EPA-600/2-76-203
July 1976
Environmental Protection Technology Series
FLOW AND GAS SAMPLING MANUAL
Environmental
Office of Research anrf
U.S. Environmental Protection
Research Triangle Park, North Carolina 27711
-------
EPA-600/2-76-203
July 1976
FLOW
AND GAS
SAMPLING MANUAL
by
E.F. Brooks andR.L. Williams
TRW Systems Group
One Space Park
Redondo Beach, CA 90278
Contract No. 68-02-1412, Task 13
ROAPNo. 21ACX-AE
Program Element No. 1AB013
EPA Project Officer: W. B. Kuykendal
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, NC 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
-------
FOREWORD
This manual describes techniques used to measure volumetric flow and
to extract representative gas samples from process streams. It was
prepared under Task 13 of EPA Contract No. 68-02-1412, "Quick Reaction
Technical Services in Air Pollution Sampling Acquisition and Analysis,
Process Instrumentation, Process Research and Process Evaluation."
This work was conducted under the technical direction of Mr. W. B. Kuykendal,
EPA, Task Order Manager, and the administrative direction of Dr. L. D. Johnson,
Environmental Research Center, Research Triangle Park, North Carolina. The
Advanced Instrumentation Department and Applied Chemistry Department, Applied
Technology Division, TRW Systems and Energy, Redondo Beach, California were
responsible for the work performed on this program. Dr. E. A. Burns, Manager,
Applied Chemistry Department, was Program Manager, and the Task Order Manager
was E. F. Brooks. Computer analyses and simulations were performed by
R. L. Williams.
tit
-------
TABLE OF CONTENTS
Page
Foreword i i i
List of Figures v
List of Tables vii
Sections
1. Introduction 1
2. Derivation of Equations 4
2.1 Conservation of Mass-Mathematical Representation 4
2.2 Conservation of Mass-Engineering Representation 7
3. Error Analysis 14
3.1 System Random Errors 15
3.2 Systematic Errors and Mistakes 25
3.3 Summary 26
4. Sampling Methodology 27
4.1 Manual Sampling 27
4.2 Continuous Sampling 41
5. Hardware 55
5.1 Velocity Measurement 55
5.2 Gas Sampling 68
5.3 Prototype Continuous Monitoring System 74
6. Prototype Continuous Monitoring Procedures 80
References 90
Glossary 91
iv
-------
FIGURES
No. Page
1. Flow measurement plane 5
2. Summation of area segment approximation to flow equations 9
3. Normalized measurement error versus number of sampling points 24
for volumetric flow and gas sample traverses
4. Minimum number of traverse points - EPA Method 1 29
5. BSI rectangular duct mapping scheme 29
6. Normalized velocity distribution - Exxon run 1 31
7. Normalized 02 concentration distribution - Exxon run 1 32
8a. Sixteen point array for manual velocity traverses 35
8b. Arrays for manual gas sample traverses 35
9. Locations of sample points for tangential and LogrLinear 38
methods - 16 point traverse
10. Circular duct continuous monitoring locations 46
11. Normalized velocity distribution downstream of an elbow - 48
TRW pilot scale test
12. Point locations for Row Average Method 49,
13. Recommended sample plane locations downstream of a 51i
rectangular elbow
14. Typical plot of normalized row averages 511
15a. Pi tot-static probe 56
15b. S probe 56
16. S probe calibration factor as a function of velocity 57
17. S pi tot probe orientation sensitivity data 59
18. Ramapo Fluid Drag Meter 61
19. Annubar 63
20. Velocity accuracy for Flare Gas Flow Probe 65
-------
FIGURES (Cont'd.)
No. Page
21. Single point velocity accuracy based on differential pressure 69
measurement accuracy versus differential pressure for three
pressure measurement methods
22. Illustration of multiple point gas sampling misconception 71
23. Multiple point gas sampling probes 73
24. Schematic of continuous monitoring system for a combustion 76
process stream
vr
-------
TABLES
No. Page
1. Process stream flow and gas sampling equations 12
2. Nominal uncertainties for measured parameters at a single point 18
3. Nominal uncertainties for process stream measurement systems 26
4. Tangential and Log-Linear sample point locations for 16 point 38
traverse in percent of duct diameter from inside wall to
sample point
5. Comparison of deviations for actual and normal distributions 42
6. Yaw characteristics of pitot-static and S probes 58
7. Estimated continuous measurement system hardware costs 77
8. Estimated continuous measurement system labor requirements 78
vii
-------
1. INTRODUCTION
This manual deals with current technology for the measurement of
total volumetric flow in process stream ducts and for the extraction of
representative gas samples. The purpose of the manual is to discuss
techniques and hardware to be used to obtain optimum measurement accuracies
while minimizing measurement system complexity and labor requirements.
Both manual traverses and continuous measurement systems are considered.
Modifications to current methods are suggested where applicable, along
with the reasoning for the proposed modifications.
The background for the manual was taken from EPA program 68-02-0636,
"Measurement Techniques for Control System Evaluation - Total Gas Flow
Rate," sponsored by the Process Measurements Branch of the Industrial and
Environmental Research Laboratory. Program results are documented in
References 1 and 2. This manual was prepared along with a shorter document,
"Guidelines for Stationary Source Continuous Gas Monitoring Systems,"
Reference 3, and under the "Quick Reaction" program cited in the Foreword.
Program results indicate that composition and flow measurements in large
process stream ducts can be routinely made with accuracies on the order of
5% to 10% on a continuous basis using available hardware and techniques.
This represents a significant improvement over commonly accepted accuracies,
especially for single point sampling, of 20% to 30%. Secondary program
results relative to the accuracy and efficiency of manual measurements have
resulted in several suggestions for improvement of standard methods, such as
optimization of the number of measurement points for flow measurement and
gas sampling.
The manual is organized as follows:
Section 2. Derivation of Equations - To provide a proper framework,
flow and gas sampling equations are derived from the basic principle of
conservation of mass, and the simplifying assumptions used to produce
the final engineering relations are identified.
Section 3. Error Analysis - A standard error analysis is performed on
the derived equations and individual error sources are discussed in order
1
-------
to identify the critical ones. Error sources fall into two broad categories:
single point errors are those due to inaccuracies inherent in the instruments
used; methodology errors are those due to the inadequacy of sampling schemes
to properly account for flow and compositional stratifications and temporal
flow variations. Either category may be responsible for the largest system
errors. The most critical sources are identified in this section and
suggestions as to how to deal with them are given in Sections 4 and 5.
Section 4. Sampling Methodology - Techniques for manual and continuous
sampling are presented. The chosen techniques are designed to optimize
accuracy, operational efficiency and cost. Evidence is presented to sug-
gest that flow proportional gas sampling is not required, and that the
optimum number of sampling points for manual traverses is sixteen. The
Row Average Method, using nominally eight measurement points, is recommended
for continuous monitoring applications, and single point sampling is found
to be generally unacceptable for continuous measurements.
Section 5. Hardware - State of the art velocity measurement instru-
ments are discussed, a need and prototype designs for a multiport continuous
gas sampling probe are presented, and a prototype continuous monitoring
system is shown. The relative merits of the pi tot-static probe and S
probe for point velocity measurement are presented, and a clear preference
for the pitot-static probe is demonstrated so long as an accurate instrument
is used to measure differential pressure.
Section 6. Prototype Continuous Monitoring Procedures - The material
from the previous sections is summarized in the form of general procedures
for installation and calibration of continuous monitoring systems. This
section is a complete unit in itself in that it contains figures, tables,
and equations necessary for system implementation.
There have been major improvements in recent years in process stream
instrumentation, and several of these are documented in the report. To
make full use of these improvements, ,the techniques for flow measurement
in process streams must also be improved. A major purpose of this report
is to indicate where commonly used techniques such as the EPA Federal
Register methods are potentially inaccurate or impractical, and to suggest
modifications to simplify these common procedures and/or make them more
accurate.
-------
Discussion of gas sampling 1s limited to the problem of how to
extract a representative gas sample from a process stream. Sample analysis
techniques for the extracted sample are very well documented in other
sources, such as References 8 and 11. In addition, the growing use of
continuous gas analyzers has resulted in a greater emphasis on how to
deliver a representative sample to the analyzer.
