Environ. Sci. Technol. 2006, 40, 4739-4745
Evaluation of the Dekati Mass Monitor for the Measurement of Exhaust Particle Mass Emissions ATHANASIOS MAMAKOS, LEONIDAS NTZIACHRISTOS,* AND ZISSIS SAMARAS Laboratory of Applied Thermodynamics, Mechanical Engineering Department, Aristotle University Thessaloniki, P.O. Box 458, GR-54124 Thessaloniki, Greece
The Dekati mass monitor (DMM) is an instrument which measures the mass concentration of airborne particles in real time by combining aerodynamic and mobility size particle classification. In this study, we evaluate the performance of the DMM by sampling exhaust from five engines and vehicles of different technologies in both steadystate and transient tests. DMM results are found higher than the filter-based particulate matter (PM) by 39 ( 24% (range stands for ( one standard deviation) for 62 diesel tests conducted in total and 3% and 14% higher, respectively, in two gasoline tests. To explore whether the difference occurs because of the different measurement principles of DMM and filter-based PM, the DMM operation is replicated over steady-state tests by combining an electrical low-pressure impactor (ELPI) and a scanning mobility particle sizer (SMPS). The correlation of ELPI and SMPS derived mass and filter-based PM is satisfactory (R2 ) 0.95) with a mean deviation of 5 ( 15%. For the same tests, the correlation of DMM with PM was also high (R2 ) 0.95), but DMM exceeded PM by 44 ( 23% on average. The comparison of ELPI and SMPS and DMM results reveals that the latter overestimates both the geometric mean diameter and especially the width of the particle massweighted size distribution. These findings demonstrate that the statistically significant difference between the DMM and the filter-based PM cannot just originate from the different measurement principles but also from the actual implementation of the combined aerodynamic-mobility measurement in the DMM. Optimizing the DMM will require changes in its design and/or the calculation algorithm to improve the resolution and width of the aerodynamic size distribution recorded.
Introduction Diesel particulate emissions are traditionally regulated in terms of mass with a procedure that requires the gravimetric determination of particulate matter (PM) by means of filter collection. While this method has the advantage of being traceable and reproducible for regulatory purposes, a single value is obtained over the whole test cycle. This is a drawback in engine research, where time-resolved information is required to identify high emission events during transient operation. * Corresponding author phone: +30 23 10 99 62 02; fax: +30 23 10 99 60 19; e-mail:
[email protected]. 10.1021/es052302c CCC: $33.50 Published on Web 06/21/2006
2006 American Chemical Society
Alternative quasi real-time techniques for mass determination include the tapered element oscillating microbalance (TEOM, Rupprecht and Patashnick) and the quartz crystal microbalance (QCM). Both instruments measure the reduction in resonant frequency of an oscillating element (tapered hollow tube in TEOM, quartz crystal in QCM) as aerosol particles accumulate on it. The application of TEOM and QCM in diesel research studies revealed several deficiencies. Dickens et al. (1) tested a Booker Systems’ QCM and found its response to be dependent on the total mass loading. Witze et al. (2) found TEOM recordings to be sensitive on pressure and temperature fluctuations, typically encountered in vehicle exhaust. Similar findings were observed by Khalek (3) for a second generation QCM (by Sensors). More importantly, the operation of both TEOM and QCM was found to be strongly affected by the adsorption/desorption of water and organic vapors onto the collected material (1-4). The contribution of such artifacts becomes more significant in modern diesel particle filter (DPF) equipped engines, where PM emissions are dominated by semivolatile species. Evidently, the gravimetric filter method is long recognized to suffer from similar phenomena, hence the recent attempts of the regulatory authorities to remedy the situation, primarily by keeping the filter and the sample flow within a narrow temperature range (47 ( 5 °C) (5). Techniques which determine particle mass without suffering from adsorption/desorption effects include the aerosol particle mass analyzer (APM) by Kanomax (6) and the mass monitor (DMM) by Dekati, Finland. The former classifies individual aerosol particles according to their mass-to-charge ratio. McMurry et al. (7) and Park et al. (8, 9) successfully combined an APM, a differential mobility analyzer (DMA), and a condensation particle counter (CPC) to measure the mass and effective density of atmospheric, diesel exhaust, and laboratory-generated particles. However, this technique requires long sampling times and cannot be used for transient engine testing. On the other hand, DMM can record the mass distribution of airborne particles in quasi real-time, with a combined aerodynamic and mobility particle size distribution measurement. Lehmann et al. (10) first employed a prototype DMM for particle exhaust measurements from two heavyduty engines and reported fast response times, good repeatability, and relatively consistent results with PM (∼20% overestimation). Khalek (3) and Kittelson et al. (11) applied a DMM in their comparative studies and also found a satisfactory correlation with filter-based PM. In this study, a DMM was employed to measure exhaust particles from five engines and vehicles, including a small gasoline one, to further assess its operation. Exhaust particle mass measurements were in parallel conducted using the conventional filter collection, the DMM, and a combination of an electrical low-pressure impactor (ELPI) and a scanning mobility particle sizer (SMPS) to replicate the DMM operation principle.
