See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/343569175

Spacecraft Smoke Detector Characterization with Reference and Smoke Aerosols

Conference Paper · August 2020

CITATIONS READS
0 132

8 authors, including:

Xiaoliang Wang John G Watson

Desert Research Institute Desert Research Institute
144 PUBLICATIONS   3,173 CITATIONS    882 PUBLICATIONS   39,261 CITATIONS   

SEE PROFILE SEE PROFILE

Marit E. Meyer
NASA Glenn Research Center, Cleveand Ohio
32 PUBLICATIONS   133 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Layers View project

Lunar Dust Aerosols View project

All content following this page was uploaded by Marit E. Meyer on 24 August 2020.

The user has requested enhancement of the downloaded file.

 ICES-2020-340

Spacecraft Smoke Detector Characterization with Reference

and Smoke Aerosols

Xiaoliang Wang 1, Judith C. Chow 2, and John G. Watson 3

Desert Research Institute, Reno, NV 89512, USA

Marit E. Meyer 4 and Gary Ruff 5

NASA Glenn Research Center, Cleveland, OH 44135, USA

John Easton 6
Case Western Reserve University, Cleveland, OH, 44106, USA

Gordon Berger 7
Universities Space Research Association, Cleveland, OH 44135, USA

and

Paul D. Mudgett 8
NASA Johnson Space Center, Houston, TX 77058, USA

Performance testing of consumer smoke detectors requires specific facilities and

experiments with smoldering and flaming emissions from different fuels. Smoke detectors for
use in spacecraft are tested using similar setups with representative fuel materials. To simplify
smoke detector testing, this study explored the use of laboratory-generated reference aerosols
as transfer standards to evaluate smoke detector performance. Among the three tested
reference aerosols, mineral oil particles were reproducibly generated with a Gemini smoke
detector tester, dioctyl sebacate (DOS) particles were generated with a wide concentration
range and flexible size distributions using an atomizer, while polystyrene latex (PSL) particles
were difficult to produce with the high concentrations needed for smoke detector testing.
Reference aerosols generated from 1.5%–100% DOS solutions and mineral oil covered the
response range of six types of smoke aerosols generated by oxidative pyrolysis of spacecraft-
relevant materials. Although no single reference aerosol can be used to simulate the response
of different smoke detection technologies to different smoke aerosols within ±10% error, the
relationship between reference and smoke aerosols derived from this study can be used to
predict smoke detector responses to combustion aerosols.

Nomenclature
ASTM = American Society for Testing and Materials
CO = Carbon monoxide
CO2 = Carbon dioxide

1
Research Professor, Division of Atmospheric Sciences, 2215 Raggio Pkwy, Reno, NV 89512.
2
Research Professor, Division of Atmospheric Sciences, 2215 Raggio Pkwy, Reno, NV 89512.
3
Research Professor, Division of Atmospheric Sciences, 2215 Raggio Pkwy, Reno, NV 89512.
4
Research Aerospace Engineer, Low-Gravity Exploration Technology Branch, 21000 Brookpark Road, Cleveland,
OH 44135.
5
Project Manager, Spacecraft Fire Safety Demonstration Project, 21000 Brookpark Road, Cleveland, OH 44135.
6
Associate Staff Scientist, Mechanical and Aerospace Engineering, 10900 Euclid Ave., Cleveland, OH, 44106.
7
Physical Scientist, MS 110-3, 21000 Brookpark Road, Cleveland, OH 44135.
8
Environmental Scientist, M/C SK4, 2101 NASA Pkwy, Houston, TX 77058.
COV = Coefficient of variation
CSP = Commercial Space Prototype
DMA = Differential mobility analyzer
DOS = Dioctyl sebacate
Dp = Particle mobility diameter
GASP = Gases and Aerosols from Smoldering Polymers
GRC = Glenn Research Center
Hz = hertz, frequency unit
ISS = International Space Station
NASA = National Aeronautics and Space Administration
PMMA = Poly(methyl methacrylate)
PSL = Polystyrene latex
PTFE = Polytetrafluoroethylene
SMPS = Scanning mobility particle sizer
STS = Space Transport System
N = Particle number concentration
PM2.5 = Particulate matter with aerodynamic diameter ≤2.5 µm
UL = Underwriters Laboratories
WCPC = Water condensation particle counter

I. Introduction

I N the United States, performance of consumer smoke detectors is regulated according to the Underwriters
Laboratories Standard for Smoke Alarms (UL217),1 which outlines tests, facilities, and experiments with
smoldering and flaming emissions generated from different fuels. Although there is no established standard for
spacecraft smoke detectors, they have been tested using similar methods with fuels representative of materials used in
spacecraft. To reduce the level of effort required to test smoke detectors and predict performance of future smoke
detector designs, laboratory-generated reference aerosols can be used as transfer standards. Compared to smoke
aerosols, reference aerosols can be generated with more precisely controlled chemical composition, size distribution,
concentration, and optical properties. They can be used to challenge smoke detectors with controllable and
reproducible test conditions. Once relationships between smoke detector responses to reference and smoke aerosols
are established, smoke detector responses to reference particles can be used to simulate their responses to smoke
aerosols. Specific objectives of this study are to: (1) evaluate a method to test smoke detectors using reference aerosols
with controlled size distributions and concentrations, (2) characterize smoke detector responses to reference and
combustion aerosols, and (3) determine the feasibility of using reference aerosols to predict smoke detector responses
to different types of smoke aerosols.

