Design of a Compact Dilution Sampler for Stationary Combustion Sources

ABSTRACT

The dilution sampling method simulates the rapid cooling and dilution processes after hot flue gas have left the stack. This allows gases or vapors to nucleate both homogeneously and heterogeneously, and to condense on preexisting particles in processes analogous to those that occur in the ambient environment. Using this method the authors can collect filterable particulate matter (PM) and condensible PM, that is, primary PM, simultaneously. In order to make this method more suitable for field investigation, a compact dilution sampler was developed. The sampler enhances mixing of dilution air with the stack gas, and thus shortens the length of the mixing section. The design decreases the nominal flow rate through the aging section, and accordingly reduces the size of the residence chamber. The decreased size of the sampler is suitable for field test. Sampling gas is pressured into the residence chamber, and air pressure in the chamber is micro-positive. Uncollected redundant gas is automatically discharged through unused sampling ports, which keeps the unit stable. Performance evaluation tests demonstrate that the design is reasonable. The sampler has been applied to characterize PM emissions from various combustion sources in China.

IMPLICATIONS

Particulate matter (PM) emitted from stationary combustion sources is an important contributor to urban ambient PM, especially for PM_2.5 (PM ≤2.5 μm in aerodynamic diameter). Characterizing PM emissions from various combustion sources and providing reliable emission data are important for identifying their contributions to urban ambient PM and for designing corresponding control measures. This research developed a compact dilution sampler, which is a reliable tool for characterizing PM emissions from stationary combustion sources in China. The data obtained by the field measurements using the dilution sampler will be helpful for source apportionment and emission inventory development.

INTRODUCTION

Particulate matter (PM) emitted from stationary combustion sources is an important contributor to urban ambient PM, especially for PM_2.5 (PM ≤2.5 μm in aerodynamic diameter). Contribution to urban ambient PM from combustion sources can generally be categorized as primary PM and secondary PM.1 Primary PM is emitted directly from sources and includes the compounds that are solid or liquid at stack conditions, and the compounds that are gas or vapor at stack conditions. The latter compounds become solid or liquid upon exiting the stack where they are rapidly cooled and diluted with ambient air to ambient temperature and pressure conditions.1–3 Characterizing PM emissions from various combustion sources and providing reliable emission data are important for identifying their contributions to urban ambient PM and for designing corresponding control measures.

According to source-level sampling methods, primary PM from a source can be classified as filterable PM and condensible PM. U.S. Environmental Protection Agency (EPA) Method 5 and Method 17 are the most common approaches to source-level sampling for filterable PM. In theses methods, solid and liquid particles present at a constant temperature (typically 120 °C) or at the stack temperature are captured on the filter.4,5 Chinese national standard GB/T 16157-1996 also provides guidance for measuring filterable PM.6 Condensible PM is frequently defined as the amount of material collected in a series of impingers in an ice bath downstream of a filter. Following the filtration, the sample gas is quenched in cold water to condense some of the inorganic and organic vapors. U.S. EPA Method 202 is designed to determine the condensable PM.7 Combination of the filter method and the impinger method can both collect filterable PM and condensible PM, that is, primary PM. However, the impinger method is generally thought to overestimate condensible PM emissions (indicating more condensible PM than is actually emitted) because noncondensible gases may react in the solution to form condensible PM, which would not form in the stack. The positive bias is mainly from the oxidation of SO₂ to form SO₄ ²⁻. The SO₂ dissolves in water to form H₂SO₃, which may oxidize to form H₂SO₄. The SO₄ ²⁻ would then be counted as condensible PM.2

