A Source Dilution Sampling System for Characterization of Engine Emissions under Transient or Steady-State Operation

Abstract

With increasingly stringent regulations being placed upon engine emissions, an in-depth understanding of engine exhaust composition will be necessary to assess environmental and health impacts and to develop better methods of reducing emission levels. Source sampling systems have commonly utilized partial sampling techniques, which may introduce the potential for particle loss, and steady-state operation, which is incapable of imitating transient engine conditions. In order to simulate atmospheric dilution and aging conditions, maintain proportional sampling throughout temperature excursions during transient operation, and minimize particle loss, while representatively taking multiple samples from the exhaust flow, a full-partial-full source dilution sampling system was developed. The system consists of a critical flow venturi-constant volume system for primary dilution, a thermophoresis-resistant secondary micro-diluter, a residence time chamber, isokinetic sampling probes, multiple sampling trains, and control systems. Further, the system was designed to prevent particle loss and secondary reactions by reducing wall effects, implementing inert materials, optimizing smooth flow transition, and minimizing electrostatic forces. An examination of the system was conducted by performing a tracer study and by determining the variations in flow rates, gaseous concentrations, and PM mass measurements. Further, a statistical analysis was performed to study the capability and performance of the sampling system. The results suggest that the system was capable of collecting similar samples among the sampling trains and displaying sufficient agreement with a certification system. Therefore, the source dilution sampling system developed in this study is suitable for collecting representative samples from an exhaust flow under transient or steady-state conditions.

INTRODUCTION

Regulatory agencies worldwide have been lowering the allowable levels of engine emissions. To date, regulations have focused mainly on criteria pollutants such as particulate matter (PM) and nitrogen oxides (NO_x). As regulated emission levels continue to decrease, a new systematic approach, that may combine the optimization of internal combustion processes with the integration of aftertreatment systems, will be required (Liu et al. 2003). It is known that diesel fuel consists of large amounts of aliphatic hydrocarbons that contain 10–25 carbon atoms (Schauer et al. 1999). These components and their thermally altered byproducts, which can be created by either in-cylinder combustion or aftertreatment regeneration, may be found in engine-out emissions. Although refined lubricating oils contain only trace amounts of n-alkanes, Rogge et al. (1993) attributed the emission of heavy molecular n-alkanes in diesel exhaust to thermal breakdown of n-alkyl hydrocarbons at temperatures greater than 250°C. Some of these compounds were placed on the U.S. EPA's list of Mobile Source Air Toxics (MSAT) (Code of Federal Regulations 2001).

Chemical composition of PM and volatile and semi-volatile organic compounds emitted from diesel engines has been analyzed in the past. Hildemann et al. (1989) developed a partial-full-partial dilution stack sampler to collect organic aerosol emissions from combustion sources. Using a monodisperse aerosol generator, it was determined that their source dilution system had a 93% collection efficiency for particles with a 1 μ m diameter and an 85% collection efficiency for particles with a 2 μ m diameter. However, the ultra-fine particle loss due to thermophoretic and electrostatic forces was not investigated. The sampling system was then used to collect primary fine organic aerosol emissions from urban sources (Hildemann et al. 1991). Schauer et al. (1999) modified the sampling system to allow single compound quantification of gases and semi-volatile organic compounds and then measured C₁ through C₃₀ organic compounds from medium-duty diesel trucks. The sampling system was further augmented by Kleeman et al. (2000) to measure the size-resolved chemical composition of fine particle emissions from motor vehicles. Further, Kweon et al. (2002) designed a similar source sampling system to analyze detailed chemical composition and particle size distribution from a single-cylinder diesel engine under steady-state operation. The system was later used to determine the effect of engine operating conditions on particle-phase organic compound emissions (Kweon et al. 2003). Many other researchers including Lowenthal et al. (1994), Westerholm et al. (1991), and Wilson et al. (1995) also examined specific volatile and semi-volatile gas-phase polycyclic aromatic hydrocarbons (PAH) from diesel emissions using source sampling techniques.