II. Material and Methods

A. Experimental Setup
The experiments were conducted in the Gases and Aerosols from Smoldering Polymers (GASP) Laboratory at the
National Aeronautics and Space Administration (NASA) Glenn Research Center (GRC).2 A schematic diagram of the
experimental setup is illustrated in Figure 1, which includes aerosol generators, a smoke chamber, smoke detectors,
and aerosol characterization instruments.
Reference aerosols were generated by three aerosol generators: a constant output atomizer (Model 3076, TSI Inc.),
a six-jet atomizer (Model 9306, TSI Inc.), and smoke detector tester (Model 501B, Gemini Scientific Corporation).
When particles were generated from a liquid suspension or solution, the solvent vapor was removed by a diffusion
dryer (filled with silica gel to remove water vapor or activated charcoal to remove alcohol vapor). A dilution bridge,
consisting of one path with a filter and a valve to remove particles and another path with a valve to adjust the undiluted
aerosol flow rate, was used to adjust the aerosol concentration. Dried and diluted (when necessary) particles were
injected into a 326-L acrylic-walled smoke chamber, where a mixing fan promoted concentration uniformity.
Smoke particles were generated by a tube furnace (Lindberg/Blue M). Fuel samples were loaded in a 9-cm-long
alumina boat at room temperature. The boat was then placed inside a quartz tube and inserted to the optimal heating
zone of the tube furnace. The temperature of the furnace was ramped to preset values chosen for thermal oxidative
2
International Conference on Environmental Systems
decomposition of each fuel material. Filtered air flowed through the tube furnace at 4.4 L/min to flush the pyrolysis
products into the smoke chamber.
Smoke obscuration level in the chamber was measured with an obscuration meter based on American Society for
Testing and Materials (ASTM) E662-17A.3 The light source and detector covered a broadband wavelength from 400
to 2200 nm. A Commercial Space Prototype (CSP) smoke detector based on light scattering was co-located with a
commercial ionization smoke detector (referred to as IonZG), and their analog output voltages were logged at 10 Hz.
Two consumer smoke alarms, one ionization type (Model I12010S, Kidde) and one light scattering type (Model
P3010H, Kidde), were used for comparison. Because their raw analog signals were not accessible, only the alarm
voltages were logged.
Portions of the test aerosol were extracted from the chamber through two separate sampling lines and directed to
several instruments. A DustTrak DRX (Model 8534, TSI Inc.) measured size-segregated particulate matter (PM) mass
concentrations (i.e., PM1, PM2.5, PM4, PM10, and PM15) based on laser (655 nm) light scattering.4-6 Most combustion
generated aerosols are in the PM2.5 size fraction. Due to its fast response (every 1 s), high sensitivity (1 µg/m3), and
wide dynamic concentration range (0–600 mg/m3), the DRX reading was used as an aerosol concentration reference
to compare smoke detector responses to different aerosols. Note that the mass concentrations reported by the DRX
can differ from gravimetric mass concentrations depending on particle size distribution, composition, and refractive
index. During polydisperse reference aerosol tests, a scanning mobility particle sizer (SMPS; Model 3938, TSI Inc.)7
measured particle size distributions every 135 s. The aerosol and sheath flow rates were 0.52 and 3 L/min, respectively,
resulting in a measured size range of ~0.01–0.7 µm in 108 equally spaced channels on a logarithmic scale. During
monodisperse reference aerosol tests, a differential mobility analyzer (DMA; Model 3081, TSI Inc.)8 classified
particles based on electrical mobility and a water condensation particle counter (WCPC; Model 3787, TSI Inc.)9
measured the monodisperse particle number concentrations. Due to fast changing particle size distributions, the SMPS
was not used during smoke tests. A second aerosol stream was directed to a light scattering smoke detector used in
the International Space Station (ISS) and an ionization smoke detector used in the Space Transport System (STS)
fleet.10 Neither the ISS scatter nor STS ion detector is equipped with a pump; a fan at the end of the sampling line
drew the aerosol through these two detectors. Due to potential aerosol concentration inhomogeneity in the smoke
chamber, different time for aerosols to reach detectors, and different detector response time, all detectors responded
to aerosol concentration changes at different speeds. To account for these differences, time series of detector responses
were aligned by maximizing the correlation coefficients of regression between the detector and DRX responses.

Tube Furnace for

Smoke Aerosol
Generation
HEPA
Dilution Bridge Purge
Air
Electrostatic
Classifier
(TSI Model 3080) Neutralizer Smoke Chamber
SMPS (TSI Model 3938;
Size distribution:
Filter ~0.01-0.7 µm)
Photo-
Vent Mixing Fan diode
Valve
Dryer
Obscuration WCPC (TSI Model 3787;
Meter Particle number
concentration)

TSI 3076 Atomizer

IonZG DustTrak DRX (TSI
Vent CSP
Model 8534 ; PM1, PM2.5,
Dryer PM4, PM10, PM15)

Kidde Kidde
TSI 9306 Six- Scatter Ionization ISS Scattering
Jet Atomizer Smoke Detector

Smoke Detectors STS Ionization

Smoke Detector
Pressure
Gemini 501B Balance
Smoke Detector
Tester HEPA
Vent to Roof

Figure 1. Experimental setup for smoke detector evaluation.

