SENSORY EVALUATION
OF DIESEL EXHAUST ODORS
f
U. S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE
Public Health Service
Environmental Health Service
-------
SENSORY EVALUATION
OF DIESEL EXHAUST ODORS
Amos Turk, Ph. D.
Professor, Department of Chemistry
The City College of the City University of New York
Janet T. Wittes
Department of Statistics
Harvard University
and
L. R. Reckner and R. E. Squires
Scott Research Laboratories, Inc.
Prepared under Public Health Service Contracts
No. PH 27-66-96 and No. CPA-69-528
U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE
Public Health Service
Environmental Health Service
National Air Pollution Control Administration
Raleigh, North Carolina
February 1970
For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington. D.C., 20402 - Price 70 cents
-------
The AP series of reports is issued by the National Air Pollution Control
Administration to report the results of scientific and engineering studies,
and information of general interest in the field of air pollution. Information
reported in this series includes coverage of NAPCA intramural activities
and of cooperative studies conducted in conjunction with state and local
agencies, research institutes, and industrial'organizations. Copies of AP
reports may be obtained upon request, as supplies permit, from the Office
of Technical Information and Publications, National Air Pollution Control
Administration, U.S. Department of Health, Education, and Welfare, 1033
Wade Avenue, Raleigh, North Carolina 27605.
National Air Pollution Control Administration Publication No. AP-60
-------
FOREWORD
Exhaust gases emitted by diesel engines are characterized by odors
that are offensive in varying degrees to many members of the general
public. The increasing use of diesel-powered trucks and buses in urban
environments has resulted in widespread public awareness of the diesel
exhaust odor problem.
It has not been established at this time •whether any health hazards are
involved; nevertheless, because the odors are unpleasant and irritating to
people, studies of their origin and possible elimination are a necessary part
of the over-all program of automotive air pollution research conducted by
the National Air Pollution Control Administration (NAPCA).
Several research organizations, working under NAPCA sponsorship,
are conducting programs of engine testing and chemical analysis designed
to establish the identity and relative concentrations of all products of
incomplete combustion present in diesel engine exhaust.
To correlate these chemical analyses with the presence of odors, one
must devise means of rating the odors, both as to quality and intensity, on
some sort of numerical scale. Satisfactory correlations, once accomplished,
would make it possible to evolve criteria for readily enforceable standards
for control of diesel exhaust odors standards based on chemical composi-
tion rather than odor ratings and hence less subject to human error or bias.
This publication consists of two parts: (1) a revision of Selection and
Training of Judges for Sensory Evaluation of the Intensity and Character of
Diesel Exhaust Odors (PHS No. 999-AP-32), a report prepared in 1967 under
Public Health Service Contract No. 27-66-96, and (2) a subsequent report
prepared by Scott Research Laboratories, Inc. , under Public Health Service
Contract No. CPA-69-528.
Part 1 outlines the development of training methods and chemical odor
standards by •which human panelists can measure the quality and intensity of
diesel exhaust odors. These techniques are being applied in much of the
PHS-sponsored research involving air pollution by diesel engines.
Part 2 summarizes further work undertaken to develop and extend the
earlier studies. It describes the physical arrangement of an exhaust dilution
and panel exposure system and presents instructions for conducting odor
evaluation tests. It also discusses design of test programs and statistical
analysis of test data.
-------
ABSTRACT
Exhaust gases emitted by diesel engines are characterized by offensive
odors, which can be rated numerically by human judges. Correlation of such
ratings with the chemical composition of diesel exhaust will aid in (1) estab-
lishing Federal standards for diesel exhaust emissions and (2) developing
methods of diesel odor control.
Part 1 presents instructions for the training of judges to rate the odors
in terms of intensity and quality standards provided to them for reference.
Part 2 describes the physical arrangement of an exhaust dilution and panel
exposure system. It describes the performance of odor evaluation tests,
design of test programs, and statistical analysis of test data.
Appendices describe (A) the theoretical basis for air purification re-
quirements in test chambers, (B) composition and makeup of diesel odor
standards, and (C) mathematical derivations of the statistical procedures.
-------
CONTENTS
PART 1. SELECTION AND TRAINING OF JUDGES FOR SENSORY
EVALUATION OF THE INTENSITY AND CHARACTER
OF DIESEL EXHAUST ODORS 1
INTRODUCTION 3
ENVIRONMENTAL CONDITIONS FOR SENSORY TESTING .... 5
SELECTION OF JUDGES FOR DIESEL EXHAUST ODOR
STUDIES 7
The Triangle Test 7
The Intensity Rating Test 10
The Multicomponent Odor Identification Test 13
Demonstration of Satisfactory Test Behavior 15
TRAINING OF JUDGES FOR DIESEL EXHAUST ODOR
STUDIES 17
Initial Improvement of Performance 17
Exposure to Diesel Exhaust 18
Learning the Diesel Odor Standards 19
ACKNOWLEDGMENTS 20
PART 2. SENSORY EVALUATION OF DIESEL EXHAUST ODOR
INTENSITY 21
INTRODUCTION 23
EXHAUST DILUTION AND PANEL EXPOSURE SYSTEM 25
TRAINING OF JUDGES FOR DIESEL EXHAUST ODOR-
INTENSITY MEASUREMENTS 29
Learning the Diesel Odor Intensity Standards 30
Program to Test Judges' Performance 31
Exposure to Diesel Exhaust 32
PANEL PROCEDURES DURING TEST PROGRAMS 35
DESIGN OF TEST PROGRAMS 37
STATISTICAL ANALYSIS OF TEST DATA 39
ACKNOWLEDGMENTS 41
APPENDIX A. Concentrations of Odorous Vapors in Test
Chambers 45
APPENDIX B. Sensory Standards for Diesel Exhaust Odor Study ... 51
APPENDIX C. Statistical Derivations 59
-------
PART 1.
SELECTION AND TRAINING OF JUDGES
FOR SENSORY EVALUATION
OF THE INTENSITY AND CHARACTER
OF DIESEL EXHAUST ODORS
Amos Turk, Ph. D.
Professor, Department of Chemistry
The City College of the City University of New York
Janet T. Wittes
Department of Statistics
Harvard University
Prepared under Public Health Service
Contract No. PH 27-66-96
1967
-------
PART 1.
SELECTION AND TRAINING OF JUDGES
FOR SENSORY EVALUATION
OF THE INTENSITY AND CHARACTER
OF DIESEL EXHAUST ODORS
INTRODUCTION
When a person is exposed to diesel exhaust on the street or highway,
he may sense odor, chemical irritation in the nose or eyes, sound, velocity
pressure of moving air, temperature gradients, impaction of small particles
and the sight of smoke, moving vehicles, and other associated actions.
When the aggregate of such experience becomes sufficiently unpleasant, the
threshold for individual action is exceeded and the affected person may
voice a complaint or react in some other relevant manner.
Odor is undoubtedly the prime sensory attribute of diesel exhaust under
the typical circumstances of human exposure. The sensory evaluation of
diesel exhaust odor under different conditions of dilution, engine type,
engine operation, fuel, fuel additives, and exhaust control devices will
therefore provide part of the basis for specifying permissible conditions of
exposure.
The sensory evaluations will be made by human judges. Taken as a
group to increase the precision of the evaluations, the judges constitute a
sensory odor panel. In serving on the panel, the individuals will be called
upon to rank diesel exhaust samples to which they are exposed according
to odor intensity and according to the quality, character, or type of odor.
Depending on dilution of the exhaust, some samples will be chemically
irritating (in the sense of "pungent") and odorous at the same time, even
though the senses of irritation and odor are physiologically distinct. In
some cases, therefore, the evaluation continuum will, in effect, comprise
both types of sensations.
The judges in these tasks express their evaluations of intensity and
quality in a quantitative way; they do not assume the role of individuals who
express their personal preferences of "like" or "dislike." The following
general requirements are imposed for the selection and performance of
such judges.
-------
1. They must have satisfactory sensory ability to distinguish among
odors of different intensities and to discriminate among different odor
qualities.
2. They must be emotionally receptive to tasks that involve quantita-
tive and discriminatory judgments without expressions of preference.
3. They must be trained for the specific tasks to be accomplished.
4. A nondistracting environment must be provided for the sensory tests
5. Standards should be established to provide a quantitative basis for
sensory measurements and to provide replicable anchoring points that
will facilitate interlaboratory comparisons.
EVALUATION OF DIESEL EXHAUST ODORS
-------
ENVIRONMENTAL CONDITIONS FOR SENSORY TESTING
A space for sensory testing should be free of competing distractions.
In-and-out visits and socialization by non-participants should be restricted,
except in demonstration exercises. The area for preparation and coding of
test samples or make-up and dilution of exhaust gas streams should be
separated from the panel members and not visible to them. Extraneous
sounds, especially the sounds of operation of test vehicles, should be in-
audible during the tests. Color schemes should be neutral.
The test area should have some means for space odor control, espe-
cially for odors introduced from the samples being tested. This may be
accomplished either by local exhaust of vapors from the test samples, or
by general purification of the room air through a recirculating device, pre-
ferably an activated carbon unit, or by some combination of both methods.
If a considerable volume of air containing odorous diluted vehicle exhaust
is spilled into the space during the tests, then the effectiveness of odor
control by general recirculation through an air purifying device will be
limited. The nature of this limitation is described in detail in Appendix A.
When the tests are to be conducted and reported on an individual basis
•without communication among judges, it is desirable to have separate booths
or enclosures for the individual panel members. When the tests are to be
cooperative, with discussion and comparisons among the judges, then a
conference table setup is convenient.
The panel moderator or chairman should be able to communicate
readily and conveniently •with all of the judges. The general atmosphere
should be comfortable and relaxing, but also should encourage the judges
to be attentive and serious.
Several simple rules for conduct should be imposed. No smoking,
eating, or drinking that is not associated with tests should be allowed
in the test area. Judges should be discouraged from the use of per-
fumes or perfumed cosmetics just prior to a test session. A period of
half an hour or more should be allowed after smoking, eating a meal, or
drinking coffee, before participation in a test exercise.
-------
SELECTION OF JUDGES
FOR DIESEL EXHAUST ODOR STUDIES
There is no magic number of panel members. Many panels number
five to fifteen judges, and a number close to ten is probably suitable for
diesel exhaust sensory •work. A large number of panelists should be avail-
able for testing so that no interruptions need be caused by temporary absen-
teeism or personnel turnover. The panel members should not be divided,
however, into "regulars" and "standbys. " Some rotation scheme should be
set up so that no trained panel member is allowed to go "stale. "
The major tasks that the odor judges will be expected to perform are:
1. To judge the relative intensity of diesel exhaust odors at dif-
ferent dilutions.
2. To discriminate among the different qualities of diesel exhaust
odors.
3. To combine (1) and (2) to give a composite profile.
The judges will be expected to follow instructions, and at the same time
to render independent judgments reflecting their own sensations and report
their findings .
Four main tests for selection of judges are recommended:
1. The triangle test.
2. The intensity rating test.
3. The multicomponent odor identification test.
4. Demonstration of satisfactory test behavior.
These tests are described in detail in the following sections. Statisti-
cal derivations of these tests are presented in Appendix C.
THE TRIANGLE TEST
Three test samples are presented at the same time. Two are identical,
the third is different. The candidate is requested to identify the different
or "odd" sample.
In dealing with untrained candidates, the moderator should avoid strange
or objectionable materials. It is convenient to use dilute aqueous solutions
of food flavors. These are easy to administer, and they produce a minimum
of initial shock to the candidate. The solutions may be smelled or tasted;
in both cases, olfactory discrimination is actually being used to make the
identification. Materials that are recommended are:
-------
Flavorant Approximate concentrations
Vanilla extract To one quart of water add 1 or
Lemon extract 2 drops of extract to achieve
Pineapple extract a detection threshold at a level
Almond extract such that about 75 percent of
Rum extract the triangle scores will be
Rose extract correct answers.
