FINAL REPORT

ELECTRONIC NOSE: NEW TECHNOLOGY INFUSION

Advanced Environmental Monitoring and Control for

Space Station
April 30, 1999

M. A. Ryan
Mail Stop 303-308
Jet Propulsion Laboratory

(818) 354-8028
mryan@[Link]

JET PROPULSION LABORATORY

California Institute of Technology
Pasadena CA 91 109
1 . INTRODUCTION

2. OBJECTIVE
2.1 Success Criteria for Ground Testing
2.2 Success Criteria for Shuttle Flight Experiment

3. SENSOR DEVELOPMENT
3.1 Sensing Films
3.2 Laboratory Testing
3.3 Laboratory Test Input to Data Analysis

4. ELECTRONICS DEVELOPMENT: DATA ACQUISITION AND DEVICE CONTROL

4.1 Computer-Controller
4.2 Power and Clock
4.3 Sensor Substrate- Co-fired Ceramic Chips
4.4 Data Acquisition Circuit
4.5 Device Control

5. DATA ANALYSIS SOFTWARE DEVELOPMENT

5.1 Response pattern extraction
5.1.1 Noiseremoval
5.1.2 Baseline drift accommodation
5.1.3 Gas event occurrence determination
5.1.4 Resistance change calculation
5.2 Pattern Recognition --data analysis technique(s)
5.2.1 PrincipalComponentAnalysis
5.2.2 Differential Function Analysis
5.2.3 NeuralNetworks
5.2.4 Linear Algebra
5.2.5 Differential Evolution
5.2.6 LM Nonlinear least square methods
5.3 Sensor Characterization and Calibration
5.3.1 Single gas responses
5.3.2 Single gas response patterns
5.3.3 Mixed gas responses
5.4 Software Development Results
5.4.1 Training data analysis
5.4.2 Flight data analysis

6 . COORDINATION WITH JSC

6.1 ToxicologyBranch
6.2 Early Human Test Experiment
6.3 Flight Manifest and Flight Qualification
6.4 Experiment Definition and Crew Training
6.5 Flight Experiment

I
7. FLIGHTUNIT DESIGN AND FABRICATION
7.1Pneumatic Design
7.2 MechanicalDesign
7.3 Electronics Fabrication for Data Acquisition and Electronic Control
7.3.1 HP200LX Control
7.3.2 CircuitBoards
7.3.3 Microcontroller
7.4 Flight Unit Assembly

8. STS-95 FLIGHTEXPERIMENT RESULTS

9. SUCCESS CRITERIAFORELECTRONIC NOSE TESTING

9.1 Success Criteria for Ground Testing
9.2 Success Criteria for Shuttle Flight Experiment

10. CONCLUSIONS AND FURTHER WORK

1 1 . REFERENCES

12. PAPERS AND PRESENTATIONS

13. TECHNICAL STAFF

APPENDIX A: Conference Proceedings Paper

M.A. Ryan, M.L. Homer, M.G. Buehler, K.S. Manatt, B. Lau, D. Karmon and S.
Jackson; "Monitoring Space Shuttle Air for Selected Contaminants Using an Electronic
Nose,"presentedatThe 28th InternationalConference on Environmental Systems,
Danvers MA , July 12-16, 1998.

APPENDIX B: Conference Proceedings Paper

M.A. Ryan, M.L. Homer, M.G. Buehler, K.S. Manatt, F. Zee, and J. Graf; "Monitoring the
Air Quality in a ClosedChamber Using anElectronicNose,"presentedatThe 27th
InternationalConference on Environmental Systems, LakeTahoe NV, July 14-17,
1997.

APPENDIX C: Crew protocol for flight experiment

APPENDIX D: ENose STS95 flight data analysis result, LMNLS method

APPENDIX E: Report on analysis of Grab Sample Containers from STS-95

1 . INTRODUCTION

This isthefinalreport for theprogram"ElectronicNose:NewTechnology Infusion"

fundedunderAdvancedEnvironmental Monitoring and Control for Space Station,
NASA Code UL. This program extended the proofof concept work done on polymer-
based sensors for an Electronic Nose, done by Prof. N. s. Lewis at California Institute
of Technology. Prof. Lewis' work identified and demonstrated the operation of polymer
based sensors for use in sensing arrays [[Link] program at JPL built on Prof. Lewis'
work to build a prototypical miniature Electronic Nose. That ENose was flown on STS-
95 in anexperiment to determine its utility as anairqualityhncident monitor in crew
habitat on a spacecraft. This program was successful in building a prototypical model,
designingdataacquisitionandcontrol,designingdataanalysissoftware,and in
monitoring crew quarters air for a group of compounds in concentrations greater than
the 1 hour Spacecraft Maximum Allowed Concentration (SMAC). The list of compounds
and their SMACs are found in Table 3.1.

Therationale behind this program is thatthere are, atpresent, no miniature gas

sensors which are capable of detectingandidentifying a broadsuite of compounds.
Analytical instruments such as a gas chromatograph-mass spectrometer are capable of
highly accurate analysis at very low levels, but require significant crew time to operate
and maintain. Contaminants in the breathing air in an enclosed space such as the crew
quarters of the space shuttle or space station are difficult to detect because of the lack
of such a monitor; however, the need for such a capability is clear. As the air in the
space station cannotbe easily replaced, and because there are contaminants which
are likely to build up overtime, it is important for crewhealththatairqualitybe
monitored at levels ranging from the 7 day SMACs to greater than the 1 hour SMAC.
Transientcontaminantsmay build up in thefiltering systems,andcompoundswhich
cannot be as easily removed by filtering will remain in the breathing air and are cause
for concern. In addition,toxiccontaminantsmaybereleased as a result of episodic
events such as fires or spills, as well as by outgassing. Present plans for air quality
monitoring on the space station include
an ion mobility spectrometer and
electrochemical sensors for specific combustion products. Such devices are limited by
power and weight restrictions as well as by the number of compounds detectable. In
addition, extensive crew time is required to operate such monitors. The need, then, is
for low power gas monitors which can be operated with a minimum of crew time and
which can be networked into space station operating protocols.
The ultimategoal of this new technology infusion is to create an integrated
environmental monitoring andcontrolsystem for theInternational Space Stationand
beyond. Reaching this goal will involve the rapid,earlyinsertion of advanced, novel
technologies in three steps. This program addressedStepOne, which is the
development of an Autonomous ElectronicNoseand flying it as an experiment on a
shuttle mission. The purpose of this step is the demonstration that the technology can
be developed to be used as an air-quality or incident monitor in crew habitat. Step Two
involvesfurtherminiaturizingthedevice so that a network of autonomouselectronic

1
noses can be integrated on the Space Station. In such a network, multiple sensors will
functionseparatelywhile a centralsystemcollectsandanalyzes [Link] Three
involves
the
integration of environmental
control with the monitoring functions
developed in steps one and two.
Thereareseveral sensors which arespecific to particularcompounds or classes of
compounds, but using compound-specific sensors to monitor changes in breathing air
is a taskwhich would involvetheuse of severalhundreds of sensors. In addition,
compound-specific sensorsare subject to interference from molecules of structure
similar to [Link] as gas chromatography-
massspectrometry(GC-MS)havemulti-compoundcapability, but are impractical for
continuous monitoring. Work on theelectronicnosewasdesigned to fill thegap
between individual, chemically specific devices and analytical instruments such as GC-
MSwhich have multi-compound capability.

Multi-compoundsensingmaybeaccomplished using anarray of partiallyspecific

sensors, or an electronic nose. In anelectronic nose, the distributed response of an
array of several sensor heads is used to identify the changes in the composition of a
gaseous environment. This is accomplished by comparing the response of the array to
a baseline response of known, “good” air. Sensors are not specific to any one gas; it is
in the use of an array of sensors with a different sensing medium on each that gases
and gas mixturescanbeidentified by the pattern of response of the array. To date,
sensorarrayshavebeenused only in fairlyrestrictedapplications,such as the
controlledatmosphere of qualitycontrol for beer or [Link]
commercially available electronic noses, which are neither miniature nor low power.

A commonlyusedsensingmedium in other electronic noses is Sn02 or othermetal

oxide thin films, which present some difficulties in reversibility and reproducibility as well
as high power consumption. Conductive polymer sensing media have also been studied
and have been used with some success, the films tend to have short lifetimes [4,5]. In
this program, the sensor heads are conductometric sensors whichuse a thin film of
insulating polymer loaded with a conductive medium such as finely divided carbon as
the sensing layer, and operate ator near room temperature. When new compounds are
present in the previously established baseline, the resistance of the conducting polymer
film changes, and the pattern of response of the sensors makes it possible to identify
thecompound or compoundsandconcentrationsresponsible for thechange in
response.
JPL has developed, built and demonstrated a low power, miniature gas sensor which
has the capability to distinguish among and identify various gas species whichmaybe
present in the recirculated breathing air of the space shuttle or space station. This effort
includeddevelopmentandoptimization of sensing films, development of electronic
interfaces between sensors and a portable computer, and development of data analysis
software to identifyandquantifyselectedcontaminants singly or in a mixture of
contaminants. The JPL ENose device which was demonstrated on STS-95 weighs 1.4

2
kg including the computer for control, hasa volume of 1700 cm3 (18.5 cm x 1 1.5cm x 8
cm), and uses an average power 1.5 W (3 W peak power).

2. OBJECTIVE

Theobjective of this work was to develop, build anddemonstrate a miniature gas

sensor, an Electronic Nose, for experiment on a shuttle flight. The prototypical sensor
was to becapable of monitoring thecabinenvironment for thepresence of 12
contaminantsattheone-hourSpacecraft Minimum AllowableConcentration(SMAC)
levels or below. The target compounds were:

methanol ammonia formaldehyde indole

ethanol benzene Freon 1 13 methane
2-propanol co2 hydrazine toluene
In addition to thesecompounds,theENosewas built to becapable of detecting
changes in humidity.
Two compounds were eliminated from the list after the program began. Carbon dioxide
was eliminated because the polymer sensors used were insufficiently sensitive to that
[Link] because of safetyconcerns in handlingthat
compound in the ENose lab at JPL. One target compound was added to the list; that
compound is the 2-propanol wipe which was used to verify device operation during the
flight. The exact composition of the wipe is not known.

TheENosefabricatedwas a completedevice,consisting of 32 sensors in anarray,

electronics to allowcontrol of the array as well as to read the response and transfer
data from thearray to a computer,and a computer for instrumentcontrolanddata
acquisition and storage. Design of data analysis and softwareto allow identification and
quantification of the target compounds list was included in the program. Although the
original
intention of theprogramwas to developreal-timedataanalysis,after
consultation with theco-investigator, Dr. John James of theToxicologyBranchat
Johnson Space Center, it was determined that post-flight analysis would be preferable
to running the risk of false positive data analysis in real time. Analysis was done after
the flight.
In pursuit of the objective, thework to be done was divided into development tasks and
integration tasks. Thedevelopmenttasksincluded work on sensors, electronics, and
analysis software. There weretwo integration tasks in this program: coordination of the
flight experiment with JSC, and integration of the three development tasks in order to
design and fabricate theflight unit.
Success criteria were set for the overall program in consultation with the co-investigator
at JSC, with theconditionthattheENoseexperiment would not takeplace in flight

3
unless 80% of the success criteria for ground testing had been met before the flight.
Sufficientcriteriaweremetbefore flight, andtheexperimentwent as [Link]
success criteria for ground test and for flight are listed below, and are discussed in the
variousrelevantsections of thereportandthecriteriaand how theyweremetare
summarized in Section 9.

2.1 S u c c e s s Criteria For Ground Testing

1. Target compounds can be detected, identified and quantified at +/- 50% the
one-hour SMAC level or lower by applying the analysis software developed
for the purpose.
2. Mixtures of 2 & 3 targeted gases can be detected, identified and quantified
at +/- 50% theone-hour SMAClevel or lower by applyingtheanalysis
software.
3. Contaminants in a pre-mixed sample of 4 gases, can be identified and
quantified at +/- 50% the one-hour SMAC level or lower by applying the
analysis software.
4. ENose data analysis segregates compounds which are not on the target list.

2.2 S u c c e s s Criteria For Shuttle Flight Experiment

1. Successful delivery and acceptanceto flight.
2. Device operates continuously (while turned on) as programmed, and data
are retrieved after flight.
3. Sensor responses correlate with the shuttle logged events of sufficient
concentration.
4. Data analysis software correctly identifies and quantifies planned events.
5. Data analysis software correctly identifies targeted compounds ator above
the one-hour SMAC level and quantifies them at +/-!joy0, confirmed by the
GC-MS analysis of the Grab Sample Container contents.
6. Dataanalysissoftwareclassifies as “unknown” compoundswhich are
detected and are not on the target list.

3. SENSORDEVELOPMENT

3.1 SensingFilms

Successfuldevelopment of anelectronicnoserequiresdevelopment of anarray of

sensors which are partially specific to the compounds of interest. The sensing media in
theENosedeveloped in this effortwereinsulatingpolymer films made conductive by
dispersion of carbon particles in the film. These sensing films respond to a change in air
composition with achange in [Link] in thearray is
characteristic of the compound causing it. For this phase of ENose development, only
the magnitude and pattern of response were considered in film selection, as the kinetics

4
of sensor response were obscured by the time for target compounds to be delivered to
the sensors by the gas delivery system.
Several steps areincluded in the development of the film. These steps include selection
of polymers, selection of sensor geometry, design of sensor head, fabricationof sensor
chips, development of a laboratory data acquisition system, and characterization and
testing of sensing films and sensor head.
Before the start of this program, the research group of Prof. Nathan S. Lewis at Caltech
haddoneextensive work in identifying polymers from which to make films for
application in the electronic nose. Prof. Lewis and his group were involved in this stage
of the development effort by providing the data necessary to select the polymer films to
be used. Prof. Lewis' group providedJPL with response data for 85polymersmixed
with carbon particles, From those data, statistical analysis determined the polymerfilms
with thegreatestdifference in magnitudeandpattern of response for eachtarget
compound. A set of 25 polymers was generated from this analysis. Tests of solubility of
the polymers in organic solvents and of the ability of the polymer solution to disperse
carbonandproduce a homogeneous film were used to further narrow the list to 16
polymers. The polymers selectedfor use in the flight unit of the ENose are listed below.
The order in which they are listed corresponds to the order in which the data for that
polymer are reported in the response files.
C7 1 Poly(2, 4, 6-tribromostyrene),66%
A poly(4-vinylphenol)
Q poly(ethy1ene oxide)
C38 Polyamide resin
c7 Cellulose triacetate
C58 Poly(2-hydroxyethyl methacrylate)
c90 Vinyl alcohol/ vinyl butyral copolymer, 80%vinyl butyral
E15 Poly(capro1actone)
E3 Poly(vinylch1oride-co-vinyl acetate)
E4 Poly(viny1 chloride-co-vinyl acetate) 1O%vinyl acetate
E6 Poly(viny1 acetate)
E5 Poly(N -vinylpyrrolidone)
C88 Styrene/isoprene, 14/86 ABA Block copolymer
C80 Poly(viny1 stearate)
c22 Methyl vinyl ether/ maleic acid 50/50 copolymer
c20 Hydroxypropyl methyl cellulose, 10/30

The polymers used in this program were selected from the set tested by Prof. Lewis'
lab, but the pool of polymers was not optimized for response to the target compounds.
Further developmentof the ENose will focus on determining an approach to selection of
the set of polymers for a set of targetcompounds which is based in the theory of
chemical interaction of the analytes and the sensingfilms (see Section 10).

5
In addition to selection of thepolymers with which to make the sensing films, it was
necessary to selectthegeometry of theelectrodepairs across which resistance is
measured. Before the start of this program, work at JPL had developed a test chip to be
used for determination of thegeometry of the sensor. Thetestchipwasmade by
screen printing thirteen sets of gold-palladiumelectrodes on a co-firedceramic
substrate; polymer films were then deposited on each set of electrodes in order to make
a chip with several sensors. The prior work at JPL had also developed a concept for the
data acquisition hardware and a bread-board for data acquisition had been made.
Using the data acquisition hardware and software already developed at JPL, several
testchipsweretested. Films weredeposited on the sensors on thetestchip,and
response to infusions of target compounds in air were studied. Response magnitude,
reproducibilityandnoiselevelwereconsidered in order to selectthebestsensor
geometry.
The design of the test chip is shown in Figure 3.1. The chip is 25 mm X 10 mm. The
various geometries were tested for magnitude of response and noise level to determine
whichwasthebest for theENose. In thecourse of these experiments, it was
determined that the U-bend shape (sensors S4-S7) gave the most reproducible, lowest
noise response. SensorS6 was chosen for the flight design.

Figure 3.7. Design of the co-fired ceramic test chip with 73 Au-Pd electrode pairs.
Polymer films were deposited on the electrode pairs and resistance measured across
the film.