3
International Conference on Environmental Systems
 Each sampling day started with a ~5-minute background measurement after the smoke chamber was purged with
filtered air to concentrations ≲20 particle/cm3. Instrument baselines were also measured at the beginning of each test
before aerosols were delivered into the smoke chamber. After the aerosol test was completed and aerosol generation
ended, the chamber was purged with clean air. In most tests, data from all instruments were logged until the
concentration returned to the baseline.

B. Reference Aerosol Materials and Generation

Three types of reference aerosols were generated using three different generation techniques:
• Refined mineral oil (Rudol®) has a nominal density of 0.861 g/cm3 and a refractive index of 1.470. It is used
in the Gemini 501B smoke detector tester for smoke detector inspection and sensitivity test. Polydisperse
mineral oil particles were generated by the Gemini 501B on three different days using similar settings to test
reproducibility of particle generation and smoke detector responses.
• Dioctyl sebacate (DOS) particles are spherical with a density of 0.915 g/cm3 and a real refractive index of
1.448. The size distribution of DOS particles can be adjusted by changing the DOS solution concentration.
Four DOS solution concentrations (i.e., 1.5%, 5%, 40%, and 100%) in isopropyl alcohol were used to
generate polydisperse DOS particles with different mode diameters by the TSI 3076 atomizer. Monodisperse
DOS particles were also generated by size selection using a DMA. Monodisperse 100 nm and 200 nm
particles were generated from the 1.5% DOS solution, while 300–700 nm particles were generated from the
40% DOS solution. However, it was difficult to generate high concentrations of <500 nm monodisperse
particles to produce smoke detector signals.
• Polystyrene latex (PSL) spheres are commonly used as particle size standards. PSL particles have well
defined sizes, density (1.05 g/cm3), and refractive index (1.590). Attempts were made to generate
monodisperse PSL particles (300 nm, 500 nm, and 900 nm) by the TSI 3076 and 9306 atomizers with and
without DMA size classification. However, the particle concentrations were too low to generate reliable
smoke detector signals. Therefore, the PSL data are not presented in this paper.

C. Smoke Aerosol Materials and Generation

Six types of materials relevant to spacecraft safety were tested and their photographs before and after pyrolysis
are shown in Figure 2.
• Cotton lamp wick represents
cellulose materials such as paper,
wood, and fabrics. Thermal
decomposition of cellulose emits
carbon monoxide (CO), carbon
dioxide (CO2), tar, and char. Three
lamp wick tests utilized 0.75–1 g
mass each with the furnace
temperature set to 200–240 °C.
• Kapton, a polyimide film with
chemical formula (C22H10N2O5)n, is
stable across a wide range of
temperatures (~-269–400 °C).
Kapton is used in thin-film heaters,
wire insulation, space suits, tape and
multi-layer insulation. Kapton
thermal decomposition results in
liquid aromatic products and Figure 2. Photograph of several fuel samples before and after
spherical particles. Three Kapton pyrolysis in the tube furnace.
tests utilized 1–2 g mass each with
the furnace temperature set to 450–500 °C.
• Nomex® is a heat and flame-resistant woven textile used for cargo bags, thermal blankets, acoustic insulation,
and pressure suits. When heated, it liberates CO, CO2, hydrogen cyanide, and organic species, leaving a

4
International Conference on Environmental Systems
 carbon-rich char. Three Nomex tests utilized 0.5–1 g mass each with the furnace temperature set to 350–400
°C.
• Teflon is a brand name of polytetrafluoroethylene (PTFE, (C2F4)n), a fluoropolymer commonly used for wire
insulation, water storage bladders, sampling bags, suits, and cargo liners. Teflon releases polymer fragments
during thermal decomposition which grow into particles through nucleation, condensation, and coagulation.
Three Teflon tests utilized ~8 g mass each with the furnace temperature set to 400–500 °C.
• Poly(methyl methacrylate) (PMMA; also known as Plexiglas or acrylic glass), with a chemical formula of
(C5O2H8)n, is a window material for spacecraft components. During smoldering combustion, PMMA
undergoes thermal oxidative decomposition, generating odorous monomers and other volatile organic
compounds (VOCs), as well as CO and CO2. One PMMA test was conducted using 2 g mass with the furnace
temperature set to 200 °C.
• Fluoropolymer (FP)/polyimide (PI) wire insulation is a potential source of smoke in spacecraft due to thermal
degradation caused by wire overheating. One wire insulation test was conducted using 2 g mass with the
furnace temperature set to 450 °C.

III. Results and Discussion

A. Polydisperse Mineral Oil Particles and Smoke Detector Signal Reproducibility

Polydisperse mineral oil particles were generated on three different days (5/14/2019, 5/17/2019, and 5/23/2019)
to evaluate the reproducibility of particle size distributions and smoke detector responses. Figure 3a shows that mineral
oil particles generated on two separate days had similar number size distributions (expressed as normalized
concentration per diameter channel width on a logarithmic scale [dN/dlogDp] as a function of diameter [Dp]). The
mode diameters (385 nm and 346 nm) differed by ~10% whereas the geometric standard deviations (1.37 and 1.36)
were nearly identical. The Gemini 501B was able to generate mineral oil particles with reproducible size distributions.