Mint extract
The samples should be presented at ambient temperature in small
throw-away paper cups.
Instructions are given as follows: "The object of this test is to distin-
guish differences between food flavors that are very weak in intensity. The
samples given to you consist of 1 or 2 drops of a common food extract,
such as vanilla, almond, or mint, in a quantity of pure drinking water. You
may smell or taste the samples, or both. Each test consists of three sam-
ples, two of which are the same and one of which is different. Your task
is to pick out the different or odd one. You do not have to identify the
flavor, just choose which one is different from the other two. Enter on
your score sheet the code number of the sample that you choose as the odd
one. "
Present a trial test to familiarize the candidate with the procedure.
This trial should be conducted at a higher level of intensity so that the
distinction is easy. Do not ask for a score. State the correct answer
verbally to the candidate. Answer any questions regarding procedure. The
purpose of this trial is to concentrate most of the learning into the first
experiment and to make the subsequent tests independent of each other to
a greater degree. Now administer and score five triangle tests.
Select the candidates on the following basis:
List the candidates and their scores (number of correct answers) in
order, highest score first.
Refer to Table 1 to determine which differences between scores are
significant.
Decide how many panel members you should select from the candidates
available. Determine from the table at what levels there are statistically
significant differences between candidates and make your selection accord-
ingly. For these rough screening operations, it is recommended that con-
fidence limits be set at about 25 or 30 percent, rather than the usual 5 per-
cent. This will have the effect of initial rejection of most of the candidates
that are likely to be unsuitable. &
*The procedure of this and some later examples does not describe a series of
independent selections between randomly chosen pairs of candidates. Instead,
the candidates among whom we are discriminating are ranked according to
their scores. Consequently, tables of the distribution of difference between
any two scores are not strictly appropriate to choosing among more than two
candidates. Exact significance levels should be based on the distribution of
the range of scores. This would require separate tables for each number of
candidates being screened. Such a procedure would be unwieldly, arid we
feel that the tests based on the difference between two scores serve as satis-
factory approximations to the exact significance levels.
EVALUATION OF DIESEL EXHAUST ODORS
-------
Example 1. There are ten candidates, who take five triangle tests each.
We need about five or six panel members from among these candidates.
Their scores are:
Candidate
A
B
C
D
E
F
G
H
I
J
Number of correct answers
5
2
3
0
0
2
3
5
5
4
Table 1. SIGNIFICANCE OF DIFFERENCES AMONG
SCORES IN TRIANGLE TESTS
Number
of
tests
1
2
3
4
5
Difference between
two candidates in
number of correct
answers
0
1
0
1
2
0
1
2
3
0
1
2
3
4
0
1
2
3
4
5
Probability that at
least this difference
could have been
obtained by chance, %
100
44
100
59
10
100
66
19
2
100
71
26
5
0-5
100
74
31
9
2
0.1
Step 1. Choose an acceptable significance level. In this case we -will
choose 0.20 to 0. 30 (70 to 80 percent confidence levels).
Step 2. List the candidates in order of the number of correct answers
they gave.
Candidates
Number of correct answers
A, H, I
J
C, G
B, F
D, E
5
4
3
2
0
Selection of Judges
-------
Step 3. Check Table 1 to determine how much of a difference in the
scores of two candidates is needed to give confidence that they are signi-
ficantly different from each other. In this example, the number of tests
equals 5; therefore, a difference of 2 or more in score indicates a signi-
ficant difference in performance.
Step 4. Choose the candidates required. Starting from the top, A, H,
and I have a score of 5. But the scores of candidates A, H, I, and J differ
by less than 2; therefore this difference is not significant, and we can con-
sider these candidates to be "equal. " If we need five or six panelists and
have no other candidates, we must add C and G to the panel. (J, C, and G
can be considered "equal. ") The panel •will then consist of the first six:
A,H,I, J,C, and G.
Note that we could establish higher confidences in our selections by
giving more tests, say 20 or 25 instead of five. This is not recommended
because the extra effort can be more profitably expended in the screening
of additional candidates.
THE INTENSITY RATING TEST
A series of dilutions of an odorant in an odorless diluent is set up.
One sample is removed from the series. The candidate is asked to replace
it according to its odor intensity in the position from which it was taken.
For use in selecting candidates, the odorant should be fairly strong
•when it is in pure form, and not toxic, unpleasant, or strange. Any of a
number of fruity or fragrant odors is acceptable. Some possibilities are
amyl acetate, eucalyptol, oil of wintergreen, and heptaldehyde.
In the example that follows, the odorant is amyl acetate in propylene
glycol (low-odor perfumer's grade). Amber 2-ounce glass bottles with
plastic screw caps are used.
Concentration
(fraction by volume
of amyl acetate in
Bottle No. Procedure the solution)
1 Add 10 ml amyl acetate. 1
2 Add 10 ml amyl acetate plus 1/2
10 ml propylene glycol. Mix.
Using a 10 ml pipet, remove
10 ml of the mixture and transfer
to bottle 3.
3 Contains 10 ml of material from 1/4
bottle 2. Add 10 ml propylene
glycol. Mix. Remove 10 ml of
mixture and transfer to bottle 4.
4 Contains 10 ml of material from 1/8
bottle 3. Add 10 ml propylene
glycol. Mix. Remove 10 ml of
mixture and transfer to bottle 5.
10 EVALUATION OF DIESEL EXHAUST ODORS
-------
Continue this procedure until 20 bottles are prepared. The odor should
be detectable by a sensitive person as far as Bottle 19'or 20 in the dilution
series .
If the bottle number is designated n, then the concentration in any
bottle equals 21"11.
The bottles must appear identical and must be consecutively numbered
so that the numbers are not visible to the candidates. Labels can be put on
the bottoms of the bottles or on the face of the bottle if the label is obscured
by wrapping with aluminum foil during the test. The bottles are lined up in
sequence in front of the candidate.
The following instructions are given: "The 20 bottles lined up in front
of you all contain solutions of amyl acetate, which is a synthetic banana oil.
They differ from each other in odor strength; the most intense odor is on
the left, and the intensity gradually decreases from bottle to bottle toward
the right. The last bottle on your right has so little banana oil odor that it
may not be detectable at all. In the test you are about to perform one of
these bottles will be removed from the series, from a position unknown to
you. The task •will be to replace it in the proper location in the series, the
location from which it was taken. If it is replaced in its proper position, it
•will smell stronger than the bottle on its right and -weaker than the bottle on
its left. Proceed as follows. First familiarize yourself with the odors of
the bottles in the series. Start from the right (number 20) remove the cap,
sniff gently and recap. Then do the same for every other bottle going to
your left. Remember, the odors will be getting stronger toward the left and
you will fatigue your sense of smell temporarily if you sniff too long at the
more intense odors. You need not smell every bottle in the series at this
stage of the test. " (After a brief time for familiarization . . . ) "Now leave
the room. " (The tester removes one bottle, and rearranges the others to
obscure the gap. The removed bottle is placed in front of the remaining 19).
"Now come back into the room and replace the bottle in its proper place in
the series."
Statistical scoring procedure: The same procedure, consisting of four
tests, is used for each candidate. The bottles are successively removed in
the standard order: 12, 8, 16, 3.
Step 1. Score as follows:
Positions removed Score (higher number
from correct location is worse score)
0 0
+ 1 1
+ 2 4
+ 3 9
+_4 or more 16
Step 2. Refer to Table 2 to determine which differences between
scores are significant.
Selection of Judges 11
-------
Table 2. SIGNIFICANCE OF DIFFERENCES
AMONG SCORES IN INTENSITY RATING TEST
Differences between
scores of any two
candidates
18
20
23
25
29
34
44
56
probability that at least
this difference could have
been obtained by chance, %
30
25
20
15
10
5
1
1/10
Example 1. A candidate performs as follows:
Missing bottle Replaced in position
12
8
16
3
12
7
6
5
What is the candidate's score?
Bottle
number
12
8
16
3
Positions removed from
correct location
Score
12
8
16
3
12 =
7 =
6 =
5 =
0
1
10
-2
0
1
16
4
Total score 21
Example 2. Two candidates show the following performances.
better than the other? What confidence is there in the selection?
Is one
Candidate A
Bottle missing Replaced in position
Candidate B
12
8
16
3
12
8
16
3
15
13
12
1
15
10
13
3
12
EVALUATION OF DIESEL EXHAUST ODORS
-------
A's score Score
12 15 = -3 9
8 13 = -5 16
16 12 = 4 16
31=2 4
45
B's score Score
12 15 = -3 9
8 10 = -2 4
16 13 = 3 9
33=0 0
22
A - B = 45 22 = 23
Therefore B scored better than A. From the significance table (Table 2),
this could have been due to chance alone 20 percent of the time. We have,
therefore, a confidence of 80 percent in our selection of B over A. If we
demand a confidence of 90 percent before making a decision, then we cannot
rule out A as a possible panel member before further testing.
THE MULTICOMPONENT ODOR IDENTIFICATION TEST
This test presents three mixtures to the candidate. These mixtures
contain, in sequence, 2, 3, and 4 odors out of a possible total of 8 known
standards. The candidate is told how many components to look for, and is
asked to identify them.
The following materials are recommended as standards:
1. Oil of cade (burnt)
2. Cassia (cinnamon)
3. Eucalyptus
4. Amyl acetate (banana)
5. Clove oil
6. Orange oil
7. Almond oil (benzaldehyde)
8. Vanillin or vanilla extract
Instructions are given as follows: "You have in front of you eight
labeled common odors. They are (state which). These will be your refer-
ence standards and you may smell them whenever you wish. Now, you will
be given an unknown odor sample. This sample contains a mixture of two
of the standard odors. Write down on your score sheet which two odors are
present in the test sample." When this test is finished, repeat the instruc-
tions but state that the new sample contains a mixture of three standards.
Selection of Judges 13
-------
Then repeat the instructions again for the third test explaining that the sample
contains a mixture of four odors.
The multicomponent odor identification test is scored as follows:
Step 1. Administer the same test sequence, consisting of three tests,
to each candidate.
Test 1. Two components
Test 2. Three components
Test 3. Four components
Step 2. Calculate the score of each individual candidate according to the
following procedure.
/Number correct V /Number correct\ , /Number correct^
Total score = J +^
^ test
-n tegt
Step 3. Refer to Table 3 to determine which differences between scores
are significant.
Table 3. SIGNIFICANCE OF DIFFERENCES BETWEEN
SCORES IN MULTICOMPONENT ODOR
IDENTIFICATION TEST
Difference between
scores of any two
candidates
6
7
8
9
11
15
Probability that at least
this difference could have
been obtained by chance, %
35
SO
15
10
5
1
Example 1. Candidates A and B scored as follows in the multicomponent
odor identification test. Are the scores significantly different?
Test
1
2
3
Number c
Candidate A
2
3
1
:orrect
Candidate B
0
1
4
Score of A = 22 + 32 + I2 14
Score of B = O2 + I2 + 42 = 17
Difference = 3. The probability that this difference could have been
obtained by chance is greater than 25%. Therefore, we do not attach any
14
EVALUATION OF DIESEL EXHAUST ODORS
-------
signifance to this difference, and the two candidates should not be differen-
tiated from each other.
