Particle Loss in Splitters and Dilutors

United Stales

Environ mental Prulotlion

Agency


-------
Particle Loss in Splitters and Dilutors

Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency

Prepared for EPA by
ICF Incorporated, LLC
EPA Contract No. EP-C-16-020
Work Assignment No. 1-03

Acknowledgements

Jacob Swanson, Adjunct Assistant Professor
Department of Mechanical Engineering
University of Minnesota

David B. Kittelson, Professor
F.B. Rowley Chair in Mechanical Engineering
Director, Center for Diesel Research
Department of Mechanical Engineering
University of Minnesota

John Gorman, Research Associate
Department of Mechanical Engineering
University of Minnesota

NOTICE

This technical report does not necessarily represent final EPA decisions or
positions. It is intended to present technical analysis of issues using data
that are currently available. The purpose in the release of such reports is to
facilitate the exchange of technical information and to inform the public of
technical developments.

rnA United States	EPA-420-R-21-005

Environrrntntal Prolotliun

\/LI Agency	January2021


-------
Contents

Table of Figures	1

Table of Tables	2

Introduction	3

Executive summary of results	3

Methods - splitter measurements	5

Results - splitters	8

2-way splitter, V2 x 3/s	8

2-way splitter, 3/s x 3/s	11

4-way splitter, 3/s x 3/8X3/8X3/8	14

Methods - dilutor measurements	16

Results - dilutor measurements	18

Conclusions	21

Table of Figures

Figure 1. Representative aircraft exhaust particle size distribution with a geometric mean

diameter of 30 nm and geometric standard deviation of 1.8	3

Figure 2. Results from three flow split conditions for the 1/2" x 3/8" splitter. Each graph includes

all experimental data, a best-fit line, and a theoretical line based on diffusion loss theory	4

Figure 3. Results from three dilutor operating conditions. Each graphs includes all experimental

data, the AIR penetration requirements, and a best-fit line	4

Figure 4. Apparatus for splitter loss measurement with DOS	6

Figure 5. Apparatus for splitter loss measurement with silver particles	6

Figure 6. Photographs of experimental apparatus for splitter evaluation	7

Figure 7. Particle size distributions of each particle generation source	8

Figure 8. Flow split: 5 L/min and 5 L/min	8

Figure 9. Flow split: 8.5 L/min and 1.5 L/min	9

Figure 10. Flow split: 1.5 L/min and 8.5 L/min	9

Figure 11. Flow split: 12.5 L/min and 12.5 L/min	9

Figure 12. Flow split: 1.5 L/min and 23.5 L/min	10

Figure 13. Flow split: 23.5 L/min and 1.5 L/min	10

Figure 14. 1.5 L/min and 8.5 L/min split condition showing poor agreement between model and

experimental results (compare to Figure 10)	11

Figure 15. 1.5 L/min and 23.5 L/min split condition showing poor agreement between model and

experimental results (compare to Figure 12)	11

Figure 16. Flow split: 5 L/min and 5 L/min	12

Figure 17. Flow split: 1.5 L/min and 8.5 L/min	12

Figure 18. Flow split: 8.5 L/min and 1.5 L/min	12

Figure 19. Flow split: 12.5 L/min and 12.5 L/min	13

Figure 20. Flow split: 23.5 L/min and 1.5 L/min	13

Figure 21. Flow split: 1.5 L/min and 23.5 L/min	13

Figure 22. Flow split: 2.5, 2.5, 2.5, 2.5 L/min	14

Figure 23. Flow split: 1.5, 1.5, 1.5, 5.5 L/min	14

Figure 24. Flow split: 5.5, 1.5, 1.5, 1.5 L/min	15

1


-------
Figure 25. Flow split: 6.5, 6.5, 6.5, 6.5 L/min	15

Figure 26. Flow split: 1.5, 1.5, 1.5, 20.5 L/min	15

Figure 27. Flow split: 20.5, 1.5, 1.5, 1.5 L/min	16

Figure 28. Apparatus for dilutor loss measurement with DOS	16

Figure 29. Apparatus for dilutor loss measurement with silver particles	17

Figure 30. Apparatus for dilutor loss measurement with NaCl (salt)	17

Figure 31. Schematic of heated dilutor with temperature setpoints indicated	18

Figure 32. Picture of heated dilutor (insulation used but not shown so that heating can be more