The dilution sampling method is used to characterize fine particle emission from stationary combustion sources because it simulates the rapid cooling and dilution processes after hot flue gas have left the stack. This allows gases or vapors to nucleate both homogeneously and heterogeneously, and to condense on preexisting particles in processes analogous to those that occur in the ambient environment.8 With this method we can collect filterable PM and condensible PM, that is, primary PM, simultaneously using a single filter, thus simplifying sampling procedures. Moreover, to sample and analyze the diluted gas, we can use ambient air methods, which provide ambient-comparable PM speciation profiles from stationary sources. The dilution sampling method has been widely used to develop fine particulate emission factors and speciation profiles from various stationary sources.19–13 Currently, U.S. EPA treats the dilution sampling method as a “conditional test method”,14 and the American Society for Testing and Materials (ASTM) and International Organization for Standardization (ISO) are considering adding dilution sampling as a new test method for determining PM_2.5 and PM₁₀ (PM ≤10 μm in aerodynamic diameter) mass in stack gases.15,16 Recently, a few studies have investigated effects of dilution and sampling conditions (including dilution ratio, residence time) on the fine particle emissions from various combustion sources.12,17–19 Lipsky and Robinson observed that PM_2.5 emission rate decreased largely with increases in dilution for the diesel engine operating at low load and wood combustion and considered that this was caused by changes in partitioning of semivolatile organics (SVOCs).18 Robinson et al. reviewed recent developments in the understanding of contribution of emissions from combustion process to fine PM mass, focusing on organic aerosols and updated the conceptual model for fine particle mass emissions from combustion systems.1 A recent review paper by Maricq and Maldonado summaries the spirit of two Coordinating Research Council PM Measurement workshops and one of the consensus is that dilution sampling can provide comparable results in laboratory and real-world applications.20

The bulkiness of early dilution samplers was a major barrier to their widespread application for field source test. Much effort has gone into developing more compact and portable dilution samplers for their use in the field.21,22

Data on PM₁₀ and PM_2.5 emissions from stationary sources in China are scarce. At the present time, no standard method is available for measuring PM₁₀ or PM_2.5 emissions from stationary sources in China. The most common method to measure air pollutant emissions from stationary sources, that is, GB/T 16157-1996, was designed for filterable PM and gas emissions.6 Only a few researchers conducted studied on PM₁₀ or PM_2.5 emissions from limited sources.23–26 Official air pollutant emission statistical data reported PM, but did not refer to PM₁₀ and PM_2.5.27 Zhang et al. estimated Chinese PM (including PM₁₀ and PM_2.5) emissions.28 Because of the lack of available measurement data, in their calculation, a considerable amount of PM₁₀ and PM_2.5 emission factors are adopted from U.S. AP-42 database and the Regional Air Pollution Information and Simulation (RAINS)-PM Model, which may cause high uncertainty in emission estimate. Therefore, a compact dilution sampler was developed to characterize PM_2.5 emissions from stationary combustion sources. This paper discusses the design and performance of the sampler, and reports the results of its application to biofuel and coal-fired combustion systems.

EXPERIMENTAL DESIGN

The design of the dilution sampler follows the basic principles of the widely cited Hildemann design,8 but is more compact and portable. An outline of the sampling system is shown in Figure 1. It consists of four main sections: sampling inlet, dilution section, residence chamber, and sampling section. The sampler is described in more detail in the following subsection.

Figure 1. Schematic diagram of the dilution sampling system.

In the sampling inlet section, flue gas is withdrawn from the stack through an in-stack cyclone with a PM₁₀ cut point, followed by passage through the heated inlet line, and is then passed to the dilution section. A nozzle is attached to the cyclone inlet to achieve isokinetic sampling of the exhaust. The cyclone, which is typically inserted into the stack, can also be installed outside the stack in the case where the stack is small and the blockage effect causing by its insertion is larger than 6%, or where the sampling port in the stack is not large enough to insert it. In order to reduce particle thermophoresis losses, the inlet line is heated to at least 5 °C above the stack gas temperature or a specified temperature.

The dilution section consists of two stage diluters. The operation principle of the first diluter is based on an ejection type dilution (Dekati Ltd., Tampere, Finland).29 Purified pressurized dilution air (typical pressure value is 2 bar) flows at high speed around an ejector nozzle and causes a pressure drop, which draws a sample through the nozzle. The raw sample is instantaneously diluted as it mixes with the dilution airflow. The nominal dilution ratio (DR₁) for a single Dekati diluter is 8:1. The diluter is made of stainless steel with the length of 36 cm. A regulator valve, attached at the bypass pipe of the first diluter, is used to the regulate gas flow rate at the outlet into the second diluter. About 10–50 L/min can be drawn from the outlet of the diluter to the second diluter according to research requirements. A venturi flowmeter is equipped to record the gas flow rate at the outlet of the first diluter. An oil-free air compressor supplies dilution air. Before entering the diluter, the pressurized dilution air is desiccated and then purified by a high-efficiency particulate air (HEPA) filter and activated carbon to remove fine particle and volatile organic compounds (VOCs).