In this study, an improved source dilution sampling (SDS) system has been developed to include a critical flow venturi-constant volume system (CFV-CVS) tunnel, a secondary micro-diluter, a residence time chamber (RTC), isokinetic sampling probes, multiple sampling trains, and control systems. The system was designed to simulate atmospheric dilution and aging conditions, maintain proportional sampling throughout temperature excursions during transient operation, and minimize particle loss, while representatively taking multiple samples from the exhaust flow to gain a more comprehensive understanding of engine emissions. The system was validated using statistical methods to analyze the variations in flow rates, gaseous concentrations, and PM mass measurements among the sampling trains. The system capability and performance were also statistically evaluated by comparing the PM mass measurements from the SDS system to a regulatory certification system.

SOURCE DILUTION SAMPLING SYSTEM

A SDS system was designed to have the ability to operate under transient or steady-state conditions, simulate atmospheric dilution and aging, and maintain proportional sampling during changes in temperature. In order to ensure representative sampling from engine emissions, the system needed to prevent particle loss and secondary reactions by reducing wall effects, implementing inert materials, optimizing smooth flow transition, and minimizing thermophoretic and electrostatic forces. Further, the system needed to allow ample time in order for the volatiles and semi-volatiles to nucleate, condense, and coagulate. Figure 1 illustrates the process for dilution and collection of the engine exhaust samples. The full exhaust flow was piped into the CFV-CVS tunnel where it was mixed with filtered, conditioned dilution air. The CFV-CVS tunnel combined primary dilution and aging in one stage to allow particles to grow into the size range encountered in the atmosphere (Kittelson et al. 2002). A partial sample of the diluted engine exhaust was isokinetically extracted into the secondary micro-diluter where more dilution air was added. The diluted sample was fully introduced into the RTC where the individual samples were taken through isokinetic sampling probes. The samples were collected with filters, tubes, cartridges, or adsorption materials to be analyzed for chemical composition.

FIG. 1 Diagram of source dilution sampling system and engine bench setup.

Full-Partial-Full Sampling Method

When partial sampling is performed, the possibility exists for particle loss and size distribution change due to aspiration, particle bounce, or transmission (Hangal and Willeke 1990). For these reasons, a SDS system should implement a full sampling method whenever possible. Previous systems, such as the system described by Hildemann et al. (1989), have used a partial-full-partial sampling method in which a partial sample was extracted from the exhaust flow and transported to a dilution tunnel where the full sample was mixed with the dilution air. Once the sample had passed through the dilution tunnel a partial sample was withdrawn into a RTC. The SDS system described in the present study, however, utilized a full-partial-full sampling method in order to reduce the effects of partial sampling. It is believed that performing fewer partial sample extractions would provide better collection efficiency.

CFV-CVS Tunnel

The use of the CFV-CVS primary dilution tunnel allowed for the engine to be operated under transient, in addition to steady-state, conditions by maintaining proportional sampling during temperature excursions. Dilution air entering the tunnel was filtered with pre-, activated carbon, and HEPA filters to ensure that the samples collected are free from particulate and gaseous contamination (Liu et al. 2003, 2007). The turbulent flow present within the tunnel allowed the exhaust gas and dilution air to mix homogeneously prior to sampling. Also, the larger tunnel diameter and flow-rate reduced particle deposition onto the CFV-CVS surface and aided representative sample collection.

Prior to the SDS system, another sampling port, connected to a PM mass measuring apparatus, extracted a sample of the diluted exhaust flow from the CFV-CVS tunnel. The distance between the two sampling ports was sufficient to ensure that any wake effects from the first sampling port did not affect the sample collected at the second port. The PM mass measuring apparatus consisted of a secondary micro-diluter and a mass collection filter. The sampling method was consistent with the transient Federal Test Procedure (FTP) (Code of Federal Regulations 2001). The results collected from this regulatory certification sampling apparatus were then used as a basis of comparison for examining the capability of the SDS system developed in the present study.