Example 2. Choose the best four candidates from the following:
Test Number correct obtained by each candidate
ABCDEFGHI
1021110202
2332010113
3424134023
The scores are computed on the basis of the sum of the squares of the
correct answers and are then listed in order of magnitude. The order is
Candidate Score
A 25
I 22
C 21
B 17
F 16
E 11
G,H 5
Now, referring to Table 3, we can say with a confidence of 80 percent
that any candidates whose scores differ by no more than 7 points are con-
sidered "equal. " By this criterion of confidence, candidates A, I, and C
could be considered to be equal and could be selected for the panel. Candi-
date B differs from Candidate A by more than 7 points and is, therefore,
ruled out at this confidence level. At a confidence level of 90 percent, how-
ever, which implies a lower probability (10 percent) that the observed dif-
ferences of scores could have been obtained by chance, we could have
selected all of the candidates except the last three.
DEMONSTRATION OF SATISFACTORY TEST BEHAVIOR
During the phases of candidate selection, the panel leader or moderator
should be attentive to various aspects of the candidates' test behavior.
These aspects are:
Speed. The best behavior is purposeful and deliberate, neither exces-
sively hasty nor slow.
Interest level. The candidate should feel challenged and motivated.
Candidates who find the work distasteful or uncomfortable should not
be selected.
Domination. In group testing, the candidate should be helpful when
asked, but should not try to push his opinions on others.
Independence. The candidate should be willing to consider the sugges-
tions of others, but should not be influenced to change score against
his own judgment.
Honesty. Candidates •who try (successfully or not) to decode labels or
peek under bottles should be rejected.
Selection of Judges 15
-------
The panel leader should observe the candidates' behavior in these mat-
ters carefully but unobtrusively. No specific numerical scoring system is
recommended; the leader should rely on his own judgment with regard to
the sensitive question of whether to reject, on the basis of poor test behav-
ior, any otherwise acceptable candidates.
16 EVALUATION OF DIESEL EXHAUST ODORS
-------
TRAINING OF JUDGES
FOR DIESEL EXHAUST ODOR STUDIES
Training should be scheduled according to the foil-owing procedure.
1. Improve the performance of the judges in the same tests that were
administered for panel selection. (Estimated time: 3 to 4 days.)
2. Expose the judges to diesel exhaust under the conditions to be used
in the testing and under conditions of random ambient air dilution, to
diesel fuel, and to components of diesel exhaust condensate if such are
available. Discuss the four diesel odor descriptors (see Appendix 3)
during each of these exposures. (Estimated time: 2 to 3 days. )
3. Introduce the judges to the kit of diesel odor reference standards.
Explain how the standards will be used in measuring intensity and quality
of diesel exhaust odor. Train the panel members in recognizing
individual samples in the kit and in recognizing multicomponent samples.
Establish minimum requirements for these tasks. During this training,
intersperse exercises that relate the diesel quality standards to the
sensory quality of diesel exhausts. (Estimated time: 1 to 2 weeks.)
4. Eliminate panel members as necessary in phases A, B, and C for
reasons of poor performance or motivation, emotional problems, exces-
sive domination or dependence, or other difficulties apparent to the
moderator.
5. The panel may now operate according to whatever experimental
design is adopted as the original research program. The continued
conduct of the work will, in effect, continue the training of the panel.
The panel members should be checked periodically to insure that the
level of the performance with the kit components obtained in phase 3 is
maintained.
Some of these phases of panel training are considered in more detail
in the following sections.
INITIAL IMPROVEMENT OF PERFORMANCE
The tests used for the selection of panel members should be repeated,
with one important difference. In training, no errors are ignored. As
soon as a trainee or group of trainees scores a test, the correct answers
are disclosed; the trainee repeats the exercise with attention focused on
elimination of the error, with the help of the leader.
Example 1. Trainee: "I detect eucalyptus, banana, clove, and cinna-
mon in this mixture. "
Leader: "The mixture contains eucalyptus, banana, almond, and cin-
namon. That means you misidentified one element. Refer to the almond
and clove standards, recheck them, and then smell your mixture again. "
17
-------
(Note — The leader must not be deprecatory. Say "misidentified" to
express the facts, not "confused" or "made an error" or "goofed" to express
disapproval. )
Trainee: "Well, that almond is overpowered by the other odors. It
didn't come through for a while. "
Leader: "That's all right, it may happen that way. A four-component
mixture is a difficult task. Focus your attention on one component at a time,
take only short sniffs so you do not get fatigued, and refer to the standards
whenever you have to. "
During training, allow trainees to work together and help each other by
suggestions and exchange of comments and samples. Keep records of the
trainees' progress.
Training should be continued until there is a noticeable levelling-off in
performance. Prolonged continuation of training on known mixtures is
wasteful, because the diesel work will involve new materials and, of neces-
sity, new training.
An important aspect of training, and, later, of panel utilization is the
question of whether the panel should work on an individual or on a group
basis. The individual basis eliminates social influences; each panelist's
score is purely his own judgment. The group basis allows for cooperative
suggestions that •will call a panelist's attention to something that he himself
senses only after his attention is properly focused. The decision between
the two methods is best made on a statistical basis. It will be appropriate
to use the t-test to determine whether the difference between the average
panel scores obtained on an individual and on a group basis is significant.
EXPOSURE TO DIESEL EXHAUST
The panel members must be exposed to diesel exhaust under various
conditions that are likely to be experienced in the proposed testing program
and also under typical conditions expected on the street or highway. The
panelists should also be exposed to diesel fuel and/or fuel components,
diesel exhaust condensate, and other relevant odor sources. Explain to the
panel members that the quality of an odor can be described in terms of
quality components. The explanation can be an expansion of remarks such
as the following:
"You have identified the components of mixtures of two, three, or four
odorants in the tests you have been conducting during the last few days.
This identification is also a description of odor quality. You may say that
the odor of a given sample is a mixture of lemon and vanilla. This statement
describes the quality of the odor mixture, just as the same statement would
be a description of flavor if the sample were a piece of cake. We are going
to describe diesel exhaust in terms of four qualities:
1. Burnt/smoky. This is the quality in which diesel exhaust is related
to the odors of other products of burning or combustion.
2. Oily. This quality is the oiliness related to the presence of the odor
of the heavy components of unburned fuel.
18 EVALUATION OF DIESEL EXHAUST ODORS
-------
3. Pungent/acid. This is the quality associated with pungency in high
concentration, gradually changing to an acid or sour quality with greater
dilution.
4. Aldehydic/aromatic. This is the quality related to what may be
thought of as the "fragrant" aspect of dies el exhaust. It may also be
thought of, in most instances, as that quality not represented by smoki-
ness, oiliness, or pungency."
LEARNING THE DIESEL ODOR STANDARDS
(Refer to Appendix B for details of the composition and preparation of
the diesel odor kit. )
The following introductory remarks may be appropriate:
"'You will be given a kit of standard odors to which you will refer in
making your odor judgments. Standard odors D-l through D-12 represent
the intensity levels of diesel exhaust odor, without regard to quality. Stand-
ards B-l through B-4 represent four intensity levels of burnt odor; O-l
through O-4 represent intensity levels of oily odor. (The moderator at this
point should explain the entire kit.) Before using these standard odors to
measure diesel exhaust, it will be necessary for each of you to learn the
various standards and to get some familiarity with mixtures of the stand-
ards. "
The training for recognition of odor standards should now proceed along
the following lines.
1. The judges should be trained in the intensity levels of the D-l through
D-12 series. Judges should be able to identify any unknown sample to
•within +_ 1 or 2 intensity levels.
2. The judges should learn each of the quality components at the
"extreme" concentration (B-4, O-4, P-4, and A-4). They should then
learn to identify each intensity of each quality, 16 bottles in all. They
should be able to identify all unknowns as to quality and intensity about
90 percent of the time.
3. The compositions of standard mixtures are shown in Tables 4 and 5.
The leader should specify to the judges the number of components in each
training mixture. The judges should practice identifying two- and three-
component mixtures until performance levels off. They should reach a
level at which they identify mixtures correctly about 80% of the time.
Correct identification is much more difficult when the number of compon-
ents is unspecified. The probabilities of obtaining correct identifications
by chance are shown in Table 6.
Table 4. COMPOSITION OF COMPONENTS
FOR ODOR STANDARD MIXTURES
Quality
Burnt
Oily
Pungent
Aldehydic
Code (t = training)
B-t
0-t
P-t
A-t
Composition
B-4 (See Appendix 3)
100% octylbenzene
P-4
1 % A in mineral oil
Training of Judges 19
-------
Table 5. COMPOSITION OF ODOR STANDARD
MIXTURES FOR TRAINING
Number of
components
2
3
4
Code (t = training)
AB-t
AO-t
AP-t
BO-t
BP-t
OP-t
ABO-t
ABP-t
AOP-t
BOP-t
ABOP-t
Composition
1 : 1 mixtures of the
solutions of Table 4.
About 14 to 1 ml of
solution is placed in a wad
of cotton in a plastic
squeeze bottle.
1:1:1 mixtures as above.
1:1:1:1 mixture as above.
Table 6. PROBABILITY OF CORRECT IDENTIFICATION OF TRAINING
STANDARDS BY CHANCE ALONE
Number of components
stated to be in
mixture
1
2
3
4
Not disclosed
Number of training
standards
4
6
4
1
15
Probability per test
of correct identification
of standard by chance alone
1/4
1/6
1/4
1
1/15
ACKNOWLEDGMENTS
The folio-wing chemists participated in the development of the odor
standards: Louis Reckner, Robert Squires, and Z. Tomaras, all of the
Scott Research Laboratories; Stanley Mehlman and Elizabeth Wolf, both
of the City University of New York; and Jonathan Turk, of Brown University.
20
EVALUATION OF DIESEL EXHAUST ODORS
-------
PART 2.
SENSORY EVALUATION
OF DIESEL EXHAUST ODOR INTENSITY
L. R. Reckner and R. E. Squires
Scott Research Laboratories, Inc.
Prepared under Public Health Service
Contract No. CPA-69-528
1969
-------
PART 2.
SENSORY EVALUATION
OF DIESEL EXHAUST ODOR INTENSITY
INTRODUCTION
The need for a reliable method to measure the intensity of diesel
exhaust odor has increased in recent years as greater attention has been
given to the reduction of air pollution from motor vehicles. Accurate, repro-
ducible, chemical or physical tests -would be the ideal answer to this need.
However, the validity of any instrumental method for odor measurement
must first be verified by comparison with results obtained from human
responses. Odor-panel techniques are, therefore, a prerequisite to this
ultimate goal, as well as a tool that can be used to advantage until adequate
instrumental procedures have been established.
The primary considerations in developing panel sensory techniques
are the presentation of the stimulus in a manner that stimulates realistic
environmental exposure and the objectivity of the resulting panel responses.
The five basic requirements for achieving these goals in diesel exhaust odor
studies are as follows:
1. Construction of an exhaust dilution and panel exposure system that
simulates on-the-street exposure.
2. Selection of panelists with superior olfactory ability.
3. Training of the selected panelists to rate the intensity of dilute
diesel exhaust.
4. Formulation of diesel odor intensity standards.
5. Design of test programs.
These requirements are discussed in detail in succeeding sections.
23
-------
EXHAUST DILUTION AND PANEL EXPOSURE SYSTEM
A system that can be used to dilute exhaust and present it to the odor
panel is essential for conducting experimental and research programs
involving the measurement of diesel exhaust odor intensity. The system
should meet the following criteria:
1. The system should present diluted exhaust to the panelists in a manner
simulating on-the-street exposure.
2. The concentration of exhaust odorants should not be altered, except
by dilution, between the exhaust source and the odor judge. Condensa-
tion, absorption, chemical reaction, or oxidation of exhaust odorants
•within the system must be minimized.
3. Additional dilution of the exhaust should not occur during the odor
appraisal.
4. The system should be suitably equipped to control and monitor a
broad range of exhaust concentrations.
The system illustrated in Figure 1 meets the above criteria and can be
used to perform all of the programs described herein. It has been success-
fully employed to rate exhaust intensity at a number of engine, fuel, dilution,
and steady-state operating-condition combinations. This system is described
in detail to demonstrate the precautions necessary to maintain the integrity
of exhaust' odor. Other systems that fit specific needs and situations may be
substituted if they meet the stated criteria.