The manual contains no completely new and radical approaches to the
problems of flow and composition measurement, but it does discuss signifi-
cant and possibly controversial changes to existing procedures. These
recommended changes are the end result of extensive testing, both in the
laboratory and in the field. Analysis of the accumulated data resulted in
the proposed modifications — it was never the case that a modification
was proposed first and then data were found to support it. Also, it is
certainly recommended that a broader data base be established for techniques
such as spatial gas sampling in order to identify their limitations.
Environmental monitoring is presently the area of greatest applica-
tion for the technology discussed in this manual. As the accuracy and
availability of continuous monitoring instrumentation increase, it is
expected that continuous monitors will also find much use in the area of
energy conservation, since precise information about process stream con-
ditions can be used to optimize plant operating conditions, thus mini-
mizing fuel and power requirements. The accuracy of a measurement system
is a function both of the instruments used and the methods by which they
are used. This manual is intended to acquaint the reader with some of the
best available hardware for process stream measurements and with methodology
developed specifically to make the best use of the hardware.
-------
2. DERIVATION OF EQUATIONS
In this section, the principle of conservation of mass is used to
derive process stream flow equations in terms of actual measured parameters.
The equations of interest are those for total gas volumetric flowrate,
total gas mass flowrate, species gas mass flowrate, and average gas
species concentration. Simplifying assumptions used to obtain the final
equations are identified. The derived equations serve as a basis for the
error analysis presented in Section 3, and as background for Sections 4
and 5.
2.1 CONSERVATION OF MASS-MATHEMATICAL REPRESENTATION
This manual is concerned with measurements in ducted process streams.
In all cases, measurements are assumed to be taken in a plane normal to
the local duct axis. This will be refered to as the measurement or sample
plane and is illustrated in Figure 1. Using the principle of conservation
of mass as described in Reference 4, we can represent the net mass flux
through the plane as:
m =pti ' n dA (1)
where A
m = total mass flow rate, gm/sec
3
p = local stream density, gm/cm
if = local velocity vector, m/sec
"n = unit vector normal to measurement plane, dimensionless
2
A = area of measurement plane, m
In order to transform this equation into a usable engineering formula,
the following qualification is being applied: the local density, p, is
taken to be that of the gaseous stream constituents only. The qualifica-
tion means that liquid and solid particles entrained in the stream are
not being considered, so that the flowrate, m, is the total gas flowrate.
The qualification is being made since the instruments to be used for this
• •-ft
-------
BUCT
MEASUREMENT
PLANE, A
MEAN FLOW DIRECTION
At any point in the plane, the velocity vector u will be at an angle 0
with respect to the plane's unit normal n, so that the mass flow dm
through the plane at a point with infinitesimal area dA is given by
dm = (pif • n)dA
which is then integrated over the entire plane to give
m = /1 pu • r\ dA
Figure 1. Flow measurement plane
-------
application respond to the gaseous stream constituents and not the liquid
and solid particles. It also permits use of the perfect gas law and
Bernoulli's Incompressible flow relation.
We may now begin the transformation of equation 1 into a form con-
taining measurable parameters. From the definition of the scalar product
of two vectors, we have
u • n = U cos e (2)
where
U = |u| , the magnitude of the velocity vector, m/sec
e = angle between if and ri (see Figure 1)
From Bernoulli's incompressible equation (Reference 4), we have
P0 • PC, + (V2)pU2 (3)
where
P0 = local stream stagnation pressure, torr
p^ = local stream static pressure, torr
so that
— (4)
f *
and
it- n - (cos e) ° - = u (5)
where
u = U cos e = velocity component normal to measurement plane, m/sec
From the perfect gas law (Reference 5), we have
p M
' (6)
-------
where
M = local average molecular weight,
HIV I t
2
R = universal gas constant, 8314.32 —^-^-
mole sec °K
T^ = local static temperature, °K
Substituting into equation 1, we get
. rr /2(p -pJPooM
m = I / (cos 6) I/ —-*—, dA
JJ f Kloo
A
Similarly, for individual gas species flowrate we have
f2(pn-P )P
m. =f
where
L = property relative to gas species i
m.j = mass flow rate of species i, gm/sec
y = local mole fraction of species i, mo °
Mi = molecular weight of species i
also
I
M=Z^iM
1=1
where
I = total number of gas species in the stream
Equations 7 and 8 represent total and individual species gas flowrate
through the measurement plane, using standard measurable parameters.
2.2 CONSERVATION OF MASS-ENGINEERING REPRESENTATION
In some flow systems, the total mass flowrate can be measured directly,
as in the case of a water system which empties into a tank so that the
7
-------
water accumulated over a specified time period can be weighed. The follow
ing approximations apply to point measurements in process streams where
measurements which involve processing the total flow are impractical for
reasons of duct size or geometry. In this situation, the integrations
specified in the above equations are approximated by a summation of point
measurements, as given below for equations 7 and 8:
N
•v
6
> /
n'\
R(U,
n=l
and
2[(p ) -(pj ](pj
"
where
( )n = mean value of the parameter in the area segment AA
2
AAn = area segment of control plane, m
N = total number of area segments
This representation is illustrated in Figure 2.
For an ideal case, the number of area segments, N, would be very large,
and each of the flow parameters would be measured at each sampling point.
In practice, this is usually not found to be feasible, and further simpli-
fying assumptions are made. In standard techniques such as the EPA Federal
Register methods (Reference 6), the simplifications deal with the number of
sampling points, which affects N and AAn, and with the variation of flow
parameters such as (peo)n and M , where simplification is achieved by
assuming a constant value in the measurement plane for the parameter.
Another type of assumption which can be made deals with flow proportion-
ality. The form of equation 11 dictates that the product of the velocity
and species mole fraction be computed at each point and then averaged to
produce m.. . In practice the technique of obtaining a gas sample at a
rate proportional to the local velocity is termed flow proportional gas
sampling, and is analogous to isokinetic sampling for particulate material.
The most common alternate form of gas sampling is spatial gas sampling,
in which samples from different area segments are considered equally, and
-------
DUCT CROSS
SECTION, A
SAMPLE POINT-
PARAMETER TO BE MEASURED:
DESIRED RESULT: X" N
RELATION USED: XA =
n=l
ACCURATE RESULTS ARE OBTAINED WHEN EACH Xn IS REPRESENTATIVE OF THE
AVERAGE VALUE OF X IN AREA SEGMENT AAn> GOOD ACCURACY IS ACHIEVED WHEN
AAn IS SMALL (LARGE N) AND/OR THE SAMPLE POINT LOCATION IN EACH SEGMENT
IS CHOSEN TO GIVE THE CORRECT VALUE OF Xn.
Figure 2. Summation of area segment approximation
to flow equations
-------
the average concentration is the arithmetic mean of the individual concen-
trations without regard for velocity effects. There are significant
simplifications which can be made to measurement systems if spatial gas
sampling is acceptable from an accuracy standpoint, so the assumption is
worth considering.
To represent spatial and flow proportional average concentrations or
mole fractions, we must first consider the concept of velocity at standard
conditions, given as
us = £- u (12)
where
( ) = value of parameter at standard conditions, defined to
be Tm = T = 293.16°K and p = pc = 760 torr
s °° s
Mathematically, flow proportional and spatial average mole fractions may
then be defined as follows:
N
y, -i^ o^H- (13)
TP N u$
£~^4 s n
n=l
and N
n=l
where
y . = flow proportional average mole fraction, moles of i
TP moTe
y"i = spatial average mole fraction, n
(~) = average (mean) value
The average velocity at standard conditions is also used to calculate total
volumetric flowrate, from
10
-------
= v
where
tf = total volumetric flowrate at standard conditions
It is often desirable to calculate total volumetric flow relative to the
mean actual pressure and temperature rather than at standard conditions,
which is of course perfectly acceptable. Whenever volumetric flowrate of
a gas is given, the pressure and temperature on which the calculation is
based must be specified for the term to have meaning. Standard temperature
and pressure are used throughout this manual.