clearly seen)	18

Figure 33. Dilutor conditions: P = 2 bar, DR = 8.16	19

Figure 34. CFD results for P = 2 bar, DR = 8.55	19

Figure 35. Dilutor conditions: P = 3 bar, DR = 8.4	20

Figure 36. Dilutor conditions: P = 4 bar, DR = 9.13	20

Figure 37. Dilutor conditions REPEAT: 2 bar, DR = 8.16	21

Table of Tables

Table 1. Total penetration of "aircraft exhaust particles" through a 1/2" x 3/8" two-way splitter.

Results are rounded to two significant figures	4

Table 2. Total penetration of "aircraft exhaust particles" through a Dekati DI-1000 dilutor.

Results are rounded to two significant figures	5

Table 3. All experimental condition for splitter evaluations. Each condition was evaluated with a
different aerosol (silver and DOS). All flowrates are given in standard conditions	5

2


-------
Introduction

The EPA provided University of Minnesota researchers three "splitter" devices and a Dekati DI-
1000 dilutor for the purposes of 1) experimental characterization and 2) modeling of particle loss
in each device. The UMN used three aerosol types (salt, silver, DOS) of various size ranges to
challenge the splitters and dilutors to determine loss. Commercial computation fluid dynamics
(CFD) software was used to model results.

Executive summary of results

The main body of the report describes the measurements and results in detail. This section
summarizes all results and estimates the impact of losses in the context of an aircraft exhaust
particle size distribution. A representative aircraft particle size distribution (Figure 1) was
multiplied by representative solid particle penetration curves measured for the 1/2" x 3/8" two-
way splitter operated at various flow split conditions and the Dekati DI-1000 dilutor operated at
various pressures. Figure 2 shows results from the three flow split conditions considered in this
analysis, two of which are the "extreme ends," including: 1) very low loss (equal split and high
flow in each branch), 2) low loss (higher split ratio then in #1, low flow in one branch), and 2)
low loss (very high split ratio, low flow in one branch). Figure 2 shows curves in red that are fits
to all experimental data for each test condition. All data was used to provide the widest size
range for the fit. The physics-based fit curve was based off the Gormley / Kennedy diffusion
equations. The "theory" curve is described more later but in brief, it is the diffusion loss
calculated using laminar flow diffusion loss equations and ignoring fluid effects (e.g.
recirculating flow). Similarly, Figure 3 shows curves in red that are fits to all experimental data
for three pressure conditions for the Dekati dilutor (2, 3, and 4 bar compressed air pressure). The
experimental data includes data for the dilutor operated both "hot" and "cold." The dilutor "fits"
were based off the same diffusion equation as the splitters multiplied by a constant*particle
diameter that represents in effect of impaction, the mechanism responsible for particle loss for
larger particles and higher velocities that are present in the dilutor. The Aerospace Information
Report (AIR) minimum dilutor penetration requirements curve (from 15 to 100 nm) is shown for
reference.

Dp, nm

Figure 1. Representative aircraft exhaust particle size distribution with a geometric mean
diameter of 30 nm and geometric standard deviation of 1.8.

3


-------
1.50
1.30
11.10

ro
Id

0.90

CL

0.70
0.50

1/2x3/8

5 L/min and 5 L/min

Theory

1.50
1.30

1110

1 0.90

0)

CL

0.70
0.50

1/2 x 3/8



1.5 L/min and

3.5 L/min

Theory

- __ '





Fit

All data



10	100

Dp, nm

1000

10	100

Dp, nm

1000

1.50
1.30
11.10

ro
%

0.90

CL

0.70
0.50

1/2 x 3/8

1.5 L/min and 23.5 L/min

Theory
Fit

All data"

1	10 Qpj pm 100	1000

Figure 2. Results from three flow split conditions for the 1/2" x 3/8" splitter. Each graph includes
all experimental data, a best-fit line, and a theoretical line based on diffusion loss theory.

1.50
1.30

c

-1.10

ro

oj

0.90
CL

0.70
0.50

Dilutor 2 bar

AIR penetration requirement
 All data

1.50
1.30
H.10

3
J

j 0.90
0.70
0.50

Dilutor 3 bar

Fit 5..