(a) (b)
3.5 2.5
Normalized by Total Concentration

Normalized by Total Concentration

Dioctyl sebacate (DOS)

Particle Number Size Distribution

Particle Number Size Distribution

Mineral Oil
3.0
2.0 1.5% DOS
5/14/2019
dN/dlogDp, unitless

dN/dlogDp, unitless

2.5 5% DOS
5/17/2019 40% DOS
1.5
2.0 100% DOS

1.5
1.0

1.0
0.5
0.5

0.0 0.0
10 100 1000 10 100 1000
Particle Diameter Dp, nm Particle Diameter Dp, nm
Figure 3. Normalized particle number size distribution for: (a) polydisperse mineral oil particles generated by
the Gemini 501B smoke detector tester on two different days (size distributions were not measured on
5/23/2019); and (b) DOS particles generated from four DOS solutions with different concentrations.

Figure 4 compares smoke detector responses as a function of mineral oil particle concentrations on three different
days. Figure 4a shows that the CSP detector signal increased linearly with DRX PM2.5 concentrations over a wide
range of ~1–160 mg/m3, with linearity down to < 1 mg/m3. The regression slopes on 5/17/2019 and 5/23/2019 were
almost identical, but were ~20% higher than the slope for the 5/14/2019 test. The slightly different particle size
distributions in Figure 3a might have contributed to the slope differences between 5/14/2019 and 5/17/2019. Within
experimental uncertainties, the three slopes are deemed equivalent, with average ± standard deviation values of
0.0108±0.0013. The linearity and high correlation between the CSP detector and DRX are expected because both
devices are based on light scattering. The CSP detector is shown to generate repeatable responses to mineral oil
particles during tests on different days.
5
International Conference on Environmental Systems
 (a) Polydisperse Mineral Oil Particles: (b) Polydisperse Mineral Oil Particles:
CSP vs. PM 2.5 Obscuration vs. PM 2.5
3.2 30

26 y = 0.1356x - 0.1486

Obscuration, %/m
2.8
y = 0.0094x + 0.954 y = 0.1492x - 0.0388
22
2.4 y = 0.0112x + 0.9316 y = 0.1329x - 0.2125
CSP, V

y = 0.011x + 0.9615 18
2.0 14

5/14/2019 10
1.6
5/17/2019 6 5/14/2019
1.2 5/23/2019
2 5/17/2019
5/23/2019
0.8 -2
0 50 100 150 200 0 50 100 150 200
DRX PM 2.5 Mass Concentration, mg/m3 DRX PM 2.5 Mass Concentration, mg/m3

(c) Polydisperse Mineral Oil Particles: (d) Polydisperse Mineral Oil Particles:
IonZG vs. PM 2.5 ISS Scatter vs. PM 2.5
5.0 8.0

4.5 5/14/2019
5/17/2019 6.0
4.0

ISS Scatter, V
Ion ZG, V

5/23/2019
3.5
4.0
3.0

2.5 5/17/2019
2.0
5/23/2019
2.0

1.5 0.0
0 50 100 150 200 0 50 100 150 200
DRX PM 2.5 Mass Concentration, mg/m3 DRX PM 2.5 Mass Concentration, mg/m3

Figure 4. Variation of smoke detector responses as a function of DRX PM2.5 mass concentration for
polydisperse mineral oil particles generated by the Gemini 501B smoke detector tester on different days.

Figure 4b shows that obscuration

increased linearly with DRX PM2.5 DRX PM2.5 Mass Concentration, mg/m3
concentrations of ~5–200 mg/m3 for
mineral oil particles. The obscuration 0.08 0.8 8 80
1200 1.86
CSP Signal-to-Noise Ratio

meter signal-to-noise ratio decreased for

concentration ≲5 mg/m3. The linear 5/23/2019
1000
regression slopes from tests on three 5/17/2019 1.68

CSP Signal, V
different days were similar, with an 5/14/2019
800
overall slope of 0.145±0.011 and a 1.50
coefficient of variation (COV) of 6.3%.
600
Figure 4c shows that the IonZG
smoke detector signal decreased 1.32
400
nonlinearly with increasing DRX PM2.5
concentrations. Similar responses were 1.14
200
found for the three tests. The radioactive
sources in ionization smoke detectors
0 0.96
ionize air and generate an ion current,
0.01 0.1 1 10
which decreases when smoke particles
are present. Working on the same Obscuration, %/m
principle, the STS ion detector signal Figure 5. The CSP detector signal-to-noise ratio as a function of
also decreased nonlinearly with obscuration and DRX PM2.5 mass concentration for polydisperse
increasing PM2.5 (not shown). mineral oil particles.