3.2 Laboratory Testing

Sensing films developed were tested in the laboratory to determine the characteristics
of the film response to contaminant. These characteristics include noise level, the limit
of response, recovery time, the pattern of response across the polymer film array, and
the
linearity of response with concentration. Film developmentrequiredseveral

6
iterations of deposit and test to determine the best conditions for making the film. It was
determinedthat a film with a baselineresistance (in cleanair) of 5-50 kQ gave a
response with the best signal to noise ratio and the fastest recovery time. All films were
deposited with resistance in this range. Thickness was not controlled; each polymerfilm
had a different thickness for it to fall within the desired resistance range.
In this phase of the program, it was found that the baseline resistance in many of the
polymer films is significantly affected by changes in temperature of as little as 0.1'C.
This change in baseline is seen in the data as baseline drift. In order to minimizethe
data analysis task, R u 0 2 heaterswereincluded in the manufacturing process on the
back of each ceramic chip to provide a constant temperature on the sensing films. The
electronic control was programmedto heat the chipsin temperature steps of 4'C, at 24,
28, 32, and 36'C. Thus, if room temperature is between 24 and 28 'C, the chips will be
heated to 28'; if it moves above 28', the chips will step in temperature to 32'. Such a
temperature change is much easier to deconvolute from the resistance data than a slow
change in temperature as would be found in a spacecraft cabin. Testing was then done
at 28' or 32'.

The films were initially tested for response using the data acquisition unit built under a
previous program. A unit capable of testing up to four chips, the number chosenfor the
flight unit, was built early in theprogram. As theelectronicsdevelopmenttask
progressed, refinements were made to the test unit, until a brassboard of the flight unit
was made and provided for all further testing. This data acquisition unit was controlled
using a PC computerand a programwaswritten in Labviewtocontrolthe unit and
acquire data. In order to deliver known concentrations of analyte to the sensors, a gas
handlingsystemwas built and a program for control of the gas handlingwritten in
Labview. A schematic of the gas handlingsystem is shown in Figure 3.2. Thedata
acquisition and gas control programs were linked so that time, concentration of analyte
in air, humidity, temperatureand flow ratewerecontrolledandrecordedalong with
resistance of each sensor at programmable intervalsfrom 10 seconds to one hour.
Theairwhich is used in the gas handlingsystem is cleanedanddehumidified using
molecular sieve and desiccant. The cleaned, dehumidified air can be humidified to a
controlledlevel by directing it through a reservoir of distilled water. The gas handling
system was built to handle liquid and solid analytes through the use of sparge tubes.
The sparge tubes direct cleaned air at a controlled flow rate into the liquid or over the
solid; air which is saturated in the analyte then leaves the sparge tube whereit is mixed
with cleaned, humidifiedair. By controllingtherelative flow rates of air going into the
sparger and the carrier air, concentration of analyte in cleaned, controlled humidity air
can be selected. Analytes in the gas phase can be plumbed in to the system to bypass
the sparge tubes and mixed with cleaned, humidified air.
After the gas handling system was built, using stainless steel valves, glass containment
and static mixers, Teflon@ and stainless steel compressionfittings and Tygon@tubing, it
was tested for accuracy of delivery of methanol using a hydrocarbonanalyzerwhich
had been calibrated on methanol. Full saturation of the air leaving the sparge tube was

7
confirmed by spargingairat a known flow rate through ethanol for 12 hours and
comparingtheweight of ethanol lost overtheperiod with thecalculatedweight loss.
Weight lost was within 2% of the calculated loss.
Afterseveral months of using the gas handling system, it was found thatthedata
analysissoftwareroutineswerequantifyingtheanalytessubstantiallylowerthanthe
presumed delivered concentration, and were reporting single gases as combination. It
was also seen that sensor recovery was significantly slower than reviously observed.
Investigation of the gas handlingsystemshowedthatthe Tygon! c tubing used in the
plumbing was reactin with the analytes and absorbing some analytes into the walls of
the tube. The Tygon8wasremovedandreplaced with Teflon@tubing. Several months
of data were rejected as contaminated or possibly contaminated, and the training sets
retaken. In addition, a regular schedule of gas system calibration using the hydrocarbon
analyzer was instituted.

-
3.3 Laboratory Test Input to Data Analysis Training Sets

Afterthecompletion of sensor geometry and polymer selection and test phases, two
sets of sensor chips weremade. Each setconsisted of four substrates, with eight
sensorseach. Four polymer films weredeposited on eachsubstrate, to make two
sensors of each polymer on the chip. In this way there were 32 sensors made of 16
polymers, so that if any sensor stopped working, there was another sensor of the same
material available. Two sets of chips were made; one was arbitrarily labeled "Flight" and
the other "Flight Spare"sets.
Training sets for use by the software analysis task were made using both sets of chips.
The data from these training sets were used to determine which method or methods of
data analysis would beused on the flight data. Thetraining sets werealsoused to
determine the JPL Electronic Nose's limit of detection in this program for each of the
target compounds. For compounds with relatively high 1-hour SMACs, such as ethanol
(SMAC = 2000 ppm), sensor response was not pushed to the lower limit of detection; a
moderate concentration for the gas handling system was selected as thelower limit.
Table 3.1 lists the limit of detection for each compound measured, as well as the
compound's SMAC.

8
 exhaust

3 Sccm

mfc
- exhaust

T
Sourc
I
I test chamber
Air
filter

solenoid valve

check valve

6 flow rate
control valve

water
25 slm

mfc

Figure 3.2 Gas handling system builtto be able to deliver combinations of

three gases and select humidity and [Link] by a PC using
Labview.
TheDataAnalysis subtask required a database of sensor array responses to use for
extracting a signature, or fingerprint, pattern for each gas. The flight and spare chips
were exposed a series of gas events requestedby the software [Link] this series
of events, the sensorswere exposed repeatedly to different concentrationsof the target
compounds. Those data were then used by theDataAnalysistask to determine the
fingerprints for each gas. A typical single gas event series would expose the sensor
array to a series of concentrations of two gases.

9
Concentrationschangeswere not organized in eitherascending or descending
concentrations. The concentration ranges to which the sensor arrays were exposed for
each gas is shown in Table 3.1, and a typical single gas exposure event sequence is
shown in Figure 3.3. Theconcentrationrangesweredeterminedprimarily from the
rangesaccessible to the gas handlingsystem for eachcompound,based on the
saturated vapor pressure of the compound and the flow rates achievable in the system.
Results from earlytraining sets anddataanalysisarepresented in thepaper
“Monitoring Space Shuttle Air for Selected Contaminants Using an Electronic Nose,”
presentedatthe 28th InternationalConference on Environmental Systems, Danvers
MA, July 12-16 , 1998. This paper is included in this report as Appendix A.

Table 3.1 Target compounds for electronic nose shuttle experiment and JPL limits of
detection. CO, and hydrazine were deleted fromthe list; see Section 2.
Compound
Detected
on 1 hr SMAC Detection
Testing
shuttle
(ppm) [6] (ppm) [7,8] Limit (ppm)
Range
(PPW

alcohols
methanol c1 30 5 5 - 300
ethanol .5 - 5 2000 50 10 - 130
2-propanol .4 - 4 400 50 30 - 160
ammonia 0 30 20 10 - 50
benzene c .I 10 10 10 - 150
co2 320 13000 * *

indole 0 1 0.03 .006 - .06

* *
hydrazine 0 4
methane 1-10 5300 3000 1000 - 7000
formaldehyde 0 .4 25 25 - 510
Freon 113 .I - 1 50 20 20 - 600
toluene .4 - 4 16 15 15 60 -
* compound eliminated from target list; not tested at JPL

10
 350 1 100

300
80
h
E 250
.-Lx
W
P
P
200
60 E
.-c0
150
C
Q)
0
c 100
0
0
50

0 0
0 250 500 750 1000 1250 1500
Time (min.)

Figure 3.3 Typical gas exposure event sequence for training sets. U p to three
gases can be delivered simultaneously, and relative humidity can be varied.
Training events weretypically 30 minutes of compound with 60 minutes of
clean air between them.
An exhaustive set of mixed gas training sets was not run because of thetime
constraints after changing the tubing in the gas handling system (see Section 3.2).The
Data Analysis task was provided with data which made it possible to judge the additivity
of the gases andthecapability to deconvolutethesignals. Mixed gas resultsare
discussed in Section 5.3.3.

11
4. ELECTRONICS DEVELOPMENT: DATA ACQUISITION AND DEVICE CONTROL

Device and data control for the ENose consists of controlling the pneumatic system, the
pump and solenoid valve, and interrogating the sensors at a known time and interval.
Dataacquisitionconsists of measuringthesensorvoltageunder known current
conditions and convertingit to resistance, computing the sensor resistance change, and
storing theintermediateandfinalresults onto a flashmemorycard for ground base
[Link] for
the laboratory testing of sensor substratesand polymer films described in Section 3.

4.1 Computer-Controller
At the beginning of the program, the plan was to fabricate data acquisition and control
hardware with memory, and to include a serial port for transfer of the stored data. Early
work wasdone using a desk-topcomputer for control,and a NationalInstruments
DAQpad to send commands to open and close valves and to acquire resistance data
from the sensors by measuringthevoltageat a known currentprovided by the
[Link] into availablehardwareshowedthat it would save considerable
timeandcost to useanoff-the-shelfHewlettPackard 200LX palm-topcomputer for
control. Using this computer for control required designing a circuit to operate the pump
and the solenoid valve, and to acquire resistance data from the sensors. The HP200LX
can be used with a flash card for memory and is equipped with an infra-red port for data
transfer. For this application, the HP200LX was equipped with 6 MB of flash memory
and 2 MB of built in RAM, sufficient for 15 days of monitoring. Largermemorycanbe
obtained..
Thecircuitrydesigned for use with the HP200LX wasfabricatedandtested in the
laboratory using a desk top or lap top PC for convenience of [Link]
hardware developed for sensor development work was also operated in the laboratory
using a desktop computer.

4.2 Power and Clock

Development of theelectronics for dataacquisitionanddevicecontrolincludedthe
ENose power up sequencing and provision of power for the device. The intention at the
beginning of theprogramwas to providepower with batteries. The safety and flight
qualification issues associated with the use of batteries made it difficult to include them
on a shuttle experiment so the orbiter 28 V dc power bus was used to power the device
and the computer. Because of safety considerations, the batteries were not used in the
palm-top computer. A power supply and circuitry designed and built for this application
wereused to boot thecomputerand to transferthe flight software from theflash
memory to RAM.
The computer is able to keep accurate track of time elapsed, but there was no way to
record the time of turning on the device except by manual recording by a crew member.
To avoidhaving to depend on a crewmember for accurate timekeeping, a real-time
clock was added to the device, so that when the HP200LX was turned on, the clock

12
would set the time in the computer, The real-time clock was set at JPL before delivery
to JSC. One LiS0Cl2 battery was includedin the device, to power the real-time clock.

-
4.3 Sensor Substrate Co-fired Ceramic Chips
As discussed in Section 3.1, a gas sensor test chip was designed to allow experiments
with sensor size and geometry, in order to determine the best combination of size and
geometry for this application. The design of the test chip and its original circuit placed
each sensor on the chip sequentially in the feedback loop of an operational amplifier
using multiplexing circuitry. Five requirements were met in the design:
a) Common bussing was used to conserve pins
b) Kelvinvoltagesensingwasused in the surround circuitry to avoidanalog
switch resistor voltage drops
c) Electrode guarding was used to eliminate leakage currents between sensors.
d) All sensors except the one being interrogated were grounded.
e) Structuralgeometryvariationswereused to separate contactand sheet
resistance
The first four characteristics were kept in the flight chip design.
25 mm

Y
Au-Pd Electrodes 'bolyrner /
Films Ground Ring
I
Dielectric Layer

Figure 4.1 Design of the flight chip with 8 U-bend electrode sets in Au-Pd. The
unwired terminalson thechip go to Ru02 heaters and the controllingheading
thermistor on the backof the alumina substrate

It was found that the co-fired ceramic technology used to make the sensor substrates
could result in pinholes in the insulating glass, especially at the vias, and the method of
depositingpolymersolutions would thenresult in cross-talkbetween sensors. In
addition, in some cases, polymer could bridgethe space between sensors. To avoid
cross talk, two approaches were taken: The design of the sensor was changed slightly
to move the vias up into a region where solution would not touch them, and grounding
ringswereplacedaroundall sensors to ground any current flowing between sensors.
Further work in sensor film development led to depositions which did not bridge, and
work with the substrate fabrication process led to pin-hole free chips,but the grounding
rings werekept as a safety measure. The design of the substrate used to makethe
flight unit sensors is shown in Figure 4.1.

13
4.4 Data Acquisition Circuit
At the outset of this program, it wasclearthatanapproach
s
to data ac uisition that
would be capable of resolving resistance changes on the order of 1 in 10 or 1 in 1O6
would be necessary. As the relationship of resistance change to gas concentration was
not known at the beginning of the program, the data acquisition task assumed that the
most difficult case would be that 10 ppm change in gas concentration would translate to
10 ppm resistance change. In addition, as the task was to miniaturize the entire system,
an approach to measurement that did not require extensive instrumentation as well as
an approach that would be cost effective for this program was necessary. In practice,
the typical resistance change for 10-50 ppm of contaminant is on the order of 2 ~ 1 O - ~
(200 ppm resistance change),and and may be as small as 1x1 Oe5 (1 0 ppm resistance
change).

12-bit
"*I$

L/ Vout

I VSEN

GAS
SENSOR
CHIP

v1

Figure 4.2 Breadboard: 12 Bit Dual OffsetNulling Amplifier

Small changes in resistance are measured using a 12 Bit DualOffset Nulling Amp, in
which a known current is put through the sensor resistor Rx by V l i and fixed resistor
RO. The resistance measurement breadboard circuit diagram is shown in Figure 4.2. It
is designed to allowthemeasurement of film resistancechanges as 1 in lo5, to
eliminatecrosstalkbetween sensors, and to minimize sensor chip pin [Link]
current setting voltage for each sensor, V1 i , is determined by a 12-bit binary search in
the range of 0-2.5V. The search is performed by setting V2i and V l i to approximately
midrange, and determining whether the output, Vout, is high or low with respect to V2i.

14
V l i is rangedhigher or lower for 11 additional steps so thatthe U 2 voltage (Vout)
approaches the linear region of U2. This ensures that the output of U1 is within its linear
range. Subsequently, V2i is ranged in a smiliarmanner with Vli fixed so that Vout is
within the linear range of U2. With the circuit operating in this linear fashion, R ~ E can
N
be determined.
Thevoltage across the sensor is determined with precision by subtracting V2i from
[Link] difference is then multiplied by a fixed gain, (R2/R1) + 1 , where R2 and R1 are
fixed resistors. For eachmeasurement,the DACs and ADC are locked to the same
voltagereference,where DAC is Digital to AnalogConverter (12 Bit MAX538 and
MAX539), and ADC is Analog to Digital Converter (12 Bit LTC1286).

The architecture of the sensor substrate,the shaded region in Figure 4.2, indicates that
one side of each sensor, RsENi is connected to a common node which is connected to
theinverting input terminal of OperationalAmplifier U 1 , Resistance caused by cross-
talkand not by polymer bridging, RcrOSS,is eliminated by grounding allunused sensor
nodes on either side of the sensor under test. This feature became unnecessary as
sensor development proceeded,but it was not removed from the circuit.
In practice, detection of changes in polymer resistance of 1 in lo5 was achieved. The
ENose circuit implementation is unique in that the circuit works with the HP200LX in the
loop. Since the resistance changes are relativelyslow, V l i and V2i data for each sensor
film are stored in microcontroller memory and used for later measurements. Only if Vout
is out of range arethese [Link] is, the HP200LX dynamically
checks Vout. I f Vout is out of thelinearrange of U2, the HP200LX requeststhe
microcontroller to reset V1 i and V2i to match Vout to within the ADC 12 bit resolution.
This amplified remainder is digitized with the 12 bit Analog-to-Digital Converter (ADC)
and signal averaged 8 times. The reported resistance change has the equivalent of 18-
20 bit resolution.

4.5 Device Control

ENose device control was implemented to be as autonomous as possible, to minimize
theneed for crewinteraction in theultimatedesign of such a device as anincident
monitor. Nevertheless,amplefeedback is available to theoperator to verifyproper
operation. Upon power up, the ENose goes through a 60 second power up sequence,
which includes self-test, and guarantees a clean reset. During the power up sequence
the solenoid is toggled on and off to verify valve operation. The HP200LX then boots,
requeststhetime from therealtimeclock,and enables the gas flow system pump.
Upon verifying the action of all the control parts, the computer commands the ENose
microcontroller to find the operating points of the 32 individual sensors. Upon verifying
the action of all the control parts, the computer commands the ENose microcontroller to
find the operating points of the 32 individual sensors. These operatingpoints are the V1
voltageandthe DAC valueproportional to the sense current for eachsensor. The
HP200LX then requests data periodically by serial command. Data are cataloged and
tagged, and sensors are rescaled if drift or response makes it necessary. Instrument
activity is indicated by LEDs on the side of the enclosure, one for "Power & Measure"

15
andone for "Baseline". Closed loop temperaturecontrol is done by theENose
hardware using ananalogfeedback loop. (See discussion on need for temperature
control in section 3.2).