The system delivers freshly diluted diesel exhaust to a sniff box at man-
ually controlled dilution ratios. Stainless steel and Teflon* lines are used
throughout. Exhaust flows from the center of the engine exhaust pipe into
the dilution air stream through a short length of heated tubing containing a
valve that controls the exhaust flow rate. An extended stem permits the
operator to control the flow rate while observing the dilution monitoring
equipment. A back-pressure valve in the engine exhaust pipe is used,
•when necessary, to create pressure (6 inches H^O) to drive the exhaust into L
the dilution air stream. Ambient air from a compressor is passed through
condensate, charcoal, and silica gel traps and used as the dilution air at a
constant flow rate of 2 cubic feet per minute.
The diluted exhaust is transferred to the sniff box through 1-inch-
diameter tubing. Heated stainless steel is employed in the vicinity of the
mixing point while the remainder is Teflon, which minimizes hangup of
odorous compounds. A baffle plate is used to diffuse the diluted exhaust
entering the sniff box. The box is constructed of stainless steel with a trans-
parent Mylar back, •which prevents the panelist from feeling closed-in while
*Mention of company or product names does not constitute endorsement by
the National Air Pollution Control Administration.
25
-------
SILICA
GEL TRAP
ROTAMETER
Figure 1. Exhaust dilution and panel exposure system.
he is sniffing. The box is illuminated by a. 25-watt bulb in the adjacent
exhaust chamber. A sheet of paper, held in place by cellophane tape, is
used to seal the face opening when the box is not being used. The box has a
volume of 1/2 cubic foot. The transfer time from the exhaust pipe to the
sniff box is approximately 4 seconds, and the residence time in the box is
15 seconds.
The odor room is constructed in accordance with the criteria specified
in Part 1. The temperature of the room is controlled at 72° +1° F. Char-
coal-filtered air is blown into the room to maintain a slight positive pressure
(0. 02 inch H2O), which prevents the diluted exhaust in the sniff box and other
contaminated air from entering the room.
The exhaust dilution ratio is continuously measured and recorded using
nondispersive infrared analyzers sensitized to carbon dioxide (003). The
exhaust-gas-analysis system is shown schematically in Figure 2. The CO,
concentration in the diesel exhaust pipe is measured by a 0 to 10 percent
analyzer that is spanned with a close-tolerance (+2 percent) analyzed gas
mixture containing 10 percent CO2 in nitrogen. The CO2 in the diluted ex-
haust stream is measured differentially with an analyzer having a nominal
range of 0 to 100 parts per million. A dual range (0 to 100 and 0 to 500 ppm)
instrument is recommended. Dilution air flows through the reference cell
while diluted exhaust (dilution air plus exhaust) flows through the sample cell.
In this differential analysis, the instrument response decreases as the CO?
in the cells increases even though the difference in concentrations between
the two cells remains constant. For example, a concentration of 100 ppm in
one cell and 0 in the other gives approximately twice the response given by
500 ppm in one cell and 400 ppm in the other. For this reason, calibration
26
EVALUATION OF DIESEL EXHAUST ODORS
-------
FLOWMETER
FLOWMETER
FLOWMETER
Figure 2. Dilution measuring system.
without a representative amount of CO£ in the reference cell is not valid.
Calibration curves are obtained for this instrument by adding low flows of a
calibration standard through a calibrated rotameter into the dilution air
stream at a point near where exhaust is ordinarily added. This simulated
actual use conditions. The readings from both infrared analyzers are
recorded on a dual-channel recorder.
Panel Exposure System
Z7
-------
TRAINING.OF JUDGES FQRJJIESEL EXHAUST
ODOR INTENSITY MEASUREMENTS
The dies el odor panel should consist of approximately ten judges. At
least 40 candidates should be screened in order to obtain a panel with the
required level of olfactory acuity.
Three tests are recommended for the selection of judges for diesel
exhaust odor intensity measurements.
1. The triangle test.
2. The intensity rating test.
3. Demonstration of satisfactory test behavior.
Statistical derivations and detailed descriptions of each test are given
in Part 1.
Part 1 also gives valuable information on the training of judges for
sensory evaluations, with details on the composition and preparation of the
odor standards. The procedures for training judges for intensity measure-
ments are presented here with additional detail based on experience acquired
after the initial work was published.
Training for intensity measurements is accomplished according to the
following procedure.
1. Improve the performance of the judges by repeating the tests that
were administered for panel selection. No errors should be ignored.
As soon as a test is scored, the correct answer is disclosed, and the
trainee repeats the exercise with attention focused on elimination of
the error. Training should be continued until there is a noticeable
leveling in performance.
(Estimated time: two 2-hour sessions)
2. Introduce the judges to the kit of diesel odor intensity standards.
Explain how they will be used. Train the panel members to recognize
the 12 standards in the series.
(Estimated time: six to eight 2-hour sessions)
3. Test the judges' ability to identify the standards.
(Estimated time: three 1-hour sessions)
4. Introduce the judges to dilute diesel exhaust. Train the judges to
rate the exhaust odor in terms of the intensity standards.
(Estimated time: four to six 2-hour sessions)
5. Eliminate panel members as necessary for reasons of poor per-
formance or motivation, emotional problems, or other difficulties
apparent to the moderator during Phases 1, 2, 3, and 4.
29
-------
6. The panel may now operate according to the experimental design
adopted. Individual performance during each program should indicate
when retraining is necessary. Performance may also be monitored
periodically by employing the test administered during Phase 3 above.
Some of these phases of training are considered in detail in the follow-
ing sections.
LEARNING THE DIESEL ODOR INTENSITY STANDARDS
The following introductory remarks may be appropriate:
"In front of each of you are 12 plastic bottles, each of which contains
an odor standard. The odorant in each bottle is the same, but the concen-
trations are different. Each bottle has a number on its base. The bottle
numbered 1 contains the weakest odor, and the bottle numbered 12 contains
the strongest. During the next few weeks you will be trained to identify each
of the standards by number. Following this phase of training, you will learn
to rate the odor intensity of dilute diesel exhaust in terms of the 12 odor
standards."
Training should now proceed along the following lines:
1. Train the judges to handle the standards. The routine handling pro-
cedure is to remove the standard from its rack, shake it gently, remove
the cap, place the bottle 1 to 2 inches below the nose, squeeze the bottle
gently and sniff the gas expelled, replace the cap, and return the stan-
dard to its original position in the rack. Care must be exercised to
prevent spillage or contact of the bottle with the nose. The same proce-
dure must be used regardless of the concentration of the standard being
sniffed.
2. Familiarize the trainees with the standards by administering inten-
sity ranking tests using the standards as the odorant. No errors should
be ignored, and emphasis should be placed upon association of the odor
intensity with the number on the base of the bottle.
The procedures are nearly identical to those employed during trainee
selection. The standards are lined up with the weakest (No. 1) on the
left and the strongest (No. 12) on the right. One bottle is selected at
random and removed from the series. The remaining bottles are
shifted to obscure the gap. The trainee is instructed to find the proper
location of the bottle that was removed. When the proper location is
found, the bottle -will smell stronger than the one on its left and weaker
than the one on its right. The trainee counts from left to right and
records the number of the standard. After everyone has recorded his
answer, the correct identity is revealed, and judges in error repeat the
exercise with attention focused upon elimination of the error.
This exercise is repeated at 3-minute intervals until the trainees can
consistently identify any unknown standard to within + 1 intensity unit
with confidence and without having to sniff more than two bottles in the
series before making the identification.
3. Train the judges to identify the intensity standards to +_! unit from
memory. Remove the odor standards from in front of the trainee.
30 EVALUATION OF DIESEL EXHAUST ODORS
-------
Randomly select standards and present them as unknowns at 3-minute
intervals. After each judge has recorded his response, reveal the
correct identity of the unknown. Instruct judges in error by greater
than one unit to re-sniff the samples. Continue this exercise until 90
percent of the responses are correct to +_! unit.
PROGRAM TO TEST JUDGES' PERFORMANCE
The judges' performance should be checked following initial training
and periodically thereafter to ensure that the performance level obtained
during training is maintained. The recommended test program consists of
three sessions, held on separate days. Two intensity standards are pre-
sented as knowns at the start of each session. The 12 intensity standards
are then presented in random order as unknowns. The standards are pre-
sented at 5-minute intervals to prevent olfactory fatigue. Each judge works
independently, and the identities of the unknowns are not revealed during the
test.
At least 90 percent of the responses obtained during this program
should be within +_! unit of the true value. Responses in error by greater
than one unit should be randomly distributed among the judges and
intensity standards. Failure to achieve this level of performance should
result in re-training the judge or judges in the area indicated necessary by
the data analysis.
Example: Ten odor judges each evaluated 12 intensity standards that
were presented as unknowns on 3 separate days. The data were tabulated
in Table 1 below, and the percentage of responses in each class was calcu-
lated. Over 90 percent of the responses were correct to within +1 unit of
the true value, indicating satisfactory overall panel performance.
Table 1. DISTRIBUTION OF RESPONSES
Error, odor units
0
±1
±2
±2
Total
Frequency
223
119
18
0
360
Total response, %
62
33 9b
5
0
100
Errors greater than one unit were tabulated in a matrix (Table 2) to
determine their distribution among judges and intensity standards.
A relatively high number of total errors •were recorded in the D and F
judge columns and in the 4 and 5 intensity standard columns. Judge D failed
to identify standards 4 and 5 within +_! unit on two of the three evaluations.
Judge F recorded two errors for standard 5 and one error each for standards
4 and 6. These errors account for the relatively high totals. This shows
that judges D and F should be re-trained in the area of standards 4, 5, and
6.
Training of Judges
31
-------
Table 2. DISTRIBUTION OF ERRORS GREATER THAN ONE UNIT
Standard
number
1
2
3
4
5
6
7
8
9
10
11
12
Total
Judge
A
X
X
2
B
X
X
2
C
0
D
X
XX
XX
5
E
0
F
X
XX
X
X
X
6
G
0
H
X
X
2
I
0
J
X
1
Total
0
1
1
5
4
2
1
1
1
0
1
1
18
EXPOSURE TO DIESEL EXHAUST
When the judges have been trained to indentify the intensity standards
presented as unknowns to within +_! unit 90 percent of the time, they are
ready to rate diesel exhaust. A variety of exhaust concentrations and engine
operating conditions should be presented. Emphasis should be placed upon
identification of the exhaust-odor intensity in terms of the odor standards.
An expansion of the following remarks is an appropriate introduction
to this phase of training:
"We are about to begin the final phase of training. The procedures
employed at the termination of this phase will be.used during future evalu-
ation'programs. The odorant will be dilute diesel exhaust presented in a
sniff box. Your task will be to sniff the exhaust and record the number of
the odor standard that has the same apparent intensity as the exhaust
sample. "
Training may now proceed along the following lines:
1. Familiarize the judges with the exposure procedure. Dilution air
is passed through the sniff box. The trainee is instructed to approach
the box, lift the door, press his face snugly against the box, and sniff
the contents in a normal manner. The judges approach the box in a
sequence specified by the panel moderator until everyone has sniffed
the sample. The entire group should sniff the .dilution air at least
twice using the same sequence and with the moderator answering any
questions that arise.
2. Introduce the judges to dilute diesel exhaust. The procedures of
step 1 above are repeated with two exceptions. First, dilute exhaust
instead of dilution air is passed through the sniff box. Second, the
judges are instructed to record the number of the standard that has
the same apparent odor intensity as the sample. The panel moderator
determines the average response for the group and presents that stand-
ard to the judges for reference. The judges re-evaluate the sample
and when all answers have been recorded, they are displayed on a
32
EVALUATION OF DIESEL EXHAUST ODORS
-------
blackboard. Judges with responses two units or more from the panel
average are instructed to re-sniff the exhaust and record their final
impressions.