The second diluter is an enclosed cylinder with a perforated cone inside. The sample gas from the first diluter is introduced into the inside of the cone. The second dilution air is forced through the apertures of the cone into the inside of the cone where it mixes with the sample flow instantaneously. The diluter was modified from a previous Southern Research Institute (Birmingham, USA) dilution cone design.30 It is made of stainless steel with a length of 48 cm. A venturi flowmeter is equipped to record the second dilution airflow. A vacuum pump supplies dilution air for the second diluter. The dilution air is as purified as that in the first diluter. A cooling unit is added optionally to cool the dilution air such that the diluted sample temperature is less than 30 °C. Due to the application of the desiccator, the humidity of the diluted sample can be lowered to less than 50%. The second diluter can supply a dilution ratio (DR₂) from 1:1 to 10:1. The total dilution ratio (DR_t) of the two-stage diluters ranges from 10:1 to 80:1.

All of the diluted gas is transferred to the residence chamber. The residence chamber is constructed completely from stainless steel and is 45 cm in diameter and 70 cm in length. It allows for an aging time of 60–70 sec for the diluted gas with a flow rate of 100 L/min. The temperature, relative humidity, and pressure in the chamber are monitored.

The sampling section is attached to the end of the residence chamber and has eight sampling ports that can be connected to the particle or gas measurement instruments. In our previous field measurements of emissions from rural household biofuel combustion,31 three sampling ports were connected with three parallel PM_2.5 cyclones with filter packs operated at 16.7 L/min to collect fine particles for mass measurement and chemical speciation (elements, ions, organic carbon, elemental carbon, and speciated organic compounds). One port is connected to an electrical low-pressure impactor (ELPI; Dekati; operated at 10 L/min) to measure real-time particle size distribution.32 Another port was connected to a vacuumed canister for VOC analyses. The diluted gas in the chamber is under a small positive pressure and extra gas can be automatically discharged from the unused sampling ports.

The sampler weighs ∼50 kg, plus ∼100 kg ancillary equipment (the air compressor, sampling pumps, and ELPI, etc.). The overall height is ∼170 cm. The sampler is relatively easy to convey to the sampling location and suitable for assembly on-site at the platform.

The main features of the design are as follows:

1.	The dilution section consists of two-stage diluters. Their operation principles are ejection dilution and mixing cone design, respectively, which enhances the mixing of dilution air with the stack gas, and thus shortens the length of the mixing section. The size of the residence chamber is reduced by decreasing the nominal flow rate through the aging section to 100 L/min. (The design flow rate of 100 L/min provides sufficient sampling volume for determining PM2.5 mass and chemical speciation.) The decreased size of the sampler is suitable for field test.
2.	The sampling gas is pressured into the residence chamber and the air pressure in the chamber is micro-positive. The uncollected redundant gas is automatically discharged through unused sampling ports, which will keep the unit stable. Moreover, some aerosol instruments that are driven by a low-power pump for ambient measurement, which cannot be used in traditional designs with residence chamber under negative pressure, can be used in our design.
3.	Unlike in Hidemann's design, where a fraction of the total flow (typically 20%) is drawn through the residence chamber, total flow is transferred to the residence chamber in our design. This design avoids the error caused by not fully mixing the sample gas and dilution air.

RESULTS AND DISCUSSION

Performance Evaluation

Mixing Performance in the Sampler

The mixing between stack gas and dilution air must be thorough enough to ensure that gas withdrawn from each sampling port is the same.

A computational fluid dynamics modeling software (Fluent33) was used to evaluate the mixing performance of the second diluter. An example of computational flow model results for mixing in six cross-sections is shown in Figure 2. The results indicate that complete mixing can be achieved immediately after the gas leaves the cone. It implies that the second diluter can be shortened in future designs.

Figure 2. Computational flow model results for mixing in the second diluter.