5. DATA ANALYSIS SOFTWARE DEVELOPMENT

Dataanalysis is anintegralandimportant step in developingtheENose. Without

adequate data analysis, the data acquiredby ENose hardware will be meaningless. High-
level data analysis tools are needed to deconvolute the gases and gas concentrations
from theresponsepatternsacrossthesensorarray. During this program,thedata
analysistaskfocused on development of a dataanalysismethodthatcancorrectly
identify and accurately quantifya gas event of single or mixed gases.
For anelectronicnosesystem,thetask of identifyingandquantifying a gas event is
roughly a two-step procedure:
1) pattern extraction, where the response pattern of a gas event is extracted
for further analysis, and
2) pattern recognition, where the gas event is identified and quantified on the
response pattern extracted.
Before these steps can be taken, a database of expected gases must be compiled; this
database was taken in the ENose lab as described in Section 3.3.
The work of designing a series of software routines to go through the steps described
below was done using MATLAB, from Mathworks, Inc. Several routines from MATLAB
andother sources wereuseddirectly or modified for this application. MATLAB is a
flexible program, and thus appealing for development of software, but it runs relatively
slowly. For future use, wherereal-time or quasi-realtimeanalysis is desired, the
routines can be translated into C and run on a desk top or lap-top computer.

5.1 R e s p o n s e PatternExtraction
In order to extracttheresistanceresponsepatternaccurately from raw time-series
resistancedata it must be pre-processed. This conversion is important because for
sensingmedia such as theconductingpolymer/carbon films used in this program,
relativeresponsechangeshavebeen found to bemorereliablethantheresponse
shapes. Theexactmethod of extractingtheresponsepatternmaybeapplication
dependent, but in general it will involve four sequential steps: 1) Noiseremoval, 2)
Baseline drifting accommodation, 3) Gaseventoccurrencedetermination,and 4)
Resistance change calculation. Figure 5.1 shows the results of pre-processing.
5.1.1 Noise removal A sensor's response to a gas event might beburied in noise,
especially at the gas concentrations targeted in this program (1 - 100 ppm). Themain
source of this noise is responsefluctuation in thesensing films. Somepolymer films
were noisier than others; that noise could have been caused by high sensitivity of the

16
film to small changes in pressurecaused by air flow. Non-uniformities in the film
thicknessandcarbondispersion could alsoberesponsible for noise. In general, this
fluctuation is fast compared to the response to a gas event. It is expected that a less
responsivesensingmedium will have a [Link] first step in the
preprocessing is therefore to filter out this high frequency fluctuation using appropriate
digital filtering.

Themethod for noiseremovalusedhere is zero-phase forwardandreversedigital

[Link] of thefiltermaybedifferent for different sensors andcanbe
determined by trial and error. Figure 5.2 shows a representative result of the effect of
this filtering to one sensor.

I I '

6)
0
8
Y
v1
32
'CI
.d
v1

2
1 10 20 30

Time
(a)
Figure 5.1 Preprocessing converts the recorded time-series resistance data shown in
(a) to resistance response pattern, shownin (b).

5.1.2 Baseline drift accommodation Baseline drift is one of the most difficult
problems to besolved in analyzingResistance vs. Time data from [Link]
causes for thebaseline drift canbemultiple,andincludevariations in temperature,
humidity, or pressure, aging of the sensors, and sensor saturation. However, there is at
present no clear understanding of the underlying mechanism of any one of the causes,
which makes drift compensating attempts very difficult.

In general, the baseline drift is slowly-varying in nature compared to the response time
of a detectablegasevent, whetherthe drift is caused by temperature,pressure or
humidity variations, or some other cause. This difference in time scale enables the use
of a long-lengthdigitalfilter to determinetheapproximatebaseline drift andthen
subtract it from theraw data. Figure5.3a shows the baseline drift determined in one
sensor; Figure 5.3b shows the result of processing the data by with both high and low
frequency filters. Although this will not accommodate the drift totally, it will eliminate the
effect to a manageable degree.

17
 X lo4 Oct291, sensor 9
8.54 I

8.515 I
0 100 200 300 400 500 600 700 f IO
start at Nov 4 , 6:34 (GMT) (=26549674).in rninites

Figure 5.2 Green line:raw resistance [Link] line: smoothed resistance

5.1.3 Gas event occurrence determination A scheme forautomateddetermination

of whether and when a gas event occurs has been developed. It is based primarily on
thresholdcalculation, in whichtheresistance change over acertaintimeinterval is
calculated, and a time-stamp is registered if the change exceeds a pre-set threshold.
This routine can detect most gas events; however, it was also found that it tends to
falsely identify baseline drift or noise as gas events. Consequently, in this experiment,
the determination of a gas event in practice was largelydone by visual inspection of the
events selected by the routine. In this way, effort was focused on gas identification and
quantification.

5.1.4 Resistance change calculation The calculation of t h e resistance change may

notbeobvious, because atlow gasconcentrationrangetheresponsesdonothave
consistentcharacteristic shapes. However, it is expectedthatdifferentsensors have
relatively
different
response strengths, It is this relative responsiveness which
determines t h e fingerprint of that gas - the response pattern. To preserve this relative
responsiveness it is important that the any calculation method of the resistance change
should be taken at t h e same time stamp after the initial presentation of a gas.

-
5.2 Pattern Recognition Data Analysis Techniques
Although response patterns such as the one shown in Figure 5.1b can provide hints to
identify different gases, more quantitative knowledge can be obtained only with the help
of dedicateddataanalysistools to identifyandquantifydetected gases. Various
approaches to response patternanalysisandclassificationhavebeen studied and
[Link] its advantages anddisadvantages. In thefollowingsections
some of t h e dataanalysismethodswhichhave been evaluated in this program are
reviewed.

18
 X lo4 [Link] 9
8 541

8.515
0 100 300200 400 500 600 700

8 53

8 525

852
8.515'
0
t 100 300200 400 500 600 700
I
800
start at Nov4, 6.34 (GMT) (=26549674),in mlnites

Figure 5.3 a) Blue line: baseline drift. b) Red line: smoothed and baseline drift
corrected resistance

5.2.1 Principal Component Analysis PrincipalComponentAnalysis (PCA) is one of

the most widelyusedmethods for electronic noses and other array-based sensors. It
performs linear
projection of multivariate data sets into a few informative and
independent axes. The new axes are the principal components of the data stream, and
the separations between data samples are readily visualized in this transformed data
space. Nevertheless it is not optimal for classification since it ignores the identity (class
label) of the gas examples in the database.

5.2.2 DifferentialFunctionAnalysis Differential Function Analysis (DFA) is another

commonly used statisticalapproach to analysis of arrays of data. Its goal is to find
projections(hyperplanes)thatmaximizethedistancebetweenexamples of different
groups yet minimize the distance between examples of the same groups. Compared to
PCA, DFA tends to do better with a projection that contains subtle, but possibly crucial,
gas-discriminatory information, while PCA may do better with a projection that contains
high-variance random noise. Unlike PCA, DFA looks at the class label of each example
and is therefore more appropriate for classification purposes.

19
5.2.3 NeuralNetworks NeuralNetworks (NN) are another popular approach to data
analysis for arraybased sensors. Neuralnetworkscanbetaught to simulateany
function. For this application, N N software would try to find a best-fit function (linear or
nonlinear; no models needed) thattakesrelativeresistancechanges as inputs and
outputs the gas classification. N N is good for generalization of functions to cases
outside the training set and is more suitable than DFA when the sensor signatures of
two gases are not separable by a hyperplane ([Link] gas has a signature surrounding
thesignatures of another gas.) N N generallyrequires "clean" trainingdataand is
thereforelessamenable to noise or drift in thetraining data than most other
approaches. N N s are inferior to DFA in classifying data sets which may overlap.
5.2.4 Linear Algebra Neither DFA nor N N s are wellsuited to recognizing the sensor
signatures from combinations of more than one gas. For this scenario, one can use a
linear algebra (LA) based approach. LA tries to solve the equation y=Ax, where vector y
is anobservation (a responsepattern), vector x is the cause for theobservation
(concentrations of a gas or combinations of gases), andmatrix A describessystem
characteristics(gassignaturesobtained from training data). Thereare two ways to
solvetheequation:directmethod or matrixpseudo-inverse,andleast-squares fitting
method. For ENose data analysis where the response pattern can be noise corrupted,
so theremayexist no exact solution, theleast-squares fitting method is preferred.
Among N N , DFA, and LA, LA has been found to perform best; however, LA is suitable
only if the training data are linear.

5.2.5 DifferentialEvolution Differential Evolution (DE) representssomerecently

emerged so-called genetic algorithms. It is a parallel direct search optimization tool. It
begins with an initial randomly-chosen population of parameter vectors, adding random
vector differentials to the best-so-far solution in order to perturb it. A one-way crossover
operation then replaces parameters in the targeted population vector with some (or all)
of theparametervalues from this "noisy" best-so-farvector. In essence it imitates
principles of geneticsandnaturalevolution by operating on a population of possible
solutions using so-calledgeneticoperators,recombination,inversion,mutationand
[Link] to the optimum solution arecheckedandinformationabout
[Link] is simple,theconvergence is fastandthe
required human interface is minimal: no more than three factors need be selected for a
specific application. However the last advantage is also its disadvantage: limited control
for ENose data analysis.

5.2.6 LM NonlinearLeast Squares Method For nonlinearmodelsthetechnique of

choice for least-squares fitting is the iterative
damped least-square method of
Levenberg and Marquardt, hereafter referredto as LM-NLS. Similar to LA, LM-NLS tries
to find the best-fit parameter vectorx from an observation vector y, which is related to x
through a known linear or nonlinear function, y=f(A,x),e.g. Y = A I x + A ~ x, where
~ A1 and
A2 are system characteristics obtained from training data. This method begins from a
given starting point of x, calculates the discrepancyof the fit:
residual =( computed-observed)lo,

20
where 0 is the standard deviation, and updates with a better-fitted parameter x at each
step. LM-NLS automaticallyadjuststheparameterstep to assure a reduction in the
residual:increasedamping(reducestep) for a highly nonlinearproblem, decrease
damping (increase step)for a linear problem. Because of this ability to adjust damping,
LM-NLS is adaptive to both linearandnonlinearproblemsand is thedataanalysis
method used in this ENose program.
5.3 Sensor Characterization and Calibration
Whatever the method selected, before it can be used to analyze unknown gas events, it
must betrained or [Link], first and
foremost,providesthebasis for lateraccurateidentification of the gas typeand
quantification of gas concentration. It also provides important feedback information for
the selection and developmentof the data analysis method itself.

During this ENose project,

lab training data of known compounds at known
concentrations of about 550 gas events of singletarget gases andseveralpairs of
mixtures have been collected. The following summarizes what has been observed with
these data.

5.3.1 Single gasresponses In thecourse of this work, it was found thatthe

response of the films to the target compounds is linear with concentration only within a
[Link] testsdone in Prof. Lewis'labweredoneatsubstantiallyhigher
concentrations than the SMACs for this program, and found responses to be linear. The
data analysis approach used in Prof. Lewis' work relies on linearity,andcould not be
used here. The nonlinearities in the training data generated for this program appear to
be of low order, but successful identification and quantification of gas events must take
the
nonlinearities into [Link] 5.4 shows the sensor responses vs.
concentration to a single gas, ethanol.
To obtain sensor characteristics without furtherknowledge of sensor nonlinearities, a
secondorderpolynomial fit wasused to [Link] sensor
response to each gas, the program finds the best-fit coefficients A1 and A2 (in the least-
squares sense) to the following equation:
resistance change = A&, + A&:
where Cg is gas concentration. The fit is constrained to pass through the origin. A, and
A2 are 13x13 matrices characterizing the sensors' response to ten targeted gases plus
water, humidity change, and the propanol wipe.

5.3.2 Single gas response patterns Because of thenonlinearity of response with

concentration,there is no single signature or fingerprint for one gas allat
concentrations. In general, the response patternsfor one specific gas remain similar for
the concentration range of interest. Figure 5.5 shows representative response patterns
for all sensors to the ten target gas compounds at a median concentration level.

21
 20 I00 4 10

0 50 2 0

-20 0 0 -10
0 100 200 0 100 200 0 I00 200 0 0
5 10 20 2

5 0 0

0 0 -20 -2
0 IO0 200 0 100 200 0 100 200 0 100 200
2 2 5 40

0 0 0 20

-2 -2 -5 0
0 I00 200 0 100 200 0 I00 200 0 100 200
5 2 20

0 10

0 -2 0
0 I00 200 0 100 200 0 100 200 0 100 200

Figure 5.4 Sensor resistance changes (vertical axis)vs, gas concentrations (horizontal)
for ethanol. Onepair of sensors in each subplot. Low-order nonlinearity was observed.
If+"and "0":experimental data points; lines: least-squares fitting curves

It is clear from the signature patterns in Figure 5.5 that similar response patterns were
observed for somedifferent gases. For example,ethanolandmethanolhavesimilar
signature patterns. Regression analysis also pointed out linear dependency to certain
degrees. This means that signature pattern of one gas could be expressed as a linear
combination of the response pattern generated by some other target gases. To reduce
this similarity,the sensors' raw resistanceresponses must bemodified by different
weights in the data analysis procedure.

22
 benzene ethanol freon indole
1 , ,, I 1 I 1

0.5 0.5 0.5 0.5

n n 0
0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30
methane methanol propanol toluene
1 1

0.5 0.5 0.5

n 0 0
0 10 TO 30 0 10 20 30 0 10 20 30 0 IO 20 30
ammonla formaldhyde wipe
1 1

0.5 0.5 0.5

n n 0
0 10 20 30 0 10 20 30 0 10 20 30

Figure 5.5 Representativesignatures of tentargeted gas compoundsplus wipe

generated by ENose sensors. Notice the similaritybetween ethanol and methanol.

5.3.3 Mixed gas responses Deconvolution for identificationandquantification of

mixtures relies on the additivity of the sensor responses. Here,additivity means that the
strength of the response to mixtures of gas A at level CA and gas B at level cB equals
the response of gas A at level CA plus the response of gas B at level cB; (cA + c B ) =
c A + c B . An exhaustive set of gas pairs was not run because of time constraints after
re-plumbing the gas-delivery system with PTFE tubing, so it was necessary to test the
additivity of gas pairs on a selected group of mixtures. For this relatively small pool of
data, additivity holds for the gas combinations run: ethanol + formaldehyde, ethanol +
ammonia, ammonia + benzene, methanol + toluene, methanol + benzene, benzene +
propanol, benzene + methanol, as shown in Figure 5.6.

5.4 Software Development Results

5.4.1 Training-data analysis Initialsoftwaredesignfocussed on using Principal
ComponentsAnalysisandLinearAlgebra for identificationandquantification of
[Link] adequate, but not optimum to [Link]
reported in the paper found in Appendix A were analyzed using those methods, and it
was through analysis with PCA and LA thattheproblems in the Tygon tubing were
found. Afterchanging out the Tygon tubing, andnewtraining sets were run, possible
approaches to data analysis were reconsidered to check on linearity.
The LM-Nonlinear Least Squares method was chosen for ENose data analysis because
it had a higher success rate for identification and quantification than any other methods

23
that were explored for this application. Here "success"means correctly identifying target
compounds and quantifying them to+/- 50%.
Forlab-controlled gas events, theoverall success ratereaches -85% fortargeted
singlesandabout -60% for mixtures. Brokendowninto individual singlesor mixture
pairs,thesuccessratesforsinglesarelistedbelow in Table 5.1. Theconcentration
in Table 3.1, where
ranges for each singlegas are also given. Lower concentrations listed
thedetection limit at JPL is listed,weredoneearly in the program with PCAand LA
analysis.

Table 5.1 Identification and quantificationsuccess rates for single gases.

The ranges shown here are ranges usedin LMNLS analysis.
ConcentrationCompound Success
Range (PPm) Rate ("Xi)
Ammonia 10 - 50 100
Benzene 20 - 150 88
Ethanol 10 - 130 87
Freon 113 50 - 525 80
Formaldehyde 50 - 510 100
Indole ,006 - .06 80
Methane 3000 - 7000 75
Methanol 10 - 300 63
Propanol 75 - 180 80
Toluene 30 - 60 50

Considering that the raw dataare often very noisy at low concentrations, nonlinear athigh
concentrations, highly correlated in some cases, andweakly additive in some mixtures,
these results demonstrate that the LM-NLS method is an effective technique for analysis
of an array of sensors. FutureworkontheENose will attempt to remove many of the
impediments to data analysis, with focus on noise and correlation.

