
SUV |
0.97 |
0.97 |
0.98 |
0.92 |
0.92 |
0.97 |
0.97 |
0.97 |
0.89 |
0.89 |
|
1) VT-Micro Model
means that HC, CO, NOX, CO2 emissions are estimated from
the fuel rates that are estimated by using the VT-Micro models and smoothed
(Moving average 5 seconds)
2) PERE means that HC,
CO, NOX, CO2 emissions are estimated from the fuel rates
that are estimated by using the PERE and smoothed (Moving average 5 seconds)
Figure 1 Second-by-Second IM240 emission test
|
|
Figure 2 Model Validation Results
Figure 3 PERE Estimated vs. In-laboratory Measured Fuel Consumption Rates
Figure 4 Estimated Emission Rates from Fuel Rates Estimated Using PERE vs. In-laboratory Measured Emission Rates
(b)
Figure 5 VT-Micro Estimated vs. In-laboratory Measured Fuel Consumption Rates
Figure 6 Estimated Emission Rates from Fuel Rates Estimated Using VT-Micro Model LDV5 vs. In-laboratory Measured Emission Rates
Derivation of Remote Sensing Cut Points for the Screening of High-Emitting Vehicles
Introduction
There is no doubt that mobile-source emissions are a large contributor to air pollution. To achieve the U.S. National Ambient Air Quality Standards (NAAQS), state departments of environmental quality have been making a tremendous and continuous effort to reduce mobile-source emissions by identifying and repairing high-emitting (HE) vehicles. Remote sensing devices (RSDs), used to identify HE vehicles, are supplementary tools to enhance existing inspection and maintenance (I/M) programs. For example, the Virginia Department of Environmental Quality (DEQ) uses remote sensing (RS) to enhance the I/M program because the DEQ determined that RSDs are the only available technology to measure on-road vehicle emissions in a cost-effective manner. Specifically, RSDs are used to identify HE vehicles, very clean vehicles, and vehicles that are operated without an emission inspection although they are subject to it [1, 2].
One concern of utilizing RSDs that researchers have identified is that the
vehicle driving condition at the time of measurement is not known, although it
is highly related to fuel consumption and emission rates. For instance, some
vehicles running on a steep roadway grade have a high engine load, which
commands the engine to operate fuel-rich and results in large increases in CO
emission and VOC rates. Consequently, several researchers proposed a
methodology to derive engine load of a vehicle passing across with given tractive
forces and resistance forces that are calculated from the information of
roadway grades, vehicle specifications, vehicle speed, and acceleration [3].
The vehicle specific power (VSP) approach was proposed by Jimenez-Palacios,
which is practically being used to avoid emission measurements under very low
and/or high engine loads [4]. However, the state-of-practice of testing the
compliance with on-road emission standards is actually to use a constant value
insensitive to the vehicle driving condition at the time of measurement as an
emission standard for a vehicle [5, 6]. Consequently, the objective of the
study presented here is to construct on-road RS emission standards sensitive to
vehicle speed and acceleration levels to enhance the effectiveness of RS.
This paper initially
presents an overview of emission measurement techniques and the framework of the
VT-Micro emission models because
the study uses these models to simulate normal-emitting and high-emitting vehicles. The following section demonstrates variability
in emission rates reported from the IM240 tests of a homogeneous group of
vehicles that are classified using the classification and regression tree
algorithm (CART). The objective of the section is to motivate this study by
demonstrating the fact that variability in emission rates exists even though
emissions are measured from a homogeneous group of vehicles. After that, the
study details the proposed
procedure for constructing remote sensing cut points sensitive to vehicle speed
and acceleration levels. Subsequently,
a comparison
between the proposed cut points and the existing cut points is presented. In addition,
sample tests utilizing the proposed remote sensing cut points are presented. Finally, the paper presents the conclusions of
the study.
Background
This section presents a brief overview of emission measurement techniques focusing on IM240, acceleration simulation mode (ASM) because the emission standards for both of IM240 and ASM tests are used for the construction of remote sensing cut points in this study. Secondly, the background, technology, and issues associated with remote sensing are introduced to provide the basics of remote sensing and to derive the research motivation. Finally, the VT-Micro emission models are introduced since the models are used to simulate normal-emitting and HE vehicles.
Measurement Techniques for
Emission Tests
There are several emission measurement techniques in use such as the Federal Test Procedure (FTP), idle, IM240, ASM, remote sensing, and on-board diagnostics tests. Among these either the IM240 or the ASM test is being utilized as a formal emission test for inspection and maintenance (I/M) programs across the United States.