1.50
1.30

c

S 1.10

ro

OJ

0.90
CL

0.70
0.50

Dilutor 4 bar

AIR penetration requirement

10	100

Dp, nm

1000

10	100

Dp, nm

1000

10	100

Dp, nm

1000

Figure 3. Results from three dilutor operating conditions. Each graphs includes all experimental
data, the AIR penetration requirements, and a best-fit line.

To determine overall "splitter" results, the representative aircraft exhaust particle number size
distribution was multiplied by each splitter penetration curve fit shown in Figure 2. The result
gives the total number penetration shown in Table 1. The particle number size distribution was
also weighted by diameter cubed to represent a volume (or mass) distribution. This distribution
was also multiplied by each penetration curve fit to give a total "mass penetration" as seen in the
right column of Table 1. Overall, losses are very low. The total number penetration is about 98%
when the flowrate is around 1.5 L/min and closer to 99% when the flowrate is higher -
regardless of the split. The uncertainty of these numbers has not been rigorously determined but
it is approximately  1%. Results show that in splitters, loss is dominated by diffusion, not
impaction, regardless of flowrate and split condition. One consequence of this is that the mass
penetration - which is impacted by larger particles that diffuse less - is nearly unity for all
conditions. As seen in the main results section, the penetration data for the 3/8" x 3/8" two-way
splitter and the 3/8" x 3/8" x 3/8" x 3/8" four-way splitter were substantially similar to these
results and the conclusions here are representative for all splitters evaluated.

Table 1. Total penetration of "aircraft exhaust particles" through a 1/2" x 3/8" two-way splitter.

Results are rounded to two significant figures.

Split condition

Number

Mass

5 L/min and 5 L/min

0.99

1.0

1.5 L/min and 8.5 L/min

0.98

1.0

1.5 L/min and 23.5 L/min

0.98

1.0

To determine overall "dilutor" results, the representative aircraft exhaust particle number size
distribution was multiplied by each dilutor penetration curve fit shown in Figure 3. The result

4


-------
gives the total number penetrations shown in Table 2. Experimental data clearly shows more
larger sized particles are lost in the dilutor compared to the splitters. The total "mass
penetrations" were calculated in the same way as described with the splitters. Overall, number
penetrations were about 88% for all operating conditions. Because some particles were lost due
to impaction in the dilutor, the mass penetration was lower - around 85%.

Table 2. Total penetration of "aircraft exhaust particles" through a Dekati DI-1000 dilutor.

Results are rounded to two significant figures.	

Operating pressure condition

Number

Mass

2 bar

0.88

0.86

3 bar

0.88

0.84

4 bar

0.88

0.86

Methods - splitter measurements

EPA provided UMN stainless steel splitters with the following inlet / outlet configuration: 1)1/2" x
3/8" (2-way), 2) 3/8" x 3/8" (2-way), and 3) 3/8" x 3/8" x 3/8" x 3/8" (4-way). Dimensions listed are
tube outer diameters. The flow split geometry met AIR requirements. As shown in Table 3, splitters
were evaluated with different flowrate and splits to evaluate particle loss for different conditions.

Table 3. All experimental condition for splitter evaluations. Each condition was evaluated with a
different aerosol (silver and DOS). All flowrates are given in standard conditions.

Flow conditions



Total flowrate, L/min

flow split, L/min

2-way (3/8 x 3/8)

10

1.5/8.5

2-way (3/8 x 3/8)

10

5/5

2-way (3/8 x 3/8)

25

1.5/24

2-way (3/8 x 3/8)

25

12.5/12.5

2-way (1/2 x 3/8)

10

1.5/8.5

2-way (1/2 x 3/8)

10

5/5

2-way (1/2 x 3/8)

25

1.5/23.5

2-way (1/2 x 3/8)

25

12.5/12.5

4-way

10

1.5/1.5/1.5/5.5

4-way

10

2.5/2.5/2.5/2.5

4-way

25

1.5/1.5/1.5/20.5

4-way

25

6.25/6.25/6.25/6.25

Figure 4 shows the apparatus to generate semi-volatile nanoparticles that atomizes a solution of
isopropyl alcohol and dioctyl sebecate (DOS). DOS is a high molecular weight organic liquid. The
atomized particles pass through a diffusion drier to remove the isopropyl alcohol. An Air Vac
Engineering dilutor is used to create a vacuum to pull the sample flow through the system as well as
increase the total flowrate. The diluted flow is pulled through the splitter under test by an Alicat

5


-------
vacuum mass flow controller. A TSI SMPS (with long column DMA and 3025 CPC) measures the
particle size distribution upstream and downstream of the splitter. Upstream and downstream sample
lines were identical in length and geometry.