5.4.2 Data AnalysisSoftwareResults Theability of thedataanalysissoftwareto

identifyandquantifysingleandmultiple gas events in cleanair was tested in the
laboratory. The targeted concentrations range for quantification was 30% to 300% of
t h e one hour SMAC for each [Link] can be seen from Table 3.1, in some cases
it waspossibleto identify and quantifysubstantiallybelow the 30% concentration;
however, in a few cases quantification was successful only as low as 100% of the one-
hour SMAC. In one case, that of formaldehyde, we were unable to reliably identify and
quantifybelowseveraltimes t h e one-hour SMAC. Identification and quantification of
single gases at the one-hour SMAC level was successful. Figures 5.7 and 5.8 show
some results of single gas identification and quantification graphically.
Single gas identification and quantification done using PCA is shown in the Figures with
the paper found in Appendix A.

24
 200
n 180
E
Q 160
Q
v
140
120
100
80

s
i.
60
40
20
0
0 20 40 60 80 100 120
140
160
180 200
Concentration delivered (ppm)

Figure 5.7 Identification and quantification o f four single gases using LM-NLS.
The shaded area is the target +/- 50% detection range.

An exhaustive set of multiple gases was not run; the software design was such that
testingseveral groups would besufficient to testtheability of thesoftware to
[Link] of mixtures in cleanairwas
moderately successful. Additive linearity holds for some combinations in concentration
rangesnearthe SMAC level of thelower SMAC-compound. The success rate for
double gases wassomewhatlessthanthat of single gases, as would beexpected.
Figures 5.9 and 5.10 show the results of double gas identification and quantification.

25
 n
E
Q 80
a
v
m
Q)
c,
$
c,
60
Q)
m
C
.g
c,
40
2
CI
C
Q)
20
0
0
0
0 20 40 60 80 100
Concentration delivered (ppm)
Figure 5.8 Identification and quantificationof three single gases using LM-NLS

160

140
P,
120
m
Q)
5Q) 100
c,
Q)
'El 80
S
0
'E 60
Y
2

0 20 40 60 80 100 120 140 160

Concentration delivered (ppm)
Figure 5.10 Identification and quantification o f sets o f double gas; benzene
plus propanol, methanol or formaldehyde. Benzene and propanol have very
similar patterns in the concentration range o f interest and so are difficult to
deconvolute.

26
 -E
P,
140

& 120
'EJ
Q)
t:Q) 100
CI
Q)
'EJ
C
80
.-
0
2 60
CI
CI
S
8S 40

8 20

0
0 20 40 60 80 100 120 140
Concentration delivered (ppm)

Figure 5.7 7 Identification and quantification of sets of double gas; ammonia

plus benzene and toluene plus methanol. Nonlinear analysis is necessary for
this range of concentrations.

6. COORDINATION WITH JSC

Thetask of coordinatingthe flight experiment with JSC wasongoing,andbegan

immediately upon starting work on this program. Coordination with JSC involved the
Toxicology Branch, which participated in the experiment, Krug Life Sciences (later Wyle
Laboratories) for IntegrationEngineering, JSC QA organization,SafetyandMission
Assurance, JSC MissionProject & Integration,andtrainingthecrewtoperformthe
experiment. At the beginning of the program, the ENose experiment was designated as
a DetailedTestObjective(DTO). JPL's interfacewasthen limited tothe Krug
integration t e a m a n d t h e JSC Institutional Review Board (IRB). IRB w a s i s s u e d in May
of 98. Eventually,theclassification of the experiment was changed to a Secondary
Payload. At that point, JPL interface with JSC intensified a n d w a s m a n a g e d b y t h e
office of Mission Projects & Integration.
6.1 Toxicology Branch
Immediatelyuponstartingworkonthisprogram,coordination withDr. J o h n J a m e s ,
Chief of the Toxicology Branch at JSC commenced. Early discussions with Dr. J a m e s
established the utility of t h e list of target compounds for detection, which had b e e n
selectedfromliteraturepublishedby Dr. James. Discussions of howtocontrolthe
experiment led toestablishment of collaboration with
Dr. J a m e s , in whichthe
Toxicology Branch provided Grab Sample Containers for daily air samples during flight

27
and ground analysis of the contents of the containers. Dr. James was included in the
flight experiment as co-investigator.

Furtherdiscussionsresulted in thedecision not to includerealtimedataanalysis or

data display during flight, A previous experiment with air quality analysis had led to false
positive identifications of contaminants, and in order to avoid a similar problem, it was
decided to do all data analysis on the ground, post-flight.

Throughout the course of this work, Dr. James and Dr. ThomasLimero of WyleLife
Sciences
were
included in discussions of the
lab work and
kept abreast of
developments in the sensors.

6.2 Early Human Test Experiment

In January, 1997, the opportunity to testanengineeringdevelopmentmodel of the
ENose during Phase IIA (60-dayexperiment) of theLunar-MarsLifeSupportTest
Project was made available to the ENose project by the Crew and Thermal Systems
Division at JSC. This experiment provided the opportunity to observe the operation of
the sensorchips, to determine how muchinterference from dailyeventscouldbe
expected, and to confirm the stability of the sensors over time. (Note: The Lunar-Mars
Life Support Test Project was previously called the Early Human Test Initiative, and the
chamber called the Early Human Test Chamber, so references to this experiment are
made under thosenames.)

AIR OUT
to ARS

Computer

1
U

AIR IN
from EHT

Figure 6.1 Design of the Engineering Development Model ENoseused in EHTC

28
An engineeringdevelopmentmodel of anelectronicnosewasfabricatedat JPL and
plumbed in totheairrevitalizationline of theEarlyHumanTestChamber (EHTC) at
NASA-Johnson Space Center for 49 days in early 1997. Flowing air (0.25 Umin) was
taken from the slipstream exiting the EHTC and directed into the ENose system. The
air, which had been heated to 30 - 34OC for EHTC testing purposes, was directed either
through anactivatedcharcoalfilter, put in line to providecleanairbaseline data, or
through a dummy glass filter, put in line to provide a pressure drop similar to that of the
charcoal filter. Solenoid valves were programmedto open the valve to the charcoal filter
andprovide 30 minutes of cleanair flow every four hours; otherwise,theairwent
through theglassfilter. Air thenenteredtheglassenclosedsensorheadchamber
where resistance was measured every 30 seconds, and then left the ENose system to
enter the air revitalization system of the EHTC.
The experiment was controlled using a personal computer and a National Instruments
DAQPad. TheDAQPadsentcommands to thesolenoids to openandcloseand
acquiredresistancedata from the sensors by measuringthevoltageat a current
provided by the DAQPad.
Thedesign of theengineeringdevelopmentmodelused in this experimentwas
somewhat different from that used in the flight unit, as it was early in the development
program. A schematic of the unit is shown in Figure 6.1.
The primary results of this experiment were confirmation that thesensor films are stable
for several months (measurementoneyearafterfabricationshowedthebaseline
resistance of the films to have changed only by a few percent), that the sensors are
sensitive to changes in the atmosphere and can be used as incident monitors to offer
early warning of leaks, and that the temperature of the sensors must be controlled to
prevent excessive baselinedrift.
Figure 6.2 shows ENose response to a spill of trimethylamine (TMA) in the showerhoilet
area of the EHTC. The significance of this figure is twofold. Comparison of the time of
thecrewreport of the odor shows that the ENose recorded the change in airquality
some 30 minutes before the crew report. Taking into account that the crew probably did
not report the odor immediately, it can be judged qualitatively that the ENose recorded
the odor severalminutesbeforethecrewwasaware of it. Thesecondpiece of
significantinformation in this figureconcernsthebaselining, or [Link]
reference cycle turned on in themiddle of the TMA event. When the reference cycle
turned off, a "virtual peak" of the odor change was created. This "virtual peak" makes it
possible to deconvolute a slow buildup of contaminant,which might otherwisebe
interpreted as baseline drift.
This experiment and other results are discussed further in the paper "Monitoring the Air
Quality in a ClosedChamber Using anElectronicNose,"presentedatThe 27th
InternationalConference on EnvironmentalSystems in LakeTahoe NV, July 14-17,
1997. The paper is included in this report as Appendix B.

29
6.3 Flight Manifest and Flight Qualification
TheElectronicNose Flight Experimentwasoriginallyplanned as a DetailedTest
Objective (DTO) on STS-91, a Mir Docking Flight in May, 1998. Coordination with JSC
Shuttle Integration engineers and with the Toxicology Branch started immediately after
[Link] of the flight unit wasscheduled for February,1998. In
February,1998,whenJPLwasready to deliverthe unit, theexperimentwas
rescheduled for STS-95, in October, 1998, so delivery was postponed until May, 1998.
To enhance the probability of remaining on the manifest, NASA Code U worked to
change the classification of the experiment from a DTO, which is easily removed form
themanifest, to a [Link], with thehelp of the Mission Project &
Integration office, the ENose was classified as a secondary payload, greatly enhancing
thechances for a [Link],theclassificationchangealso
changed the requirements and required JPL, Wyle, and JSC to perform additional tests
and enhance the flight qualification action.
Testing to pass all flight qualification and safety gates was conducted at JPL and at
JSC. The ENose met all JSCrequirements for a mid-deck payload. The tests and
certifications included:

a. Institutional
ReviewBoard (IRB)
b. Vibration
C. EMI-RFI
d. Acoustic
e. Temperaturecycling
f. Heater analysis for safety
g. Safetyanalysisandapproval of thereal-timeclock Li-ion battery
h. Toxicologyapproval for the safety of alloutgassing ofall components
i. Toxicologyapproval for thesafety of thepolymermaterials
j. Certified by JPLStructures and MaterialsReviewCommittee(SAMRC)
k. JSC
safetyapproval

30
 T
m
0 0 0 0 0 0 0 0
0 In 0 In 0 In 0 In
m cu cu 7 7 0 0 0
8 8 8 8 8 8 8 9
?
Otl/tiP
6.4 Experiment Definition and Crew Training
The ENose flight experiment was designed to provide continuous monitoring of the air
in the mid-deck of the orbiter (Le. data points every 30 seconds). The ENose sensor
responses to the air was recorded for 6 days during the STS 95 flight (Oct. 29 through
Nov. 7, 1998). In order to confirm that the ENose was operating, a crew member would
check the operating lights (LEDs) on the side of the unit daily, and provide an “event” or
dailymarker by exposing the inlet of the unit toanalcohol wipe made of 70% 2-
propanol, balance unknown. At the start of the daily check, before exposingthe sensors
to the wipe, the crew membertook a daily air sample in a grab sample container(GSC).

JPL and JSC developed the crew procedure and timeline. JPL researchersmet with the
crewtwicebefore flight, once as a video conference to familiarize the crew with the
objectives of the experiment, and once at JSC to review, practice, and demonstrate the
procedures for the experiment.
The crew protocol for the flight experiment was defined to consist of t h e following steps:
1. ENose deployed in mid-deck 1-2 days after launch (10 minutes)
a. Remove device from locker
b. Fix device to its space on top of mid-deck lockers with Velcro strips.
Space was on mid-deck lockers near an air-intake
c. Connect pre-routed28 V power cord
d. Turn on device and confirm that turn-on sequence is complete:
LEDs light, computer signature tone andpump can be heard
e. Log time and comments

2. Daily air samples (grab samples) taken for post flight analysis at JSC(5
mi n/day)
3. Daily calibration of ENose to 2-propanol done by crew member (5 midday)
a. Determine that device is not in a reference cycle (template provided)
b. Confirm that air inlet screen is not clogged with debris. Clean if necessary.
c. Open propanol wipe and hold near air inlet 30 seconds. ObserveLEDs.
d. Log time and comments

4. Turn off and stow ENose after experimental period (5 minutes)

The experimental procedure used by the crew is included as Appendix C.

After the flight the GSC air samples were returned to JSC for analysis, and the ENose
unit was returned to JPL for data analysis and post-flight calibration on the unit. The
ENosewascalibrated post-flight to confirmthatthe sensor baseline resistances and
array responses had not changed. After both JSC and JPL analyzed the data, t h e two
teams met for a data review.

6.5 STS-95 Flight Experiment

32
The
ENoseflew on STS-95, which launched on October 29, [Link] flight
experiment took place over a period of 6 days, as planned, with no unusual events. The
data and unit were returned to JPL after the flight without incident.

7. FLIGHT UNIT DESIGN AND FABRICATION

The design of the flight unit was undertaken early in the program, but many aspects of
the design were added as the development tasks proceeded. A diagram of the ENose
flight unit is shown in Figure 7.1. Photographs of the flight unit are shown as Figures 7.2
- 7.4.

AIR OUT

Shuttle
28 V in

Figure 7.1 Diagram of JPL ENose flight unit.

33
34
 our sensor chi s
P
Figure 7.3 Interior viewsof the €Nose withou the HP200LX Computer
35
36
7.1 Mechanical Design
Themechanicaldesign of the flight unit began with identifying a flight-qualifiable
container. On theadvice of JSCShuttleIntegrationengineers, acast aluminum
container was selected and anodized. After identification of the components that would
be included in the flight unit, a design for assembly was made. The mechanical design
was reviewed by the JPL SAMRC committee and approved.

7.2 Pneumatic Design

The first part of the flight unit to bedefinedwasthepneumaticsystem. It was
determined early on in the program that it would be necessary to use a miniature pump
to pull air through the sensor chamber in the flight unit in order to make measurements
on the constituents in the air. Leaving the sensors in still air would result in very slow
responses in the sensors, and possibly excessively slow sensor recovery from an
event. Consultationwith fluid mechanicians led to the selection of a diaphragm (pulsed)
pump, in order to ensure turbulent flow in [Link] flow could
result in partitioning of contaminants in the sensing chamber, and thus would result in
irreproducible responses. The pump was put behind the chamber so that air was pulled
overthe sensors, andanydead space in the pump wherecontaminantscouldbe
trapped would not result in contamination later in the experiment. The sensor chamber
wasdesignedaccordinglyandwasminimized to -15 cm3 in [Link] pump
selected is a Thomas model X-400 miniature diaphragm pump, which flows air at 0.25
Iiterdminute.

Laboratory testing and the Early Human Test Chamber Experiment showed the utility of
including a reference or baselining cycle in the data acquisition routine. The reference
cycle is used to constructthebaseline to correct for low frequency drift, and to
determine whether there has been slow buildup of contaminants.
In thereferencecycle,anactivatedcharcoalfilter is used to trap species in theair
which would not generallybe found in cleanair. At intervalsprogrammed into the
controlcomputer, a solenoidvalve turns on anddirectstheintakeair through the
charcoal filter. When the valve is off, intake air is directed through a dummy filter. When
clean air is directed through the charcoal filter the sensors' response is no different from
the response with the dummy filter. The dummy filter is made of an inert material such
as glass or Teflon beads to provide a similar pressure drop under filtered or unfiltered
conditions. Compounds on the target list are reasonably well collected by the charcoal
filter and water is passed by the filter, so there is no humidity change associated with
the reference cycle. Formaldehyde is only partially collected, but a sufficient quantity is
collected to provide baselining information. The length and width of the charcoal filter
wasdetermined by calculatingequivalentplatesandassumingcollection of SMAC
levels of all target compounds for 1 hour of operation.
In the flight unit, the filtering schemeto allow for baselining was altered slightly from the
scheme used in the EHTC model. Activated charcoal was used as the filtering material,
andTeflon@ beads wereused as the dummy filterprovided to matchtheressure
change induced by the charcoal filter. A three-way solenoid valve with a TeflonRinterior

37
was placed after the two filters to select the flow route. The reference cycle time usedin
the flight experiment was 15 minutes of filtering after 3hours of data.
The sensorchamber in the flight unit was a 15 cm3glassdome through whichthe
pumped air was passed. The sensor chips were inserted into receptacles and the glass
dome placed over them. The receptacles and dome were sealed to a copper-coated
platform using electronic grade RTV@ siliconrubberglue. All the parts of the system
were connected using Tygon@tubing; where necessary polypropylene connectors were
used. All connections were sealedwith RTV@.
Whentheproblem of Tygon@ tubing wasidentified in thelaboratory gas handling
system,the flight unit hadalreadybeendelivered to [Link] tubing for
Teflon' would have required a new mechanical design for the unit. Because the Tygon@
tubing did not developproblems with gas retention until it hadbeenoperated with
relatively high concentrations of thetargetcompounds for several months, it was
determinedthe tubing would not pose a problem for the short, 6-7 dayperiod of
operation of theENose flight unit. Afterthe unit wasreturned from the flight, it was
tested with the hydrocarbon analyzer to determine that the delivered concentration of
methanolwasexpelled from the pump. Theaidmethanolmixturethatwascollected
after the pump in the flight unit had a methanol concentration within 5% of the delivered
concentration.

7.3 Electronics Fabrication for Data Acquisition and Electronic Control

The unit was built using commerciallyavailable, mostly surface mount, components
assembled on standard4-layer FR4 printed wiring [Link] four sensor alumina
substrates accommodated the polymers on one side and a unique resistor pattern on
the back. These resistors were used to heat the substrates. Electrical connectionto the
substrates was through a small edge connector. Substrate temperature was controlled
by a JPL developed closed loop control with the computer in the loop. While there are
no new circuits in the ENose, the fusion of control and measurement circuits with the
HP200LX is novel.