IM240
The IM240 test directly measures the mass of exhaust emissions every second over a 240-second drive cycle while the test vehicle is being driven on a dynamometer following a pre-defined driving cycle. The driving cycle is equivalent to the first 240 seconds of the bag 2 phase of the FTP test, thus the vehicle is assumed to be fully warmed before the test [3, 7]. Instantaneous emission rates in grams of fuel are recorded and composite emission rates in grams per mile for phase 2 (from 49 to 239 seconds into the drive cycle) and the entire drive cycle are reported. Given the reported emission rates, the compliance to the corresponding IM240 emission standards is tested. The IM240 final standards include seven tables for light duty vehicles, high-altitude light duty vehicles, light duty trucks 1, high-altitude light duty trucks 1, light duty trucks 2, high-altitude light duty trucks 2, and heavy duty trucks. Each table includes standards in grams per mile for phase 2 and the entire test for hydrocarbons, carbon monoxide, and oxides of nitrogen for each model year classification [8].
Acceleration
Simulation Mode (ASM)
The ASM test was developed by the California Bureau of Automotive Repair (BAR) to overcome the shortcomings that reside in the IM240 test. One shortcoming is a relatively-long test time of 240 seconds per vehicle. Another is the small number of centralized test stations available and thus causing significant queuing at the test facilities [3, 9]. The ASM test measures exhaust concentrations from the vehicle that is driven on a dynamometer under a pre-specified mode. The pre-specified mode is characterized by a speed and a vehicle load. Practically, the ASM 2525 test, 25% of the maximum vehicle load encountered on the FTP at 25 miles per hour, and ASM5015 test, 50% of the maximum vehicle load encountered on the FTP at 15 miles per hour, are given to each vehicle [3, 7]. The ASM final standards consist of three lookup tables for hydrocarbons, carbon monoxide, and oxides of nitrogen. The standards can be looked up using vehicle type, model year, and vehicle weight. The vehicle types include light duty vehicles, high-altitude light duty vehicles, light duty trucks 1, high-altitude light duty trucks 1, light duty trucks 2, and high-altitude light duty trucks 2 [10].
Remote
Sensing Emissions
The first version of remote sensing instrumentation applied in the field was developed in the late 1980’s by researchers at the University of Denver, although the idea of remote sensing measurements of emissions was proposed elsewhere [3, 11]. The first attempt to develop an instrument that measures emissions was firstly made by Lockheed Missiles and Space Corporation, but it was not reported whether the device measured on-road emissions successfully. Later, it was demonstrated that the use of a gas filter correlation radiometer enables the measurement of CO plumes exhausted from passing cars, his system did not provide the parameters necessary to estimate emission rates from the measured plumes [11].
RSDs use a technology for measuring the changes in the intensity of a light
beam due to the interruption caused by a passing vehicle’s exhaust plume. The
first generation of RSDs used an infrared beam to measure concentrations of HC
and CO. Some RSDs recently use an
ultraviolet beam to measure NO [3, 7]. The most affordable reason to use RSDs
is that they can measure on-road emissions from the in-use vehicle fleet. RSDs
are installed at locations avoiding high load conditions, and providing safety
to equipment, staff, and drivers. An installed RSD measures and records the
concentration of HC, CO, CO2 and NOX emissions of a
passing vehicle as well as its speed, acceleration, and license plate. Then the
recorded data is checked to determine if it is valid and within the acceptable
vehicle specific power (VSP) range, in order to increase the detection rate of HE
vehicles because it is difficult to identify HE vehicles under high or low VSP
conditions. A comparison of the measured emissions to the RS cut points
determines whether or not a vehicle is a high emitter. If a vehicle is
identified as a high emitter, the vehicle is subjected to an official I/M
program test such as the ASM or IM240 test for confirmation [3, 12]. Meanwhile,
RS is also used to identify very clean vehicles in terms of exhaust emissions,
which is referred to as white screening. In this application, if a vehicle is
identified as a clean vehicle, the vehicle is exempted from a scheduled official
emission test [13, 14].
The literature describes several on-road remote sensing emission thresholds
for HE vehicles. For example, the Oregon Department for Environmental Quality
(DEQ) conducted an RSD study in 2003 that defined the dirty category threshold
for HC, CO, and NOX emissions standards as 220 parts per million (ppm),
1.0%, and 1,000 ppm, respectively [5]. The state of Illinois has established
its own on-road RS emissions standards to include HC and CO emission thresholds
for each model year classification [15]. These examples show that existing
on-road remote sensing emission standards are not unified and are typically
insensitive to vehicle speed and acceleration levels even though vehicle
emissions are highly affected by engine loads.
VT-Micro
Emissions Models
The VT-Micro emission models were developed from experimentation with numerous polynomial combinations of speed and acceleration levels. Specifically, linear, quadratic, cubic, and fourth degree combinations of speed and acceleration levels were tested using chassis dynamometer data collected at the Oak Ridge National Laboratory (ORNL). The final regression model included a combination of linear, quadratic, and cubic speed and acceleration terms because it provided the least number of terms with a relatively good fit to the original data (R2 in excess of 0.92 for all measures of effectiveness [MOE]). The ORNL data consisted of nine normal-emitting vehicles including six light-duty automobiles and three light-duty trucks. These vehicles were selected in order to produce an average vehicle that was consistent with average vehicle sales in terms of engine displacement, vehicle curb weight, and vehicle type. The data collected at ORNL contained between 1,300 to 1,600 individual measurements for each vehicle and MOE combination depending on the vehicle’s envelope of operation [16].