DMM Operation Principle DMM samples exhaust aerosol at a flowrate of 10 lpm, downstream of an inlet preseparator with an aerodynamic diameter (AD) cutpoint at 1.3 µm. The sample then enters a triode corona charger where a positive ion flux charges particles. A weak electric field downstream of the charger deflects particles having high electrical mobility onto an electrometer (mobility electrode), which measures the current produced by the deflected charged aerosol stream (Imob). Particles with lower electrical mobility exit the electric field and are aerodynamically classified in a six-stage cascade VOL. 40, NO. 15, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4739
FIGURE 1. Typical measurement setup applied for diesel exhaust sampling. impactor. Each impactor stage is connected to an electrometer that measures the current produced as particles release their charge. Assuming that no particles escape the impactor, the total current measured in the six impactor stages is associated with the underlying particle number distribution according to
∫
Iimp )
∞
0
(
)
dN(db) ddb ) ddb
Pn(db)‚e‚Q‚
∫
∞
0
( )
dI(db) ddb (1) ddb
where e is the electron charge (1.602 × 10-19 C), Q is the sample flowrate, and N(db) is the number concentration of particles having mobility diameter (MD) db. The product Pn accounts for the fraction (P) of particles with diameter db that exit the mobility electrode with n elementary charges per particle. This product characterizes the charger efficiency and can be derived by calibration (12). In the case of DMM, Pn is given by Lehmann et al. (10):
Pn )
{
577.62db2.294 db e 0.068µm 30.80db1.204 db > 0.068µm
(2)
The current measured in the impactor stage i (i ) 1, 2, ..., 6 with d50% decreasing with increasing index) depends on the collection efficiency (E) of this and the preceding stages (1 to i - 1), since the distribution of airborne particles continuously alters as the sample passes through each stage:
Ii )
∫
∞
0
[
]
dN(db)
Pn(db)‚e‚Q‚
ddb
i-1
‚{Ei(Feff(db),db)‚
∏[1 - E
j
j)1
(Feff(db),db)]}‚ddb (3)
Equation 3 makes use of Feff because particle charging is a function of MD, while particle impaction is a function of AD (da). These two expressions are associated via
da2‚Cc(da)‚F0 ) db2‚Cc(db)‚Feff(db) 4740
9
(4)
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 15, 2006
where F0 is the unit density (1 g/cm3) and Cc(da) and Cc(db) are the slip correction factors evaluated for the AD and MD, respectively. The DMM calculation algorithm assumes a log-normal size distribution and calculates in real time the effective density profile that gives the best agreement between the AD and MD size distributions derived by the impactor stages and the mobility electrode respectively (13). The particle mass is then determined by combining the volume distribution (based on MD) and the effective density of airborne particles.