To evaluate the degree of mixing in the system, initial tests were conducted to measure emissions from a coal-fired boiler with a stable load. A flue gas analyzer (model KM9106; Keison, Essex, UK) was inserted into the sampling ports to measure the NO_x concentration in the cross-section of the residence chamber. Figure 3a shows the location of the eight measuring points in the cross-section. Figure 3b presents the results obtained from three different DR conditions. Relative standard deviations (RSDs) of the NO_x concentrations at the eight points are less than 3%. The results indicated that stack gas and dilution gas were mixed thoroughly.

Figure 3. Mixing in the sampler. (a) Locations of the measurement points in the cross-section. (b) Results of NO_x concentrations at different measurement points at three dilution ratios.

Checking the Dilution Ratio

In the field tests, DR of the sampler was calculated as

(1)

where DR₁ was determined according to the information supplied by Dekati. DR₂ was determined by the gas flow rate of the outlet of the first diluter (Q ₁) and the second dilution air flow rate (Q ₂). DR₂ was calculated as

(2)

where Q ₁ and Q ₂ were adjusted to the standard condition.

We used two methods to check the DR of the sampler:

1.

In the laboratory, standard NO2 gas was injected through the sample inlet line. A flue gas analyzer was used to measure diluted NO2 in the chamber. DRt could be calculated in two ways, using eqs 1 and 3. (3) where (NO2)st is the standard NO2 gas concentration, (NO2)cham is diluted NO2 concentration. The relative bias errors of DRt from the two ways for a given experiments is given by eq 4, (4) where DRt[1] and DRt[2] are determined by eq 1 and eq 3, respectively. The values were less than 7.4%, which indicated that the DRt determined by two ways were comparable.

2.

In the field, a flue gas analyzer was used to measure NOx in the stack gas and diluted NOx in the chamber, respectively. DRt could be calculated in same two ways that were used in method 1. The relative bias errors of DRt from two ways were less than 10%, which showed that DRt determined by two ways were close.

Both experiments suggested that the DR of the sampler determined by eq 1 was reasonable in the field tests.

Blanks Tests

System blanks were performed by operating the sampler only on the air purified by the HEPA filter and activated carbon, and by measuring fine particle concentration using ELPI. System blanks were very low (for example, the number and mass concentrations are 2–3 orders of magnitude less than those obtained during actual experiments).

Particle Losses

Particle loss is one of the major issues in sampling fine particles from combustion sources when using dilution sampling method. To minimize loss, the following were taken into considerations:

1.	Heating the inlet line to reduce thermophoretic losses.
2.	Minimizing the use of bends and contractions to reduce inertial deposition.
3.	Use of conductive tubing material to minimize electrostatic effect.

Due to the lack of the means for aerosol calibration, particle loss inside the dilution sampler has not been conducted. Hildemann et al. evaluated particle losses in their design and results indicated that losses in their sampler were minimal for fine particle.8 In consideration of similarity with Hildemann's design, it is assumed that the new design is suitable for measurement fine particle.

Discussion

The design operates at dilution ratio from 10:1 to 80:1 with particle- and organic-free dilution air. The level of dilution can reduce the flue gas temperature to ambient levels, but is far less than the more dilute conditions encountered in the real world. Furthermore, in the real world, emissions are mixed with polluted background air, not particle- and organic-free air. Because of these differences, this design may change the partitioning of SVOCs and result in biases on measured PM emission. According the partitioning theory34 and the studies from Lipsky et al.,18,35 the possible biases could be explained at a qualitative level as follows: For high-emitting sources, such as biofuel and coal combustion, the concentration of the absorbing organic phase inside the dilution sampler are much higher than ambient levels, then too much of the semivolatile material will partition into the particle phase such that measurement will overestimate measured fine particle emissions compared to ambient levels of dilution. On the contrary, for low-emitting sources, such as natural gas combustion, the concentration of the absorbing organic phase inside the dilution sampler can be much lower than typical ambient levels, then too much of the semivolatile material will partition into the gas phase and the resulting measurement will underestimate measured fine particle emissions compared to ambient levels of dilution.