This method has a significant advantage over emission data
collected from a few driving cycles because it is impossible to cover the
entire vehicle operational regime with only a few driving cycles. Typically,
vehicle acceleration values ranged from −1.5 to 3.7 m/s2
at increments of 0.3 m/s2 (−5 to 12 ft/s2
at 1‑ft/s2 increments). Vehicle speeds varied from 0 to
33.5 m/s (0 to 121 km/h or 0 to 110 ft/s) at in increments of
0.3 m/s [16].
Additionally, the VT-Micro model was expanded by including
data from 60 light-duty vehicles (LDVs) and trucks (LDTs). Statistical
clustering techniques were applied to group vehicles into homogenous categories
using classification and regression tree (CART) algorithms. The 60 vehicles were
classified into five LDV and two LDT categories [17]. In addition, HE vehicle emission
models were constructed using second-by-second emission data. In constructing
the models, HEVs are classified into four categories for modeling purposes. The
employed HEV categorization was based on the comprehensive modal emission model
(CMEM) categorization. The first type of HEVs has a chronically lean
fuel-to-air ratio at moderate power or transient operation, which results
in high emissions in NO. The second type has a chronically rich
fuel-to-air ratio at moderate power, which results in high emissions in CO. The third
type is high in HC and CO. The fourth type has a chronically or
transiently poor catalyst performance, which results in high emissions in HC, CO, and NO.
Each model for each category was constructed within the VT-Micro modeling
framework. The HE vehicle model was found to estimate vehicle emissions
with a margin of error of 10% when compared to in-laboratory bag measurements [18].
Variability
of Vehicle Emissions
This section demonstrates variability in emission rates due to vehicle characteristics and/or engine load conditions to motivate the study of constructing remote sensing (RS) cut points. First, the study shows that emission rates for two groups of vehicles, which are classified by model year, are distinctively different. Second, it demonstrates variability as a function of load within a feasible vehicle specific power (VSP) range, 3 to 15kw/metric ton. Specifically, the description of dataset used for the demonstration is briefly presented. The procedure for classification of vehicles is presented as well as the CART algorithm because the classification of vehicles into several homogeneous groups is conducted using the CART algorithm. Finally, the variability of vehicle emissions is assessed by illustrating 95% confidence intervals of exhaust vehicle emissions in grams per second for two groups of vehicles. In addition, the coefficient of variation (CV) of each vehicle’s emission rates at different engine load levels but within the feasible VSP range is calculated to demonstrate the variability due to engine load conditions.
Description
of Data
The study utilizes a dataset of 1,001 light duty vehicles’ IM240 test results. Each IM240 test result includes second-by-second emission rates in grams over a 240-second drive cycle and emission rates in grams per mile for the entire trip for HC, CO, and NOX as well as the corresponding second-by-second vehicle speeds and vehicle characteristics including the equivalent test weight (ETW), engine size (ES), model year (MY), odometer reading, gross vehicle weight rating (GVWR), number of cylinders (Cylinder), and type of transmission (Transmission).
Vehicle Classification
The CART algorithm is a widely used classification technique that generates output in the form of trees that are easy to interpret and can be used with nonparametric and nonlinear problems. The most appropriate feature of the CART algorithm is its ability to group vehicles into homogeneous classes based on independent variables and emission and fuel consumption rates. The classifiers identified in the analysis can be considered as critical factors affecting vehicle fuel consumption and emission rates. The CART algorithm has been successfully utilized to classify vehicles into a series of homogeneous groups [17, 19]. For instance, the regression tree analysis was used to classify HEVs into several homogenous groups in the literature [19]. In this study, the FTP bag 2 emission rates for each of HC, CO, and NOX were used as dependent variables. The independent variables included vehicle characteristics such as model year (MY), curb weight, dynamometer horsepower setting, engine displacement, transmission type, fuel delivery technology, and catalytic converter type. Model year and catalytic converter type were analyzed as critical factors. A detailed description of the CART algorithm is beyond the scope of this paper but is provided in the literature [20].
For the construction of trees, the fuel consumption, HC, CO, and NOX emission rates were used as dependent variables while the Equivalent Test Weight (ETW), Engine Size (ES), Model Year (MY), odometer reading, Gross Vehicle Weight (GVWR), number of engine cylinders, and type of transmission were used as model predictors. In developing these trees, the study used the R software, which is defined as “an integrated suite of software facilities for data manipulation, calculation and graphical display” [21]. The modeler can control the size of the tree by setting a minimum number of vehicles to include in a child node and the minimum deviance within a node. In the study, the individual tree was generated so that the deviance of a node to be split is at least 1% of that of the root node. In addition, the minimum number of vehicles in a parent and a child node was set at 60 and 30, respectively in order to ensure that each group had sufficient vehicles. A total of four trees were produced for each of the measures, namely: fuel consumption and emission rates for LDVs, as illustrated in FIGURE 7.