6
International Conference on Environmental Systems
 The ISS scatter detector results were invalidated during the 5/14/2019 tests due to a leak in the transfer line, and
only the 5/17/2019 and 5/23/2019 data are plotted in Figure 4d. The ISS scatter detector showed similar variation with
DRX PM2.5 concentrations between the two tests: the signal increased linearly up to ~4 V at DRX PM2.5 concentrations
of ~50 mg/m3 and saturated at ~6 V with higher PM2.5 concentrations (≳70 mg/m3).
The CSP detector signal-to-noise ratios as a function of obscuration level is shown in Figure 5. The x-axis is plotted
on a logarithmic scale to better illustrate detector signal changes at lower obscuration levels. The signal-to-noise ratios
were ~73–110 at 1%/m obscuration and ~755–1125 at 10%/m obscuration, indicating reasonable sensitivity of the
CSP detector for polydisperse mineral oil particles. The CSP detector used a narrow voltage range for smoke detection:
~0.96 V at 0%/m obscuration, ≲1.2 V at 3%/m obscuration, and ≲ 3 V at 30%/m obscuration. Although the signal-
to-noise ratio was high (>50) at 1%/m obscuration, the signal resolution and reliability would be enhanced by
increasing the dynamic range of the CSP detector signal within the obscuration range of interest for smoke detection.
In summary, tests on three days over a 10-day period shows that the Gemini 501B smoke detector tester was able
to generate polydisperse mineral oil particles with reproducible size distributions and that the smoke detectors showed
reproducible responses as a function of the DRX PM2.5 concentrations.

B. Polydisperse DOS Particles and Smoke Detector Responses

The advantage of using an atomizer to generate DOS reference aerosols is that particle size distributions can be
adjusted by changing the solution concentration. Assuming that the atomizer generates the same droplet diameter with
different solution concentration, the conservation of solute mass indicates that the mean particle diameter would be
approximately linearly proportional to solution concentration to the 1/3 power. Figure 3b shows representative particle
number size distributions (normalized by total measured concentrations in each SMPS scan) for polydisperse DOS
particles generated from four solution concentrations. As expected, the mode (peak concentration) diameter increased
with DOS concentration. The size distributions were approximately lognormal, with geometric standard deviations of
1.6–1.7, wider than those for mineral oil particles (1.36–1.37 in Figure 3a). The mode diameters were: 175, 241, 385,
and ≳ 600 nm for 1.5%, 5%, 40%, and 100% DOS solutions, respectively. The SMPS did not cover the full distribution
for the 100% DOS, resulting in an inaccurate size distribution due to inaccurate multiple charge corrections.
Smoke detector responses as a function of DRX PM2.5 concentration for polydisperse DOS particles from the four
solution concentrations are compared in Figure 6. Figure 6a shows that the CSP detector voltage signal increased
linearly with DRX PM2.5 concentrations. The linear regression slope increased slightly with particle size: 0.0122,
0.0134, and 0.0166 for 1.5%, 5%, and 40% DOS concentrations, respectively. The slope for 100% DOS concentration
(0.0155) was somewhat lower than, but similar to, that for the 40% DOS concentration. The CSP detector linearity
covered a wide range of DRX PM2.5 concentrations from <1 mg/m3 to ~200 mg/m3. At high concentrations of >300
mg/m3, the signal started to saturate at ~5–8 V. The CSP vs. DRX relationships had higher variability for DOS particles
in Figure 6a than those for mineral oil particles in Figure 4a. This is probably due to the different optical designs (e.g.,
light source wavelength and scattering angle) of the CSP and DRX, which caused their light scattering responses to
be different for different size distributions.
Figure 6b shows that obscuration meter response was also linearly related to DRX PM2.5 concentrations. Similar
to the CSP detector, the linear regression slope increased slightly with particle size: 0.198, 0.225, 0.280, and 0.312 for
1.5%, 5%, 40%, and 100% DOS concentrations, respectively. The linearity extended to a wide DRX PM2.5
concentration range of ~5–200 mg/m3.
Similar to mineral oil particles, Figure 6c shows that the IonZG smoke detector signal decreased nonlinearly with
increasing DRX PM2.5 concentrations. It had good signal-to-noise ratios for concentrations ≳ 2 mg/m3. The response
per mass concentration was larger for smaller particles (1.5% and 5% DOS concentrations) than larger particles. This
is expected because for the same aerosol mass concentration, ionization smoke detectors are more sensitive to smaller
particles. On the other hand, the DRX based on light scattering is less sensitive to particles ≲300 nm, and it may
underestimate PM2.5 mass concentration for smaller particles. Figure 6d shows that the STS ion detector signal
decreased approximately linearly with increasing DRX PM2.5 concentrations up to ~80 mg/m3. Similar to the IonZG
smoke detector, the STS ion detector had higher sensitivity for smaller particles. The signal became nonlinear at high
concentrations (above ~200 mg/m3) and this detector had low signal-to-noise ratio for concentration ≲5 mg/m3.
Figure 6e shows that the ISS scatter detector signal increased approximately linearly with DRX PM2.5
concentrations between ~1-15 mg/m3. The data had a wider spread than other detectors, probably due to the extra
plumbing and lack of a pump causing non-uniform concentrations in the sensing zone of this detector. The ISS scatter
detector signal increased faster for larger particles than smaller particles as a function of DRX PM2.5 concentrations.
The signal saturated at ~6 V at higher concentrations.