The procedure is repeated using a number of exhaust concentrations
and engine operating conditions. The time between exhaust presenta-
tions must.be long enough to prevent olfactory fatigue. Individual
performance should be monitored closely. If a judge consistently
disagrees with the group average by two or more units, the reason
for the low or high responses must be determined and corrective
action taken.
As the trainees become familiar with the task, the deviation of indivi-
dual response from the group average will decrease until there is a
noticeable leveling in performance. At this time displaying the re-
sults and repeating exposure to the same sample is discontinued.
Standards are then selected at random and presented as unknowns
between exhaust samples. The true identity of the standard and the
average response for the exhaust samples are not revealed. This
procedure should be continued for at least one 2-hour session or until
individual responses are consistently less than two units from the panel
average.
During the final phase of training the alternate presentation of stand-
ards with exhaust is discontinued and two randomly selected standards
are presented as unknowns at the start of each session only. After the
judges have recorded their responses, the true identity is revealed
and those in error by greater than one unit are instructed to re-sniff
the standard. Exhaust samples covering a broad range of concentra-
tions and engine operating conditions are presented to the panel.
Judges work independently, and the results are not discussed. Samples
are presented until individual responses are consistently less than two
units from the panel average.
The training is complete, but the decision must be made whether the
judges should or should not physically refer to the odor standards
during formal exhaust evaluation programs. Physical reference may
result in olfactory fatigue and no improvement in panel performance.
The decision is best made on a statistical basis. The use of a t-test
is appropriate to determine whether the average panel response and
deviation between replicate responses obtained using the two methods
are significantly different. Unless there is a significant improvement
in performance, physical reference to the standards should not be
ma.de during an evaluation program.
Training of Judges 33
-------
PANEL PROCEDURES DURING TEST PROGRAMS
During formal programs, judges with colds or other disorders that
might affect their sensory ability should not participate. Judges should allow
a period of at least 1/2 hour after smoking, eating a meal, or drinking
coffee before participating in a test exercise. The use of perfumes or per-
fumed cosmetics must be discouraged.
Two randomly selected intensity standards are presented as unknowns
at the start of each session. After the judges have recorded their answers,
the true identities are revealed. Judges in error by greater than 1 unit
should re-sniff the standard.
Dilute exhaust is presented in the order specified by the program
design. Individual judgements are made at intervals sufficient in length to
prevent olfactory fatigue. Intervals of 3 to 5 minutes are recommended.
The judges work independently, and the results are not discussed during the
test session.
The judges do not sniff the odor standards during the test program
unless this has been shown to be beneficial in the final training phase. If
the judges refer to the standards during the test, it is recommended that
the reference be made only to confirm a predetermined answer and that the
sniffing of standards be minimized in order to reduce the possibility of
olfactory fatigue.
35
-------
DESIGN OF TEST PROGRAMS
Statistically valid conclusions concerning the effects of variables such
as engines, fuels, exhaust devices, operating conditions, and exhaust con-
centrations can b.e obtained only through formal test programs that are
designed to meet specific test requirements. The complexity of the statis-
tical design and subsequent data analysis is determined by the number of
test variables to be studi'ed and the statistical precision desired. The
following recommendations apply to the general design of test programs.
1. Selection of test variables. The number of variables investigated
during each program should be minimized in order to decrease the
likelihood of obtaining interaction effects that may mask the effect of
the'variables of interest. Only variables pertinent to the evaluation
should be included in the design. All test conditions should be con-
trolled at realistic levels.
2. Order of sample presentation. The order of sample presentation
may affect an individual's response to the odorant. Such influences
are commonly referred to as time error and contrast effect. These
influences, may be avoided by balancing the order of presentation so
that over the entire program each test condition will precede and
follow each other test condition an equal number of times.- In addi-
tion to the balanced order of presentation, partially balanced or
randomized-block, and paired-observation designs have been employed
with success under various test situations. When only two levels of a
single variable are of primary interest, the paired-observation design
is considered ideal because of the precision of the results obtained and
the simplicity of data analysis.
3. Number of replicate observations per test condition. Sufficiently
large numbers of replicate observations should be made in order that
conclusions concerning the test variables can be drawn with a high
degree of confidence. It is recommended that initial randomized-block
and paired-observation designs include 5 and 8 replicate observations
per test condition, respectively. Statistical procedures, beyond the
scope of this publication, may be employed to predict from initial
data the number of replicate observations that would be required to
draw conclusions at specified confidence levels during subsequent
tests.
4. Number of observations per test session. The number of observa-
tions per test session should be between 12 and 15, and should not
exceed 18. Presentation of a greater number of samples may result
in the loss of sensory ability because of either physical or psycholog-
ical factors.
A program design that has wide practical application is given in Table
3. The program was designed to test the intensity difference produced by
two levels of a variable at three engine operating conditions and one exhaust
concentration. The program follows a paired design •with a random order
37
-------
of presentation. Operating conditions are randomized within blocks.
term "block11 refers to two time periods within a test session.
The
The two levels of variable (land 2) may be two fuels of different com-
position, a fuel with and one without an additive, exhaust before and after a
device, two engines of different design, or any other pair of variables that
realistically fit the design. F, H, and I may be any conditions pertinent to
the evaluation of 1 and 2, such as three engine operating conditions.
Table 3. DESIGN FOR INTENSITY RATING PROGRAM
Day
1
2
3
4
Block
1
2
1
2
1
2
1
2
Test condition (engine condition, fuel)
(H, 1)
(H, 2)
(1-2)
(F, 2)
(F, 2)
(H, 1)
(F, 2)
(I. 2)
(H. 2)
(H, 1)
(I- 1)
(F, 1)
(F, 1)
(H, 2)
(F, 1)
(I. 1)
(F.I)
(I. 2)
(H, 2)
(I. 1)
(H, 1)
(F, 1)
(H, 2)
(F, 2)
(F, 2)
(I.D
(H. 1)
(I. 2)
(H, 2)
(F, 2)
(H. 1)
(F. 1)
(I. 1)
(F. 2)
(F, 2)
(H, 2)
(I. 2)
(I. 1)
(I. 1)
(H. 1)
(I. 1)
(F. 1)
(F. 1)
(H.I)
(I. 1)
(1.2)
(1.2)
(H, 2)
Notes: The design emphasizes definition of the difference between variables 1 and 2, but the re-
lationship between variables F, H, and I are also subject to analysis of variance.
F, H, and I may be any engine operating conditions. 1 and 2 may be any two levels of a variable
such as two fuels or two engines.
38
EVALUATION OF DIESEL EXHAUST ODORS
-------
STATISTICAL ANALYSIS OF TEST DATA
Many kinds of hypotheses concerning the odor intensity of diesel
exhaust can be made and statistically tested through the use of panel
response data. The "null hypothesis" is the only one that will be discussed
here. This hypothesis states that there is no real difference in odor inten-
sities produced during specified operating conditions. Tests are applied to
determine how reasonable the hypothesis is. This is stated in terms of
significance level (a). The significance level indicates the probability of
finding a difference when actually there is none. Significance is often ex-
pressed as a confidence level that is the complement of significance. For
example, a 95 percent confidence level corresponds to a 5 percent signifi-
cance level.
Two methods of anaylsis, analysis of variance and t-tests, are com-
monly used to test hypotheses concerning the odor intensity of diesel exhaust.
Both methods may be used to test the difference in intensity produced at two
or more exhaust conditions. However, the t-test may be used to test hypoth-
eses concerning only two exhaust conditions at one time. For example, if
it is desired to test the difference between five exhaust conditions, it is
necessary to compute ten t values. There are a number of reasons why
this is not good statistical practice. There is no restriction as to the num-
ber of conditions that may be evaluated simultaneously using an analysis of
variance. Therefore, analysis of variance techniques are recommended
when multiple tests of significance are desired.
Detailed descriptions of the above analytical techniques go beyond the
scope of this publication, and only the analytical procedure pertaining to the
program design given in Table 3 •will be illustrated. Reference to statistical
tests is recommended for those interested in obtaining a knowledge of analy-
sis of variance and/or diversified t-test techniques.
Panel results obtained during a program employing the design given in
Table 3 are shown in Table 4. The program tested the intensity difference
produced by two fuels (1 and 2) at three engine operating conditions, idle (I),
half load (H), and full load (F) and one exhaust concentration. A short
method of analysis, t-by-difference, was employed to test whether the dif-
ference between fuel intensities was significant at the 95 percent confidence
level (<*= 0. 05) for each operating condition independently. The 95 percent
confidence limits for the difference between means were also calculated. A
summary of the analyses is given in Table 5. Actual computations for the
half-load operating condition are given below to illustrate the calculation
procedure.
1. Test of significance
a. H: F£ Fj = 0. That is, the mean intensity rating for Fuel 1 =
mean intensity rating for Fuel 2. The hypothesis is accepted or
rejected by comparing t values computed from the data -with t
values obtained from tables published in statistics texts.
39
-------
Table 4. PANEL INTENSITY RATINGS OF DILUTED DIESEL EXHAUST
Day
1
2
3
4
Block
1
2
1
2
1
2
1
2
Average
I
Fuel 2
4.88
7.75
6.25
5.88
7.75
8.00
8.13
8.25
7.11
Fuel 1
5.75
6.50
5.13
5.63
5.88
5.38
6.00
6-88
5.89
Da
-0.87
1.25
1.12
0-25
1.87
2.62
2.13
1.37
1.22
H
Fuel 2
7.00
6.38
6.63
6.63
5.75
6.00
7.13
6.63
6.51
Fuel 1
5.25
5.00
5.00
5.88
4.63
5.00
5.25
5.88
5.23
Da
1.75
1.38
1.63
0-75
1.12
1.00
1.88
0.75
1.28
F
Fuel 2
5.63
6.75
7.75
6.50
S.38
6.38
6.25
6.88
6.56
Fuel 1
6.75
6.25
6.25
5.38
5.63
6.75
6.13
6.25
6.15
Da
-1.12
0.50
1.50
1.12
0.75
-0.37
0.12
0-63
0-41
= Fuel 2 - Fuel 1 panel rating.
Table 5. SUMMARY OF STATISTICAL ANALYSIS OF DATA IN TABLE 4
Engine operating condition
Fuel intensity difference (Fr> - F^)
95 % confidence interval
t for fuel difference (calculated)
^.975 (a =0-05, d.f. =7)
I
1.22
0.30, 2.14
3.13
2.36
H
1.28
0.91, 1.65
8.16
2.36
F
0-41
-0.31, +1.1]
1.39
2.36
b. a = 0.05. This is the chance of finding a significant difference
between Fj and F2 'when there is none.
c. Degrees of freedom (d.f.) = 7. That is, number of paired ob-
servations minus one.
d. Critical t - tl/2a and t± _ j/2a = +?. 365 as read from a published
t-table for the level ,of significance (a) specified and the d. f. of
the actual test.
e. Calculated t =
1.28
S/TT" 0.44/2.83
Where: d = average difference between responses,
2
S —
N-l
N
variance of the differences
between responses,
d. = difference between responses, and
N = number of paired observations.
40
EVALUATION OF DIESEL EXHAUST ODORS
-------
f. Calculated t (8.23) is greater than critical t (2. 365), and the
hypothesis of equal intensity produced by the two fuels at this
operating cpndition is rejected. Therefore, it is concluded with
95 percent confidence that Fuel 2 produced a greater intensity
than Fuel 1.