Secondary Micro-Diluter

Maintaining atmospheric dilution conditions is vital to the achievement of representative sampling (Abdul-Khalek et al. 1998; Lodge 1989). The effects of dilution conditions on particle formation and size distributions have been studied by many researchers including Harris and Maricq (2001), Ntziachristos et al. (2004), and Liu et al. (2007). In the present study, a secondary micro-diluter with adjustable dilution ratio, complemented to the primary CFV-CVS tunnel, was designed to abate further particle formation and changes caused by the physical processes of nucleation, coagulation, and growth so that a stable aerosol sample can be transmitted to the sampling ports, consistent with the work of Kittelson et al. (2002) and the certification practices. The diluter, which included a static pressure chamber, a perforated tube, and insulation materials, improved exhaust sample and dilution air mixing, reduced thermophoretic forces, and increased the overall dilution ratio of the system. To optimize mixing, the partially perforated inner tube was installed concentrically inside the diluter. This formed a static pressure chamber from which the compressed air was uniformly distributed into the mixing region. The air gap between the tube and the diluter wall also decreased the temperature gradient between the exhaust gas and the mixing chamber surfaces, which reduced particle loss due to thermophoretic forces. The compressed air was purified by PM, moisture, and absorption filters prior to entering the secondary diluter.

Residence Time Chamber

Because the nucleation and condensation of the gas-phase organics in the diluted exhaust involve the diffusion-limited transport of supersaturated vapor onto existing particles, a residence time chamber similar to Hildemann et al. (1989) was included in the SDS system. For this design, 30 seconds of residence time was used. During this time, the sample flow and concentrations within the residence time chamber become uniformly distributed before entering the sampling probes.

The sampling section of the RTC was designed to extract multiple samples of the diluted and aged exhaust gas. Eight probes were inserted into this section to perform the isokinetic sampling. The probes were made with thin walled stainless steel nozzles, and the entry edges were sharpened to a 5° taper as suggested by Belyaev and Levin (1972, 1974). To avoid interference, the probes were aligned coaxial to the flow direction and positioned equidistantly more than five times their diameter from the inner wall of the residence time chamber and the outer surfaces of adjacent probes.

Sampling Trains

Multiple sampling trains were used to facilitate the simultaneous collection of volatile and semi-volatile organic, gas-phase, and PM emission samples. The sampling trains consisted of polytetrafluoroethylene (PTFE) membrane filters, baked quartz filters, polyurethane foam (PUF) and XAD adsorber, thermal desorption (TD) tubes, and dinitrophenylhydrazine (DNPH) cartridges used for sample collection. These samples have the capability of being analyzed for PM mass, elemental and organic carbon, ions, trace elements, polycyclic aromatic hydrocarbons (PAH) and nitro-PAH, semi-volatile organic compounds, and gas-phase carbonyl and hydrocarbon species. Additionally, two of the ports provided samples for particle sizing instrumentation and temperature, pressure, and humidity sensors.

Three types of PM sampling train setups were used and are shown in Figure 2 and Figure 3. Train Setups 1 and 2 began with a PTFE-coated aluminum PM_2.5 cyclone separator. Flow through each train was controlled by downstream critical flow orifices in series with rotary vacuum pumps. A rotameter prior to each critical flow orifice continuously monitored the flow rate during sample collection. In Train Setup 1 shown in Figure 2, the cyclone was in series with two 47 mm PTFE membrane filters. The second PTFE filter was used to increase the PM collection efficiency. Three trains were run in parallel and connected downstream of the filters into a manifold joining the critical flow orifices on one side and the vacuum pump on the other. In Train Setup 2 shown in Figure 2, two 47 mm quartz fiber filters in series with a PUF cartridge followed a cyclone. The second quartz filter was used to increase PM collection efficiency and to allow for organic compound corrections. This train setup was capable of collecting particle-phase organic compounds on the quartz filters and semi-volatile organic compounds on the PUF cartridges. Three trains, run in parallel, connected downstream of the PUFs to a manifold containing the critical flow orifices. For future engines with ultra-low emission levels, Train Setup 3 (see Figure 3) was designed as a high volume sampler to take greater amounts of sample than Train Setups 1 and 2 during the same time period. The first component was a PM_2.5 PTFE-coated aluminum cyclone followed in series with two 90 mm quartz filters, a rotameter, critical flow orifices, and a rotary vacuum pump.

FIG. 2 Diagram of sampling Train Setups 1 and 2.

FIG. 3 High volume sampler Train Setup 3 and gaseous trains.