7.3.1 HP200LX Control The HP200LX computerwasattachedinsidetheENose to

the top of the ENose aluminum container by Velcro@strips. For flight the computer top
wasclosedandsecured by a tie-wrapThe HP200LX controlsthe timing of the
experimental measurements and archives the [Link] also performs the display function
and converts the voltages read at constant current to resistance. The archived data are
stored in raw format as this preserves the most information and is dense. The program
is written in QuickBasic 4.5, uses National Instrument "Lab Windows" driver calls and
Microsoft C5.1; allcompiledandlinked to create an executable file. As shown in Fig.
7.1, the computer communicates directlywith, and commands, a local microcontroller, a
PIC16C74. (See 7.3.3below).

7.3.2 Circuit Boards The ENose device includes two circuit boards: an analog board
[Link] 4 identicalchannels of
sensor measurement circuitry, and analog temperature control feedbackloops for each

30
substrate, or [Link]
four
channels
makes it possible to take
measurementsseveraltimesasecond, if necessary. Eachanalogchannelhasone
ADC andtwo DACs, theseareeach 12 bit. The temperaturecontrollerhasone 4
channel 8 bit DAC. The sockets for the sensor chipsare mounted directly to t h e analog
board (see section 7.1).
For the flight unit, the circuit shown in Firgure 4.2 was modified to enable the electronics
to be run from a single 0-5 volt supply (see Figure 7.2). This is essentially the same
circuit with the ground on the DAC Vli floating. The sensor resistance is determined by
equation:
RsENi = [(VOUt + V2i )/ (1 + (R2/R1)) + V2i) - V2i ] / (VlJRO)
Thedigitalboardcontrols each individualmeasurement as well as t h e temperature
control, pump andvalve. This boardalsohastheswitchingpowerconverterthat
isolates the circuitryfromthepower lines andallows the ENoseto run from a wide
range of voltages.

Vout

Figure 7.2 Flight Unit Data Acquisition Circuit: 12 Bit Dual OffsetNulling Amplifier

7.3.3 Microcontroller The PIC16C74 microcontroller is programmed in assembly

language to command all the ENose components. It controls the reading and writing to
the DACs and ADCs, thestoring of measurement parameters for the 32 sensors to
quicklymultiplexbetween sensors, theswitching of the pump andvalve (under HP

39
command), the monitoring of temperatures and voltage, reading and setting of the real
time clock, and initial boot and HP200LX power on and boot.

7.4 Flight Unit Assembly

The flight unit was assembled at JPL. The electronic components and electrical wiring
as well as pneumatic components which needed power (valve, pump) were installed by
the Electronic Packaging and Fabrication Section at JPL in their flight certified facility.
The tubing and sealing were done in the Electronic Nose Lab in a clean bench.

8. STS-95 FLIGHT EXPERIMENT RESULTS

TheENose is designed to monitor for common contaminants in space shuttle air. To

testthedevice'sability to detecttargetcontaminants,an in-flight experimentwas
conducted on STS-95. Afterthedevicewasremoved from storage andturned on, it
took data automatically and autonomously for 6 days. To verifythatthedevicewas
operating as tested on the ground and to confirmthetimekeeping,analcoholwipe
made of 70% 2-propano1, balance unknown, was used to create a dailyeventwhich
was used as a daily marker. Before creating the event,a daily air sample was collected
in a GrabSampleContainer(GSC)provided by JSC for laterindependentGC-MS
analysis.
Initial, visual examination of the flight data received indicates that the ENose responded
to allplannedwipe events, andthattheresponsesweretimelyanddistinguishable.
Softwareeventidentificationanddataanalysisfurtherconfirmedallplannedwipe
events. In addition to thewipe events,thedeviceresponded to a number of other
events. Software analysis identified all events which were not wipe events as humidity
changes. Many of those changes can be correlated in time with the humidity changes
recorded by the independent humidity measurements provided to JPL by JSC. Those
events identified as humidity changes but not correlated with cabin humidity change are
likely to becaused by local humidity changes; that is, changes in humidity nearthe
ENose which were not sufficient to cause a measurable change in cabin humidity. The
independent humidity monitor was located in the stairway between the middeck and the
flight deck, and so would not record any humidity changes localized around the ENose.
Figures 8.1 a&b show how cabin humidity correlates with ENose response in several
cases.
Figure 8.2 shows the similarity between the pattern for particular events in Figure 8.1,
and compares them with the patterns recorded in training sets for exposure to the wipe
and for humidity change. Software analysis of the flight data did not identify any other
target compounds, as single gases or as mixtures. The results of the software analysis
of the 46 peaks identified as events are presentedin Appendix D. The analysis program
returned an identification and quantification of each wipe event, and every other event
was identified and quantified as a change in humidity.

40
The independent GSC analysis provided by JSC confirmed that no target compounds
were found in thedailyair samples in concentrationsabovetheENosedetection
threshold. Copies of the GC-MS analysis report are included in this report as Appendix
E.

The correlation between the ground training and in-flight response patterns for both the
alcohol
wipe
and humidity change shows that the
operation of the
ENose is
microgravityinsensitive,and thus canbeused in a space-based application without
further accountingfor microgravity effects.

41
 55

50

45

40

35

30
FLT3171 d

307.3
307.2
307.1
307.0
306.9
306.8
3 06.7
T i m e (day GMT)

50

48

46

44

42

40

38

36

34

32

30
, .9

T i m e (day GMT)

Figure 8.1 a,b) STS-95 Shuttle Data. Circles are theplot of independenthumidity
measurements in the stairway from mid-deck to flight deck and colored lines plot the
response of differentsensors in the ENose, Note that at times when there is a major
excursion in cabin humidity, the ENose responds, as well. The step at time 308.58 is a
temperature change; cabin temperature approached the ENose chip set point, so the
€Nose temperature was raised 4 C . The daily marker in FLT3174d occurs at the end of
the plot,at time 308.89.

42
 -
Training 10% R H Training -
0.8 Change from 40% R H W Wipe 3000 ppm
v)
v)
2
z
8
v)
0.6 W
W U
a n
n W
0.4 !Y
Y
U
4s
s
=4 0.2
K
4
0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
SENSOR NUMBER SENSOR NUMBER

1
I Flight event
9% Humidity Change
at 306.78 in Flight
A
W W 0.8
v) 0.81 v)
z z
8
v)
8
v)
W W 0.6
K U
n n
w W
! Y 0.4
J -I
U U
I I
a K
4 4 0.2

0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
SENSOR NUMBER SENSOR NUMBER

Figure 8.2 Fingerprints, or response 1 ,

patterns, for events in the flight data show Flight event at 306.95
thesmilarity between training data and
flight data identified as humidity change or
as wipe. Thehumidity event at Time
306.78 is identified as a 9% humidity
change. The pattern is verysimilarto the
10% humiditytraining data. A wipe event
quantified at 900 ppm is similar to, but not
exactly the same as a wipe training pattern
at 3000 ppm; this difference is the non-
linearity that must be accomodated in the
software analysis. The wipe event at Time
306.95 is analyzed as a combination o f 0 5 10 15 20 25 30
humidity and wipe, and the pattern can be SENSOR NUMBER
seen to be a combination of the two
training patterns.

43
9. SUCCESS CRITERIA FOR ELECTRONIC NOSE TESTING

Success criteria for ground testing and for flight were set with the agreement of the co-
investigator, Dr. John James of JSC. Three of four success criteria for ground testing
were met before flight; criterion #3 was met partially. All success criteria for flight were
met within the limit of the events of the flight ([Link] #6 could not be met because
there were no events whichwere unknown.) The success criteria and how they were
met are discussed below.

9.1 S u c c e s s CriteriaFor Ground Testing

1 . Target compounds can be detected, identified and quantified at +/- 50% the one-
hour SMAC levelorlower by applyingtheanalysissoftwaredeveloped for the
purpose.
Nine of ten compounds on the target list can be detected, identified, and quantified ator
belowthe SMAC level. Formaldehyde can be detected, identified and quantified at 15
-
ppm (SMAC is 0.4 ppm). Identification accuracy overall is 90%.

2 . Mixtures of 2 & 3 targeted gases can be detected, identified and quantified at +/-
50% the one-hourSMAC level or lower by applying the analysis software.
Exhaustive mixture testing was not undertaken. Several mixtures containing propanol,
ethanol,ammonia,[Link],andthe
constituents detected, identified and quantifiedas required.

3. Contaminants in a pre-mixed sample of 4 gases can be identified and quantified at

+/- 50% the one-hourSMAC level or lowerby applying the analysis software.
Data were recorded for analysis, but because the softwareanalysisroutines did not
have a significantlevel of success on mixtures greaterthanthree, they were not
analyzed. Mixtures of two and in some cases three gases were successfully analyzed,
but those data were not taken from pre-mixed samples

4. €Nose data analysis segregates compoundswhich are not on the target list.
Analysis of single and multiple gases successfully segregates unknowns, >go%.

9.2 S u c c e s s Criteria For Shuttle Flight Experiment

7. Successful delivery and acceptance toflight.
Delivered to JSC May 19, 1998.
2. Deviceoperatescontinuously(whileturned on) as programmed,and data are
retrieved after flight.
Operatedcontinuously for 6 days during flight with no [Link]
with no corrupted files.

3. Sensor responses correlate with the shuttle loggedevents of sufficientconcentration

44
No shuttle logged events for correlation. Responses do correlate to rises and falls in
cabin humidity as recorded by independent humidity monitor.

4. Data analysis software correctly identifies and quantifies planned events .

All plannedevents(dailymarker of propanolwipe)[Link]
eventswerequantified in the 500-1000 ppm range, which is within range found for
similar tests done in the laboratory.
5. Data analysis software correctly identifies targeted compounds at or above the one-
hour SMAC level and quantifies them at +/-50%,confirmed by the GC-MS analysis of
the Grab Sample Container contents.
There were no compounds on the target list found at the one-hour (or even 24-hour)
SMAClevel during thetimetheENosewas running. TheENose did not record false
positives.
6. Data analysis software classifies as “unknown”compounds which are detected and
are not on thetarget list.
There were no false positive events; neither were there events during the flight which
could not be identified. All events which were not 2-propanol daily marker events were
identified as humidity changes.

10. CONCLUSIONS AND PLANS FOR FURTHER WORK

In January, 1999 John James, Tom LimeroandHelenLane of JSC, and M. A. Ryan,

MargieHomer,DarrellJanandHanying Zhou of JPL met to reviewtheENose flight
data and the independent ground analysis (GC-MS) of theair samples taken during
flight. In thatmeeting,severalconclusionsregardingthe flight experimentwere
discussed. The participants agreed that the ENose Flight Experiment was a successful
[Link] to using insulatingpolymersloaded with a conductivemedium
whichwasdevelopedatCaltechwastaken from a laboratorydemonstration to
development of a miniaturedemonstrationdevice which operatedcontinuously in a
microgravity environment. The demonstration was successfulin that all events recorded
were analyzed by the software, and in many cases could be correlated to independent
measure of cabin events. The primary conclusions reached in that meeting were that
theENoseoperated without incident,thatallplannedevents(dailymarkers)were
detected and that there were no inconsistencies between the data analysis from the
ENose and from the GC-MS within detection limits of the techniques.
While the hope in an experiment such as this one is that there will be several events
which test the ability of the device, such events would certainly be anomalous events in
the space shuttle environment. The fact that the shuttle air is largelyuneventfuland
clean is, of course, desirable for the crew, but made for an uneventful experiment and
did not significantlychallengeeithertheresponsecharacteristics or thesoftware
analysisroutines of theENose. It is not surprising thatthe only unplannedevents

45
recorded by theENosewere humidity changes, and it is becauseevents were not
expected that the experiment included the relatively uncontrolled daily marker events.
Locating the ENose near the intake for the air revitalization system was a good choice
of location, and unremarkable events (such as use a crew member passing by or using
a product such as deodorant)were not recorded. Also, metabolicproductssuch as
methanebelow 3000 ppm andhydrogen did not result in responses in the sensors.
Localizedeventsweresufficientlydiluted as not to causeresponse in theENose
sensors. There wereno events reported by the crew that would be expected to induce a
response in the ENose. However, further work on the ENose, especially if the plan is to
usemanysmall,distributed sensorheads, will have to takelocalizedevents into
account.
For the two compounds seen, water in quantitiessufficient to constitute a change in
relative humidity (a humidity change from 35-40% at 3OoC is approximately 3000 ppm),
and the alcohol wipe, the response of the ENose in microgravity was not significantly
different from the response on the ground. This result leads to the conclusion that the
device is microgravity insensitive.
The shuttle air was warmer than was expected from the EHTC experiment, and so the
ability of the on-chip [Link]
temperature of the chips steady, but if the air had been warmer than it was, generally
around 3OoC, thechipheaters would [Link]
ENose will take the temperature into account. It has been found in ground testing that
heating the chips rather than the air can lead to excessive noise in the response; further
investigation of heating the chamber or eliminating the need to heat altogether will be
planned for the next phase.

For the flight experiment,thesizecharcoalfilterusedwascalculated to fit the

experiment. Using PTFE beads in the dummy filter in the flight experiment was not as
effective as using glass beads in the dummy filter in the EHTC experiment. The need
for filtering for baseline control will need to be further considered.
The events seen, wipe in particular, but in some cases humidity change, were shorter
thanthetraining events. Thenoisefilteringused in thesoftwarethencouldlead to
obscuring some [Link], future work will use shorter training events.
The limitations of the flight experiment were primarily in the lack of events for analysis.
As it is unlikelythat it will be possible to stage a number of different events during a
flight to challengetheabilities of theENose,the group thatmet in January, 1999
concluded that in the absence of such a flight experiment, resources would be better
applied to extensive ground testing, including blind testing of the software analysis in a
relevant environment as well as inclusion of the ENose in future module tests such as
[Link] group alsoagreed that thebestapplication of theENose for Space
Stationuse would be twofold: as anincident monitor to warncrewmembers of
anomalous events such as leaks, spills, air filter breakthrough and incipient fires; and to
monitor theprogress of [Link]

46
widedynamicrangeaccessible to theENosemakes it anexcellentdevice for
application to these uses. It does not, and probably will not, have the capability to be
used as an analytical instrument or to analyzed all the constituents in a vapor sample,
but can be used in concentration regimes where a GC-MS cannot.

Further work with the ENose should take account of the limitations of the experiment
done in this program. While the experiment was controlled to the extent that daily air
samples were taken and daily confirmation of the device's operation was made, it was
not controlled in thatanevent which occurredseveral hours away from thetime of
sampling would not be picked up by the sample. Truly testing the ENose as an incident
monitor will requirecontrolledrelease of targetcompounds,mixtures of target
compounds,and unknowns. This scenario is not a likelyone for use in a flight
environment, as it will pose risk to crew health.
A proposal for further work to follow this final report will propose a program in which the
ENose developed in this program will be further developed and tested extensively on
the ground, with the goal of producing an incident monitor, with a detection range from
single ppm to hundreds of ppm. The major points of this proposal will include:
1. The device will be further miniaturized
2. The list of target compounds will be expanded in consultation with JSC.
3. The number of polymers will be expanded to 32.
4. The target concentrationsfor contaminants will be pushed to the 24-hour
SMAC.
5. Extensive ground testing with controlled release of contaminants in an
environment similar to that of the space shuttle will be undertaken.
6 . Data analysis softwarewill include the capabilityfor real-time analysis.

47
1 1 . REFERENCES

1. M.S. Freund and N.S. Lewis, "A Chemically Diverse

Conducting
Polymer-Based
Electronic Nose," Proc. Natl. Acad. Sci., 92, 2652 (1995).
2. M.C. Lonergan, E.J. Severin, B.J. Doleman, S.A. Beaber, R.H. Grubbs, and N.S. Lewis,
"Array-BasedVapor Sensing Using Chemically Sensitive, Carbon Black-PolymerResistors,"
Chem. Materials, 8, 2298 (1 996).
3. [Link], R.D. Sanner, E.J. Severin, R.H. Grubbs, and N.S. Lewis, "Use Of
CompatiblePolymerBlends To FabricateArrays Of CarbonBlack-PolymerComposite
Vapor Detectors,"Anal. Chem., 70, 2560 (1 998).
4. H.V. Shurmer and J.W. Gardner, "Odour Discrimination with anElectronic Nose,"
Sensors and Actuators, B8, 8, (1992).
5. J.W. Gardner, T.C. Pearce, S. Friel, P.N. Bartlett and N . Blair, "A Multisensor System for
Beer
Flavour
Monitoring Using and Array of Conducting Polymersand
Predictive
Classifiers,"Sensors and Actuators, B18-19, 240, (1994).
6. J.T. James, et a/., Aviation, Space Environ. Med., 65, 851 (1 994).
7. Spacecraft Maximum Allowable Concentrationsfor Selected Airborne Contaminants,
Vols. 1 & 2, National Academy Press, Washington, DC (1 994).
8. Space Physiologyand Medicine, A.E. Nicagossian, C.L. Hunton & S.L. Pool, eds., Lea
and Febiger, Philadelphia (1 994).