A group of expressions for m, m., and V based on various assumptions
is given in Table 1. The spatial average mole fraction, u.,^is calculated
from equation 14, and the flow proportional average mole fraction,'V- »
FP
can be calculated from equation 13 or from
m.M
y, = — (16)
TP mMi
The remainder of the manual deals with the evaluation of these equations
from the standpoint of accuracy, methodology, and hardware.
11
-------
Table 1. PROCESS STREAM FLOW MEASUREMENT AND GAS SAMPLING EQUATIONS
Equation
Number
17a,b,c
18a,bl,c
18b2
19a,b,c
20al,bl,cl
20a2,b2,c2
Assumptions
Hono
Constant ?„ , uncoupling of
average molecular weight from
velocity
1
i
'• Same, plus uncoupling of mole
. fractions from velocity
\
i
' e » o, constant ?„ and H,
i uncoupling of temperature from
) velocity
i
1 Constant Pro , uncoupling of
\ average molecular weight from
| velocity, averaging Mn before
taking square root; uncoupling
of mole fraction from velocity
in 20bl
Same, plus assumes capability
of Annubar to sense representa-
tive velocity
Applications.
Rigorous sampling - each param-
eter measured at each sampling
point
TRW field test reference tra-
verse equations
EPA Federal Register methods
TRW recommended equations for
continuous monitoring: al ,bl
and cl assume point velocity
sensor array, a2,b2 and c2
assume use of Annubar as
velocity sensor. All equations
assume a point gas sampling
array. K = calibration factor.
Ideally K % 1.
Total Gas Mass Flowrate, m, gm/sec
A v« (PM . ) /2(iPn)(P»'aMn
"VTtt "V ITJ"
1 N 1 N /
Jzy-lT £V^ £
-------
Table 1 (Cont). PROCESS STREAM FLOW MEASUREMENT AND GAS SAMPLING EQUATIONS
Relation for Parameter to Be Calculated
Species Gas Mass Flowrate, m.,, gm of i/sec
3
Total Volumetric Flowrate, V ,m /sec
AM,
N
E
n=l
Mi)n(cos en;
n=l
N IN
AM
' N
E
n=l
n=l
AT
AM. \
N
n=l
AT,
n=l
N
c
n=l
AM,
-V2,
8
E n
. n=l
n=l
-1/2
n=l
5'
n=l
AT .8'
8
|n=l
,1/2
13
-------
3. ERROR ANALYSIS
Three types of errors will be considered in the following discussion:
random errors, systematic errors, and mistakes. Working definitions for
each type of error are as follows, based on discussion in Reference 7:
t Mistake - an incorrect action, such as misplacing a decimal
point when writing down data.
• Systematic error - a reproducible error such as a shift in
output of an electronic device due to ambient temperature
changes.
t Random error - a nonreproducible error which may usually
be described by a normal error distribution; the error
associated with an accurately calibrated instrument is
a measure of the instrument's random error.
It is not always clear which category a particular error belongs in. If
an instrument operator makes an error in reading a dial because of parallax,
it may be a mistake. If he continually makes the same error, say by
always having his head in the same position and always looking through
his right eye, the error may become systematic. If he moves his head
around randomly and takes readings with either and/or both eyes, the
error may be termed random, for purposes of this report, it is being
assumed that mistakes are due to test personnel carelessness, systematic
errors are due to physical effects, either known or unknown, and random
errors are unavoidable errors which are due to the limitations of the
systems being used.
Since mistakes and systematic errors can be either eliminated or
reduced to a minimum through the application of proper test procedures,
estimates of the random error associated with a particular system give
the best indication of achievable system accuracy. A system error analysis
performed as a preliminary task prior to installation of a continuous moni-
toring system can help to assure optimum system accuracy by pointing out
14
-------
the largest error sources, which then allows for proper allocation of
money and manpower to the most critical system areas.
The following example is presented to illustrate error terminology
to be used in the remainder of the manual. Consider a typical pressure
measurement where the true and measured values are:
True pressure: 500 Torr
Measured pressure: 600 +_ 6 Torr
The systematic error for the measurement would be 100 Torr, or 20% of the
true value. The random error, which is identical to the uncertainty of
the measurement, would be 6 Torr, or 1% of the measured value. If we now
calibrate the pressure instrument to eliminate the systematic error, we
will have
Corrected pressure: 500 +_ 5 Torr.
The systematic error is now zero, which it should be for a properly
calibrated instrument, while the random error, or uncertainty, remains
at +_ 1% of the reading. We may now say that the pressure measurement
device has an accuracy of +_ 1%, which means that the random error, or
uncertainty, in the measurement is +_ 1%. In the remainder of the manual,
the terms "accuracy," "random error," and "uncertainty" are used inter-
changeably, while "systematic error" and "mistake" are not used inter-
changeably with any other terms. Following is a random error analysis of
the equations derived in Section 2. It in turn is followed by a brief
discussion of systematic errors and mistakes.
3.1 SYSTEM RANDOM ERRORS
3.1.1 Single Point Measurement Accuracies
The following technique is taken from Reference 7, "The Analysis of
Physical Measurements," Chapter 11. Assume that it is desired to calcu-
late parameter G, which is a known function of variables H,, H2, . . . H ,
given by
G = f(H1§ H2 Hr) (21)
15
-------
where f denotes the functional relationship. Define the error in the
measurement of variable Hr as er< The standard deviation of the measure-
ment of M is then given by
N
° - E " (22>
where
ar = standard deviation of Hr
N = number of measurements
By derivation, the standard deviation of G is then given as
,2
Og = I TiT~ <3l I +
/9f \2 + +/af \
V'V2/ " VVV
For purposes of the present discussion, it is most convenient to
analyze equations 17-20 in terms of single point measurement accuracies,
and then to make a determination of the effects of number and location of
sampling points. Thus the following single point equations will be con-
sidered first:
V2(Ap)p M
5r (24)
RT
00
. = AM.y.(cos 0) \1Mr-^ (25)
VA^(cose) , -^
/
V
T ZUPMPR
(27)
These equations, when subjected to the above analysis, result in the
following equations in the form of equation 23:
16
-------
m _ s _ /+_M2fi\ 2
---
(28)
This means that the uncertainty is the same for rii as for V . For the
other parameters, we have
. 2 , 2
°m 2 a
m. m
2
2
2
The standard deviation, a, is a measure of the uncertainty in the accuracy
of the associated parameter. For the case of a normal error distribution,
there is a 68.3% probability that the error associated with a single
measurement will be less than +^a, and a 95.5% probability that the error
will be less than +2o. The +2a band is normally used for most engineering
applications, so that if an instrument is said to have an accuracy of +2%,
it means that 2a = 2% or a = 1%. Equations 28 and '29 are given in a form
which allows for convenient discussion in terms of accuracy expressed as
percent of reading.
The e term in Equation 28 may be confusing at first glance, since
2
tan e becomes infinite at e = 90°. The explanation is as follows: as e
O
approaches 90°, m approaches zero and tan 6 approaches infinity, which
makes both sides of the equation approach infinity. However, the problem
*2 • 2
disappears if we multiply through by m (or V for that form) and then
17
-------
substitute for m (or V$) from equation 24 (or 26), as all terms become
finite. The special case of e = 90° 1s discussed in more detail 1n
Section 5, along with a discussion of alignment errors in general. For
the purposes of this section, it is adequate to note that in general for
a well designed probe, empirical data show that
(31)
|(tan e)oe| < .025
The following may be concluded from the form of equations 28 and 29:
• The uncertainty in the determination of total gas mass flow
is the same as for total gas volumetric flow.
• The uncertainty in the measurement of species mass flow rate
cannot be less than that for total mass flow rate.
• A given percentage uncertainty in cross-sectional area
results in a larger system uncertainty than would the
same percentage uncertainty in pressure, temperature, or
average molecular weight.
The first two are self explanatory; the third is mentioned because in
many cases, very little attention is devoted to obtaining a good measure-
ment of cross-sectional area.
Nominal achievable accuracies for each of the parameters in equations
28 and 29 are given in Table 2:
Table 2. NOMINAL UNCERTAINTIES FOR
MEASURED PARAMETERS AT A SINGLE POINT
Parameter
P»
Ap
T
CO
M
A
yi
Achievable Uncertainty, Percent
o
1
2
.5
1
1
1
2a
2
4
1
2
2
2
18
-------
Substituting these values into equations 28 and 29 gives, using .025 for
(tane)a., we obtain
o
!to = _L=3.o% (32)
m Vs
and
= 3.1% . (33)
which corresponds to 2a uncertainties on the order of +6% for each of the
three parameters.