Field Test

The sampler has been applied to measure the PM emissions from coal-fired industrial boilers, coal-fired power plants, and biofuel combustion in China.31,36 Emission samples were diluted by purified air and cooled to close to the ambient temperature. The dilution ratios for all tests ranged from 17 to 50, with a typical ratio of 20. The diluted air was aged for 60 to 80 sec to allow for condensation, coagulation, and rapid reactions to occur prior to being collected. The sampling time for coal-fired industrial boilers and coal-fired power plants is about 2 hr. The time for biofuel combustion is approximately 30–50 min.

PM_2.5 samples were collected in three parallel PM_2.5 cyclones with filter packs operated at 16.7 L/min. One filter pack was a 47-mm Teflon-membrane filter used to determine PM_2.5 mass and element analysis. The second and third filter packs were 47-mm quartz fibers for organic carbon (OC), elemental carbon (EC), and ion analyses. These analysis methods have been described in our earlier papers36,37 and are briefly summarized here. Teflon-membrane filters were conditioned for 24 hr at about 40% relative humidity (RH) and 25 °C in an air-conditioned room and weighed on a microbalance with a resolution 10 μg. Elements were analyzed by inductively coupled plasma optic emission spectrometry (ICP-OES; IRIS Intrepid II XSP, Thermo, Waltham, USA). OC and EC were determined, with a thermal/optical carbon analyzer (DRI, model 2001, Reno, USA) using the Interagency Monitoring of Protected Visual Environments (IMPROVE) protocol. Ionic species including chloride, nitrate, sulfate, and ammonium were quantified using an ion chromatograph (Dionex-600, Sunnyvale, USA).

PM_2.5 chemical source profiles for coal-fired power plants, coal-fired industrial boilers, wood burning stoves, and maize residue burning stoves are illustrated in Figure 4. Chemical abundance for each species is expressed as a percentage of each chemical concentration relative to the PM_2.5 mass concentration. The data are given as the average and standard deviation of each chemical abundance. Si, OC, Al, Fe, and Ca were abundant in PM_2.5 from the coal-fired power plant. The result was similar to the typical profile of coal-fired power plants, except for SO₄ ²⁻ and S.32 The possible explanation is that PM_2.5 was collected at the outlet of the wet limestone scrubber and most of SO₄ ²⁻ was captured. Coal-fired industrial boiler emissions contain high levels of SO₄ ²⁻, OC, Na, S, Si, and EC. Ge et al. observed that high percent of OC, EC, Si, Al in PM_2.5 was emitted from a coal-fired stocker/chain boiler in China and they did not present the data of SO₄ ²⁻ and S.23 Our data are comparable with that study. Wood burning stove emissions consist primarily of carbonaceous components (84% ± 13%, OC plus EC), whereas maize residue burning stove emissions were enriched in OC, Cl⁻, K and EC. Those species are the typical components of PM_2.5 emitted from biomass burning.38,39 Currently, there are few data on PM_2.5 emissions from various stationary combustion sources in China. These data obtained by the field measurements using the dilution sampler will be helpful for source apportionment and emission inventory development. These tests also demonstrate that the dilution sampler is suitable for use in the field experiments.

Figure 4. PM_2.5 chemical source profiles. (a) Coal-fired power plants; (b) coal-fired industrial boilers; (c) wood-burning stoves; (d) maize-residue-burning stoves.

CONCLUSIONS

A compact dilution sampler was developed to characterize PM emissions from stationary combustion sources. The sampler enhances mixing of dilution air with the stack gas, and thus shortens the length of the mixing section. The design decreases the nominal flow rate through the aging section, and thus reduces the size of the residence chamber. The decreased size of the sampler is suitable for field test. The sampling gas is pressured into the residence chamber, and the air pressure in the chamber is micro-positive. The uncollected redundant gas is automatically discharged through unused sampling ports, which will keep the unit stable. Performance evaluation tests, including mixing performance, DR check, and blank tests, demonstrate that the design is reasonable. The sampler has been applied to measure PM emissions from coal-fired industrial boilers, coal-fired power plants, and biofuel combustion in China.

ACKNOWLEDGMENTS

This work was funded by the National Natural Science Foundation of China (grant no. 20921140095) and the National High Technology Research and Development Program (863) of China (grant no. 2006AA06A305). We thank Mr. Jerry M. Davis of the EPA for his great help in revising this paper.