12. PAPERS AND PRESENTATIONS

"Results From The Space Shuttle STS-95 Electronic Nose Experiment"

M. A. Ryan, M .G. Buehler, M .[Link], K. S. Mannatt, [Link], S. Jackson, and H. Zhou,
Proceedings of the Micro/NanoTechnology99 Conference,Pasadena CA (1999).
"Monitoring Space Shuttle Air for Selected Contaminants Using an Electronic Nose, M. G.
"

Buehler, M. A. Ryan, M. L. Homer, K. S. Manatt, B. Lau, S. Jackson, and H. Zhou, 5th

InternationalSymposiumonOlfactionandElectronic Noses(Baltimore, MD), September
1998.
"Monitoring Space Shuttle Air for Selected Contaminants Using an Electronic Nose"
M. A. Ryan, ML.. Homer, M. G. Buehler, K. S. Manatt, F. Zee, B. Lau, D. Karmon, V. Ryan,
A. Kisor, S. Jackson, Proceedings of the 28th International Conference on Environmental
Systems (1998).
"Development and Testing an Electronic Nose at JPL"
M. L. Homer, M. A. Ryan, M. G. Buehler, K. S. Manatt, S. Jackson and A. Kisor,NASA
Workshop on Advanced Environmental Monitoring, October 1997.

48
“Test Structures Appliedto the Rapid Prototyping of Sensors”
M. G. Buehler, L.-J. Cheng,and D. P. Martin, Proceedings of the /€E€ lnternational
Conferenceon Microelectronic Test Structures, 10, 21 2-21 8 , March 1997.
“Temperature and Humidity Dependence of a Polymer-Based Gas Sensor”
M. G. Buehlerand M. A. Ryan,SPlEConference on Electro-OpticalTechnology for
Chemical Detection and Identification, 21 April 1997.
“Monitoring the Air Quality in a Closed Chamber Using an Electronic Nose”
M. A. Ryan, M. L. Homer, M. G. Buehler, K. S. Manatt, F. Zee, J. Graf, Proceedings of the
27th lnternational Conferenceon Environmental Systems (1 997).
“Electronic Nose Technology atJPL”
M. A. Ryan, M. G. Buehler, K. S. Manatt, V. Shields, L.-J. Cheng, D. Karmon, F. Zee; NASA
Workshop on Advanced Environmental Monitoring, September 1996.

13. TECHNICAL STAFF

The ENose project was divided into fivemain tasks, as described above. The Principal
Investigator for theprojectwas Dr. Margaret A. [Link] for the
project was Dr. Dan Karmon.

Severaltechnicalstaffmemberswereinvolved in morethanone task, but are listed

here in their primarycapacity. This is not anexhaustive list of staff involved in this
work, but lists those people who were responsible for major areas of work.

Sensor Development and Test Coordination with JSC

Margie L. Homer DanKarmon
Darrell L. Jan Karan L’Heureux
Margaret A. Ryan

Data Acquisition and Control Flight Unit Design and Fabrication

Martin G. Buehler Shannon Jackson
Kenneth S. Manatt Roy Packard
Yuki Salinas
Data Analysis Software
BrianLau
Frank Zee
Hanying Zhou

Undergraduate students who participatedin this program

Roger H. Cortez, Occidental College
Adam K. Kisor, University of California at San Diego
Victoria Ryan, California State University at Long Beach

49
APPENDIX A: CONFERENCE PROCEEDINGS PAPER

"Monitoring Space Shuttle Air for SelectedContaminants Using anElectronicNose,"

M.A. Ryan, M.L. Homer, M.G. Buehler, K.S. Manatt, B. Lau, D. Karmon and S.
Jackson; presented at The28thInternationalConference on Environmental Systems,
Danvers MA, July 12-16, 1998.

50
Proceedings of the 28" International Conference on Environmental Systems, Danvers MA; July 12 - 16, 1998

981 564
Monitoring Space ShuttleAir for Selected
Contaminants Using an Electronic Nose
M.A. Ryan, M.L. Homer, M.G. Buehler, K.S. Manatt, B. Lau, D. Karmon and S . Jackson
Jet Propulsion Laboratory, California Instituteof Technology
Pasadena CA 91 109

ABSTRACT A baseline of clean air is established,and

deviations from that baseline are recorded as changes in
A miniaturizedelectronic nose
has been resistance of the sensors. Thepattern of distributed
constructed at JPL in collaboration with Caltech. This response of the sensors maybedeconvoluted,and
array of conductometric sensors has beentrainedto contaminants identified and quantifiedby using a software
detect and quantify the presence of vapors in the air; the analysis
programsuch as pattern
recognition
and/or
compounds detected have been found as contaminants neural network.
in shuttle air. This device has potential application as a
miniature, distributed device monitoring
for and At present,thebest realtime,broadbandair
controlling the constituentsin air. quality monitor available in space habitats is the human
nose. It is limited by human factors such as fatigue and
INTRODUCTION exposure [Link] sensors are
designed to detect specific
molecules.
Array-based
Theabilitytomonitor theconstituents of the sensing uses non-specific sensors in which the pattern
breathing air in a closed chamber in which air is recycled andmagnitude of response are usedtoidentifyand
is importantto NASAfor use in closedenvironments the presence of [Link]-based
quantify
such as the space shuttleandthe space station. At sensors are based on a biologicalmodel of "sniffing",
present, air quality is determinedafterthefact by detecting changes in odor, and can be trained to detect
collecting samples and analyzing them on the ground in new patterns.
laboratory
analytical
instruments such as a gas
chromatograph-mass spectrometer (GC-MS). The The overall goal of the program at JPUCaltech is
availability of a miniature, portable instrument capable of the development of a miniature sensor which may be used
identifying contaminants in the breathing environment at tomonitorthebreathingair in the international space
part-per-million
levels
would greatly enhance
the station,
and
which
may be coordinated with the
capability for monitoring the quality of recycled air as well environmental control system to solve air quality problems
as providingnotification of thepresence of potentially withoutcrewintervention. Progress towardthatgoal will
dangerous substances from spills and [Link] dependonthe success of this portion of theElectronic
instrument is Electronic
the Nosenow
under Noseprogram,which is the development of a prototype
development at JPL and Caltech[l-31. systemwhich will be the subject of an experimental test
during a space shuttle flight in 1998.
An electronicnose is anarray of non-specific
chemical sensors, controlled and analyzed electronically,
whichmimics theaction of the mammaliannose by THE ELECTRONIC NOSE DEVELOPMENT MODEL
recognizing patterns of response to vapors. The sensors
usedhere are conductometricchemical sensors which The
conductometric sensors used in the
changeresistance when exposed to vapors. The ElectronicNose (ENose) built at JPL are polymer films
sensors are not specific to any one vapor; it is in the use deposited ongold [Link] films are madefrom
of anarray of sensors, each with a differentsensing polymers in which a conductive medium, carbon black,
medium, that gases and gas mixtures can be identified by has been dispersed [2].
thepattern of response of [Link]
havebeendiscussed by severalauthors,andmaybe Presence of a contaminant in air is measured as
applied to environmental monitoring and quality control in achange in resistance in the polymer films. Sensor
such
wide
fields as foodprocessing,
and
industrial response is expressed as a ratio of change in resistance
environmental monitoring [4,5]. at time t to resistance at time t=O, (Rt-Ro)/Ro. Data are
acquired on a Hewlett PackardHP200 LX palmtop
 computer using a programwrittenfor this purpose in 8. vinyl alcohol/ vinyl butyral copolymer
Labwindows. 9. poly (a -methylstyrene)
10. poly (vinyl chloride-co-vinylacetate)
DESIGN OF SENSOR HEAD 11. poly (vinyl acetate)
12. poly ( N -vinylpyrrolidone)
Substrate
and
Sensinu Films The sensor 13. styrene/isoprene, block copolymer
head of theElectronicNoseused in this experiment 14. poly (vinyl stearate)
consists of 32 sensor positions arranged on 4 15. methyl vinyl ether/ maleic acid
substrates, each with 8 sensor positions. The substrates 16. hydroxypropyl methyl cellulose
were made using hybrid microelectronic cofired ceramic
(alumina) processes.
Electrodes
and
contacts
were
deposited as thick films using screen printing. The These polymers were selected by statistical
substrate layoutandfabrication has been discussed in analysis of the responses of 100 polymers to the set of
detail elsewhere [9]. A guard ring is used around each contaminantslisted in Table 1. Datafor thestatistical
sensor to minimize cross-talk and sensor interaction. A analysiswereprovided by Caltech [ll]. Theanalysis
sketch of a sensor chip is shown in Figure 1. selected the set of polymerswhichwouldresult in the
maximum difference in patterns of response.
25 rnrn
I I DeDosition of Films160 mg of each polymer
wasdissolved in 15 mL of [Link]
usedwere
tetrahydrofuran (THF), acetone, dichloro
methane,toluene or a mixture of solvents. 40 mg of
carbon black was added to t h e solution, and dispersed
by sonication. 1 - 3 pLof solution was pipetted onto the
sensor area and allowed to dry in flowing, clean dry air
while the sensor chip was held at 28'C. The resistance
of the resulting films was in the range 1 -50 kQ. Solution
cover
polymer
contacts
guard wasadded in increments of 1 pL until thedesired
ring film layer resistancewasreached. The use of polymer films as
sensing media in an electronic nose has been discussed
Figure 1: Sketch of theceramicsubstratechip in detail by several authors, including t h e Caltech group
containing eightsensors working with JPL on this project [l-51.

The polymeric sensor films were made by

depositing a solution of polymer mixed with carbon black
tomake a film 1 - 5 micronsthick in contact with gold ENOSE SYSTEM
electrodes.16polymerswere used in this experiment,
four
on eachceramic substrate. Eachpolymerwas A diagram of theENosesystemused in this
deposited in 2 positions on each chip, with the positions experiment is shown in Figure 2.
side-by-side to ensure that each sensor would see the
same vaporenvironment. A thermistor is includedon AIROUT
eachchipandonthe sensor headfortemperature
monitoring.

Temperaturecan be controlled
from
room
temperature to36'C using R u 0 2 heaters deposited on
the back of each chip in the ceramic fabrication process.
The sensor resistance is sensitive to changes in
temperature, so the ENose is operated with the sensors
held at a constant temperatureof 28'C.
-
carton filter
for basdiriq

The 16 polymers used are: U

1. poly(2, 4, 6-tribromostyrene)
poly(4-vinylphenol)
2.
3. poly(ethyleneoxide) Figure 2: Diagram of ENose system.
[Link] resin
5. cellulosetriacetate Flowing air (.25 Umin) is pumped from the room
6. poly(2-hydroxyethylmethacrylate) into the sensor chamber of the ENose using a Thomas
7. poly (caprolactone)
modelX-400miniaturediaphragm pump. Theair is MAX538 andMAX539),and ADC is AnalogtoDigital
directed either through an activated charcoal filter, put in Converter (12 Bit LTCl286).
line toprovidecleanairbaseline data, orthough a
dummy Teflonbeadfilter, put in line to
provide a Data are stored in flash memory in the HP 200,
pressure [Link] and are analyzed later using software designed for the
valves are programmed to open the path to the charcoal task.
filter and provide 30 minutes of clean air flow every four
hours; otherwise, the air is directedthroughtheTeflon Data
Analysis
The
goal of ENose
the
bead filter. Air then enters the glass enclosed sensor development is the construction of an air quality monitor
head chamber where resistance is measured every 30 capable of identifying the target compounds in Table 1 at
seconds, and then is returned to the room. less than SMAC levels. Toaccomplish this goal, we
have developed data analysis software which recognizes
The experiment is controlled using a HP200 LX the patterns of response of the target compounds. The
palm
top
computer and a circuit designed for
the dataanalysissoftwareforms,fromtraining data,a
purpose [9,10]. The circuit is commanded by the HP 200 characteristic vector of sensor responses for each target
to operate t h e pump, toopenandclosethesolenoid [Link]
valve, and to acquire resistance data from the sensors compound is expressed as a linearcombination of the
by measuring the voltage at a current provided. characteristic vectorsof the target compoundsvia a least
squares solution using pseudo inverses computed by the
Sensor Resistor
singular value decomposition algorithm. The result is a
RS
listing of whatquantities of whichtargetcompounds
compose the unknown compound. At present, unknown
compounds are expressed as a combination of up to four
contaminants. In the case of exposure to asingle
V a R,/R, contaminant,additionalverification of the analysis is
obtained by astandardbackpropagation-trainedneural
network and by linear discriminant functions.

Table 1 shows the minimum concentration

(R,IR,) + 1 = gain R,
detected for each of the targetcompounds using the
software analysis program, and Figures 4 a-d show t h e
linear region of concentration detection. The goal of this
>L
+ To ADC
program wasto
quantify
contaminants +/- 50% of
deliveredconcentration;the shaded region in theplots
describes the 50% error region. Humidity was controlled
from 20-60 YO relative humidity during [Link]
Figure 3: Diagram of ENose measuring circuit. content of the air raises the resistance of the sensors,
and can either be deconvoluted from the response as a
DataAcauisitionandControlDataacquisition separate vapor,orzeroedout of themeasurement if
and
device
control are accomplished using a PIC humidity has not changed since the last baseline.
16C74Amicrocontroller. The HewlettPackard HP 200
LX palmtopcomputer is programmedtodirectthe CONCLUSIONS
microcontroller to open or close the solenoidvalve which
controls access to thecharcoalorTeflonfilterandto TheminiaturizedENosedesignedand built at
record sensor resistance. Typical resistance chan e for JPL has the
capacity
to
detect
limited
a suite of
10-50 ppm of contaminant is on t h e order of2x1 0-P (200 contaminantsat 1 hour SMAC levels with +/- 50%
ppm resistance change), and may be as small as x ~ O - ~ . accuracy. Combinations of four or fewer vapors can also
Thedataacquisitioncircuit is shownschematically in be detectedanddeconvolutedforidentificationand
Figure 3. Small changes in resistance are measured quantification. The ENose experiment will be performed
using a 12 bit dual offset nulling amp, in which a known on a shuttle flight in 1998 toverify its operation. The
current is put through t h e sensor resistor R, by DAC-VO experiment will consist of several steps:
andfixedresistor Ro. The voltage across the sensor is a. ENose response will be recorded over 5 - 8 days
measured with precision by subtracting DAC-V1, an
experimentally determined voltage, from Vs, the voltage b. Daily calibration of ENose to 2-propanol done by
drop across the sensor resistor, Rs. Thedifference is a crew member
then multiplied by a fixed gain, (R2/R1) + 1, where R2 and c. Daily air samples ("grabsamples")taken for post
R1 are fixed resistors. For each measurement, the DAC flight analysis at JSC
and ADC are lockedto t h e same voltagereference,
where DAC is Digital to
AnalogConverter (12 Bit
 Compound
Detected SMAC (ppm) ['y81 Detected
at JPL
on shuttle
(ppm) 16] 1 hr (PPm)

alcohols
methanol <1 30 25
ethanol .5 - 5 2000 50
2-propanol .4 - 4 400 50
methane 1-1 5300 3000
ammonia 0 30 20
benzene < .1 10 10
formaldehyde 0 0.4 10
Freon 113 .1 - 1 50 50
indole 0 < 1 0.03
toluene .4 - 4 16 15

Table I: Compounds detected by the ENose, Spacecraft Maximum Allowable Concentration of each
compound , and minimum concentration detected by this model ENose.

d. Post flight analysis of data using software 1. M.S. Freund and N.S. Lewis, Proc. National Academy
developed under this program of Science, 92, 2652, (1995).
e. Post flight analysis of grabsamples by gas
2. M.C. Lonergan,[Link],[Link], S.A.
chromatography/mass spectrometry Beaber, R.H. Grubbs and N.S. Lewis, [Link],
8, 2298 (1996).
f. Post flight comparison of analyzedENose 3. M.A. Ryan, M.L. Homer, M.G. Buehler, K.S. Manatt,
response to ground analysisof grab samples. F. Zee, and J. Graf, Proceedings of thelnternational
Conference Environmental
on Systems, Society of
At its present level of maturity, the ENose is not Automotive Engineers, (1 997).
an analytical instrument, but can be used to monitor an 4. H.V. Shurmer and J.W. Gardner, Sensors and
environment against a baseline, which is determined at Actuators, B8, 8, (1992).
intervals using filtered air. 5. J.W. Gardner, T.C. Pearce, S. Friel, P.N. Bartlett and
N . Blair, Sensors and Actuators, B18-19, 240, (1 994).
Long-duration space flight requires a high level 6. J.T. James, et a/., Aviation, Space Environ. Med., 65,
of crew productivity in tasks other than
habitat 851(1994).
maintenance. Decentralization of habitat control and de- 7. SpacecraftMaximumAllowableConcentrations for
coupling of spacecraft fromgroundcontrolrequires a Selected AirborneContaminants, Vols.1 & 2, National
moveto a distributed network of small sensors and Academy Press, Washington, DC (1994).
[Link] be programmedtomonitor 8. Space PhysiologyandMedicine, A.E. Nicagossian,
habitatairforthe presence of contaminants which C.L. Hunton & S.L. Pool, eds., Lea and
Febiger,
Spacecraft
exceed
the Maximum
Allowable Philadelphia (1994).
Concentration (SMAC) and to sound an alarm or actuate 9. M.G. Buehlerand M.A. Ryan, Proc. lnt'l Conf. on
remedialaction,aform of [Link] Microelectronic Test Structures (March 1996).
sensors lend themselves to distribution of several 10. M.G. Buehlerand M.A. Ryan. [Link] Conf. on
miniaturearrayslinkedto a commoncomputerfor Electro-OpticalTech. for ChemicalDetection, (April,
[Link] presence of severalarrays 1 997).
distributed about the habitat will allow early identification 11. N.S. Lewis, California Institute of Technology,
of areas requiring remediation private communication.