The nominal accuracies given in Table 2 are discussed further in
Section 5. The key results of this section are equations 28 and 29,
which show the relationships among the various parameters involved in terms
of accuracy. Also, the nominal achievable single point accuracy of +6%
becomes a reference point to which other error sources can be .compared.
Following is a discussion of multiple point sampling errors.
3.1.2 Multiple Point Measurement Accuracies
Single point measurements can be represented by exact equations,
which makes them directly amenable to a standard error analysis as was
performed above. Multiple point measurement equations are not exact
(e.g. a summation is used instead of an integration) and also involve
simplifying assumptions. For present purposes, we can represent the
total system random error as follows:
2 2, 2. 2. 2 ,,.,
aSE = aSPE + aAE + °ME + °TE (34'
where
OSE = total system uncertainty, percent
°SPE = Sin9^e P°int measurement uncertainty, percent
19
-------
a.E = assumption uncertainty, percent
OME = mapping uncertainty, percent
o-E = temporal uncertainty, percent
The single point measurement uncertainty was treated above, resulting in
°SPE % ^ ^or a norni'na^ achievable case. Assumption uncertainties are
due to mathematical simplifications such the assumption of constant static
pressure and average molecular weight in the measurement plane. Mapping
uncertainties deal with the number and location of measurement points.
Temporal uncertainties occur when there is a significant time interval
between measurements which could allow conditions to change in the
measurement plane. Each of the latter three error sources is considered
below.
The following discussion is based on empirical data obtained from
several coal fired power plants, and may be expected to apply in general
to process streams with the following characteristics:
• Cross-sectional shape: round or rectangular
t Stream static pressure: local atmospheric pressure +10%
• Bulk composition: air, combustion products and water vapor
• Flow velocity: < 30m/sec.
3.1.2.1 Evaluation of Assumptions
This section deals with the equations in Table 1. In that table,
equations 17 are presented for reference as the most rigorous form which
involves no simplifying assumptions. Equations 18-20 involve various
assumptions concerning stratification of and interaction among the param-
eters. The following evaluation makes use of the data base from References
1 and 2 along with a general knowledge of process streams to determine the
effects of the assumptions on system"accuracy.
3.1.2.1 Stratification/Interaction Considerations
Static pressure, p^: of the variables in equation 17, static pressure
will vary the least across the sampling plane. The static pressure
20
-------
variation will rarely be greater than +0.2% of the mean value unless there
is a large leak to or from the atmosphere at the sample plane location.
Thus the assumption of constant pressure in equations 18-20 helps to sim-
plify instrumentation and data reduction without compromising system
accuracy.
Static Temperature, T^: Temperature stratification itself has not
been examined as much as have velocity and compositional stratification.
However, it can be intuitively recommended that temperature be measured
at every point where velocity is measured, and the data reduction per-
formed in the "coupled" sense of equation 17 rather than the "uncoupled"
manner of equation 19. Since the stream velocity must go to zero at the
wall of the duct and most processes take place at higher than ambient
temperature, it is to be expected that both velocity and temperature
will be at a minimum near the wall, and at a maximum away from the wall.
Since both parameters vary in the same manner, they may be said to be
coupled, and so should be treated that way in the data reduction scheme.
In a practical sense, temperature is perhaps the easiest parameter to
measure, so taking simultaneous velocity (AP) and temperature measurements
does not present a problem. Since there is no significant "cost" involved
in handling the temperature data in the rigorous form, the temperature
should be handled as in equation 17 and 18 rather than in the uncoupled
manner of equation 19.
Average Molecular Weight, M: In combustion streams, the predominant
gaseous components are N2> C02 and CO, 02> and H20. EPA Federal Register
Methods 2, 3 and 4 rely upon the validity of this assumption, and the
assumption is supported by data in Reference 8 for fossil fuel power plants,
municipal incinerators, and a number of industrial processes. It is also
the case that most of these streams involve the use of excess air. In
most of these streams, then, the gas will be composed of 70-80% N2, which
strongly limits the variation which will be observed in M. During a
typical TRW field test, for a 49 point traverse the maximum deviation from
the mean for M was +_!%, although for the same traverse, individual gas
species concentrations differed by as much as +25% from the mean concen-
tration. Since the available data, as discussed below, indicate that it
21
-------
is acceptable to decouple the measurement of local concentration (mole
fraction) from velocity, it would certainly be appropriate then to
decouple average molecular weight from velocity. In addition, the low
observed stratification of M lends support to the single point determination
of M 1n equation 19,
Mole Fraction (concentration), y^: In Reference 2, twenty six manual
traverses were analyzed to determine the extent of coupling between con-
centration and velocity. Fourteen of the traverses were conducted by TRW
personnel, eight were obtained from Exxon Research, and four from a report
by Maiden Research (Reference 9). The average species concentrations were
calculated using the coupled Cflow proporttonal) method of equation 19bl
and using the uncoupled method of equation 19b2. For the 26 runs, the
average difference between the two calculations was .83%, and the worst
case difference was 3.7%, despite the fact that the average gas stratifi-
cation level (discussed and defined in Section 4) was +15%. This led to
the conclusion that spatial gas sampling is acceptable, and will not
result in a significant increase in system random error. This conclusion
has very strong hardware and methodology implications for continuous
monitoring systems, as discussed in Sections 4 and 5.
3.1.2.2 Mapping Errors
This section applies to manual traverses. Number and location of
sampling points for continuous monitoring are discussed at length in
Section 4. As a general statement, it can be said that the number of
sampling points should be maximized to obtain the best accuracy (minimum
OME). This statement must be tempered by the labor costs involved in
taking data at a large number of points, and by the fact that the traverse
should be performed over a short period of time to minimize the effect of
temporal flow variations. A more practical approach is to select the
minimum number of points which can be used without sacrificing accuracy.
A computer analysis has been performed on the 26 runs mentioned above
to determine the effect of varying the number of sampling points. All
runs were performed in rectangular ducts. In addition, eighteen velocity
traverses were also considered. For each of the traverses, the test data
22
-------
were input to a program which could compute local velocity or concentra-
tion at any point using curve fitting techniques. Average flowrate or
concentration was then computed for n x n centroid of equal area arrays
where l
-------
CO
1,0
.8
MEASUREMENT ERROR NORMALIZED WITH
RESPECT TO SINGLE POINT ERROR
o
FOR EACH TRAVERSE, MEASUREMENT ERRORS OBTAINED
FOR THE SPECIFIED NUMBER OF TRAVERSE POINTS WERE
NORMALIZED WITH RESPECT TO THE SINGLE POINT ERROR.
AVERAGE
NORMALIZED
ERROR
GAS SAMPLE TRAVERSES (26)
VELOCITY TRAVERSES 08)
0
10
15 20 25 30
NUMBER OF TRAVERSE POINTS
35
40
45
50
Figure 3. Normalized measurement error versus number of sampling
points for volumetric flow and gas sample traverses
-------
response so that a^ will be negligible. The problem is that the cost
of a continuous system which used as many measurement points as a manual
traverse would be prohibitive. Use of as few as one measurement point
can result in very large mapping uncertainties, aME» which often become
the largest contribution to total system error. The recommended relations
given in Table 1 for continuous monitoring systems (Equation 20), are
based on the use of a single gas sampling probe containing eight ports,
and as few as one velocity sensor. These are described in Section 5.
Such a system has been successfully demonstrated in the field as described
in References 1 and 2. For manual measurements, it is certainly reasonable
to use enough measurement points so that the mapping uncertainty OME is
small compared to the single point uncertainty OSPE. For continuous
systems, keeping a..r about the same as aspc represents a reasonable
tradeoff between accuracy and cost.
3.2 SYSTEMATIC ERRORS AND MISTAKES
Systematic errors are most likely to occur as instrument calibration
shifts, inaccurate calibration gases, and other quality assurance related
phenomena. Like mistakes, they must be minimized through precise, correct
test procedures and proper training of personnel. One of the most diffi-
cult to control systematic error sources is the calculation of the duct
cross sectional area, A. As shown in equation 28, uncertainty in the
cross sectional area has a relatively strong effect on system accuracy,
but ordinarily less effort is devoted to this parameter than to any other.