In terms of model predictors, the most critical vehicle
characteristics affecting vehicle exhaust emissions and fuel consumption rates
are identified as the model year of the vehicle and the engine size,
respectively, as illustrated in FIGURE 7.
However there are some differences in the classifiers from the second level downwards
depending on the selection of the dependent variables. For example, while the MY, ES, and
Odometer reading were used in constructing the trees for either HC or CO
emissions as the dependent variable, the MY and GVWR were only used in constructing the
tree for NOX
emissions as the dependent variable. Consequently, the critical vehicle characteristics affecting
vehicle exhaust emissions are MY, ES, Odometer reading, and GVWR. Also ES and number
of engine cylinders are identified as the critical characteristics affecting
fuel consumption when compared to ETW, Transmission, Odometer reading,
and MY. In
addition, the results of the analysis demonstrates it is reasonable that the
IM240 and ASM test emissions standards are developed as a function of model
year, engine size, and vehicle weight.
Illustration of Vehicle Emission Variability
First, vehicles that fall into the category of either TN (Terminal Node) 1 or TN7 for HC emissions, illustrated in FIGURE 7(a) are utilized for illustration purposes. The reason that the algorithm selects the categories of TN1 and TN7 is to demonstrate the differences in emission rates between the vehicle classified based on their model year. A total of 274 vehicles and 31 vehicles are categorized into TN1 and TN7, respectively. The average of their second-by-second emission rates in grams are plotted along with the 95% confidence interval over time, as illustrated in FIGURE 8 (a). As can be seen, two confidence intervals do not meet over time. This means that the emission rates for the two groups of vehicles are distinctively different at a 95% confidence level. Given that the engine load conditions vary over time, it is clearly demonstrated that one of the major sources of variability of vehicle emissions is engine load conditions as well as the critical vehicle characteristics affecting vehicle emissions, model year in this illustration. In order to demonstrate the variability due to engine load condition more clearly, the CV of each of the 305 vehicles’ HC emission rates at different engine load conditions is calculated. Specifically, the second-by-second emission rates over a 240-second drive cycle are checked if their engine loads are within the feasible VSP range and then the CV of the valid emission rates is calculated. Subsequently, a set of 305 CVs for the vehicles is given and the mean and median of the CVs is 1.48 and 1.26, respectively, with a standard deviation of 0.79. The minimum and maximum of the CVs is 0.47 and 4.50, respectively. The distribution of the CVs is illustrated in FIGURE 8 (b). Additionally, the CV between the vehicles is calculated to compare to that for different engine loads. Since the CV for TN1 and TN7 is 1.36 and 0.83, respectively, which is less than that for different loads. Consequently, the results show that the variability due to engine load conditions is significant even though the engine load conditions are all within the feasible VSP range. It supports strongly a need to construct RS cut points sensitive to vehicle engine load levels.
Methodology
for developing rs cut points
The methodology presented in this study to develop RS cut points for the screening of HE vehicles can be broadly divided into three sub-processes, as illustrated in Figure 9.
First, HE cut points in grams per second (g/s) were developed as a function
of a vehicle’s speed and acceleration levels using the VT-Micro emissions
models for a representative LDV and LDT. Second, the HE cut points in grams per
second were converted to concentration emissions cut points in parts per
million using the carbon balance equation. Third, scale factors were computed
using either ASM ETW- and model-year-based standards or engine-displacement-based
standards. The following section describes the proposed procedure in more
detail.
Process
1: Developing HE Speed and Acceleration Cut Points
In developing HE grams per second cut points, the VT-Micro emissions models for LDV4 and LDT1 categories were utilized to simulate normal-emitting vehicles. LDV4 includes vehicle models of 1990 and greater, with an average weight of 2,760 lbs and an average engine displacement of 2.0 liters. Alternatively, the LDT1 class includes vehicle models of 1993 and greater, with an average weight of 4,061 lbs and an average engine displacement of 2.5 liters.
First, LDV4’s and LDT1’s IM240 emissions in grams per mile were computed
using the VT-Micro model, as illustrated in Figure 10. Second, the scale factors for LDV4 and LDT1 were
calculated as the ratio of the IM240 threshold to the total drive cycle
emission rate, as summarized in Table 4. The literature details the IM240 standards [8].
Third, emission tables for the LDV4 and LDT1 vehicles were developed as a
function of speed and acceleration levels. The speed and acceleration were
varied from 0 to 120 km/h and −1.7 to 2.8 m/s2,
respectively. Finally, HE cut point tables for the LDV4 and LDT1 vehicles were
calculated by multiplying the tables by scale factors.