Experimental Section Measurement Setup. The DMM employed in this study was specifically provided by the manufacturer for this evaluation, which was conducted using the setup shown in Figure 1. Sampling was performed directly from raw exhaust in case of engine (steady-state) tests, while in vehicle (mostly transient) tests sampling was performed at primarily diluted exhaust from the constant volume sampler (CVS). Exhaust particles in both cases were sampled with a Dekati fine particle sampler (FPS-4000) operating at a nominal dilution ratio of 12:1, and the sample was further diluted with calibrated ejector-type dilutors (14), using conditioned air at ambient temperature. To investigate the operation of DMM with solid particles, a thermodenuder (TD) operating at 250 °C was employed in some tests to remove volatile and semivolatile exhaust components. The particulate mass (PM) concentration was determined gravimetrically, using PTFE-coated glass-fiber filters. Engine exhaust samples were collected on Pallflex 70 mm TX40H120WW filters at a flowrate of 70 lpm. The filters were placed in a holder located downstream of the FPS which was kept below 52 °C (35-45 °C) in all measurements, as required by the regulations. Vehicle exhaust samples were collected on Pallflex 47 mm T60A20 filters at a flowrate of 15 lpm, following the CVS dilution. The exact flowrates in each case were monitored on-line with a TSI-4040 hot-wire flowmeter. Aerosol samples were also collected with a TSI 3936L SMPS, which provided the number-weighted size distribution
TABLE 1. Information on the Vehicles and Engines Employed in the Study emission std engines vehicles
a
Euro 2 Euro 3 Euro 2 Euro 3 Euro 3
engine principle diesel diesel diesel diesel gasoline
make
model
capacity (lt)
fuelling system
aftertreatment
VW PSA VW Renault Daewoo
TDI TW12TED4 Golf Laguna Matiz
1.9 2.2 1.9 1.9 0.8
rotary pump common rail rotary pump common rail port-fuel injection
no no noa oxidn precatala three way catal
The two diesel vehicles were originally equipped with oxidation catalysts which were removed in the particular measurements.
FIGURE 2. DMM and PM measurements over transient driving cycles and steady-state tests for two diesel vehicles. in the range 7.64-289 nm MD (nominal sheath/monodisperse flowrates, 10/1 lpm; upscan time, 90 s; retrace scan time, 15 s; inlet impactor, 0.0508 cm). A TSI 3010 CPC was used over transient tests to monitor the total particle number concentration. An ELPI was implemented to measure the AD number size distribution in real time. The ELPI operated with dry (not oil-soaked) sintered plates and a filter stage that extended the lower cutpoint to ∼7 nm MD (15). DMM, ELPI, and SMPS sampled aerosol from the same location. Test Procedures. The DMM operation was evaluated on two light-duty diesel engines, two diesel passenger cars, and one gasoline car (Table 1). All measurements were conducted using fuels (diesel or gasoline) with a sulfur content of 50 mg/kg. The engines were tested on an engine dynamometer over several steady-state operation conditions to obtain different particle mass concentrations and size distributions. The DMM transient response was evaluated by running the cars on a chassis dynamometer over driving cycles, including the New European driving cycle (NEDC), the urban part of the NEDC (UDC), and three real world driving cycles developed in the framework of the Artemis project (16). These cycles correspond to urban, rural, and motorway driving conditions (average speeds of 17.5, 57.5, and 99.5 km/h, respectively). The Euro 3 diesel vehicle was also tested at steady speeds of 50 and 120 km/h under road load. At least two repetitions were conducted for each operating condition or test cycle employed.
Results and Discussion Comparison of the DMM with the Gravimetric Measurement. Figure 2 compares the filter-based PM and DMM measurements for all diesel vehicle tests. On average, DMM was 38 ( 10% (mean deviation ( standard deviation) higher than the filter-based PM. This difference was only slightly affected
FIGURE 3. DMM and PM measurements on steady-state tests of two diesel engines: (a) PM-filter and DMM sample both solid and volatile particles; (b) DMM samples downstream of a thermodenuder while the filter samples both solid and volatile particles. by the individual vehicles and test cycles, as indicated by the relatively good correlation between PM and DMM (R2 ) 0.93). The comparison of the steady-state engine tests is given in Figure 3a. Again, DMM was consistently higher than the filterbased PM (except for one measurement), by 42 ( 34% on average. Also, the correlation of DMM and PM was somehow lower (R2 ) 0.68) than the transient vehicle tests. It will be shown in a following section that the correlation of DMM and PM depends on the shape of the underlying size distribution. Hence, the averaging effect of a vehicle transient VOL. 40, NO. 15, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4741