This is usually due to the inability of the test crew to measure A
directly, so that it must be determined from blueprints. It is definitely
recommended that the cross-sectional area of a designated test section
be physically measured in new sources after construction is completed,
and estimates made concerning how the area may change during actual
operation due to temperature and pressure effects. The same procedure
should be performed if possible at existing sites during a non-operating
period, which would also allow assessment of the effects of scale build
up on the walls. It is not difficult to imagine the actual cross-sectional
area of a duct differing by 5% from the value indicated by a blueprint.
Such an error would have a very adverse effect on the accuracy of emission
measurements taken at that location.
25
-------
Discussion of specific technique details which tend to result in
systematic errors or mistakes when not performed properly is given in
Section 4.
3.3 SUMMARY
The error analysis results are summarized in terms of achievable 2a
uncertainties (random errors) in Table 3.
Table 3. NOMINAL UNCERTAINTIES FOR PROCESS STREAM
MEASUREMENT SYSTEMS
Parameter
• . ••
m, m.t Vs
^1
Types of Measurement
Manual traverse
Continuous
Manual traverse
Continuous
Uncertainty, Percent
20SE
7
9
4
5
2aSPE 2aAE 2aME 2aTE
6232
6 2 60
2222
2240
The above figures apply in general to process stream measurement systems
utilizing good quality available hardware and proper procedures. In order
to get an idea of the achievable accuracy for a proposed measurement
system, equation 28 (or an appropriate variation thereof which is correct
for the proposed instruments) should be used to calculate the single point
accuracy, as was done to calculate the above nominal figures. If proper
methodology is applied and care taken to avoid systematic errors, system
accuracies approaching the calculated value should be achievable in the
actual application. The following sections deal with methodology and
instrumentation to achieve optimum system accuracies.
26
-------
4. SAMPLING METHODOLOGY
This section deals with site selection, probe placement, and other
procedural details involved in flow and composition measurement in
stationary source process streams. Since procedures for manual traverses
are well established, most of the section deals with methodology for con-
tinuous monitoring systems. In all cases, the objective is to provide
optimum accuracy at minimum cost. Specific hardware items are discussed
as required; details of hardware selection and evaluation are presented
in Section 5. It is assumed throughout the sectipn that the parameters
• . •
to be determined are total gas mass flow, nv, species mass flow, m., total
• I
volumetric flow, Vg, and/or average mole fraction (concentration),.y"..
4.1 MANUAL SAMPLING
4.1.1 Site Selection
Very often, site selection is beyond the control of the sampling
team — testing must be performed using existing sample ports, which may or
may not be ideally located. The following rules are a guide for site
selection if there is a choice of locations, and are presented in order of
priority:
1. Select a location which allows the greatest length (in
effective duct diameters) of ducting ahead of the sample
plane, allowing for 20% of the straight run to be downstream
of the sample plane.
2. Select a location in a circular duct rather than in a rectang-
ular duct unless there is a strong swirl pattern in the
circular duct.
3. Avoid sampling immediately downstream of areas where sig-
nificant leakage into the duct occurs, unless a purpose of
testing is to determine the extent of the leakage.
These are standard suggestions for site selection, and are all subject to
the common sense considerations of accessability, working conditions, etc.
-------
Similar suggestions are given 1n References 8, 11, and 12. The reasoning
behind these rules follows directly from the error analysis of Section 3.
Rule 1 helps to insure minimum flow angularities (e 1 0). As shown in
equation 28, the error due to flow angularity is a minimum at e = 0.
Since flow angularity presents a difficult instrumentation problem, as is
discussed in Section 5, all efforts should be made to minimize it.
Past experience has shown that for a given number of sampling points,
more accurate velocity data can be taken in a circular duct than in a
rectangular duct. In addition, it is usually easier to take measurements
in a circular duct since fewer sampling ports are involved. The caution
to avoid swirl patterns, often intentionally produced in stacks, is based on
flow angularity considerations.
Areas of leakage should be avoided to minimize both flow angularity
and compositional stratification. It is intuitively obvious that a
representative gas sample can best be obtained when there is minimum
compositional stratification in the sampling plane.
4.1.2 Number and Distribution of Sampling Points
EPA Federal Register Method #1 (Reference 6) calls for the number of
sampling points to be selected as a function of the distance from the
sampling site to the nearest upstream and downstream disturbances
(Figure 4), with the minimum number of points being 12. The most obvious
deficiency of,Method 1 is that it does not state how many points to use if
the sampling site is less than two diameters downstream or less than half
a diameter upstream of a disturbance. For velocity traverses, the British
Standards Institute recommends the method shown in Figure 5 (taken from
Reference 8), where the number of subdivisions is based solely on the cross-
sectional area. The BSI method also requires additional sampling points
in the subdivisions adjacent to the walls. If we take as an example a
3.05 m x 3.05 m (10 ft x 10 ft) duct, the BSI method would require a
minimum of 560 sample points in 400 subdivisions, and would require about
60 sampling ports to gain access to the points.
The objective in a manual traverse is to obtain representative velo-
city and/or gas sampling data in each of the specified area segments, so
that the summation of all data points gives a value representative of the
28
-------
50
40
0.5
Oil—
uj z
CO HH
so
=3 O.
§£20
10
NUMBER OF DUCT DIAMETERS UPSTREAM
1.0 1.5 2.0 2.5
T
I
I
FROM ANY DISTURBANCE
_L
_L
I
I
I
8
10
23456 7
NUMBER OF DUCT DIAMETERS DOWNSTREAM*
Figure 4. Minimum number of traverse points - EPA Method 1
•4-
4-4-4-
4- 4- '-f
4- .+ +
•4-
4
4-
4-
4-
4-
4-
-f
^
4-
4-
+ + +
4- -f +
+ J ^
Diagram shows distribution of points
in area segments at and away from
walls. The minimum number of area
segments in 16, and the maximum area
segment size is .023 m2 (36 in2).
All area segments
are of equal size.
Figure 5. BSI rectangular duct mapping scheme
-------
total flow through the cross section. The probability of obtaining a
representative measurement in a given area segment increases as the var-
iation (stratification) of the parameters to be measured in the area
segment decreases. Typical normalized velocity and concentration profiles
are shown in Figures 6 and 7, where the contours represent percentage
deviations from the average velocity and average concentration. The
computer program used to compute the profiles is described in Reference 2.
The raw data were from 48 point (6 x 8) traverses performed by Exxon
Research and Engineering Company under EPA contract 68-02-1722 at a TVA
coal fired power plant. It is clear from Figures 6 and 7 that the smaller
the area segment, the more uniform the flow pattern is within the segment.
As the number of sample points increases, then, the representativeness of the
measurement in each area segment also increases.
There is, however, a conflicting accuracy consideration: as the
number of data points increases, the time required for the traverse also
increases, which gives use to the problem of temporal flow variations.
When a large number of sampling points is used, data at each point may be
representative for the respective area segment at the time the measurement
was performed while the total accumulated data are in error due to flow
changes during the period of the traverse. The problem with temporal
variations is that they are very difficult to assess in a quantitative
manner, and so the usual procedure is to simply ignore them. It is our
belief that better accuracy can be obtained by taking a relatively small
number of points during a traverse and then repeating the traverse one or
more times to check for consistency of the data than by performing a
single traverse which involves a very large number of points. This belief
led to the analysis which produced Figure 3, which was discussed briefly
in Section 3. The 26 gas sample traverses used are considered representa-
tive, and were obtained by three completely independent testing groups.
The velocity traverses were obtained under representative to worst case
conditions: most of them were obtained at a location approximately 0.4
effective diameter downstream of a large disturbance and 0.1 diameters
upstream from a large disturbance. The analysis performed had two major
results with respect to manual sampling:
30
-------
DUCT DIMENSIONS
8.27m
3.18m
I
•CO,
PROFILES ARE VARIATION IN 10% INCREMENTS FROM MEAN VELOCITY
MEASURED MEAN VELOCITY =6.81 m/sec
Figure 6. Normalized velocity distribution -- Exxon run 1
-------
DUCT DIMENSIONS
3.18m
8.27m
CO
to
PROFILES ARE VARIATION IK 5% INCREMENTS FROM MEM CONCENTRATION
MEASURED MEAN CONCENTRATION = 4.36% (DRY GAS)
figure 7. Normalized O concentration distributton--Exxon run 1
-------
1. No significant increase in accuracy was noted for traverses
involving more than 16 points.