ACKNOWLEDGMENTS

The research reported in this paper was carried

out by the Jet Propulsion Laboratory, California Institute
of Technology under contract
a with the National
Aeronauticsand Space [Link]
supported by NASA Code UL.

REFERENCES
 160.00

140.00
E
Q
Q
W 120.00
B5
-
Q
100.00
C
Q 80.00
C
.-0
+ 60.00
E
CI
C
40.00
8c
0
0 20.00

0.00
0.00 20.00
40.00 60.00 80.00 100.00
120.00
140.00
160.00
concentration delivered(ppm)

Figure 4a: Identification and Quantification of Benzene, Toluene and Ammonia. SMACs for each
compound are marked on the plot. The shaded area is the +/- 50% target zone.

300.00

n
E 250.00
a
W
Q

1 200.00
-
r
(21

5
C
150.00
.-0
c,

!
c,
c 100.00
Q)
0
C
50.00
8
0.00
250.00
200.00
0.00 150.00
100.00
50.00 300.00
concentration delivered(ppm)

Figure 4b: Identification and Quantification of Ethanol, Propanol and the Propanol Wipe to be used
as calibration. SMACs for each compound are marked on the plot. The shaded area is the +/- 50%
target zone.
 300.00

250.00

200.00

150.00

100.00

50.00

0.00
0.00 50.00 100.00 150.00 200.00 300.00
250.00
concentration delivered(ppm)

Figure 4c: IdentificationandQuantification of Indole,Freon 113 andFormaldehyde. SMACs for

each compound are marked on the plot. The shaded area is the +/- 50% target zone.

300.00

250.00

200.00

150.00

100.00

50.00

0.00
0.00 50.00 100.00 150.00 200.00 250.00 300.00

concentration delivered(ppm)

Figure 4d: IdentificationandQuantification of Methanol. The SMAC is markedon the [Link]

shaded area is t h e +/- 50% target zone.
APPENDIX B: CONFERENCE PROCEEDINGS PAPER

“Monitoring the Air Quality in a Closed Chamber Using an Electronic Nose,’’

M.A. Ryan, M.L. Homer, M.G. Buehler, K.S. Manatt, F. Zee, and J. Graf; presented at
The 27th InternationalConference on Environmental Systems, LakeTahoe NV, July
14-17, 1997.

58
Proceedings of the 27th International Conference on Environmental Systems, Lake Tahoe NV July 14-17, 1997

97ES-84
Monitoring the Air Quality in a Closed
Chamber Using an ElectronicNose
M.A. Ryan, M.L. Homer, M.G. Buehler, K.S. Manatt, F. Zee
Jet Propulsion Laboratory, California Instituteof Technology
Pasadena CA 91 109

J. Graf
Johnson Space Center, NASA
Houston TX 77058

ABSTRACT andmaybeappliedtoenvironmentalmonitoringand
quality control in such wide fields as food processing, and
An ElectronicNose is being developedatJPL industrialenvironmentalmonitoring [3,4]. A baseline of
andCaltechfor use in environmentalmonitoring in the "healthyair" is established,anddeviations fromthat
International Space Station. The Electronic Nose baseline are recorded as changes in resistance of the
(ENose) is anarray of 32polymer film conductometric sensors. The pattern of distributedresponse of the
sensors; the pattern of response may be deconvoluted to sensors may be deconvoluted, and contaminants identified
identify
contaminants in the environment. An and quantified by using a software analysis program such
engineeringtestmodel of the ENosewas used to as pattern recognition and/or neural network.
monitor the air of the EarlyHuman Test experiment at
Johnson Space Center for 49 days. Examination of the The overall goal of the program at JPUCaltech is
data
recorded by t h e ENose shows
that
major the development of a miniature sensor which may be used
excursions in the resistance recorded in the sensor array tomonitorthebreathingair in the International Space
may be correlated with events recorded in t h e Test Logs Station,
and
whichmay
coordinated
be with the
of the Test Chamber. environmental control system to solve air quality problems
without
crewintervention. Progress to that
goal will
INTRODUCTION depend on the success of this portion of theElectronic
Noseprogram,which is the development of a prototype
Theability to monitor theconstituents of the systemwhich will be the subject of an experimental test
breathing air in a closed chamber in which air is recycled during a space shuttle flight in 1998.
is importantto NASA for use in closedenvironments
such as the space shuttleandthe space station. At The Electronic Nose
(ENose) now
under
present, air quality is determinedafterthefact by development at JPL and Caltech is designed to monitor
collecting samples and analyzing them on the ground in forcommoncontaminants in space [Link]
laboratory
analytical
instrumentssuch as a gas ENose is notintendedtobeananalyticalinstrument
chromatograph-mass spectrometer (GC-MS). The such as a GC-MS, although it is being developed with
availability of a miniature, portable instrument capable of the capacity to identify a limited number of contaminants
identifying contaminants in the breathing environment at and metabolic products and to quantify them within 10-
part-per-million
levels
wouldgreatly enhance the 20% of their concentration. The goal of this portion of the
capability for monitoring the quality of recycled air as well program is to develop the ability to detect the 12 target
as providingnotification of thepresence of potentially compounds listed in Table 1 at levels approximately 1/2
dangerous substances from spills [Link] the SMAC (Spacecraft MaximumAllowable
instrument is Electronic
the Nose now
under Concentration) levels. For most compounds, SMACs are
development at JPL and Caltech[1,2]. in the single to tens of parts-per-million (ppm) regime.
Success of t h e ENose to detect and
quantify
An electronic nose is an array of conductometric contaminants will be confirmed by traditional analyses of
chemical sensors which change resistance when air samples with GC-MS.
exposed to vapors. The sensors are not specific to any
one vapor;it is in the use of an array of sensors, each with In theexperimentdescribed in this paper,an
a different sensing medium, that gases and gas mixtures early developmental modelof the ENose was attached to
can be identified by the pattern of response of the array. the air recycling line of the Early Human Test Experiment
Electronic Noses have been discussedby several authors, NASA-Johnson
at Space Center in January
 Compound
shuttle
Detected
on m"(PPW
( P P ~ )[51 1 hr / 7day [6,7]

alcohols
methanol <1 301 7
ethanol .5 - 5 "-
2-propanol .4 - 4 400 / 60
methane 1-10 5300 15300
ammonia 0 301 0
benzene < .1 101 0.5
CO2 320 13000 1700
formaldehyde 0 .4/0.4
Freon 1 13 .1 - 1 50 I 50
hydrazine 0 4 / .04
indole 0 1 I .05
toluene .4 - 4 161 16

Table 1 : Target compounds for electronic nose shuttle experiment

1997. This experimentofferedanearlyopportunityto were made using hybrid microelectronic cofired ceramic
determinesome of theoperatingparameters of the (alumina) processes. Electrodesand
contacts were
ENose and to observe its operation in a closed deposited as thick films using screen printing. The
environment similar to that of the space shuttle or space substrate layoutandfabrication has been discussed in
station.
The
version of theENose used in this detail elsewhere [8].
experiment was an early developmental model, and the
analytical software for identification and quantification of In the EHT model,24polymeric sensors were
contaminants is not yet completed; thus, only qualitative made by depositing a solution of polymermixed with
interpretations of events can be made. carbon black to make a film 1 - 5 microns thick in contact
with gold electrodes.12 polymerswereused in this
EXPERIMENTAL experiment,fouron each of [Link]
wasdeposited in 2 positionson each [Link]
The
Electronic Nose Engineering
Development chippositionwasoccupied by resistors included in the
Model: JSC Experiment ENose sensor head for
electronic
reference. A
thermistor
was
included
on
the sensor headfor
An engineeringdevelopment
model of an temperature monitoring. The 12 polymers used are:
electronic nose was fabricated atJPL and plumbed in to
theairrevitalizationline of the EarlyHumanTesting A poly(4-vinylphenol)
Chamber (EHT) at NASA-Johnson Space Center for 49 B poly(styrene-co-allyl
alcohol)
days in early 1997. Air whichwasto be recycled in the C poly(vinylch1oride-co-vinyl acetate)
EHT Air Revitalization System (ARS) was heated to 30 - E poly(viny1 acetate)
34 O C before being passed through the electronic nose H poly(styrene)
chamber. After leaving the ENose chamber, the air was I poly(styrene-co-maleic
anhydride)
directed to the ARS. Theairwhich passed through the J poly(su1fone)
ENosechamber bypassed the EHT testing equipment, K poly(methy1methacrylate)
then rejoined the air stream tobe revitalized. M poly(viny1
butyral)
P poly(ethy1ene-co-vinylacetate)
Presence of a contaminant in air is measured as Q poly(ethy1eneoxide)
achange in resistance in the polymer films. Sensor
response is expressed as a ratio of resistance at time t=t Deposition of Films 160 mg of each polymer
to resistance attime t=O (deItaR1R).Theelectronic wasdissolved in 15 mL of [Link]
circuit for data acquisition has been previously described include
tetrahydrofuran (THF), acetone, methylene
in detail[9]. Data were acquired on a personal computer chloride,tolueneor a mixture of solvents. 40 mg of
using a program written for this purpose in LabView. carbon black was added to the solution, and dispersed
by sonication. 1 - 3 pLof solution was pipetted onto the
Desiqn of Sensor Head
The sensor sensor area and allowed to dry in air. The resistance of
head of theElectronicNoseused in this experiment the resulting films was in the range 1 -100 kL2. Solution
consisted of 32 sensor positions arranged on 4 wasadded in increments of 1 pL until thedesired
substrates, each with 8 sensor positions. The substrates
[Link] of polymer films as 25 mm
sensing media in an electronic nose has been discussed
in detail by several authors, including the Caltech group
working with JPL on this project [l-41.

ENoseSvstem
The
ENose
developmental
system used in this experiment is shown in the diagram
in Figure 1 .

AIR OUT
I' :: ..
I
to ARS ekctrodes &ymer films cover layer

Figure 2: Sketch of the ceramicsubstrate chip

containing eight sensors

Data
Analvsis The goal of ENose the
development is construction of anairqualitymonitor
capable of identifying t h e target compounds in Table 1 at
less than SMAC [Link] this goal,data
analysis
software
which recognizes the patterns of
response of the target compounds is beingdeveloped.
This software has not been completed, so the data from
the EHT experiment may be used only qualitatively.

U 0.0090
AIR IN
from EH'I 0.0080

0.0070
Figure 1 : Diagram of ENosesystemused in EHT
0.0060
experiment e
;
d
0.0050

4 0.0040
Flowing air (.25 Umin) was taken
from the
slipstreamexitingthe EHT chamber, anddirectedinto 0.0030
t h e ENose system. Theair,whichhadbeenheatedto
0.0020
30 - 34 "C for EHT testing purposes, was directed either
through an activated charcoal filter, put in line to provide 0.0010
150 0
100 50 200 250 300
clean air baseline data, or though a dummy glass filter, Time (min.)
put in line toprovide apressure dropsimilartothe
[Link] Figure 3: Partialtraining set forENoseanalysis
open the valvetothecharcoal filter andprovide 30 software
development.
Methanol response of 4
minutes of clean air flow every four hours; otherwise, the polymers.
airwentthroughthe glass filter. Air thenenteredthe
glass enclosed sensor head chamber where resistance ENose data analysis development
requires
wasmeasuredevery 30 seconds, andthenleftthe recordingtraining sets of polymer response atthe
ENose system to enter t h e air revitalization systemof the concentrations desired. A sample of a partial training set
EHT. is shown in Figure 3, where response of 4 polymers to
150,100, 50 and 25 ppm methanol in air is plotted
The experiment was controlled using a personal againsttime. As may be seen in this figure,different
computerand a [Link] polymers havedifferent
magnitudes of response to
DAQPad sent commands to the solenoids to open and methanol;polymer Q responds veryweakly whereas
close, and acquired resistance data from the sensors by polymer A hasa very strongresponse. It is these
measuringthevoltageatacurrentprovided by the differenceswhichallowsidentification of compounds;
DAQPad. each compound has a unique pattern of response, and
magnitude of response may be correlated to
A sketch of the sensor chip is shown in Figure 2. concentration of compound. The response to a mixture
Eachchip consists of eight sensors approximately 1.5 of compounds has been shown in most cases to be the
mm x 2 mm. Polymer films weredeposited on the linearcombination of theresponse to the individual
sensor positions as described above. compounds [lo].
ENOSE RESPONSE INEHT CONCLUSIONS

Responses of theENosehavebeencorrelated Operation of the ENose for 49 days in the JSC

to several events recorded in the test logs of the EHT. EHT experimentairstreamshowedthatthepolymer
Two
of the mostnotablecorrelations are shown in sensors will respond to events in thebreathingair
Figures 4 and 5. [Link] of ENose sensor spiking
in temporal correlation with GC observed C02 spiking is
In Figure 4, spiking in C02 levels, as determined especially
encouragingfor use of t h e ENose in
by C02 measurements made by on line monitoring of t h e [Link] of C02 rise
air to be recycled, correlates in time with spikes in ENose was not sufficient to be a danger to crew members, but
sensor response. The magnitude of response for the ENose was able to observe the rise, and thus could
different polymers was different; as shown in the figure, beused to monitorandquantify the C02 levels in an
Polymer E respondsstronglyandPolymerCrather enclosed space.
weakly to the changes.
As may be seen in both Figures 4 and 5, there is
C02 concentration fluctuated between .46% and littlebaseline drift overtheperiod of 12hours. As the
.60% as apercent of chamberair (measured by GC); sensor is zeroed on cleanairprovided by thecarbon
this is reflected as a fluctuation of a similar magnitude in filter, moderate baseline drift, defined as a change over
the mostresponsivepolymer (E). On thosedays in time in the resistance of the polymer in the presence of
which C02 spiking is found not in the EHT the same atmosphere, will not affect the operation of the
measurements, spiking is not found in the ENose data. [Link],comparison of the responses ofthe
polymers to cleaned air show that the carbon filter was
Figure 5 showssignificantchange in sensor not sufficiently large to give good cleaning for 49 days.
resistance when the EHT chamber door was opened to The limit of the filter used is about 15 days of continuous
the outside and untreated air was allowed to enter the operation.
Future
investigations will include
filter
chamber. The event in Figure 4 which begins about 8:OO materials capable of longer term operation.
may be correlated to the JSC Test Director Log entry at
8:03 AM (3/14/97). Careful study of the datacurves in Figure 4
show a cycling which corresponds to the carbon filter on-
Transfer in of Dr. Z and a blood draw technician off cycle.
Temperature measurements showedthat
through
outerlock
door. Mylar screen turning on the valve to direct air through the carbon filter
implemented to secureair integrity. heated t h e air by 1-2 'CMany
. of thepolymers are
strongly responsive to temperature changes of 1-2 'C.
Subsequent rises in sensor resistance correlate Future work also focus on calibration of the temperature
with openingthe EHT chamber, and a fall in response response as data which must beincluded in thedata
correlates with closing the EHT chamber. In spite of the analysis software and on design of a system which is not
use of a Mylar screen, there was substantial change in highly responsive to temperature changes in the
constituents in chamber air at the time the outerlock door environment.
was opened. There is no independent analysis of the air
constituents at thattime, so it is possible only to observe Figure 5, which shows t h e response of the
the change and correlate it with events recorded in the sensor to outer door activity, shows that the ENose will
logs. As theidentificationsoftware is not yet complete, respond to leaks in a chamber. Further development of
the sensor responses to opening the chamber cannot be thedataanalysissoftwareand the low concentration
deconvolutedto identify the mixture of compounds monitoring capacity of the device will lead to a miniature
inducingthe response. instrument
for
environmental monitoring
and
early
detection and identification of changes in environmental
Several other events recorded in the logs or the constituents.
onlinemonitoring of the EHT werecorrelated with
response in theENose,such as change in humidity, Development of the low concentration monitoring
water spills andcrewreport of anodor in theshower capacity of the ENose will be accomplished by improving
area. the polymer response and by diminishing the noise in the
electronicsresponse. Polymer response improvement
includes
development of applicationmethods
and
determination of optimum carbon
loading
and film
thickness.
0.0040
0.0035

0.0030

0.0025

0.0020

0.0015

0.0010

0.0005

0.0000

-0.0005

6:OO 8:OO 1O:OO 12:oo 14:OO 16:OO

Time of Day (CST)
Figure 4: Spiking in EHT air CO, levels may be correlated to spikes in Enose
sensor response at the same time. Dotted vertical lines show the 30 minute
reference cycle with a corresponding dip in sensor response.