7
International Conference on Environmental Systems
 (a) Polydisperse DOS Particles : (b) Polydisperse DOS Particles:
CSP vs. PM2.5 Obscuration vs. PM2.5
2.4
30

Obscuration, %/m
2.1 26
22
CSP, V

1.8 18
14
1.5
100% DOS 10
100% DOS
40% DOS 6
1.2 40% DOS
5% DOS
2 5% DOS
1.5% DOS 1.5% DOS
0.9 -2
0 20 40 60 80 0 20 40 60 80

DRX PM2.5 Mass Concentration, mg/m3 DRX PM2.5 Mass Concentration, mg/m3

(c) Polydisperse DOS Particles: (d) Polydisperse DOS Particles:

IonZG vs. PM2.5 1.8 STS Ion vs. PM2.5
5.0
1.6
4.5
1.4
Ion ZG, V

STS Ion, V
4.0
1.2
3.5
1.0
3.0 100% DOS 100% DOS
0.8
40% DOS 40% DOS
2.5 5% DOS 0.6 5% DOS
1.5% DOS 1.5% DOS
2.0 0.4
0 20 40 60 80 0 20 40 60 80

DRX PM2.5 Mass Concentration, mg/m3 DRX PM2.5 Mass Concentration, mg/m3

(e) Polydisperse DOS Particles:

ISS Scatter vs. PM2.5
7.0

6.0

5.0
ISS Scatter, V

4.0
Figure 6. Smoke detector responses as a function of
3.0
DRX PM2.5 concentration for polydisperse DOS
100% DOS
2.0
40% DOS
particles generated by the TSI 3076 atomizer using
1.0 5% DOS
four different DOS solution (in isopropyl alcohol)
1.5% DOS concentrations. Only up to 80 mg/m3 PM2.5
0.0
0 20 40 60 80 concentrations are shown to more clearly illustrate
DRX PM2.5 Mass Concentration, mg/m3 the 5% DOS case.

C. Smoke Detector Response to Smoke Particles

An example of instrument responses to smoke evolution from smoldering lamp wick is illustrated in Figure 7. All
instruments responded to smoke concentrations after the smoke filled the chamber at around 1000 s. The signal of
light scattering-based instruments (i.e., DRX, obscuration meter, CSP detector, and ISS scatter detector) increased
linearly with smoke concentration, whereas those of ionization-based instruments (i.e., IonZG and STS ion detector)
decreased nonlinearly with smoke concentration. The STS ion and ISS scatter detectors saturated at higher
concentrations. The light scattering smoke alarm (Kidde P3010H) was triggered earlier and ended later than the
ionization smoke alarm (Kidde I12010S).
To evaluate the feasibility of using reference aerosols to predict smoke detectors’ responses to combustion aerosols,
Figure 8 compares the slopes for CSP detector, obscuration meter, and ISS scatter detector when linearly regressed
against the DRX PM2.5 concentrations. Because these four measurements are based on light scattering or extinction,
signals were linearly correlated with correlation coefficients > 0.95. Mineral oil particles had the lowest slopes among
8
International Conference on Environmental Systems
all tests, while the DOS particles from the four solution concentrations covered the slope range for most tested smoke
aerosols. The regression slopes between the CSP detector and the DRX had lower variations (COV = 13%) among
different aerosols than those for the obscuration meter (COV = 26%) and ISS scatter detector (COV = 35%). While
the optical design of the CSP detector is proprietary, its optical measurement is probably similar to that of the DRX,
which detects light scattered at an angle range of 90 ± 62° relative to the laser beam.4 The larger COV for the ISS
scatter detector is likely caused by its different optical design from the DRX: it uses a near infrared laser and measures
forward light scattering (30°). Furthermore, its signal saturated at lower concentrations.

Figure 7. Example time series of instrument responses to the evolution of smoke particles from a smoldering
lamp wick test. Around 100 s, one gram of lamp wick was added to the tube furnace which was fast ramped up
to 200 °C. All instrument responded to the smoke concentration changes. The fuel boat was taken out around
1500 s and the smoke chamber was purged slowly starting from ~1560 s.

The regression slope ratios (smoke over reference aerosols) for each pair are summarized in Table 1. For the CSP
detector (Table 1a), its responses to lamp wick particles were similar to those for 5% or 100% DOS particles; its
responses to Kapton and PMMA particles were similar to those for 1.5% or 5% DOS particles; and its responses to
wire insulation particles were similar to 5% DOS particles. The CSP detector’s response to mineral oil particles were
lower than all smoke aerosols except Teflon. Polydisperse DOS particles from 5% solution were able to predict
responses of different smoke particles within an error of ±10% except for ~20% overestimation of Teflon smoke. On
the other hand, mineral oil reference particles underestimated most smoke particle responses by 20–35%, but with a
good representation for Teflon smoke particles.
The obscuration meter had linear response with the DRX over a wide PM2.5 concentration range of ~5–200 mg/m3.
However, it had low signal-to-noise ratios for PM2.5 concentration ≲5 mg/m3 (~1%/m obscuration). Therefore, the
DRX was used as the reference. As shown in Table 1b, the responses of the obscuration meter to lamp wick smoke
could be predicted by 40% or 100% DOS particles; responses to Kapton and Nomex smoke could be predicted by 1.5%
DOS; and responses to PMMA and wire insulation smoke could be predicted by 40% and 5% DOS, respectively.
Teflon particles generated the highest slope among all tested materials (Figure 8b), and its slope was 16% higher than
100% DOS particles. Mineral oil particles would underestimate the responses of all smoke particles by factors of up
to 2.5.
The responses of the ISS scatter detector to smoke aerosols differed from reference aerosols by more than 10%
except that PMMA particle signals could be predicted by 5% or 40% DOS particles within 6% error (Table 1c). Within
±20% error, 1.5% DOS particles was able to predict Kapton particles; either 5% or 40% DOS could predict Nomex,
PMMA, and wire insulation particles; and 100% DOS could predict lamp wick and Teflon particles. Similar to the
CSP detector and obscuration meter, ISS scatter detector responses to smoke particles would be underestimated by a
factor of over three if mineral oil particles were used as the reference aerosol.
9
International Conference on Environmental Systems
 Reference Aerosols Smoke Aerosols