Process
2: Converting to Concentration-Based Cut Points
Given the mass emissions of HC, CO, NOx, and CO2 from process 1, the emission concentrations were computed by first estimating the mass emissions of N2 through the use of the combustion equation, which can be cast as
where CH1.9 represents gasoline; O2 + 3.76 N2 represents air composed of 21% O2 and 79% N2 (with argon and other non-oxygen components lumped with N2); combustion is assumed to be complete with an equivalence ratio of 1; and formation of minor species such as NO and CO can be neglected relative to the amount of major species such as N2 and CO2 emitted in the exhaust.
Consequently, the mass ratio of N2 to CO2
can be computed as
.
The N2 emissions in grams per second are then computed as
.
The volumetric concentrations of HC, CO, NOx,
and CO2 can be computed as
,
,
, and
.
Process
3: Computation of Scale Factors
The HE cut point tables as a function of vehicle speed and acceleration levels were computed assuming no differences across different vehicle model years. The next step was to adjust the HE cut points to reflect the fact that ASM standards vary depending on vehicle model year, ETW, and engine size. In the calculation of scale factors, either ASM ETW-based cut points or ASM engine-displacement-based cut points were used because the Virginia DEQ utilizes both cut points. Specifically, the ASM ETW-based cut points are used for emission tests for LDTs and model year 1995 or older LDVs. The ASM engine-displacement-based cut points are used for model year 1996 or newer LDVs. Therefore, both of the ASM standards are used to adjust the scale factors to account for vehicle model year and characteristics. The details of both calculations are presented in the following section.
ETW- and Model-Year-Based Scale Factor
First, the ASM ETW-based cut points for LDV4 and LDT1 were used as the base factor within the lookup table. Table 5 shows a portion of the ASM cut point lookup table. As can be seen in Table 5, the ASM ETW-based cut points for LDV4 are 120 ppm (HC ASM 2525), 124 ppm (HC ASM 5015), 0.67% (CO ASM 2525), 0.70% (CO ASM 5015), 894 ppm (NOX ASM 2525), and 989 ppm (NOX ASM 5015).
Second, the
ASM ETW-based scale factors were calculated by dividing the ASM ETW-based cut
point table by the cut points for the LDV4 and LDT1 vehicles. An example of the
calculation of the ASM ETW-based cut points is shown in Table 6.
As can be seen in Table 6,
the scale factors for ASM 2525 and ASM 5015 were averaged.
For the convenience of the calculation of the ASM
ETW-based scale factors, regression models were constructed for each model year
category, as illustrated in Figure 11. In developing the regression models, the
scale factors were regressed against ETW. Figure 11 illustrates a sample regression model for the LDV scale
factors. All regression models for LDVs and LDTs are summarized in Table 7.
Engine-Displacement-Based
Scale Factor
ASM engine-displacement-based standards are easily calculated as:
.
Where, Disp is engine
displacement in liters. HCcoefficient,
COcoefficient, and NOcoefficient is the
coefficient for each emission, which is available in TABLE 8. HCcutpoint and NOcutpoint are the cut points
for HC and NO emissions in ppm. COcutpoint
is the cut point for CO emissions in percent.
As can be
seen in the equations, cut points vary depending on engine displacement rather
than model year and ETW. Therefore, the engine-displacement-based scale factor
for a vehicle is calculated by dividing the engine displacement of LDV4 (2.0) by that of the vehicle to be tested.
Comparison
to the Current Standards
The Virginia DEQ developed RS cut point tables by multiplying RS cut point factors and ASM HE thresholds. These RS cut point tables are similar to the ASM HE thresholds and are insensitive to vehicle speed and acceleration levels. For example, the DEQ’s RS cut points for LDV4 described in TABLE 9 (RS column) are computed by multiplying the RS cut-point factor of 4 and the average of the ETW-based cut points for ASM 2525 and ASM 5015 (sample calculation: RS for HC emission = (120+124)/2*4 = 488) or the average of the engine-displacement-based cut points.
It should be
noted that the DEQ’s RS cut points are utilized when the VSP of LDVs and LDTs
ranges between 3 to 15 kW/metric
ton for model year 1968 to model year 1995 and from 3 to 22 kW/metric ton for model year 1996 and
newer. The reason that uses the given VSP ranges is the Virginia DEQ is utilizing
those figures. The
variation of VSP as a function of vehicle speed and acceleration levels is illustrated
in Table 10. The
shaded cells show valid VSP speed and acceleration measurements for use with
the current procedures.
Sample
Tests
This section presents sample tests using the proposed RS cut points. For the test, the VT-Micro HE vehicle emission models were used to simulate HE vehicles’ emissions as a function of speed and acceleration levels. As discussed in the overview of the VT-Micro emission models, four categories of high emitters were incorporated into the VT-Micro models: HE-1 has high NOX emissions; HE-2 has high CO emissions; HE-3 has high HC and CO emissions; and HE-4 has high HC, CO, and NOX emissions.