The configurations for the gaseous sampling trains are shown in Figure 3. A PTFE membrane filter prevented particles from entering the gaseous sampling media. Downstream, two DNPH cartridges in parallel had different flow rates to account for the variability in the chemical concentrations of exhaust from different engines. Additionally, two TD tubes allowed for further collection of gaseous samples. Flow rates were regulated by two separate mass flow controllers before a rotary vacuum pump.

Recent studies have shown a widespread interest in particle sizing information. A scanning mobility particle sizer (SMPS) was connected to the residence time chamber to measure particle size distributions and number concentrations. However, other PM or gaseous instrumentation can also be attached to this port for size resolved PM or gaseous composition analyses.

The system was also carefully designed to minimize vapor and particulate losses. For example, conical sections were used at the entrance and exit of the residence time chamber to promote smooth flow transitions. The chamber diameter was also maximized to prevent particle diffusion on the wall while remaining small enough to be portable and to fit into crowded engine test cells. Also, bends in the sampling trains were given generous radii to reduce particle impaction. To prevent secondary reactions, the components of the system were made of inert materials such as stainless steel, PTFE, and PTFE-coated aluminum. Electrically non-chargeable materials found in this system such as 304, 316, and 316 L stainless steels can also reduce electrostatic deposition of the charged particles that are typically polarized during the combustion processes. Oils, greases, rubbers, and other materials that could outgas organics were completely eliminated from the sampling system to avoid contamination.

VALIDATION PROCEDURE

In order to ensure repeatability and reproducibility, a precise set of procedures were followed before, during, and after testing. Before tests were conducted, the SDS system components were cleaned, assembled, and tested for leaks. A calibration was performed to check the accuracy of the flow rates through each of the critical flow orifices and also to check the inline flow meters. During testing the engine was operated under conditions specified by standard test procedures. PM substrates were conditioned in a climate controlled (temperature, humidity, drift, and electrostatics) weighing room for 24 hours before and after each sample collection.

The SDS system was cleaned prior to testing to minimize pre-existing organic and metal compounds, including rinsing with deionized water to remove large particles, double vapor degreasing with trichloroethylene to remove the oils and greases, and heating to a temperature of 200°C for 14 hours to remove residual trichloroethylene left from the degreasing process. The collection trains, including the stainless steel piping and sampling nozzles, were cleaned with soap and 18 Mohm water. The trains and piping that would be used to collect organics were then sonicated in hexane to cleanse impurities and rinsed in acetone to remove hexane and expedite drying. The trains and piping that would be used to collect trace metals were rinsed in a 5% HNO₃ solution to detach contaminants and then rinsed with 18 Mohm water followed by acetone.

Once the sampling system was reassembled, a leak test was performed to ensure minimal contamination and alteration from the surrounding environment. Military leak inspection MIL-PRF-62048C section 3.4.6 was conducted, holding the chamber at a vacuum of 12.5 kPa. To pass, a leak rate of under 0.5 m³/h must be observed. However, the flow rate of the system was much lower than the required leak rate, so a percentage of less than 0.5% of total flow was used to determine pass/fail (ISO 8178-1, 2006) and a leak rate of under 0.3% was always found. After the leak test was completed, the source dilution system was turned on, and a Gilibrator bubble meter was used to measure the flow rate produced by the orifices and controllers. As each train was checked, an inline rotameter was calibrated for in-test flow rate monitoring.

A 2001 Cummins ISM, 325 kW, six cylinder, direct injection diesel engine was used for the validation tests, and was controlled by a General Electric direct current dynamometer, which was capable of simulating most of the U.S. and European transient and steady-state cycles on an engine producing less than 525 kW of power. The oil used to lubricate the ISM was 15W40, and the fuel was Amoco ECD, which was guaranteed to contain less than 30 ppm sulfur and tested at 2 ppm. The engine was run under the FTP Heavy Duty cycle (Code of Federal Regulations 2001) and ISO modes 1, 8, and 11 (ISO 8178-4, 1996). The FTP cycle is 20 minutes long and consists of four phases, simulating a series of New York and Los Angeles freeway and non-freeway driving conditions. The steady-state modes chosen for the validation testing were ISO modes 1, corresponding to 2100 RPM and 100% load; 8, corresponding to 1313 RPM and 50% load; and 11, corresponding to 650 RPM and 3% load.