FIGURE 4. DMM and PM measurements over transient tests for the gasoline vehicle. The percentage figures correspond to the deviation of DMM from the distance-averaged filter-based PM. driving cycle on particle size distribution, as opposed to steady-state engine tests, could explain the better agreement between cycle-averaged DMM and PM results. Lehmann et al. (10) also identified a discrepancy between DMM and PM and attributed this solely to desorption of volatile material in the filter measurement. To test this assumption, a TD was placed upstream of the DMM in some tests to remove volatile and semivolatile species. Hence, DMM measured only the mass of solid particles, while the filter continued to sample both solid and nonsolid PM. In this case, one would expect that PM should exceed DMM because of the contribution of volatiles to the total PM. Additionally, 22% of solid particles by number (17) are lost in the TD, which should further reduce the DMM signal by some amount. Indeed, DMM gave lower concentrations by 17 ( 14% on average for the Euro 2 engine tests (Figure 3b). When testing the more recent Euro 3 engine though, DMM continued to exceed filter measurement by 36 ( 20%. This clearly shows that the higher DMM measurement cannot be attributed to volatile species desorption from the filters. Last, Figure 4 compares DMM and PM for the gasoline vehicle tests. In this case, a single PM filter was used over several driving cycles, to collect sufficient PM mass for filter weighing. The mass emission rate determined with the DMM for each test cycle, and the distance-averaged DMM demonstrate a reasonable agreement in both measuring days; the DMM slightly exceeded the filter measurement by 3 and 14%, respectively. The results for diesel exhaust are consistent with similar research works, where DMM has been consistently found to overstate PM mass. Lehman et al. (10) found DMM mass concentrations to be about 20% higher than the filter-based ones for conventional diesel exhaust. Similarly, Kittelson et al. (11) established a correlation between DMM and filterbased PM with a slope of 1.35, when condensates were removed downstream of a catalytic stripper. DMM was only found to understate the filter-PM measurement when the latter was associated with a high volatile fraction, which obviously does not contribute to airborne particle mass. Specifically, Kittelson et al. (11) reported that the slope decreased to 0.98 in the case of untreated diesel exhaust, which frequently exhibited a distinct nucleation mode (evidence of high concentration of volatile material), while Lehmann et al. (10) reported DMM to significantly understate PM levels in the exhaust of a DPF-equipped heavy duty engine. Similarly, Khalek (3) found DMM mass concentrations to be about 20% lower than filter-based PM for a diesel engine with low overall emissions (70% of the US2007 standards) and a high organic fraction content (∼45% over hot FTP). 4742
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 15, 2006
FIGURE 5. Comparison of measured and calculated electrical currents for the DMM and the ELPI, based on the steady-state Euro 3 diesel engine data. Assessment of DMM Operation. The discrepancy between the DMM and filter-based PM may either have to do with the different measurement principles involved in each case, i.e., gravimetric vs combination of aerodynamic-mobility size distribution measurement, or with the implementation of the latter in the DMM, or with erroneous operation of the particular instrument used in the comparison. To identify the actual reasons of the discrepancy, we first examined the operation of the DMM with regard to its charging and collection mechanisms to check whether the instrument operates according to the manufacturer calibration. Second, the transformation of current to mobility size distribution was examined to check whether one of the primary assumptions in data interpretation is realistic. Finally, the DMM operation principle was replicated by combining the ELPI and SMPS signals to identify whether the discrepancy is bound to the difference in the measurement principles or to the particular implementation of the aerodynamic-mobility measurement by DMM. Charging and Collection Mechanisms. The performance of the charging and collection mechanisms of the particular DMM utilized in this study were evaluated by comparing the current recorded to that expected for the underlying particle population, calculated by means of eq 1 with the SMPS number-weighted size distribution (calculated current). The same approach was applied to examine the operation of ELPI, which utilizes a similar diode-type corona charger with the efficiency given by Marjama¨ki et al. (12). Figure 5 compares the calculated and measured currents from DMM and ELPI, corresponding to the Euro 3 engine as an example. There was a remarkable agreement between the two instruments in interpreting the SMPS distributions to measured current. The same picture was obtained for all engines and vehicles tested (not included to improve clarity). The similarity in the response of the two instruments verifies that they perform according to their manufacturer calibration. Figure 5, though, shows that there is a deviation between the calculated and the measured currents, which was 26 ( 16% for ELPI and 29 ( 17% for DMM. This could partially be attributed to the limited SMPS measuring range (