2. For the 26 gas sample traverses and 18 velocity traverses
examined, the average mapping error for p.. was l.U and
for V$ was 1.8%.
The first results mean that a 16 point traverse is acceptable from a
qualitative standpoint because use of a larger number of points does not
represent a notable improvement; the second results meaa that a 16 point
traverse is acceptable from a quantitative standpoint since sampling
errors of that magnitude would not seriously degrade system accuracy.
It is recognized that a sixteem point traverse is always less rigorous
than the BSI method, and usually less rigorous than EPA Method 1. The
16 point traverse recommendation was arrived at after detailed examination
of a sizable body of empirical data from three different sources, and the
analysis is believed to be sound. In addition, the 16 point traverse must
result in smaller temporal errors than traverses involving a larger number
of points. It is recognized that a larger body of data must be examined
before the 16 point traverse can be seriously considered as a standard
technique, and we strongly encourage that such an examination be performed.
A 16 point velocity traverse can be performed in about 20 minutes; a
gas sample traverse would require a somewhat longer time depending upon the
type of instrumentation used. A gas sample traverse using a continuous
gas analyzer or analyzers should take from 40 minutes to an hour. Sample
times on this order should be acceptable for use in a large number of
industrial process streams. Traverses involving more than 16 points are
recommended only when it is desirable to obtain flow and concentration
maps for purposes such as identification of stratification patterns. In
this type of situation, traverses involving 36-64 points are recommended.
Once the number of sampling points has been selected, their location
must be determined. The optimum location for velocity sampling is not
necessarily the same as for gas sampling. Rectangular ducts will be con-
sidered first. There is one known fact about the velocity distribution in
any duct: the velocity must be zero at the wall. In a rectangular duct,
this means that the total velocity variation must be the same between each
pair of walls — from zero at each wall to the maximum velocity, wherever
33.
-------
it occurs. This leads to the recommendation that each side of the duct
should be divided into an equal number of segments for point sampling,
2
producing an n x n sampling matrix containing n sampling points, as illus-
trated in Figure 8a. This arrangement helps to insure minimum velocity
stratification in each area segment. The velocity should be measured in
the center of each segment unless there is good reason to pick a different
location.
For gas sampling in rectangular ducts, the situation is different
because there are no Known boundary conditions for gas concentration. In
this situation, the sample points should be dispersed to maximize the
distance between the points. For a given number of sample points, this
occurs when the area segments are as close as possible to being squares,
as illustrated in Figure 8b.
It is recognized that there is not always complete freedom to select
the shape of the sample array - existing sample ports must often be used
out of necessity, or it may not be practical to use different array shapes
for velocity and gas sampling. In such cases, array shapes should be
selected according to the following criteria:
1. For an n, x n^ array, n, and r\2 should each be greater
than or equal to four for a velocity traverse and not less than
two for a gas sample traverse. The product n-, x n^ should be
greater than or equal to 16.
2. For a velocity traverse, it is preferable to have n, = n2-
For a gas sample traverse, it is preferable to have the
ratio /2^1 ^e as close as possible to one, where ii is the
length of the longer side of the area segment, and ^ is the
length of the shorter side. If a single array is to be used
for both velocity and gas sampling, the array should be
selected according to the velocity traverse criteria.
3. All area segments should be of equal size and shape, and
the sample points should be located as close as possible
to the center of each segment.
For traverses in circular ducts (pipes, stacks), all common techniques
involve measurements along two orthogonal diameters. The sixteen point
34
-------
1
• 1
i.
• !
• l
i
T
• 1
i
i
• I
• i
• !
T
• 1
1
• ! •
_ —— —
• ! ^
• \ •
l
r
• i •
i
MEASUREMENT
POINT
Figure 8a. Sixteen point array for
manual velocity traverses
'I,
i i
i ,
• • ! •
- — \-— \—
. i • .
• i • i •
j i —
i i
• i • i •
i i
i i
1 !
1 ^ »
. , . .-
1 l
+ j _rO.
1 . ' . **
ii
J— I—A
1 . i .
l • i • •
! ! !
/
SAMPLE
POINT
ARRAY
SHAPES
4x4
4x5
3x6
3x7
2x8
CHOOSE ARRAY FOR WHICH
CLOSEST TO UNITY
IS
Figure 8b. Arrays for manual gas
sample traverses
35
-------
traverse Is also recommended for use in circular ducts, although the
analysis which resulted in the sixteen point recommendation involved only
rectangular duct data. Past TRW experience in pilot and full scale
facilities has led to the conclusion that for velocity measurement, a
given number of points distributed in a circular duct will result in con-
sistently better accuracies than the same number of points in a rectangular
duct. This contention is supported by a recommendation in Reference 12 to
add ducting to a rectangular duct to change it to a circular cross section
if optimum accuracy is required. The conclusion is that if a sixteen point
array is adequate in a rectangular duct, it will be at least as adequate
in a circular one.
The tangential, or centroid of equal areas, method specified in EPA
Method 1 is recommended for gas sampling in circular ducts, but not for
velocity measurement. The tangential method assumes that nothing is known
about the distribution of the parameter to be measured, so the sample
points should be located at the centroids of the area segments. This same
philosophy is used above in locating points in rectangular ducts. There is
a very large body of knowledge about velocity distributions in circular
ducts, and sampling methods have been developed on the basis of this
knowledge which demonstrate better accuracy than the tangential method.
The best of these methods appears to be the Log-Linear technique, presented
in Reference 10 and also discussed at some length in References 8 and 12.
The nane Log-Linear is descriptive of the basic assumption used — that
many velocity distributions can be closely approximated by the family of
curves
u = c1 + c2 log (£/D) + c3 (y/D) ^ (35)
where
u = velocity, cm/sec
y = distance from duct wall, cm
D = duct diameter, cm
C1***2'C3 = constants> cm/sec
36
-------
The method involves dividing the duct into a number of equal area segments,
and then differs from the tangential method by placing the sampling points
not at the centroid of each segment, but at locations which give a repre-
sentative velocity for each segment. The method has been tested exten-
sively in TRW facilities, as described in Reference 1, and has been found
very acceptable.
Sixteen point traverses are illustrated for the tangential and Log-
Linear methods in Figure 9, and point locations are given in Table 4.
4.1.3 General Sampling Techniques
Procedures for manual sampling are given in Reference 3. The purpose
of this section is to indicate technical details which can have a signifi-
cant effect on measurement accuracies and testing efficiency. The follow-
ing should be given appropriate attention:
• Measurement of Duct Cross Sectional Area - This was discussed
in some detail in Section 3. In most cases, not all dimen-
sions can be checked, and it is often necessary to rely upon
blueprints. At a minimum, the dimensions from the sampling
ports to the opposite wall should be physically measured
and appropriate corrections made in the calculated area.
These measurements must also be used to assure correct
placement of the probes, since distances can differ among
ports in a rectangular duct or along diameters in a circular
one. Also be sure that the proper area is calculated for
the instruments being used. If the duct is tapered in the
region of the sample plane and an instrument such as a pi tot-
static probe is used to obtain velocity data, the proper area
will not be the area at the sampling ports, but slightly
upstream of them where the head of the probe is actually
located.
• Probe Positioning and Alignment - Proper probe placement is
a critical item for velocity sensors, especially in circular
ducts. This requires accurate measurement of the line or
diameter along which the probe must be located, as well as
the height of the sampling port and connective fittings, etc.