For "salt" particle measurements, the DOS solution was replaced by a solution of NaCl in deionized
water. All other components of the apparatus were the same.

Silica gel diffusion dryer

' ^ Vent

Flow controllers

2-way splitter

1000 ppm dioctyl sebecate
(DOS) in n-propanol

JL f

I	iDilutor

3-way valve

-KX-

~

SMPS

1 fll

Ai-

Bi_

~

T-.

Figure 4. Apparatus for splitter loss measurement with DOS

Figure 5 shows the apparatus to generate silver nanoparticles using an evaporation/condensation
technique. The furnace temperature was 1050C and the flowrate through the furnace was 1.5 L/min.
An Air Vac Engineering dilutor is used to create a vacuum to pull the sample silver particle flow
through the furnace as well as increase the total flowrate. The diluted flow is pulled through the
splitter under test by an Alicat vacuum mass flow controller. A TSI SMPS (with long column and
3025 CPC) measures the particle size distribution upstream and downstream of the splitter. Upstream
and downstream sample lines were identical in length and geometry.

T-.

Figure 5. Apparatus for splitter loss measurement with silver particles.

Figure 6 is a photograph showing both of the particle generation apparatuses and the mass flow
control system.

6


-------
Figure 6. Photographs of experimental apparatus for splitter evaluation.

Figure 7 is a graph of the typical size distributions produced by each technique. Particle loss in each
splitter was determined by measuring the penetration of particles for each condition. Penetration
based on three repeat measurements was calculated as the ratio of these concentrations and error bars
are calculated based on the standard deviation. Preliminary work showed number of scans (between
three and nine scans) did not affect the calculation of penetration. For all results, loss was measured
in the flow split listed first. E.g., a result labeled "8.5 L/min and 1.5 L/min" means that the
measurements were made in the 8.5 L/min branch. "Theory" curves are calculated based on diffusion
loss from the Gormley / Kennedy equations (Diffusion from a Stream Flowing through a Cylindrical
Tube P. G. Gormley and M. Kennedy, Proceedings of the Royal Irish Academy. Section A:
Mathematical and Physical Sciences, 52, 1948 - 1950, 163-169) using the length / appropriate
flowrate of the splitter. Data is corrected for diffusion loss in the few inches of tubing connecting
upstream and downstream sampling system. This correction is on the order of 1%.

Commercial CFD software was used to model the results. Details of the computer modeling,
including boundary conditions, mesh information, and software settings is not yet available.

7


-------
1.E+08

1.E+07

o

05
CL

1.E+06

CD

0
_l

1	1.E+05

1.E+04

1	10 _	100	1000

Dp, nm

Figure 7. Particle size distributions of each particle generation source

Results - splitters
2-way splitter, % x 3/s

Figure 8 through Figure 13 shows the experimental and modeled particle loss results for the 2-
way splitter with 1/2" inlet and 3/8" outlet diameters. The silver, DOS, and salt particles have
overlapping size ranges. For all cases where the flow split branch is 1.5 L/min, there is a clear
increase in particle loss for small sizes (<10 nm). For all cases, there is no indication of any loss
for particles >100 nm, indicating that particle impaction is not a loss mechanism in the splitters.
CFD results show regions of flow recirculation in the splitters for all cases, which apparently
impacts loss more in the low flowrate cases. In all cases, the modeled results appear to be in
close agreement with the experimental results. Good agreement between model and experiment
was not the case initially, as discussed in the next section. Results from Figure 8, Figure 10, and
Figure 12 were averaged and analyzed further in the "executive summary of results" section.

Figure 8. Flow split: 5 L/min and 5 L/min.

8


-------
1.50

1.30

; no

CL 0.90

0.70

0.50

1000

Dp, nm

Figure 11. Flow split: 12.5 L/min and 12.5 L/min

9


-------
1	10	100	1000 ?