RESULTS AND DISCUSSION

The system was examined by measuring the variations in flow rates and gaseous concentrations among the sampling trains and by comparing these values to measurements collected from a regulatory certification system. PM size distributions from steady-state modes were also compared between the two systems as a tracer study. A thorough statistical analysis was conducted on the PM mass measurements because it is always more difficult to achieve homogenous PM distribution, relative to gaseous distributions. Using the PM mass measurements, a capability study, to determine statistical indices which were used to assess the capability and performance of the SDS system, was performed.

Flow Rate Monitoring

Ideally, the flow rate through each train should remain constant for the entire sampling period. However, it is possible for the flow rate to decrease as the restriction across the filter increases from PM loading. The flow rate in each sampling train was measured before PM loading and after 1, 3, and 5 FTP cycles using a Gilibrator flow meter. The flow rates slightly decrease as the filters become loaded; however, the maximum error is only 2.0% after five FTP cycles. Due to concerns of sampling artifacts from extended filter exposure, it was decided that a maximum of three FTP cycles be used for sample collection. The corresponding error was found to be 1.4%. For modern ultra-low emission engines, this error is expected to be lower than the higher emission 2001 ISM engine used in this experiment.

Gaseous Distributions

The variations in gaseous distributions at the sampling section of the residence time chamber were determined by measuring the NO_x and CO₂ concentrations at six ports and the regulatory certification system with a Horiba gas bench analyzer. Gaseous concentrations were measured for ISO modes 1 and 8 and are shown in Figure 4. The NO_x concentration was uniform for both modes with maximum errors, from the certification system, of 1.37% for mode 1 and 0.51% for mode 8. The CO₂ concentration was comparable to the certification system for both modes with a maximum error of 5.67% for mode 1 and 1.58% for mode 8. The similarity between the gaseous concentrations found in the two sampling systems confirmed that the performed leak test was sufficient to prevent contamination from the surrounding air.

FIG. 4 Concentrations of (a) NO_x and (b) CO₂ from the sampling trains and regulatory certification system.

Particle Size Distributions

To further investigate the SDS system capability, a tracer study needed to be conducted to ensure that the system did not alter the samples from the CFV-CVS tunnel. Due to the high flow rate of this tunnel, chemical injection was determined to be impractical since a large amount of chemical would have been necessary. Therefore, it was decided to use diesel PM as a tracer because nuclei-mode particles are especially sensitive to dilution ratio, residence time, and sampling temperature. Although the SDS system was located significantly downstream of the certification system sampling port, both samples were collected under the same dilution, residence, and temperature conditions. Using a TSI model 3963 SMPS, consisting of an ultra-fine condensation nuclei particle counter (CPC) and a long differential mobility analyzer (DMA), the particle size distributions from the SDS and the certification systems were collected and compared. Figure 5 displays the normalized particle size distributions for ISO modes 8 and 11 at two different dilution ratios of 0:1 and 1:1. The similarity of the scans shows that the SDS system did not alter the samples from the CFV-CVS tunnel.

FIG. 5 PM size distributions with (a) no secondary dilution and (b) 1:1 secondary dilution.

PM Mass Distribution

The capability of the SDS system was analyzed in two studies using the PM mass distributions which were obtained from a statistically stable process and followed a normal distribution. During testing, mass samples were collected from six PM sampling trains (Train Setup 1) and also from the regulatory certification system for 10 FTP cycles. The first study was used to determine the ability of the SDS system to take similar samples across all of the trains. The target value was set as the average PM emission rate from the SDS system ports while the upper specification limit (USL) and the lower specification limit (LSL) were ± 15% of the target. From Figure 6a, an X-bar chart graphically shows the average emission rate per test compared to the overall average of the tests while the R-chart displays the range of the emissions collected from the trains per test compared to the average range. By examining the charts for outlying points between the USL and LSL, it was possible to determine whether or not the system was performing in a controlled manner. The charts in Figure 6a show that the process used in the present study was stable. Additionally, the Normal Probability Plot, in Figure 6a, suggests that the data collected followed a normal distribution with a p-value of 0.337, larger than the critical value of 0.05. Because the process was statistically stable and the data followed a normal distribution, a capability histogram (Figure 6b) was used to assess the capability of the system. Statistical values used to determine the system capability included the capability index (C_pk), performance index (P_pk), and defect ratio in parts per million (PPM). The C_pk is defined as