37
-------
• LOG-LINEAR
POINTS
TANGENTIAL POINTS
Figure 9. Locations of sample points for tangential
and Log-Linear methods - 16 point traverse
Point
Location
Tangential
Log-Linear
1 2 3
3.3 10.5 19.4
2.1 11.7 18.4
4
32.3
34.5
5
67.7
65.5
6
80.6
81.6
7
89.5
88.3
8
96.7
97.9
Table 4. TANGENTIAL AND LOG-LINEAR SAMPLE POINT LOCATIONS
FOR 16 POINT TRAVERSE IN PERCENT OF DUCT DIAMETER
FROM INSIDE WALL TO SAMPLE POINT
"38
-------
The accuracy of the probe location should be on the order
of +0.2% of the corresponding duct dimension. Alignment of
the gas sampling probe is not critical, but care should be
taken to insure that the velocity sensor is always lined up
parallel to the duct axis to minimize flow angularity errors.
Such errors can easily reach a magnitude of 5% or more, as
is discussed in Section 5, so proper alignment is a very
critical item.
Interference Between Velocity and Gas Sampling Probes - With
some combination velocity measurement/ggs sampling probes,
the velocity probe response is affected by whether or not a
sample is being drawn. This is easy to check for, and
should it prove to be the case that the velocity is affected,
the unit should be calibrated without a sample being drawn,
and in actual use, no sampling should be performed while a
velocity measurement is being taken. If separate velocity
and sampling probes are used, care must be taken to avoid
interference effects.
Leakage - Good seals must be provided around all probes to
minimize errors due to leakage. The necessity for good
seals increase in inverse proportion to the duct dimensions.
If the duct pressure is below ambient, leakage will create
air jets into the duct, which will disturb the velocity
profile along the line of interest, and can lead to gross
sampling errors at points near the leak. If the duct pres-
sure is above ambient, the gas composition will not be
greatly affected, but velocity measurement errors will occur
near the port. 0-ring seals are generally the most reliable
and easiest to work with. Probes with unusual cross sectional
shapes which are difficult to seal around should be avoided.
Traverse Times - As mentioned above, it should not be a
problem to conduct 16 point traverses in about 20 minutes
for velocity, and in less than an hour for gas samples.
These times must be minimized to avoid errors due to tem-
poral variations, but rushing which results in sloppy
39
-------
placement, misalignment, and leakages cannot be tolerated. Times
can be minimized by insuring that all equipment is in proper work-
ing order and all test personnel are thoroughly familiar with the
test setup and procedures. The time response of the gas sampling
system must be determined in order to avoid taking too much or
too little time at each sampling point. This response time will
be a function of the size and length of tubing used, the flowrate
and pressure drop along the sample line, and the response time
of the analyzers or other equipment used. Velocity traverses can
be performed much more quickly because the system time response
is practically instantaneous.
Proportional Sampling - Proportional gas sampling is not con-
sidered to be a requirement. This was discussed in Section 3
and is treated in detail"in Reference" 2. This allows for the
use of separate velocity and sampling probes and even separate
sampling arrays, as discussed earlier in this section. If a
combined velocity/gas sample probe is used, there is no
reason not to sample proportionally, but if separate probes
are used, it would be advantageous to conduct several velocity
traverses during the gas sample traverse, as discussed in the
following paragraph.
Repeatability Data - Sixteen point traverses, especially for
velocity, can be conducted quickly enough to allow for pro-
curement of a significant amount of repeatability data. The
proper amount of repeatability depends upon the test objec-
tives, economic and manpower factors, and the consistency
of the data obtained. If possible, at least three traverses
should be performed for each operating condition of interest.
Multiple velocity traverses performed during the course of
a gas sample traverse are a good check on the validity of
the gas sample data - if a significant change in volumetric
flow occurs during the course of a gas sample traverse, the
gas sampling data will almost certainly be non-representa-
tive. If successive traverses show changes near the sample
ports but not elsewhere, there are probably errors due to
40
-------
leakage. Multiple traverses are needed to verify constancy
of flow conditions during the period of the traverse, and
also to check for errors at individual sample points due
to mistakes.
4.2 CONTINUOUS SAMPLING
4.2.1 Single Point Sampling
Single point sampling cannot be recommended if good accuracy is
desired except in a straight circular duct when the sensors can be placed
at least twenty diameters downstream and five diameters upstream of signif-
icant flow disturbances. This type of condition is most likely to occur
in pipelines or tall stacks, but is not typical in most stationary source
process streams^ /Hong straight run is required for full development of
velocity profiles, and a long run with no chemical reaction or addition of
mass is needed to avoid gas stratification. There is little question about
the need for multiple point velocity measurement, but there is serious con-
troversy over the adequacy of single point gas sampling, so it is worthwhile
to consider it in some detail. Acceptability of single point gas sampling
depends primarily on two factors: the desired measurement accuracy and
the extent of compositional stratification in the sampling plane. It is of
course possible that a highly stratified flow could exist in which a constant
relationship would exist between composition at a specified point and the
average composition, but no evidence of this has been discovered in the
data which TRW has examined. As a rule of thumb, it can be said that if
the stratification level of the gas stream is higher than the desired
measurement accuracy, single point gas sampling will not be acceptable.
To make this assessment, a workable definition for "stratification level"
must be found, since there is none in common usage.
In order to determine stratification levels from various sources in
a consistent manner, a uniform method to determine measures of data spread
is required. Perhaps the most common" and easy to work with measure of
spread is the standard deviation for a normal distribution. The potential
problem with using standard deviation as a measure of spread is that it
is normally applied to a number of measurements of the same parameter, such
as repeated measurements of the length of a table top, or repeated
41
-------
measurement of gas composition at a single point in a uniform gas stream.
For the present case, we wish to apply it to a number of individual
measurements of concentration at separate points in a non-uniform gas
stream. For this case, there is no a priori reason to believe that the
readings can all be considered to belong to a normal distribution. Gas
traverse data have been examined to see if they approximated a normal
distribution. Data consisted of 49 point traverses for 02, C02. and NO
in a coal fired power plant. For each traverse, the mean value and
standard deviation of the gas concentration were determined as follows:
N
n=l
where
a. = standard deviation of
(36)
(37)
Results of the calculations are shewn 1n Table 5.
Table 5. COMPARISON OF DEVIATIONS FOR ACTUAL AND
NORMAL DISTRIBUTIONS
DEVIATION
a
2o
3o
NORMAL
DISTRIBUTION
68.3
95.4
99.7
ACTUAL DISTRIBUTIONS
NO,
65.3
100
100
co2
61.2
100
100
°2
71,4
89.8
100
Data are for actual distributions taken from a TRW field test.
Tabular values are percentage of data points which differ from
the mean value by less than the specified deviation.
42
-------
Results In each case are very close to what would be expected for a normal
distribution.
Given the validity of the standard deviation as a measure of spread,
we can now define a parameter to indicate the degree of stratification in
a stream. For most engineering applications, a 95% level of certainty
with regard to the accuracy of a measurement is acceptable. This corres-
ponds to a band of +_ 2a about the mean value. Consider the NO traverse
" A
used for Table 5 as an example. The mean concentration was 734 ppm with
a standard deviation of +26 ppm, or 3.5%. By definition, then, we will
say that the stratification level was +7% (i.e. 2a) from the mean, which
means that the probability is 95.4% that a point measurement taken any-
where in the sample plane will deviate from the mean concentration by
less than j7%. Stratification level is being defined as +_2a from the
mean concentration value in the sampling plane.
This definition of stratification level is desirable because it
ftts in directly with the error analysis in Section 3, which is also
based on standard deviations. This allows the effect of stratification
on system accuracy to be determined in a simple, meaningful manner. For
the 26 gas sample traverses previously cited, the stratification level,
2a, of the actual data points varied from 7% to 36% of the mean concen-
tration, with an average of 15%. By definition, the stratification level
is an indication of the accuracy which can be achieved with a single point
measurement. The most common single point gas sampling location 1s in
the center of the duct. For the 26 runs, the average 2a error for a
single point measurement at the center of the duct was 13.6%, which is
very close to the average stratification value, and very high with respect
to the other errors identified in Section 3. On the basis of
available data, then, it has been concluded that single point gas sampling
will severely degrade achievable system accuracies. If installation of a
single point continuous gas monitor is anticipated, the stratification level
at the selected site should be determined by full manual traverses. From
analysis of accumulated data, we have concluded that for accurate measure-
ments, single point gas sampling is unacceptable for a specific site
unless proven otherwise.