Dp, nm	

Figure 12. Flow split: 1.5 L/min and 23.5 L/min

1.50

1.30

| 1.10
"ro
oj

S. 0.90
0.70
0.50

1

Modeling issues

Figure 14 and Figure 15 show selected sample results from the initial modeling efforts. In each
example, the model result does not closely match the experimental or theoretical calculation. It
was determined that the size of the computational solution domain had an effect on some of the
results that included larger recirculation zones. All of the cases that appeared to have the
truncated recirculation zone were "re-run" with a larger solution domain. These re-run
simulations gave much better agreement in all cases.

10


-------
1.50

 1.10

q_ 0.90

0.70

0.50

1000

10	100

Dp, nm

Figure 14. 1.5 L/min and 8.5 L/min split condition showing poor agreement between model with
truncated solution domain and experimental results (compare to Figure 10).

1.50

1.30

 1.10

a! 0.90

0.70

0.50



Til DOS

Salt



Theory







Silver !





s



Model



I

10	100

Dp, nm

1000

 0.00
ImsMl

Figure 15. 1.5 L/min and 23.5 L/min split condition showing poor agreement between model
with truncated solution domain and experimental results (compare to Figure 12).

2-way splitter, 3/s x 3/s

Figure 16 through Figure 21 shows the experimental and modeled particle loss results for the 2-
way splitter with 3/8" inlet and outlet diameters. These results show very similar trends as
compared to the 1/2" x 3/8" splitter results with regards to increased diffusion loss for 1.5 L/min
cases and close agreement between modeled results and theory.

11


-------
I

I 2.50

I 2 00

0.00
(m sA-t]

1000

Dp, nm

Figure 17. Flow split: 1.5 L/min and 8.5 L/min

I

 0.00
ImsM]

1000

Dp, nm

Figure 18. Flow split: 8.5 L/min and 1.5 L/min

12


-------
Model jjjjy.

Salt



Theory







^ DOS



Silver T



I

6.67
I 5.72

1 0.00
(msA-il

> f

10	100	1000

Dp, nm

Figure 19. Flow split: 12.5 L/min and 12.5 L/min

10	100	1000

Dp, nm

Figure 20. Flow split: 23.5 L/min and 1.5 L/min

1.50

1.30

; no

CL 0.90

0.70

0.50

1000

Dp, nm

Figure 21. Flow split: 1.5 L/min and 23.5 L/min

13


-------
4-way splitter, 3/s x 3/s x 3/s x 3/s

Figure 22 through Figure 27 shows the experimental and modeled particle loss results for the 4-
way splitter. These results show very similar trends as compared to both the 1/2" x 3/8" splitter
and 3/8" x 3/8" splitter results with regards to increased diffusion loss for 1.5 L/min cases and
close agreement between modeled results and theory.

1.50

1.30

j 1.10

 0.90

0.70

0.50

10	100	1000

Dp, nm

Figure 22. Flow split: 2.5, 2.5, 2.5, 2.5 L/min

1.50

1.30

i 1.10

a. 0.90

0.70

0.50

Salt

Theory

1000

10	100

Dp, nm

Figure 23. Flow split: 1.5, 1.5, 1.5, 5.5 L/min

14


-------
1.50

1.30

| 1.10

cl 0.90

0.70

0.50

Salt

Model

Theory

Silver

10	100	1000

Dp, nm

Figure 24. Flow split: 5.5, 1.5, 1.5,

1.50

1.30

; no

cl 0.90

0.70

0.50

10	100	1000

Dp, nm

Figure 25. Flow split: 6.5, 6.5, 6.5, 6.5 L/min

1.50

1.30

; no

q1 0.90

0.70

0.50

10	100

Dp, nm

1000

Figure 26. Flow split: 1.5, 1.5, 1.5, 20.5 L/min

15


-------
Methods - dilutor measurements

Dekati provided UMN a DI-1000 dilutor on loan. The particle generation methods were described
previously and just briefly mentioned here. Figure 28 shows the apparatus to generate semi-volatile
nanoparticles that atomizes a solution of isopropyl alcohol and DOS. Figure 29 shows the apparatus
to generate silver nanoparticles using an evaporation/condensation technique. Figure 30 shows the
apparatus to generate larger salt (NaCl) particles that atomizes a salt / water solution. Particle loss in
the dilutor was determined by calculating the penetration of particles for each condition. Penetration
based on three repeat measurements was calculated as the ratio of these concentrations and error bars
are calculated based on the standard deviation. The dilution ratio of the dilutor was determined by
measuring the sample inlet flowrate with a Gilibrator and the outlet flowrate with a mass flow meter
accurate to within 1% of measured flowrate. The measured dilution ratio was within 10% of
manufacturer specifications.