Concentration Slope, V·m /mg

0.018

CSP Voltage vs. DRX PM2.5

3
(a)
0.016

0.014

0.012

0.010

0.008

0.40
Concentration Slope, %·m /mg

(b)
Obscuratio vs. DRX PM2.5
2

0.35

0.30

0.25

0.20

0.15

0.10
ISS Scatter Detector vs. DRX PM2.5

0.4
Concentration Slope, V·m3/mg

(c)

0.3

0.2

0.1

0.0
A
5%

5%

0%

n
n
k

ex
l
oi

ic

to

t io
f lo

M
40

om
w
1.

10

al

PM
ap
S

la
Te
S

er
O

p
S

su
K
S
O

m
D
O

in
O
D

In
La
D

M
D

ire
W

Particle Source

Figure 8. Comparison of linear regression slopes between (a) CSP detector, (b) obscuration meter, and (c) ISS
scatter detector (for signals <4 V) and DRX PM2.5 concentration for reference and smoke aerosols. The error
bars represent the linear regression slope uncertainty (standard deviation).

This study shows that while the 1.5%–100% DOS particles and mineral oil particles covered the response slope
range of most smoke particles, there was no single reference aerosol that predicted the responses of the CSP detector,
obscuration meter, and ISS scatter detector to different smoke aerosols within ±10% error. The 5% DOS particles
predicted smoke aerosol responses within ±20% error for the CSP detector, while mineral oil particles consistently
underestimated smoke aerosol responses except for Teflon particles. It is possible to use the slope ratios in Table 1 as
conversion factors to predict smoke aerosol responses by reference aerosols. Alternatively, multiple reference aerosols
can be used for different smoke detectors and different smoke particles.

10
International Conference on Environmental Systems
Table 1. Linear regression slope ratios of smoke over reference aerosols. Slopes ratios highlighted in green
indicate differences within ±10%.

Lamp Wire
(a) CSP vs. DRX Kapton Nomex Teflon PMMA
Wick Insulation
DOS 1.5% 1.20 1.06 1.14 0.89 1.03 1.14
DOS 5% 1.09 0.96 1.04 0.81 0.94 1.04
DOS 40% 0.88 0.78 0.84 0.65 0.76 0.84
DOS 100% 0.94 0.83 0.90 0.70 0.81 0.90
Mineral Oil 1.35 1.19 1.29 1.00 1.17 1.29

Lamp Wire
(b) Obscuration vs. DRX Kapton Nomex Teflon PMMA
Wick Insulation
DOS 1.5% 1.51 0.94 0.98 1.83 1.39 1.12
DOS 5% 1.32 0.83 0.86 1.61 1.22 0.98
DOS 40% 1.06 0.67 0.69 1.29 0.98 0.79
DOS 100% 0.95 0.60 0.62 1.16 0.88 0.71
Mineral Oil 2.06 1.29 1.34 2.50 1.90 1.53

Lamp Wire
(c) ISS Scatter vs. DRX Kapton Nomex Teflon PMMA
Wick Insulation
DOS 1.5% 1.97 1.10 1.28 2.11 1.40 1.66
DOS 5% 1.35 0.76 0.88 1.45 0.96 1.14
DOS 40% 1.32 0.74 0.86 1.41 0.94 1.11
DOS 100% 0.80 0.45 0.52 0.86 0.57 0.67
Mineral Oil 3.03 1.70 1.97 3.26 2.16 2.55