2. Confidence limits for the difference between two means
a. It is often desirable to express the difference between population
means in terms of confidence limits. The confidence limits
between means are
/T /T
d + '1/2* VN and d + *i 1/2. sjv
Where : d = average difference between population means,
S = square root of the variance of the difference between
population means,
N = number of paired observations,
i = level of significance, and
l/2a = a value obtained from a table for the level of
significance specified and the degrees of freedom
for the test (N 1).
b. The 95 percent confidence limits for the difference between fuel
intensities are:
d + t . S / — = 1. 28 (2. 365 x 0. 44 x 0. 354) = 0.91, and
1 - 1 / <-. a V N
_ AT
d + t, /0 SJ — = 1. 23 + (2. 365 x 0. 44 x 0. 354) 1. 65, or the
1 1 C.CL " N
intensity of exhaust produced by Fuel 2 is between 0. 91 and
1.65 odor-standard units greater than the intensity of exhaust
produced by Fuel 1 at this engine operating condition and
exhaust concentration.
ACKNOWLEDGMENTS
The authors wish to thank Dr. Amos Turk, consultant, of Danbury,
Connecticut, and Dr. S. K. Katti, of Florida State University, for their
invaluable assistance in the development of the techniques described herein.
Analysis of Data 41
-------
APPENDICES
A. CONCENTRATIONS OF ODOROUS VAPORS IN TEST CHAMBERS
B. SENSORY STANDARDS FOR DIESEL EXHAUST ODOR STUDY
C. STATISTICAL DERIVATIONS
-------
APPENDIX A. CONCENTRATIONS^ ODOROUS
VAPORS IN TEST CHAMBERS
Odor test chambers are used either for evaluation of odor-reducing
devices or to provide odor-free environments in which a jury can measure
the odors of materials, products, or foods. It is important to consider the
factors that determine the changing or equilibrium concentrations in such
chambers.
Processes that tend to increase the concentration of odorous vapors
are:
1. The generation of vapor •within (or injection into) the space.
2. The introduction of vapor by replacement of chamber air by ventila-
tion or infiltration with outdoor air of higher vapor concentration.
Processes that tend to decrease the concentration of odorous vapor are:
1. The treatment of the chamber air by a vapor-reducing device (e. g. ,
activated carbon recirculator).
2. The removal of vapor by replacement of chamber air by ventilation
or infiltration with outdoor air of lower vapor concentration.
The concentration of odorous vapor in a chamber will approach an
equilibrium point at which the rates of vapor-reducing and vapor-increasing
processes are equal. If it is assumed that air introduced into the chamber
by ventilation, infiltration, or recirculation through a treatment device is
completely and instantaneously mixed with the chamber air, then the
concentrations of vapors at any time and at equilibrium are given by the
general equations:*
-(Qi + EQr)t/V / CjQj + G
C = C°e + \Qi + EQr
-(Qi + EQr)t/vl
J
CiQi + G
00 ~ Qi + EQr
SPECIAL CASES OF THE GENERAL EQUATIONS
Consider the following possibilities:
1. Ventilation air is pure. (Ci = 0)
2. No vapor is being generated or injected. (G 0)
*See page 49 for definition of terms.
45
-------
3. The vapor reducing device is 100% efficient.
(E = 1 and Cr = 0)
4. The chamber is originally pure. (Co = 0)
5. The room is tight. (Qi = 0)
6. Combination of (1) and (2)
7. Combination of (1) and (3)
8. Combination of (2) and (5)
9. Combination of (1), (2), and (3)
For each possibility, 1 through 9, the equation for the vapor concentration
at any time (C) or at equilibrium (C°° ) is derived directly from the general
equations:
Possibility 1: C^ = 0
-(Qi + EQr)t/V
1 e
G
(Qi + EQr)
-(Qi + EQr)t/V
EQr)
Possibility 2: G = 0
-(Qi + EQr)t/V C.Q.
C =
\
and Cra —
' (Qi + EQr)
-(Qi + EQr)t/V
— (Qi + EQr)
Possibility 3: E - 1 and Cr = 0
-(Qi + Qr)t/V / c Q + G
C = Cne ' '
oc
[•
CiQi H
T\Qi+ Qr
-(Qi + Qr)t/V"|
J
h G
- Qr)
Possibility 4: Co = 0
CjQj + G [ -(Qi + EQr)t/vl
Qi + EQr [l e J
46 EVALUATION OF DIESEL EXHAUST ODORS
-------
Note that the concentration rises to the equilibrium value expressed by
the general equation.
Possibility 5: Qi = 0
-EQ t/V / \
C = C0e +-2- ( -EQrt/V\
EQr (l e j
and C«, = G/EQr
Possibility 6: Combination of (1) andi (2). Ci and G = 0
-(Qi + EQr)t/V
C = C0e
and C „ = 0
Possibility 7: Combination of (1) and (3). Q = 0, E = 1, and Cr = 0
-{Qi + Qr)t/V
C = C0e
Qr)t/V
Qr)
and
Qi + Qr
Possibility 8: Combination of (2) and (5). G = 0, and Qi = 0
-EQrt/V
C - C0e
and Co, = 0
Possibility 9: Combination of (1), (2), and (3). Ci = 0, G = 0,
E = 1, and Cr = 0
-(Qi + Qr)t/V
C = C0e
and Co, = 0
EXPRESSION IN TERMS OF AIR CHANGES
An air change is the addition to the chamber of a volume of air equal to
the volume of the chamber. Then the number (N). of air changes (dimension-
less) per unit time is given by:
N/t = Q/V, or
N = Qt/V
MIXING FACTOR
In any expression of the general form e~Qt/V or e~N. a mixing factor, m,
may be applied to account for the fact that dilution of air is not instantaneous,
and that concentration fall-off rates are actually smaller than the ideal values
given by the equations developed here. Brief- suggests that m commonly
ranges between 1/3 and 1/10.
Appendix A 47
-------
SENSORY TESTING
The efficiency of a vapor -reducing device may be measured by sensory
methods that obviate the need for determination of material concentrations.
If we select a sealed chamber in which no odor is being generated, then the
ratio of concentrations Cl/C2 corresponding to any two times tl and t2 during
the operation of the device is
(EQrm/V) (t2 ti)
Ci/C2 = e
It is possible to measure the ratio of two supra-threshold concentrations
of odorous vapor Cl/C2 by either a dilution or a matching technique. Let
Volume of air sample diluted to threshold _ .
Volume of original air sample
Then, the ratio of P values for any two concentrations Ci/C2 is:
PI/PZ =
We may therefore write:
(EQrm/V) (t2
= 0.434 (EQrm/V) (t2 t:)
Solving for E,
Z. 303 logiQPi/P2
= (Qrm/V) (t2 flT
The latter equation tells us that for a room of known volume and air
mixing characteristics, it is possible, by sensory determination of odor in
the room at different times, to measure the efficiency of a vapor-reduction
device operating within the room.
Example: In a 1000-cubic-foot sealed room, 50 cubic feet per minute of
air is recirculated by an air purifier of 60 percent efficiency. The mixing
factor in the room is 1/3. Some diesel exhaust vapor is spilled into the room
How long must the air purifier operate if the vapor concentration in the room
is to be reduced by 90 percent?
Solution: C2 = G! 0. 9Ci, and Ci/C2 = 10
t t _ Z. 303 log GI/CZ _ Z. 303 log 10
2 l ~ EQrm/V ~ 0. 60 x 50 x 1/3/1000
= Z30 min. , or 3 hr. 50 min.
48 EVALUATION OF DIESEL EXHAUST ODORS
-------
NOTATION
V = Volume of chamber
t = Time
C = Concentration of vapor in chamber at any time
CQ = Initial concentration of vapor in chamber
Coo = Concentration of vapor in chamber at equilibrium
Cj = Concentration of vapor in ventilation or infiltration air
C = Concentration of vapor delivered by the air treatment device
E = Efficiency of vapor reduction by the air treatment device
Qi = Volume rate of ventilation or infiltration
Qr = Volume rate of air delivery by the air treatment device
G = Quantity rate of generation of vapor within (or injected into)
chamber
N = Number of air changes
m = Mixing factor
REFERENCES
1. A. Turk, Measurements of Odorous Vapors in Test Chambers:
Theoretical. ASHRAE J. , Oct. 1963.
2. R. S. Brief, Simple Way to Determine Air Contaminants, Air
Engineering, Vol.2, p. 39(1960).
Appendix A 49
-------
APPENDIX B. SENSORY STANDARDS
FOR DIESEL EXHAUST ODOR STUDY
This Appendix describes the composition and method of preparation of
sensory standards for a diesel exhaust odor study. The standards are
designed to serve as sensory references for (1) the intensity of diesel
exhaust odor, without regard to the quality of the odor, (2) the quality of
diesel exhaust odor, expressed in terms of the intensities of each of four
odor quality descriptors, and (3) the intensity of odor-modifying agents
that are designed to improve diesel exhaust odor by odor masking or odor
counteraction.
The intensities of the individual odor qualities, taken together, consti-
tute a "quality-intensity profile, " or "QI profile. " The profile is not
designed so that the sum of its intensity scores should be related to the
overall diesel exhaust odor intensity.
Descriptors are used to designate diesel exhaust odor quality.
The descriptors are:
Name
Burnt/smoky
Oily
Pungent/acid
Aldehydic/aroma tic
Masking
Code
O
P
A
M
BURNT/SMOKY (B)
Burnt quality or smokiness is a typical odor component of the products
of combustion of organic matter. There is considerable variation in the
quality of burnt odors, however, among different conditions of combustion,
different materials being burned,and different states of molecular or aerosol
aggregation of the airborne combustion products. Many of the primary
combustion products are unstable and therefore unsuitable for use as sen-
sory reference standards. The chemical makeup of materials that have
a burnt odor includes products of decomposition and partial oxidation. Oil
of cade (juniper tar) is included in the B standard because this oil has a
typical burnt odor and is readily available from commerical sources.
Guaiacol and carvacrol impart the phenolic odor component that is con-
tributed in part by the oxidation of benzenoid aromatic matter. Acetylene
dicarboxylic acid is a commercially available chemical, the odor of -which
somewhat resembles that of a dilute mixture of carbon suboxide in air. Car-
bon suboxide is a likely product of partial oxidation that contributes to burnt
odor. Benzyl benzoate is an almost odorless diluent that solubilizes the
other components.
51
-------
Composition of B
Component
Oil of cade (Juniper tar)
Guaiacol
Carvacrol
Acetylene dicarboxylic acid
Benzyl benzoate
Percent by weight
Including
solvent
16. 8
0.2
0. 9
2. 1
80.0
100. 0
Excluding
solvent
84.3
1.0
4.2
10. 5
(solvent /odor ant,
4:1)
Intensity series (diluent: as noted)
Odor intensity
Slight
Moderate
Strong
Extreme
Code
B-l
B-2
B-3
B-4
Concentration (B/B + diluent)
Diluent
min. oil
min. oil
min. oil
benz. ben.
Fraction
1/720
1/180
1/45
1/4
Decimal
0. 00139
0. 00556
0. 0222
0.25
Percent
0. 139
0. 556
2.22
25. 0
Procedure
Make up the stock mixture B.
To make B-4, mix 1 part of B with 3 parts of diluent.
To make B-3, mix 1 part of B-4 with 14 parts of diluent.
To make B-l, mix 1 part of B-2 with 3 parts of diluent.
Notes: 1. The dilutions can be made on a volume basis without intro-
ducing any significant error.
2. Stock solution B should be shaken before dilutions are made
to bring into suspension any component of the oil of cade
that might have settled.
3. Note the change in diluent between B-4 and B-3.
OILY (0)
Oiliness is an odor quality generally associated with organic chemicals,
the molecular structure of which is characterized by long saturated hydro-
carbon chains. Among materials of plant origin, such substances are typi-
cally esters of long-chain fatty acids. In diesel exhaust, oily quality is
believed to be associated with the presence of high-boiling-point components
of unburned fuel, in the boiling ra'nge approximately 300° C (1 atm). The
most satisfactory standard found to represent this odor quality is
n-octylbenzene.