The
instantaneous HC, CO, and NOX emissions for high emitters were developed, varying the speed from 0
to 120 km/h in 1-km/h increments and the acceleration from −1.7 m/s2 to 2.8 m/s2 in 0.1-m/s2 increments.
Consequently, the number of the instantaneous emissions rates that were within the effective VSP range, is 1,020. At the same time, the
proposed RS cut points corresponding to the speed and acceleration levels were computed.
Finally, each instantaneous emissions rate was compared to the corresponding RS
cut point. Table 11 presents the test results, or the number of emission
rates exceeding the cut points. The results demonstrate that the
approach is effective for HE-3 and HE-4 and requires further enhancements for
HE-1 and HE-2 vehicles.
Conclusions
The study demonstrated variability in vehicle emissions as a function of speed and acceleration levels within valid VSP ranges are significant. In addition to the demonstration, model year and engine size were demonstrated to be the most critical vehicle characteristics affecting vehicle exhaust emissions.
The study proposed a procedure for constructing on-road RS emissions standards sensitive to vehicle speed and acceleration levels. The proposed procedure is broadly divided into three sub-processes. In the first process, HE cut points in grams per second are developed as a function of a vehicle’s speed and acceleration levels using the VT-Micro emissions models. Subsequently, the HE cut points in grams per second are converted to concentration emissions cut points in parts per million using the carbon balance equation. Finally, the scale factors are computed using either ASM ETW- and model-year-based standards or engine-displacement-based standards.
The study demonstrated that the use of on-road RS cut points sensitive to
speed and acceleration levels is required in order to enhance the effectiveness
of RS. In addition, the study demonstrated that the proposed standards are
effective for HE-3 and HE-4 and requires further enhancement for HE-1 and HE-2
vehicles. Consequently, further research is required to enhance the models
within the proposed framework.
Acknowledgements
The authors are greatly indebted to the financial support provided by the Virginia Department of Environmental Quality and the ITS Implementation Center in conducting the research presented in this paper.
References
[1] Virginia Department of Environmental
Quality, "Report to the General Assembly on Remote Sensing of Vehicle
Emissions in Virginia," 2003.
[2] National Research Council. (U.S.), Modeling mobile-source emissions.
Washington, D.C.: National Academy Press, 2000.
[3] T. Wenzel, B. C. Singer, and R.
Slott, "Some Issue in the Statistical Analysis of Vehicle Emissions,"
Journal of Transportation and Statistics,
2000.
[4] J. L. Jimenez, P. McClintock, G. J.
McRae, D. D. Nelson, and M. S. Zahniser, "Vehicle Specific Power: A Useful
Parameter for Remote Sensing and Emission Studies," Proceedings of the 9th CRC On-Road Vehicle Emissions Workshop, San
Diego CA., 1999.
[5] Oregon Department of Environmental
Quality, "Vehicle Emission Testing using a Remote Sensing Device (RSD):
Demonstrating the Value of Routine Vehicle Maintenance & Providing Initial
Quantification of Vehicle Emissions in Pendleton, Bend and LaGrande,
Oregon," 2003.
[6] Illinois General Assembly,
"TITLE 35: ENVIRONMENTAL PROTECTION / SUBTITLE B: AIR POLLUTION / CHAPTER
I: POLLUTION CONTROL BOARD / SUBCHAPTER K: EMISSION STANDARDS AND / LIMITATIONS
FOR MOBILE SOURCES / PART 240: MOBILE SOURCES / SUBPART G: ON-ROAD REMOTE
SENSING TEST EMISSION STANDARDS."
[7] National Research Council, Evaluating Vehicle Emissions Inspection and
Maintenance Programs: National Academy Press, Washington DC, 2001.
[8] EPA, "High-Tech I/M Test
Procedures, Emission Standard, Quality Control Requirements, and Equipment
Specifications: IM240 and Functional Evaporative Systems Tests, Revised
Technical Guidance (Draft)," 1996.
[9] T. C. Austin and L. Sherwood,
"Development of Improved Loaded-Mode Test Procedures for Inspection and
Maintenance Programs," vol. SAE Technical Paper 891120, 1989.
[10] EPA, "Acceleration Simulation Mode
Test Procedures, Emission Standards, Quality Control Requirements, and
Equipment Specifications (Final Technical Guidance)," 2004.
[11] G. A. Bishop, J. R. Starkey, A.
Ihlenfeldt, W. J. Williams, and D. H. Stedman, "IR Long-Path Photometry, A
Remote Sensing Tool For Automobile Emissions," Anal. Chem., 61: 671A-677A, 1989.
[12] EPA, "Romote Sensing: A
Supplemental Tool for Vehicle Emission Control, Fact Sheet OMS-15, EPA
420-F-92-017," 1993.
[13] EPA, "Technical Highlights - Clean
Screening in Inspection and Maintenance Program, EPA420-F-98-023," 1998.
[14] T. Wenzel and R. Sawyer, "Analysis
of a Remote Sensing Clean Screen Program in Arizona," 1998.