IV. Conclusion
Among the three types of reference aerosols (i.e., mineral oil, DOS, and PSL) tested in this study, polydisperse
mineral oil particles generated by the Gemini 501B smoke detector tester and DOS particles generated by the TSI
3076 atomizer were suitable for smoke detector testing. Repeated measurements with mineral oil particles on three
different days found that the Gemini 501B smoke detector tester generated polydisperse mineral oil particles with
reproducible particle size distributions and that smoke detectors had reproducible responses as a function of the DRX
PM2.5 concentrations. The TSI 3076 atomizer can vary DOS particle size distributions and concentrations by changing
the solution concentration and adjusting a dilution bridge, respectively. Monodisperse DOS particles of 500 nm and
700 nm can be generated with relatively high concentrations, whereas it was a challenge to produce high
concentrations of smaller monodisperse particles sufficient for smoke detector testing. It was difficult to produce high
concentrations of PSL particles, either by the TSI 3076 or 9306 atomizers.
For both reference aerosols and smoke aerosols generated from smoldering pyrolysis of spacecraft-relevant
materials, the responses of smoke detectors based on light scattering or extinction (i.e., CSP detector, obscuration
meter, and ISS scatter detector) increased linearly with DRX PM2.5 concentrations, whereas responses of ionization
smoke detectors (i.e., IonZG and STS ion detector) decreased nonlinearly with DRX PM2.5 concentrations. The CSP
detector linearity covered a wide range of DRX PM2.5 concentrations from <1 mg/m3 to 200 mg/m3, depending on
aerosol materials. While the CSP detector had high sensitivity at low smoke levels, it only used a narrow voltage range
(~0.96 V at baseline and ≲ 3 V at 30%/m obscuration), which should be increased to improve the signal resolution
and reliability.
Using the DRX PM2.5 concentrations as the smoke level reference, reference aerosols generated from 1.5%–100%
DOS solutions and mineral oil covered the response range of most smoke particles. However, no single reference
aerosol can be used to predict all smoke detectors’ responses to different smoke aerosols within ±10% error. The 5%
DOS particles were able to predict smoke aerosol responses within ±20% error for the CSP detector, whereas mineral
oil particles typically underestimated smoke aerosol responses. The relationship between reference and smoke aerosols
derived from this study can be used as conversion factors to predict smoke aerosol responses by reference aerosols.
Future work should experimentally verify the applicability of the derived conversion factors. Measuring particle size
11
International Conference on Environmental Systems
 distributions and optical properties of smoke particles, along with light scattering and absorption modeling, will help
explain the differences in smoke detectors responses and further guide smoke detector optimization and performance
prediction.

Acknowledgments and Disclaimer

This study was supported by Universities Space Research Association (USRA) Contract No. NNC13BA10B, HX5
Contract No. DRIN20D03, and NASA EPSCoR Research CAN Grant No. 80NSSC19M0152. Technical support by
and discussions with Mr. Matthew Claassen, Mr. Christopher Pestak, and Dr. David Urban are gratefully appreciated.
Trade names and trademarks are used in this report for identification only. Their usage does not constitute an official
endorsement, either expressed or implied, by the National Aeronautics and Space Administration.

References
1. UL. "UL 217: Standard for Smoke Alarms." 8th ed., Underwriters Laboratories Inc., Northbrook, IL, 2015.
2. Kulis, M. J., Meyer, M. E., Mudgett, P. D., Berger, G. M., and Pilgrim, J. S. "Heating of Printed Circuit Board, Wire
Insulation and Electronic Components for Fire Signature Sensor Evaluation," 44th International Conference on
Environmental Systems. Tucson, Arizona, 2014.
3. ASTM. "ASTM E662-17a, Standard Test Method for Specific Optical Density of Smoke Generated by Solid Materials."
American Society for Testing Materials International, West Conshohocken, PA, 2017.
4. Wang, X. L., Chancellor, G., Evenstad, J., Farnsworth, J. E., Hase, A., Olson, G. M., Sreenath, A., and Agarwal, J. K.
"A Novel Optical Instrument for Estimating Size Segregated Aerosol Mass Concentration in Real Time," Aerosol Science
and Technology Vol. 43, No. 9, 2009, pp. 939 - 950.
5. Wang, X. L., Zhou, H., Arnott, W. P., Meyer, M. E., Taylor, S., Firouzkouhi, H., Moosmüller, H., Chow, J. C., and
Watson, J. G. "Characterization of smoke for spacecraft fire safety," Journal of Aerosol Science Vol. 136, 2019, pp. 36-
47.
6. Wang, X. L., Zhou, H., Arnott, W. P., Meyer, M. E., Taylor, S., Firouzkouhi, H., Moosmüller, H., Chow, J. C., and
Watson, J. G. "Evaluation of Gas and Particle Sensors for Detecting Spacecraft-Relevant Fire Emissions " Fire Safety
Journal Vol. 113, 2020.
7. Wang, S. C., and Flagan, R. C. "Scanning Electrical Mobility Spectrometer," Aerosol Science and Technology Vol. 13,
1990, pp. 230-240.
8. Knutson, E. O., and Whitby, K. T. "Aerosol Classification by Electric Mobility: Apparatus, Theory, and Applications,"
Journal of Aerosol Science Vol. 6, 1975, pp. 443-451.
9. Hering, S. V., Stolzenburg, M. R., Quant, F. R., Oberreit, D. R., and Keady, P. B. "A Laminar-Flow, Water-Based
Condensation Particle Counter (WCPC)," Aerosol Science and Technology Vol. 39, No. 7, 2005, pp. 659 - 672.
10. Urban, D., Ruff, G., Brooker, J., Cleary, T., Yang, J., Mulholland, G., and Yuan, Z.-G. "Spacecraft Fire Detection: Smoke
Properties and Transport in Low-Gravity," 46th AIAA Aerospace Sciences Meeting and Exhibit. American Institute of
Aeronautics and Astronautics, Reno, NV, 2008, pp. 1-9.

12
International Conference on Environmental Systems

View publication stats