0.0600

0.0500

0.0400
. *

.. .-
0.0300 . -

0.0200

0.0100

7:OO 9:OO 11:oo 13:OO 15:oo 17:OO

Time of Day (CST)
Figure 5: When the EHT chamber door was opened, there is significant change in
resistance in Enose sensors. Dotted vertical lines show 30 minute reference
cycles with a corresponding dip in sensor response.
 The
goal of this portion of the ENose
developmentprogram is demonstration of the ability of
the device to recognize and quantify the12 contaminants
listed in Table 1. It is intendedthat a demonstration
experimentbeperformed with theENoseon a Shuttle
flight in [Link] of theENose’sabilityto
detectleaks
andunusual
events in the JSC EHT
experimentshowsthat with development of analysis
software, it will be possible
to build a miniature
environmental monitoring device.

ACKNOWLEDGMENTS

The research reported in this paper was carried

out by the Jet Propulsion Laboratory, California Institute
of Technology under a contract with the National
Aeronauticsand Space [Link]
supported by NASA Code UL.

REFERENCES

1. M.S. Freund and N.S. Lewis, Proc. National Academy

of Science, 92, 2652, (1995).
2. M.C. Lonergan, [Link],[Link],S.A.
Beaber, R.H. Grubbs and N.S. Lewis, Chem. Materials,
8, 2298 (1996).
3. H.V. Shurmer and
[Link], Sensors and
Actuators, 88, 8, (1992).
4. J.W. Gardner, T.C. Pearce, S. Friel, P.N. Bartlett and
N. Blair, Sensors and Actuators, 818-19, 240, (1994).
5, J.T. James, et a/., Aviation, Space Environ. Med., 65,
851 (1994).
6. SpacecraftMaximumAllowableConcentrations for
Selected AirborneContaminants, Vols. 1 & 2, National
Academy Press, Washington, DC (1994).
7. Space PhysiologyandMedicine, A.E. Nicagossian,
C.L. Hunton & S.L. Pool, eds., Lea and
Febiger,
Philadelphia (1994).
8. M.G. Buehlerand M.A. Ryan, [Link]’l Conf. on
Microelectronic Test Structures(March 1996).
9. M.G. Buehlerand M.A. Ryan. Roc. SPlE Conf. on
Electro-OpticalTech. for ChemicalDetection, (April,
1997).
10. N.S. [Link] Institute of Technology,
private communication.
APPENDIX C: CREW PROCEDURES FOR FLIGHT EXPERIMENT
ENOSE ACTIVATION

1. Remove ENose from locker

2. Remove and discard protective paper from hook and pile fasteners. Attach ENose to locker
TBD
3. Connect prerouted ENose power cable PIN 101 08-1 0082-06 to ENosePWR connector
4. Check that ENose PWR - OFF
5. Connect ENose power cable to PPDB J2 connector

NOTE
Immediately following ENose PWR ON, PWR MEASURE LED will illuminate (green).
After approx. 60 secs internal solenoid will cycle (audible click) and
BASELINE LED will briefly illuminate (green),then off.
Low audible beep, then 10 secs later audible fanfare.
Internal pump then turns on and remains on (audible buzzing sound).
Within 15 - 20 sec after power on, green PWR MEASURE LED
blinks amber approximately every15 sec

6. ENose PWR - ON
Log PWR ON MET: I
7 . 4 PWR MEASURE LED - illuminated (green)
8. 4 Air flow or buzz of internal pump audible
* *
If PWR MEASURE LED is not illuminated green
* *
andlor no airflow or buzz of internal pump not audible,
* *
perform 1.4a, ENOSE PWR MEASURE
* *
LED NOT ILLUMINATED
* *
* *
I f PWR MEASURE LED is not blinking amber
*
ENose PWR - OFF, wait 10 sec *
* *
ENose PWR - ON
* *
Log PWR ON MET: I
* *
Wait 20 sec
* *
Report to MCC, status of PWR MEASURE LED and
* *
PWR ON MET

9. Take two photos documenting deployed setup

10. Report PWR ON MET to MCC

65
ENOSE DAILY MARKER

NOTE
Mark 1 must be performed at least 2 hrs after initial ENose activation
and be complete by 3 hrs after activation, or
be performed in next available 3-hr period.
No wait is required after anysubsequent restart.

1. Remove alcohol wipe kit and one Grab Sample Container(GSC)

2. Temp stow alcohol wipe kit
3. Complete E-Nose Daily Marker Datasheet block 1.
I f illuminated, wait to perform alcohol wipe operation (step 8) until green BASELINE LED is off
(approx 1 5 min)
4. Complete ENose Daily Marker Data Sheet blocks 2-3
5. Using procedure on GSC, collect air sampleat 10to 25 cm in front of ENose air inlet
6. TempstowGSC
7. Complete ENose Daily Marker Data Sheet block 4
8. Remove one alcoholwipe and gently wave it at 5 to 10 cm in front of t h e ENose air inlet for
30 to 60 sec
9. Replace alcohol wipe in its wrapper
10. Discard alcohol wipe/wrapper away from ENose
11. Complete ENose Daily Marker Data Sheet block 5.
12. For continuous 20 sec,

4 green PWR MEASURE LED - blinking amber approx. every15 sec

* *
I f PWR MEASURE LED is not blinking amber
*
ENose PWR - OFF, wait 10 sec *
*
ENose PWR - ON *
* *
Log PWR ON MET: I
* *
Wait 20 sec
* *
Report to MCC, status of PWR MEASURE LED and
* *
PWR ON MET

13. Complete ENose Daily Marker Data Sheet block 6

14. Restow GSC, alcohol wipe kit
15. Complete ENose Daily Marker Data Sheet blocks 7 and 8.
16. Call MCC, report illumination status of green BASELINE LED, GSC completion MET

66
ENOSE RESTART

1. ENose PWR - OFF, wait 10 sec

2. ENose PWR -ON, log PWR ON MET: / :
3. 4 PWR MEASURE LED - illuminated (green)
4. 4 air flow or buzz of internal pump audible
* *
I f PWR MEASURE LED is not illuminated green
*
and/or noair flow or buzz of internal pump not audible, *
* *
perform 1.4a, ENOSE PWR MEASURE LED
* *
NOT ILLUMINATED

5. Report PWR ON MET to MCC

ENOSE DEACTIVATION

1. ENose PWR - OFF

2 . MET:
Log /
3. Disconnect ENose power cable from ENose box and PPDB, secure cable ends
4. Stow ENose

67
APPENDIX D: ENose STS95 flight data analysis result, LMNLS method
Lines which contain identification and quantification of the daily markers (wipe events) are in blue type.
1 2 3 4 5 6 7 8 9 10 11 1 31 2 I GMTtime f i l e t i m e %rsdu
ben eth f irned CH4 MeOH pro to1 amn f o r wat deltaRH wipe1 mddhrmn
_"" ""- """

0 0 0 0 0 0 0 0 0 0 0 13 332 I 10310004 102910130 7

0 0 0 0 0 0 0 0 0 0 0 19 0 I 10310025 102910151 7
0 0 0 0 0 0 0 0 0 0 0 12 0 I 10310238 102910284 7
0 0 0 0 0 0 0 0 0 0 0 8 0 1 10311319 102920135 3
0 0 0 0 0 0 0 0 0 0 9 0 1 10311633 102920329 5
0 0 0 0 0 0 0 0 0 0 8 0 I 10312147 102920643
0 0 0 0 0 0 0 0 0 0 402 1 10312232 102920688
0 0 0 0 0 0 0 0 0 10 0 I 11010128 102930076
0 0 0 0 0 0 0 0 0 9 0 I 11010155 102930103
0 0 0 0 0 0 0 0 0 7 0 I 11010252 102930160

I
""_"""""""""~2""
1 3 4 5 6 7 8 9 10 1113 1 2
_"" ".

0 0 0 0 0 0 0 0 0 0 0 6 0 I 11010601 102930349 4
0 0 0 0 0 0 0 0 0 0 0 8 0 I 11011409 102940050 5
0 0 0 0 0 0 0 0 0 0 0 9 0 1 11011601 102940161 5
0 0 0 0 0 0 0 0 0 0 0 7 0 1 11011814 102940295 5
0 0 0 0 0 0 0 0 0 0 0 9 0 I 11011855 102940336 4
0 0 0 0 0 0 0 0 0 0 0 7 330 1 11012333 102940613 4
0 0 0 0 0 0 0 0 0 0 0 10 0 I 11020104 102940704 5
0 0 0 0 0 0 0 0 0 0 0 9 0 I 11020636 102950249 5
0 0 0 0 0 0 0 0 0 0 0 5 0 I 11021039 102950493 6
0 0 0 0 0 0 0 0 0 0 0 10 0 I 11021447 102950740 7

1 2 3 4 5 6 7 8 9 10 11 1 2 13 I _ -- -------- "_
"__""""""~~""""""""""""~"""""__""""~~"~------- .

0 0 0 0 0 0 0 0 0 0 0 10 0 I 11021739 102960124 6
0 0 0 0 0 0 0 0 0 0 0 11 0 I 11021911 102960216 5
0 0 0 0 0 0 0 0 0 0 0 0 910 1 11022148 102960374 6
0 0 0 0 0 0 0 0 0 0 0 8 0 I 11022302 102960448 5
0 0 0 0 0 0 0 0 0 0 0 10 0 I 11022323 102960469 8
0 0 0 0 0 0 0 0 0 0 0 14 0 I 11030028 102960534 5
0 0 0 0 0 0 0 0 0 0 0 7 0 I 11030712 102970150 4
0 0 0 0 0 0 0 0 0 0 0 7 0 I 11031524 102970643 6
0 0 0 0 0 0 0 0 0 0 0 7 0 I 11031651 102970729 7
0 0 0 0 0 0 0 0 0 0 0 5 0 I 11031703 102970741 7

_"""""~""~"2"""
1 3 4 5 6 7 8 9 10 13
11 12
.-I I _"""".

0 0 0 0 0 0 0 0 0 0 0 9 0 1 11031846 102980058 5
0 0 0 0 0 0 0 0 0 0 0 9 0 j 11031932 102980103 6
0 0 0 0 0 0 0 0 0 0 0 8 2 9 6 1 11032213 102980264 4
0 0 0 0 0 0 0 0 0 0 0 1 2 0 I 11032230 102980282 5
0 0 0 0 0 0 0 0 0 0 0 1 0 0 I 11032303 102980315 4
0 0 0 0 0 0 0 0 0 0 0 7 6 3 I 11032324 102980336 5
0 0 0 0 0 0 0 0 0 0 0 1 0 0 I 11041152 102990298 4
0 0 0 0 0 0 0 0 0 0 0 1 0 0 I 11041424 102990449 6
0 0 0 0 0 0 0 0 0 0 0 9 0 I 11041740 102990645 6
0 0 0 0 0 0 0 0 0 0 0 8 5 1 3 I 11042128 103000087 5

1 2 3 4 5 6 7 8 9 10 113
1 12 I
"_""""""_~"~""""""""""~~~""""""""""~~""""""""""""
0 0 0 0 0 0 0 0 0 0 0 6 0 I 11042144 103000103 7
0 0 0 0 0 0 0 0 0 0 0 7 0 I 11042232 103000151 5
0 0 0 0 0 0 0 0 0 0 0 6 0 I 11042350 103000228 6
0 0 0 0 0 0 0 0 0 0 0 9 0 I 11050146 103000345 5
0 0 0 0 0 0 0 0 0 0 0 7 0 I 11050414 103000493 5
0 0 0 0 0 0 0 0 0 0 0 9 0 I 11052041 103010692 9

68
APPENDIX E: REPORT ON ANALYSIS OF GRAB SAMPLE CONTAINERS FROM
STS-95

69
 Life Sciences, Systems and Services
TO: John T. James.
Ph.D.ISD2 Technical Monitor

FROM: Thomas F. Limero,Ph.D./Toxicology Laboratory Supervisor

SUBJECT: Results from the Analyses of STS-95 Mission Air Samples

for Supporting the E-Nose Experiment

DATE: January 8, 1999

INTRODUCTION

An environmental monitoring prototype calledthe E-Nose was supported by the Toxicology

Laboratory on STS-95. Toxicology support for this experiment consisted of providing seven
instantaneous grab samplers for air sample collected near the operating E-Nose duringthe
mission and the subsequent groundbased analyses of these air [Link] report presents the
groundbased analysesof the air contaminants in the grab samples from STS-95 and quality
assurance data that verifiesthe results. The results in this report will be compared to the real time
sensor data acquiredby the E-Nose duringthe mission.

The Toxicology Laboratoryof the Medical Operations Branchat the Johnson Space Center
processed and analyzed sevenair samples from STS-95 forthe E-Nose experiment. Seven grab
sample containers (GSC)were used to collect air samples in the Discovery mid-deck forthe
period MET day 1 to METday 7. A backup GSCwas not used during the mission and was
assigned as a trip control. The samples and
the trip control were returned tothe Toxicology
Laboratory on November 10,1998 for analyses throughthe United Space Alliance Flight
Equipment Processing Center.

GRAB SAMPLE CONTAINERS (GSC)

Tracking Container Sample Location Sample Time Date

Number Number Received
AA02359 1043 MID-DECK MET1\4:40 11/10/98
AA02360 1035 MID-DECK MET2\3:10 11/10/98
AA0236 1 1046 MID-DECK MET3\4:05 11/10/98
AA02362 103s MID-DECK h E T 4\22? 11/10/98
AA02363 1020 MID-DECK MET 5\2:38 11/10/98
AA02363 1039 MID-DECK MET 6\2:05 11/10/98
AA02365 1 os0 MID-DECK MET 7\1:57 11/10/9s
AA02366 1036 TRIP CONTROL 12/08/98

Wyle Laboratories, Inc. 1290 Hercules Drive, Suile 120 Houston, TX 77058 Tel: (281) 212-1200 FAX: (281) 212-1210
The GSC air snmples and the t r i p control were analyzed using both GC and GC/MS methods
according to the Toxicology Laboratory ISO-9000 Work Instructions (WI-TOX-003)
“Measurement of VOCs i n Spacecraft Air Using Grab Sample Containers” and (WI-TOX-004)
“Gas Chromatography Analysis of Methane, Ethylene, Hydrogen, Carbon Monoxide, and
Carbon Dioxide in Spacecraft Air Using Grab Sample Containers.”GC and GUMS analyses
were initiated immediately after receipt of the samples.

Results: GSC

The analytical results are reported in Table 1. Quantitation of TO-I4/polar compounds,and GC

target compounds was derivedfrom calibration curves, whilethe quantitation of other
compounds was based upon response factors available in the literature (see WI-TOX-003 for
details). All G C M S target compounds wereidentified and quantified if their base peakarea was
greater than 0.5% of the fluorobenzene base peak area. No attempt was made to identify
nontarget compounds with peak areas of lessthan 20% of the fluorobenzene standard peak area,
as these compounds wouldhave little effect on the toxicological assessment of the Orbiter’s
breathing air quality. However, if significant
a non-target compound (i.e. above 20% of
fluorobenzene peak area) was detected in one GSC air sample, this compound was searched for
in other samples. Compound concentrations listed as “< the laboratory report detection limit”
indicate that this target compound was analyzed for, but not detected.

S u l h r hexafluoride (SF6) was detected in all STS-95 air samples, by using the GC/TCD method
for hydrogen analyses. The concentrations of SF6 were not calculated because of the low
toxicity of the compoundand the current unavailability of a standard.

OAIOC: GSC

Each sample containerwas cleaned, proofed, certified leak-free, and prepared for shipment to
Kennedy Space Centerin the JSC Toxicology laboratory according to WI-TOX-010 “Cleaning
Summa-Treated Canister” specifications. AAer the containerswere evacuated, 1Occ of surrogate
standard (14SOppb of each component prepared on 4/10/98) was injected into each using a pre-
conditioned 2Occ syringe on 9/17 /98. Following the dosing procedures, each GSC was
individually bagged to “clean room” requirements and delivered to the Flight Equipment
Processing Center for bench review and transfer to KSC. Recoveries of surrogate standards from
the STS-95 air samples and the trip control are presented in Table 2. All recoveries were within
advisory QC limits. Pressure for each GSC air sample was measured and recorded during the
sample nnalJ,sis process. These data are presented i n Table 3. All GSC air samples appeared to be
leak-free.

Verification of instrument function and adherence to Toxicology Laboratory Work Instructions

was documcntcd for CSC air sample analyses by the completed “GC Method Calibration and
Blank Summary” and “Volatile Organic GUMS Tuning and Mass Calibration-
Bromofluorobenzene (BFB)” fornls presented i n Appendix 1.

Abnormalities: GSC

None