[15] Illnois General Assembly, "TITLE
35: ENVIRONMENTAL PROTECTION / SUBTITLE B: AIR POLLUTION / CHAPTER I: POLLUTION
CONTROL BOARD / SUBCHAPTER k: EMISSION STANDARDS AND / LIMITATIONS FOR MOBILE
SOURCES / PART 240: MOBILE SOURCES / SUBPART G: ON-ROAD REMOTE SENSING TEST
EMISSION STANDARDS."
[16] K. Ahn, H. Rakha, A. Trani, and M. Van
Aerde, "Estimating vehicle fuel consumption and emissions based on
instantaneous speed and acceleration levels," Journal of Transportation Engineering, vol. 128, pp. 182-190, 2002.
[17] H. Rakha, K. Ahn, and A. Trani,
"Development of VT-Micro model for estimating hot stabilized light duty
vehicle and truck emissions," Transportation
Research Part D-Transport and Environment, vol. 9, pp. 49-74, 2004.
[18] K. Ahn, H. Rakha, and A. Trani,
"Microframework for modeling of high-emitting vehicles," Transportation Research Record. n 1880 2004,
pp. 39-49, 2004.
[19] J. Wolf, R. Guensler, S. Washington,
and W. Bachman, "High-emitting vehicle characterization using regression
tree analysis," Transportation
Research Record, pp. 58-65, 1998.
[20] L. Breiman, Classification and regression trees. Belmont, Calif.: Wadsworth
International Group, 1984.
[21] W. N. Venables, D. M. Smith, and t. R.
D. C. Team, An Introduction to R
Notes on R: A Programming Environment for Data Analysis and Graphics
Version 2.3.0 (2006-04-24), 2006.
List of Tables
TABLE 1 Calculation of Scale Factors
TABLE 2 ASM
Cut Point Lookup Table for Model Year 1991 to 1995 LDVs
TABLE 3
Calculation of the ASM ETW-Based Scale Factors for Model Year 1991 to
1995 LDVs
TABLE 4
Regression Models for the ASM ETW-Based Scale Factors
TABLE 5 ASM Engine Displacement Based Cut
Point Coefficient
TABLE 6 Virginia
DEQ’s RS Cut Point for LDV4
TABLE 7 VSP as
a Function of Speed and Acceleration Levels
TABLE 8 Sample
Test Result: Number of Emissions Rates Exceeding the Cut Points
List
of Figures
FIGURE 2 Illustration of Variability of Vehicle Emissions
FIGURE 3
Schematic for RS cut point table.
FIGURE 4 Flow
chart of process 1.
FIGURE 5
Illustration of regression models (scale factors for LDV CO emissions).
TABLE 4
Calculation of Scale Factors
|
|
HC (grams/mi) |
CO (grams/mi) |
NO (grams/mi) |
||||||
|
IM240 Emissions |
IM240 Standard |
Scale Factor |
IM240 Emissions |
IM240 Standard |
Scale Factor |
IM240 Emissions |
IM240 Standard |
Scale Factor |
|
|
LDV4 |
0.26 |
0.80 |
3.13 |
4.47 |
15.00 |
3.36 |
0.51 |
2.00 |
3.92 |
|
LDT1 |
0.11 |
1.60 |
14.64 |
2.37 |
40.00 |
16.91 |
0.51 |
2.50 |
4.90 |
TABLE 5 ASM Cut Point Lookup Table for Model Year
1991 to 1995 LDVs
|
ETW |
HC (ppm) 2525 |
HC (ppm) 5015 |
CO (%) 2525 |
CO (%) 5015 |
NO (ppm) 2525 |
NO (ppm) 5015 |
|
1,750 |
176 |
183 |
1.00 |
1.03 |
1,369 |
1,516 |
|
1,875 |
167 |
173 |
0.95 |
0.97 |
1,289 |
1,428 |
|
2,000 |
159 |
164 |
0.89 |
0.92 |
1,217 |
1,347 |
|
2,125 |
150 |
156 |
0.85 |
0.88 |
1,150 |
1,273 |
|
2,250 |
143 |
149 |
0.81 |
0.83 |
1,089 |
1,205 |
|
2,375 |
137 |
141 |