FIG. 6 (a) Six pack capability study and (b) Capability histogram of the variation across the samplers with the target value set to the average measurement.

where is the sample mean and σ is the within sample standard deviation (Pyzdek 1996). The C_pk is a measure of how well the data are centered on the target value in addition to a measure of process variation. When C_pk is greater than 1.0, it is expected that the delivered results are within set limits 99.73% of the time (Box et al. 1978; Chrysler Corporation et al. 1992), therefore, the calculated C_pk value of 1.35 indicated that the entitled measurements from the SDS system would be centered on the target value statistically. The P_pk, similar to C_pk, is used to determine the actual system performance. The P_pk is defined as

where σ* is the overall sample standard deviation (Pyzdek 1996). The P_pk, found to be 1.16 (99.94% of the time), reinforced that the real measurements would be centered on the target value statistically. Finally, the PPM was used to estimate the expected defects per million operations and was found to be 521 for this system.

A second study was used to determine the capability of the sampling trains to take accurate samples when compared to the regulatory certification system. For this analysis, the average certification system measurement was used as the target value, and the USL and LSL were set to ± 20% of this target value. The six pack capability study results were almost identical to the first study and are shown in Figure 7a. The p-value from the Normal Probability Plot once again was 0.337, indicating that the data followed a normal distribution. Although the sample mean from the ports was approximately 0.814% higher than the target value, the calculated C_pk, P_pk, and PPM values of 1.71, 1.47, and 6.25, respectively, show that the SDS system was capable of matching samples collected from the regulatory certification system. The slightly higher measurements from the SDS system may be attributed to greater gas-to-particle conversion due to longer residence time and lower gas temperature.

FIG. 7 (a) Six pack capability study and (b) Capability histogram of the variation across the samplers with the target value set to the average regulatory certification system measurement.

SUMMARY

A source dilution sampling system, consisting of a CFV-CVS tunnel, a thermophoresis-resistant secondary micro-diluter, a residence time chamber, isokinetic sampling probes, multiple sampling trains, and control systems, was developed to representatively collect multiple samples from an exhaust flow. The system was examined by performing a tracer study and by determining the variations in flow rates, gaseous concentrations, and PM mass measurements.

The multiple sampling trains were found to be capable of operating for up to 5 FTP cycles with a maximum flow error of 2.0%. The gaseous concentrations among the sampling ports coincided not only with each other but also with the regulatory certification system. A tracer study, comparing particle size distributions from the SDS and certification systems, confirmed that the SDS system including a CFV-CVS tunnel did not alter the exhaust samples. Statistical analysis of the PM mass distribution provided C_pk, P_pk, and PPM values of 1.35, 1.16, and 521, respectively, which indicated that the SDS system was capable of collecting similar samples among the sampling trains within set limits. Further, a capability study on the PM mass measurements compared to the regulatory certification system produced C_pk, P_pk, and PPM values of 1.71, 1.47, and 6.25, respectively, demonstrating that the SDS system displayed sufficient agreement with the certification system despite a slight increase in the average PM mass measurement. These results suggest that the SDS system developed in the present study is suitable for collecting representative samples from an exhaust flow under transient or steady-state conditions while simulating atmospheric dilution and aging, maintaining proportional sampling during temperature excursions, and minimizing particle loss and alteration.

Acknowledgments

The authors acknowledge Prof. David Foster of the University of Wisconsin—Madison, Dr. Glynis Lough of the U.S. EPA, Dr. Roger McClellan, Prof. John Johnson and Prof. Susan Bagley of Michigan Technological University, Dr. Doug Lawson of the National Renewable Energy Laboratory, and Prof. David Kittelson of the University of Minnesota for their advice and reviews. We also acknowledge Devin Berg, Tom Wosikowski, Joe Lincoln, Howard Tews, and Gary Frank of Cummins Filtration and Emission Solutions for their assistance with the manuscript and the source dilution sampling system.