52
EVALUATION OF DIESEL EXHAUST ODORS
-------
Composition of O
n-octylbenzene, 100 percent
Intensity series (diluent: mineral oil)
Odor intensity
Slight
Moderate
Strong
Extreme
Code
O-l
O-2
0-3
0-4
Concentration, O/(O + diluent)
Exponent
2-7
2-5
2-3
2-1
Fraction
1/128
1/32
1/8
1/2
Decimal
0.0078
0. 0313
0.125
0. 5
Percent
0. 78
3.13
12.5
50. 0
Procedure
The stock material O is pure n-octylbenzene.
To make O-4, mix 1 part of O with 1 part of diluent.
To make O-3 , mix 1 part of O-4 -with 3 parts of diluent.
To make O-2, mix 1 part of O-3 with 3 parts of diluent.
To make O-l, mix 1 part of O-2 with 3 parts of diluent.
Note: The dilutions can be made on a volume basis without introducing
any significant error.
PUNGENT/ACID (P)
Diesel exhaust in high concentration can be perceived as an irritant.
The word "irritant" is used here to denote a substance that can be detected
by the common chemical sense, as distinguished from the specific olfactory
sense. Such irritants are said to have a "pungent" quality. This type of
common chemical sensation can coexist •with odor. For example, a concen-
trated mixture of butyric acid vapor in air is both pungent and odorous. As
the mixture is diluted, the common chemical irritation diminishes and then
disappears at concentrations at which odor still persists. The sensation is
then no longer said to be pungent; instead, it is described as "acid" or "sour.
The existence of pungent, acid components in diesel exhausts is evidenced
by (1) the interpretation of infrared spectra by Scott Research Laboratories
to indicate the presence of organic acid (carboxyl function), (2) the sensory
identification of "sour" substances among the column chromatographic frac-
tions of diesel exhaust obtained by Scott Research Laboratories, and (3)
the pungency experienced by direct exposure to concentrated diesel exhaust.
The character of "sourness" or "acid odor" (as distinguished from pun-
gency) varies among different sources. Butyric and valeric acid odors are
characterized by sour odor qualities typical of rancidification of organic
matter. An acidic odor quality more closely^elated to combustion products
is associated with organic acids that have olefinic or acetylenic unsaturation.
Appendix B
53
-------
Composition of P
Component
Percent by weight
Crotonic acid
Propiolic acid
Benzyl benzoate
3.3
2. 9'
93. 8
100. 0
Intensity series (diluent: mineral oil)
Odor Intensity
Slight
Moderate
Strong
Extreme
Code
P-l
P-2
P-3
P-4
Concentration, P/(P + diluent)
Fraction
1/720
1/180
1/60
1/20
Decimal
0. 0014
0. 0055
0. 0167
0. 05
Percent
0. 14
0. 55
1.67
5. 0
Procedure
Make up the stock solution P. The mixture will have to be warmed
slightly to bring the components into solution. Allow the solution to cool
to ambient temperature.
To make P-4, mix 1 part of P with 19 parts of diluent.
To make P-3, mix 1 part of P-4 with 2 parts of diluent.
To make P-2, mix 1 part of P-3 with 2 parts of diluent.
To make P-l, mix 1 part of P-2 with 3 parts of diluent.
Note: 1. The dilutions can be made on a volume basis without
introducing any significant error.
2. The components of P-4 may have to be warmed to bring
them into solution.
ALDEHYDIC/AROMATIC (A)
Aldehydes are known to exist as components of diesel exhaust. Some of
the column chromatographic fractions of diesel exhaust obtained by Scott
Research Laboratories were characterized as having "sweet" or "spicy"
odors. A mixture used at Scott Research Laboratories for setting up odor
intensity ratings of diluted diesel exhaust contained heptaldehyde as the
major component. These circumstances, taken together, support the
selection of a quality description of diesel exhaust odor that reflects alde-
hydic and other highly odorous, somewhat fragrant components. The inten-
sity series that consists of these pervasive odorants is represented by
comparatively dilute solutions.
54
EVALUATION OF DIESEL EXHAUST ODORS
-------
Composition of A
Component
Percent by weight
n-Butylbenzene
sec- Butylbenz ene
p-Cymene
Heptaldehyde
Nonaldehyde
Salicylaldehyde
Cinnamic aldehyde
alpha -Methylcinnamic
aldehyde
p-Tolyl aldehyde
16.6
16.6
16.6
38.9
9.7
0.4
0.4
0.4
0.4
100.0
Intensity series (diluent: mineral oil)
Odor intensity
Slight
Moderate
Strong
Extreme
Code
A-l
A-2
A-3
A-4
Concentration, A/ (A + diluent)
Exponent
2-16
2-14
2-12
2-10
Fraction
1/65536
1/16384
1/4096
1/1024
Decimal
0. 00001525
0. 0000610
0. 000244
0. 000976
Percent
0. 0015
0. 0061
0. 0244
0. 0976
ppm
15
61
244
976
Preparation
Make up the. stock solution A. The components are highly odorous and
should be handled in the fume hood.
To make A-4, calibrate a pipet or dropper in milligrams per drop of
stock solution A. Pipet a measured quantity of A into a Flask. Add mineral
oil in the ratio 1. 02 grams mineral oil per gram of A. Shake the mixture to
dissolve the components.
To make A-3, mix 1 part of A-4 with 3 parts of diluent.
To make A-2, mix 1 part of A-3 with 3 parts of diluent.
To make A^-l, mix 1 part of A-2 with 3 parts of diluent.
Note: The dilutions can be made on a volume basis without introducing
any significant error.
MASKING (M)
Odor-modifying agents are designed to improve objectionable odors by
admixture with another vapor that will change the malodorous quality (masking
action) and/or reduce the intensity of the malodor (odor-counteracting action.)
Agents of this type differ among manufacturers and are proprietary. When
such agents are used for modification of diesel odor it •will be helpful to have
a general "masking" standard to use as a quality reference. Such a standard
should contain components of the type likely to resemble the composition of
Appendix B
55
-------
common diesel masking agents or to be associated in the experience of
panel members with commercial products that contain mixtures of industrial
essential oils.
Composition of M
Component
Oil of wintergreen
Terpineol
Cedrene
Bornyl acetate
Phellandrene
Precent by •weight
10
20
20
30
20
Intensity series (diluent: mineral oil)
Odor intensity
Slight
Moderate
Strong
Extreme
Code
M-l
.M-2
M-3
M-4
Concentration, M/(M + diluent)
Exponent
2-l6
2-13
2-10
2-7
Fraction
1/65536
1/8192
1/1024
1/128
Decimal
0.00001525
0.000122
0. 000976
0. 00781
Percent
0.0015
0.0122
0.0976
0. 781
ppm
15
122
976
Preparation
Make up the stock solution M. The components are highly odorous and
should be handled in the fume hood.
To make M-4, calibrate a pipet or dropper in milligrams per drop of
stock solution M. Pipet a measured quantity of M into a flask. Add mineral
oil in the ratio 0. 127 grams of mineral oil per milligram of M. Shake the
mixture to dissolve the components.
To make M-3, mix 1 part of M-4 with 7 parts of diluent.
To make M-2, mix 1 part of M-3 with 7 parts of diluent.
To make M-l, mix 1 part of M-2 with 7 parts of diluent.
Note: The dilutions can be made on a volume basis without introducing
any significant error.
DIESEL (D)
As stated earlier, the intensity of diesel exhaust odor may be measur-
ed without regard to the quality of the odor. The intensity reference stand-
ards, therefore, could consist of a dilution scale of any convenient odorant.
It is considered likely, however, that panel members will become proficient
in QI odor profile work more easily if the overall intensity reference stand-
ard is related to the individual odor quality standards.
56
EVALUATION OF DIESEL EXHAUST ODORS
-------
Composition of D
Component
B-4
0-4
A-4
Percent by volume
59.3
37.0
3. 7
100.0
Intensity series (diluent: mineral oil)
Odor intensity
Slight
Moderate
Strong
Extreme
Code
D-l
D-2
D-3
D-4
D-5
D-6
D-7
D-8
D-9
D-10
D-ll
D-12
Concentration D/(D + diluent)
Exponent
2-11
2-10
2-9
2-8
2-6
2-5
2-1
Fraction
1/2048
1/1024
1/512
1/256
1/128
1/64
1/32
1/16
1/8
1/4
1/2
1
Decimal
0. 000488
0. 000976
0.00195
0. 00391
0. 00781
0.0156
0. 0313
0. 0625
0. 125
0.25
0. 5
1
Percent
0. 0488
0.0976
0. 195
0.391
0.781
1. 56
3. 13
6.25
12.5
25.0
50. 0
100. 0
ppm
488
976
Procedure
Make up the stock solution D.
D-12 is identical with stock solution D.
To make D-ll, mix 1 part of D-12 with 1 part of diluent.
To make D-10, mix 1 part of D-ll with 1 part of diluent.
Continue in this manner until D-l is prepared.
Notes: 1. The dilutions can he made on a volume basis without intro-
ducing any significant error.
2. Stock solution B should be shaken before dilutions are made
to bring into suspension any component of the oil of cade that
might have settled down.
SOURCES AND PURITY OF CHEMICALS
All of the chemicals used must be pure enough that any impurities pre-
sent do not make a detectable contribution to odor. For highly odorous
chemicals, such as aldehydes, this requirement normally poses no problem.
A suitable odorless mineral oil is Primol 325 available from Humble
Oil & Refining Company, Hutchinson River Parkway, Pelham, New York
Appendix B
57
-------
The benzyl berizoate should not have any of the cherry or almond odor
that may be associated with the presence of some benzaldehyde impurity.
Satisfactory grades are available from Mallinkrodt Chemical and from the
Matheson Company.
The octyl benzene must not emit any of the sulfur odor that is some-
times associated with inferior samples. A satisfactory grade may be pur-
chased from the Humphrey Chemical Company, in North Haven, Connecticut.
Acetylenic compounds are available from Farchan Research Laborator-
ies, Willoughby, Ohio.
CONTAINERS
Twenty-five milliliters of each odorant reference solution is placed in a
labeled 4-oz. polyethylene squeeze bottle, fitted with a screw cap having a
conical polyethylene liner.
The entire kit comprises 32 bottles (12 of the intensity series and 4
each of the 5 components of the QI profile, including the masking standards).
These bottles should be arranged in a metal rack in a metal box that has
an internal activated-carbon panel to keep the atmosphere in the box odor-
free. Porous materials of construction like wood or cardboard should be
avoided because they absorb and retain odor. The activated carbon should
be granular material of the type commonly used for air purification.
58 EVALUATION OF DIESEL EXHAUST ODORS
-------
APPENDIX C. STATISTICAL DERIVATIONS
STATISTICAL DERIVATION OF TRIANGLE TEST SELECTIONS
A number of candidates are given m triangle tests each. On the
basis of the number of correct answers, we wish to choose the "best" candi-
dates. It is assumed (1) that the candidates act independently of each other,
and (2) that the test has been demonstrated to them beforehand so that they
are not learning during the test sequence. Each test is therefore indepen-
dent of the other tests taken by the same candidate.
If a candidate answers purely by guesswork, the chance of getting
a correct answer on any single test is 1/3. The probability of getting x
correct answers out of m is expressed by the probabilities of the binomial
distribution:
where: Px = probability that a candidate would give x correct
answers .
m = number of triangle tests given.
x = number of correct answers.