0.77 |
0.79 |
1,034 |
1,144 |
|
2,500 |
131 |
136 |
0.74 |
0.76 |
983 |
1,087 |
|
2,625 |
125 |
130 |
0.70 |
0.73 |
936 |
1,035 |
|
2,750 |
120 |
124 |
0.67 |
0.70 |
894 |
989 |
|
2,760 |
120 |
124 |
0.67 |
0.70 |
894 |
989 |
|
2,875 |
115 |
119 |
0.65 |
0.67 |
855 |
945 |
|
3,000 |
111 |
115 |
0.62 |
0.65 |
820 |
907 |
|
|
|
|
|
|
|
|
|
7,125 |
59 |
61 |
0.33 |
0.33 |
381 |
419 |
|
7,250 |
59 |
61 |
0.33 |
0.33 |
381 |
419 |
|
7,375 |
59 |
61 |
0.33 |
0.33 |
381 |
419 |
|
7,500 |
59 |
61 |
0.33 |
0.33 |
381 |
419 |
TABLE 6 Calculation of the ASM ETW-Based Scale
Factors for Model Year 1991 to 1995 LDVs
|
ETW |
HC (ppm) 2525 |
HC (ppm) 5015 |
HC (ppm) Mean |
CO (%) 2525 |
CO (%) 5015 |
CO (%) Mean |
NO (ppm) 2525 |
NO (ppm) 5015 |
NO (ppm) Mean |
|
1,750 |
1.47 |
1.48 |
1.47 |
1.49 |
1.47 |
1.48 |
1.53 |
1.53 |
1.53 |
|
1,875 |
1.39 |
1.40 |
1.39 |
1.42 |
1.39 |
1.40 |
1.44 |
1.44 |
1.44 |
|
2,000 |
1.33 |
1.32 |
1.32 |
1.33 |
1.31 |
1.32 |
1.36 |
1.36 |
1.36 |
|
2,125 |
1.25 |
1.26 |
1.25 |
1.27 |
1.26 |
1.26 |
1.29 |
1.29 |
1.29 |
|
2,250 |
1.19 |
1.20 |
1.20 |
1.21 |
1.19 |
1.20 |
1.22 |
1.22 |
1.22 |
|
2,375 |
1.14 |
1.14 |
1.14 |
1.15 |
1.13 |
1.14 |
1.16 |
1.16 |
1.16 |
|
2,500 |
1.09 |
1.10 |
1.09 |
1.10 |
1.09 |
1.10 |
1.10 |
1.10 |
1.10 |
|
2,625 |
1.04 |
1.05 |
1.05 |
1.04 |
1.04 |
1.04 |
1.05 |
1.05 |
1.05 |
|
2,750 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
|
2,760 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
|
2,875 |
0.96 |
0.96 |
0.96 |
0.97 |
0.96 |
0.96 |
0.96 |
0.96 |
0.96 |
|
3,000 |
0.93 |
0.93 |
0.93 |
0.93 |
0.93 |
0.93 |
0.92 |
0.92 |
0.92 |
|
|
|
|
|
|
|
|
|
|
|
|
7,125 |
0.49 |
0.49 |
0.49 |
0.49 |
0.47 |
0.48 |
0.43 |
0.42 |
0.42 |
|
7,250 |
0.49 |
0.49 |
0.49 |
0.49 |
0.47 |
0.48 |
0.43 |
0.42 |
0.42 |
|
7,375 |
0.49 |
0.49 |
0.49 |
0.49 |
0.47 |
0.48 |
0.43 |
0.42 |
0.42 |
|
7,500 |
0.49 |
0.49 |
0.49 |
0.49 |
0.47 |
0.48 |
0.43 |
0.42 |
0.42 |
TABLE 7 Regression Models for the ASM ETW-Based Scale
Factors
|
|
Model Year |
Regression
Model |
R2 |
P-value |
|||
|
Model |
Coefficient |
||||||
|
Intercept |
Linear |
Quadratic |
|||||
|
LDV |
|
|
|
|
|
|
|
|
HC |
1996 |
Y = 387.64*ETW-0.7846 |
0.9953 |
6.1E-54 |
4.7E-52 |
6.1E-54 |
|
|
|
1991 to 1995 |
Y = 497.09*ETW-0.7835 |
0.9957 |
5.8E-55 |
6.8E-54 |
5.8E-55 |
|
|
|
1981 to 1990 |
Y = 766.51*ETW-0.8181 |
0.9962 |
3.9E-56 |
1.5E-55 |
3.9E-56 |
|
|
CO |
1996 |
Y = 468.3*ETW-0.8089 |
0.9969 |
3.6E-58 |
2.7E-56 |
3.6E-58 |
|
|
|
1991 to 1995 |
Y = 573.79*ETW-0.8019 |
0.9956 |
1.4E-54 |
1.7E-53 |
1.4E-54 |
|
|
|
1983 to 1990 |
Y = 870.71*ETW-0.8281 |
0.9964 |
1.3E-56 |
3.7E-56 |
1.3E-56 |
|
|
|
1981 to 1982 |
Y = 2,498.1*ETW-0.8775 |
0.9971 |
6.6E-59 |
3.9E-60 |
6.6E-59 |
|
|
NOX |
1996 |
Y = 1,244.9*ETW-0.9283 |
0.9973 |
2.5E-59 |
1.2E-57 |
2.5E-59 |
|
|
|
1991 to 1995 |
Y = 1,555.3*ETW-0.9282 |
0.9973 |
1.9E-59 |
2.3E-58 |
1.9E-59 |
|
|
|
1981 to 1990 |
Y = 1,994.2*ETW-0.9404 |
0.9974 |
5.5E-60 |
2.7E-59 |
||