For some measurements, the dilutor was heated according to the schematic show in Figure 31.

Figure 32 shows a photograph of the heated dilutor.

Silica gel diffusion dryer ^ Vent



1000 ppm dioctyl sebecate
(DOS) in n-propanol

Figure 28. Apparatus for dilutor loss measurement with DOS

16


-------
Silver

Filter

Furnace

.	^

Q Flow controller

Figure 29. Apparatus for dilutor loss measurement with silver particles

Diffusion dryer

0.008 v/v salt (NaCI) in water

3-way valve

-KX

23.5 L/min

~o

SMPS

~



Figure 30. Apparatus for dilutor loss measurement with NaCI (salt)

17


-------
Heated	Active heating

T=333K

Figure 31. Schematic of heated dilutor with temperature setpoints indicated.

Figure 32. Picture of heated dilutor (insulation used but not shown so that heating can be more

clearly seen)

Results - dilutor measurements

Figure 33 through Figure 37 shows the particle loss measurement results. Figure 33 contains
modeled particle loss results based on the CFD results shown in Figure 34. It is believed that the
model result does not perfectly match the experimental measurement because of the solution
domain effect that was discussed with the splitter results. Work is ongoing to "re-run" cases with
a larger solution domain.

18


-------
1.50

1.30

.2 1/1

I

"S

a. 0.90

0.70

0.50

Silver

Model *

DOS

10	100

Dp, nm

Figure 33. Dilutor conditions: P = 2 bar, DR = 8.16

1000

<5>	^ !* #

Vetoctty

|m s*-1J

Figure 34. CFD results for P = 2 bar, DR = 8.55

19


-------
1.50

1.30

o1-10

TO
s_

0

. 0.90

0.70

0.50

Silver

10	100

Dp, nm

Figure 35. Dilutor conditions: P = 3 bar, DR = 8.4

1000

1.50

1.30

J1"10

TO

"S

q! 0.90

0.70

0.50

1	10	100

Dp, nm

Figure 36. Dilutor conditions: P = 4 bar, DR = 9.13

1000

On different days, an entire set of conditions (silver, DOS and salt particles) was run for the
dilutor with 2 bar inlet pressure. The results are shown in Figure 37. Particle loss trends are
qualitatively reproduced quite well.

20


-------
1.50

Testl

Test2

1.30

1.10

cl 0.90

0.70

0.50

10

100

1000 1

10

100

1000

Dp, nm

Dp, nm

Figure 37. Dilutor conditions REPEAT: 2 bar, DR = 8.16

Conclusions

Particle loss in splitters is very low (< 2%) - regardless of flow split ratio and flowrate. Although
other loss mechanisms (inertial, etc.) are included in the modeling work, diffusion is the only
mechanism producing any significant loss. As such, particle loss increases when the flowrate in
the splitter decreases. For example, the number penetration is about 98% when the flowrate is
around 1.5 L/min and closer to 99% when the flowrate is 5 L/min.

In nearly all cases, particle loss in splitters determined experimentally was slightly higher than
calculations considering diffusion theory (and not any fluid mechanics effects). This suggests
that particle loss in the fluid recirculation zones resulting from the flow split is an additional loss
mechanism.

Dilutor measurements showed the number penetration of representative "aircraft exhaust
particles" through the dilutor was on the order of 90%. Slightly less mass penetrates because of
some inertial loss in the dilutor at larger sizes. Loss increased with decreasing particle size (<20
nm) and with increasing particle size (>50 nm) with the least amount of loss occurring at an
intermediate size of - 40 nm. There was no significant difference between loss measured with
the dilutor "hot" or "cold." Agreement between experimental and CFD particle loss for the 2 bar
case was reasonable but not as close as with the splitter measurements.

21


-------