The classical treatment of triangle tests uses a chi-square statistic to
assess the candidate's sensory acuity for odor and flavor. However, the
significance level computed by chi-square is an approximation to the true
significance level. When the number of triangle tests is greater than 15,
the approximation is quite satisfactory. In our screening procedures, we
recommend only four or five triangle tests per candidate. Therefore, the
chi-square is not valid, and our significance levels are based upon the
following derivation.
Suppose we have two candidates, one who has x correct answers and
one who has y, and we want to know whether the difference between the two
is significant. We therefore want the distribution of |x-y|. According to
distribution theory in probability, x and y are identically, independently
distributed binomial random variables, where:
Px = -7-^- -- .7 (l/3)x (2/3)m-x and
x! (m-x) I
pv = - VT
y y! (m-y) !
The expected value (E expectation) of any function f(x) is defined as:
m
x = 0
By convention, we choose f (x) tx, •where t is a dummy variable.
59
-------
Then, m
E (tx) = V ,x
x = 0
m
"px and
E tyPy
x = 0 /
These are called the
"generating functions"
of px and py.
To find p[x-y|> the probability distribution of |x-y|, the expected value
of t(x~y) is computed.
E (tx-y) = E (tx) E (t-y)
E (tx)
tx
E txpx - 2^ " v x! (m-x) !
x = 0 x = 0
x /?./7\m-x
) (1/3P (2/3)
m
ml
E v x! (m-x)!'
x = 0
I t\x I \rn-x
I ;-1/ \ -> /
3 3
(by binominal expansion)
and similarly
(t-y)
v rn'
y = 0 Y' m Y '
,1 2 im
i r — )y (— )m-y
' v 3t ' v 3 '
(by binominal expansion)
Then,
E
To find the distribution it is necessary to expand the generating function,
gather like terms, and consider two competing candidates, A and B, -whose
respective scores are x and y. The probability that, by chance alone, the
absolute difference between the two scores is z (or, that |x-y| = z), is the
sum of the coefficients of the terms in t and -t. If two candidates take m
triangle tests each, the maximum difference between scores is +m.
Let m 0. (There are no tests. )
Then,
= t°
and PQ = 1. (The probability is 1 that the score is tied. )
60
EVALUATION OF DIESEL EXHAUST ODORS
-------
Let m - 1. (One test is carried out.)
Then,
"Y> = - — —
9 9 9t
2-1 5,0 21
Therefore,
9 '
for, the chances are — that A will win over B, — that B will win over A,
5 '
and — that A and B will tie.
Let m = 2
Then,
E(tx-Y) =
81
81
81
?T'2
Therefore,
P0 = 33/81
P! = 20/81 + 20/81 40/81
P2 = 4/81 + 4/81 = 8/81
Calculations are similar for m = 4 and m = 5.
In a series of triangle tests taken by candidates A and B, the distribu-
tion of the differences in correct scores |x-y| tells us •what the probability
is that any given difference in scores obtained by A and B is due to chance
alone. These distributions appear in Table C-l.
STATISTICAL DERIVATION OF INTENSITY RATING TEST SELECTIONS
As sumptions
When two candidates replace a sample at or near the correct position,
they are candidates who will be considered for favorable action and it is
important to make a valid distinction between them.
Candidates who replace the samples very far from the correct position
are not likely to be chosen and we do not care much about distinctions between
"very bad" and "extremely bad. "
For example: correct position - 3
Candidate A Replaces sample in position 3
Candidate B Replaces sample in position 4
Candidate C Replaces sample in position 17
Candidate D Replaces sample in position 18
Appendix C
61
-------
Table C-l. TRIANGLE TESTS: PROBABILITY DISTRIBUTION AND CUMULATIVE
DISTRIBUTION FUNCTION OF ABSOLUTE DIFFERENCE IN SCORES
Number of triangle
tests
m
0
1
2
3
4
5
Absolute difference
between scores
|x-y|
0
0
1
0
1
2
0
1
2
3
0
1
2
3
4
0
1
2
3
4
5
P|x-y|
1.0000
0.5556
0.4444
0.4074
0.4938
0.0988
0.3361
0.4774
0.1646
0.0219
0.2928
0.4512
0.2024
0.0488
0.0049
0.2629
0.4257
0.2234
0.0732
0.0135
0.0011
Cumulative distribution function
(Probability that a difference in
scores of at least |x-y| would be
obtained by chance alone.)
U-y
^2 PI
i = o
1.0000
1.0000
0.4444
1.0000
0.5926
0.0988
1.0000
0.6639
0.1865
0.0219
1.0000
0.7072
0.2561
0.0537
0.0049
1.0000
0.7371
0.3113
0.0878
0.0146
0.0011
Then A's score is slightly better than B's and we want to know how reliable
this difference is. Candidates C and D are both very poor and we do not really
care whether one is -worse than the other.
Scoring
Position of replaced
bottle
Correct position
Correct position +_ 1
Correct position +_ 2
Correct position +_ 3
Correct position +_ 4 or more
Score (the higher number
is the worse score)
0
1
4
9
16
Each candidate must be given the same program; they must not communicate
with each other.
Test number
Correct position of sample
1
2
3
4
12
8
16
3
62
EVALUATION OF DIESJEL EXHAUST ODORS
-------
Distribution of Possible Scores:
Correct
answer
3
Possible
scores
0
1
4
9
16
Probability of
each score
1/20
2/20
2/20
1/20
14/20
Explanation:
Bottle
Position:
Deviation from
correct location:
Score
-2 -1 0 +1 +2 +3 +4 +5 +6
4 1 0 1 4 9 16 16 16
etc.
20
16
Number of ways of getting a score of 0 is 1 out of 20.
Number of ways of getting a score of 1 is 2 out of 20.
Number of ways of getting a score of 4 is 2 out of 20.
Number of ways of getting a score of 9 is 1 out of 20.
Number of ways of getting a score of 16 is 14 out of 20.
Correct answer
8, 12, or 16
Possible
scores
0
1
4
9
16
Probability of
each score
1/20
2/20
2/20
2/20
13/20
Therefore, in the sum of the four tests, the lowest possible score (the best)
is 0; the highest possible score (the worst) is 64.
Table C-2 gives the probability distribution of scores in the odor inten-
sity tests .
STATISTICAL DERIVATION OF MULTICOMPONENT ODOR IDENTIFICATION
TEST SELECTIONS
A candidate is given eight known odor standards, A, B, C, D . . . H.
He is then asked to identify, in three successive tests, the components of a
two-component mixture, a three -component mixture, and a four -component
mixture.
Two-Component Mixture
Let the two components be A and B. The chance of guessing both
correctly is:
1 1 _ 1
T X 7 56
chance of guessing
A first
and
guessing
B next
Appendix C
63
-------
Table C-2. PROBABILITY DISTRIBUTION FUNCTION AND CUMULATIVE DISTRIBUTION
FUNCTION OF DIFFERENCE IN SCORES OF TWO SUBJECTS ON ODOR-INTENSITY TEST
Difference
in score
0
1
2
3-
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
Probability of
difference
0.0899
0.0289
0.0140
0.0467
0.0372
0.0374
0.0124
0.0754
0.0465
0.0274
0.0165
0.0191
0.0834
0.0123
0.0175
0.0822
0.0466
0.0119
0.0111
0.0333
0.0176
0.0096
0.0259
0.0214
0.0222
0.0051
0.0063
0.0292
0.0156
0.0046
0.0164
0.0182
0.0062
0.0021
0.0073
0.0055
0.0033
0.0037
0.0040
0.0053
0.0024
0.0007
0.0038
0.0038
0.0012
0.0015
0.0023
0.0013
0.0004
0.0005
0.0005
0.0004
0.0003
0.0002
0.0004
0.0003
0.0001
0.0002
0.0002
0.0001
0.0001
0.0001
0.0001
0.0000
0.0000
Cumulative
probability of
difference
1.0000
0.9101
0.8812
0.8672
0.8205
0.7883
0.7459
0.7335
0.6581
0.6116
0.5842
0.5677
0.5486
0.4652
0.4529
0.4354
0.3532
0.3066
0.2947
0.2836
0.2502
0.2326
0.2230
0.1971
0.1757
0.1534
0.1483
0.1423
0.1131
0.0975
0.0929
0.0765
0.0583
0.0521
0.0500
0.0427
0.0372
0.0339
0.0302
0.0262
0.0209
0.0185
0.0178
0.0140
0.0102
0.0090
0.0075
0.0052
0.0039
0.0035
0.0030
0.0025
0.0021
0.0018
0.0016
0.0012
0.0009
0.0008
0.0006
0.0004
0.0003
0.0002
0.0001
0.0000
0.0000
Nominal
significance
level
0.30
0.25
0.20
0.15
0.10
0.05
0.01
0.001
0.0001
64
EVALUATION OF DIESEL EXHAUST ODORS
-------
plus
1
8
chance of
guessing B first
and
guessing
A next
Total
The chance of guessing only one correctly is:
1
plus
chance of guessing guessing C,
A first D, E, F, G,
or H next
6 1
8
chance of guessing
C, D, E, F, G, or
H first
guessing
A next
Similarly, chance of guessing only B is:
6
2 x
56
Total
The chance of getting none correct is:
2
"56"
6
56
6
TT
12
24
"5T
chance of guessing
C, D, E/F, G, or
H first
_5_ 30
7 "56
chance of guessing
C, D, E, F, G, or
H next
Grand total ——-
56
Three-component mixture and four-component mixture are treated analo-
gously. The results are:
Number of
components
in mixture
2
3
4
Number of
components
identified correctly
0
1
2
0
1
2
3
0
1
2
3
4
Probability of identifying
this number of components
by chance
30/56
24/56
2/56
60/336
180/336
90/336
6/336
24/1680
384/1680
864/1680
384/1680
24/1680
Appendix C
65
-------
Scoring
It is assumed that it is much better to get a score of j correct than a
score of j-1. Therefore, a quadratic scoring system is used. In other
•words, there is no great trick in getting 1 out of 4 right, even if chance is
not involved, but a candidate who gets 4 out of 4 right is doing very much
better, and we therefore credit him with being not (4/1) times better but
(4/1)2 times better. Therefore, the score is taken to be (the number of
correct answers) .
The test program consists of giving each candidate three tests (the test
components are the same for each candidate).
Test 1: two-component mixture
Test 2; three-component mixture
Test 3: four-component mixture
Candidates are selected on the following basis:
Let the score of candidate 1 equal X and the score of candidate 2 equal
Y. Then the difference between scores is X - Y.
When is this difference significant? What is the basis for choice
between candidates?
The lowest possible total score isO+ 0+ 0= 0.
The highest possible total score is 2^ + 3^ + 42 = 29; hence the maxi-
mum difference (X Y) = 29 - 0 = 29.
Table C-3 lists all of the probabilities of difference between scores
from 0 to 29.
66 EVALUATION OF DIESEL EXHAUST ODORS
-------
Table C-£ DISTRIBUTION FUNCTION AND CUMULATIVE DISTRIBUTION
FUNCTION FOR ABSOLUTE DIFFERENCE IN SCORES OF ANY TWO
SUBJECTS ON MULTICOMPONENT ODOR IDENTIFICATION TEST
Difference
in score
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Probability of
difference
0.1132
0.1518
0.1025
0,1397
0.1278
0.0938
0.0647
0.0511
0.0570
0.0370
0.0139
0.0148
0.0141
0.0062
0.0032
0.0037
0.0027
0.0011
0.0008
0.0006
0.0002
0.0001
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Cumulative
probability of
difference
1.0000
0.8868
0.7350
0.6325
0.4928
0.3650
0.2712
0.2062
0.1554
0.0984
0.0614
0.0475
0.0327
0.0186
0.0124
0.0092
0.0055
0.0028
0.0017
0.0009
0.0003
0.0001
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Nominal
significance
level
0.25
0.20
0.15
0-10
0-05
0.01
0-001
0.0001
Appendix C
67
------- |