43
-------
4.2.2 Multiple Point Sampling
Multiple point continuous sampling is very distinct from multiple
point manual sampling in a practical engineering sense. Large arrays of
point sensors for continuous monitoring would be very expensive to procure
and maintain. This makes it imperative that the number of sampling points
be minimized, s6 long as large errors can be avoided. Previous discussion
has shown that from an accuracy standpoint, a sixteen point array would
be quite acceptable, and that a single point sample would generally be
inadequate. The number and location of points for a given installation
should be determined by the best balance of cost and system accuracy. The
methodology presented in this section has been selected to achieve this
optimization. The methodology also presumes the use of continuous gas
analyzers, electrical outputs for all sensors and analyzers, and the
availability of a low cost computer to perform calculations of the kind
shown 1n equation 20. This hardware 1s discussed in Section 5.
4.2.2.1 Continuous Sampling in Circular Ducts
For monitoring in circular ducts, the site selection criteria are
the same as for manual sampling, as discussed in Section 4.1.1. Extensive
laboratory testing described in Reference 1 showed a 2a mapping error
of less than 7% for average velocity using the Log-Linear method with
eight points (four points on each of two orthogonal diameters), and a
2a mapping error of less than 9% for average velocity using the Ellison
Annubar (a non-point sensor described in Section 5) as the velocity
sensor. Each of these errors is larger than the nominal 6% single
point accuracy previously cited, but not prohibitively so. In addition,
the laboratory testing was conducted using primarily "worst case"
flows, so that the expected accuracy in most field situations would be
better than the cited figures.
Gas stratification has not been studied as extensively as velocity
stratification in circular ducts. An eight point gas sample traverse
44
-------
using the tangential method is considered to be acceptable on the basis of
present knowledge, with the option that the eight points can be taken on
either one or two diameters. If one diameter is chosen, it should be the
one along which the greatest stratification is observed during preliminary
manual traverses. The most important practical point to remember about
gas sampling is that work cited in Section 3 shows that it is unnecessary
to sample proportionally. This means that the velocity and gas sampling
components can be totally independent and the gas sampling rate should
be the same at each sample point. The uncoupling of velocity and compo-
sition measurements clearly offers the prospect of much lower system cost
and complexity.
Planning and setting up a multiple point monitoring system in a
circular duct should involve the following methodological steps:
1. Select the site so that, as nearly as possible, 80% of
the local straight run is upstream of the sample plane.
2. Perform a preliminary traverse at each operating condition
of interest to check for unusual velocity patterns and to
determine compositional stratification levels.
3. Determine whether or not single point gas sampling is
acceptable.
4. Select, the number and determine the location of sampling points.
Information for locating points using the Log-Linear and tangen-
tial methods and the Annubar is shown in Figure 10.
5. Calibrate the system in place using 16 point velocity and
gas sample traverses, preferably using at least three traverses
for each operating condition. In-place calibration is essen-
tial to minimize systematic errors which must be expected
due to the small number of sampling points used for the
continuous measurements.
The proposed approach for continuous sampling in circular ducts
follows directly from the procedures used for manual sampling - the only
significant differences deal with the number of sample points involved
and the use of spatial rather than flow proportional gas sampling at the
selected points.
45
-------
diameter A
diameter B
Duct cross-section
Velocity profiles
For eight point samples using two diameters, locate points along each
diameter as follows:
Point
Location
Tangential (gas sample)
Log-Li near (velocity)
1 2
6.7 25.0
4.3 29,0
3
75.0
71.0
4
93.3
95.7
Locations given in percent of duct diameter from inside wall to
sample point
For Annubar use or eight point samples using one diameter, select
diameter along which velocity is more irregular (diameter A for the
example above), and the diameter along which composition waives the
most for velocity measurement and gas sampling, respectively. Point
locations are given in Table 4.
Figure 10. Circular duct continuous monitoring locations
46
-------
4.2.2.2 Continuous Sampling in Rectangular Ducts
Unlike the circular duct recommendations, the proposed rectangular
duct continuous sampling techniques are significantly different from the
corresponding manual methods. The differences are due to the nature of
velocity patterns in rectangular duct systems and to practical considera-
tions such as cost. A brief discussion of the former is given to serve as
an introduction to the proposed sampling techniques.
Velocity measurement methods in circular ducts may be characterized
as line averaging techniques, since they involve taking a number of
measurements along a line (diameter). The success of a given method
depends upon two factors: the ability to obtain an accurate average along
the selected line, and the relationship between the average velocity along
the line chosen and the average velocity in the entire duct. Common
techniques are concerned almost solely with the first factor, and assume
a one-to-one correspondence for the second. The basis for this assumption
is that fully developed flow in a circular duct is in fact axisymmetric,
and a circular duct will automatically condition the flow toward axisymmetry.
It would be desirable to have an analogous situation occur in rectang-
ular ducts, so that a line average across the duct would be representative
of the total flow. In practice, this desire is hampered by the fact that
in most stationary source streams, the straight runs of rectangular ducting
tend to be very short. This inhibits the development of predictable flow
patterns, which in turn makes it difficult to develop techniques to
accurately measure the flow through use of a small number of sampling
points. Fortunately, there is one situation in which the flow is developed
in a desirable manner for use of line averaging techniques, and it is a
very common one. Immediately downstream of a rectangular elbow, the flow
tends to become conditioned so that profiles taken across the duct in the
plane of the elbow are very similar as illustrated in Figure 11. The elbow
itself constitutes such a violent disturbance to the stream that the pro-
files tend to be very repeatable with respect to each other regardless of
changes in flow further upstream. This property aids in obtaining good
repeatability and minimum random error due to flow pattern changes. Due
^Eo the nature of the flow downstream of a rectangular elbow, however, the
47
-------
DUCT DIMENSIONS .41m
.30m
00
DEAD AIR REGION
PROFILES ARE VARIATION IN 10% INCREMENTS FROM MEAN VELOCITY
MEASURED MEAN VELOCITY =16.0 m/sec
Figure 11. Normalized velocity distribution downstream of an elbow
TRW pilot scale test
-------
desirable flow pattern shown in Figure 11 is very much a function of
distance from the elbow. The flow reaches an optimum pattern
about one duct width downstream of the elbow, where the duct width is the
dimension in the plane of the elbow, and then changes into a much less
desirable shape further downstream, as illustrated in Figure 6. Since
elbows are such a common occurrence, considerable laboratory development
6as been performed to make use of their known properties. This work is
described in detail in Reference 1, and the end result was the Row
Average Method, which allows the use of hardware that is not more compli-
cated than that required for measurements in circular ducts. The Row
Average Method was developed for velocity measurements, but has also been
shown to be adequate for gas sampling.
The Row Average Method is very simple in concept, but in practice
usually requires more preliminary work (i.e. manual traverses) than do
corresponding circular duct techniques. The Row Average Method is
illustrated in Figure 12.
- - - SAMPLE POINT
Flow
is into
page
Figure 12. Point locations for row average method
The first step is to select the sample plane location, which will be
influenced by the local duct geometry. The second step is to select the
appropriate direction, X or Y, in which to take the Row Average. The
third step is to select the optimum distance from the wall, d , for the
X
sampling line. Once the line has been selected, it is divided into a
number of equal segments (eight or more) and the sample points are located
at the center of each segment. The output is the average of the
individual readings.
49
-------
The advantages of the Row Average Method are practical ones. Taking
all measurements along a single straight line allows for minimum hardware
complexity, including an extremely simple approach to gas sampling dis-
cussed in Section 5. The penalties to be paid compared to a 4 x 4 array
would be a probable higher system mapping uncertainty and a need for greater
initial manual investigation to select an optimum site. It is our belief
that in most cases the system error can be made small enough to be acceptable.
The general procedure for use of the Row Average Method is as follows:
1. Site Selection - Select a location downstream of an elbow if
possible. If such a location is available, determine the
position of the sampling plane from Figure 13. Note that
there are two different locations given, depending upon
whether the Annubar or a point array is used for velocity
determination. Optimum location will produce local velocity
profiles similar to those in Figure 11. If an elbow is not
available, the sample plane should be selected so that 80%
of the local straight run is upstream of the sample plane.
2. Determination of Row Direction - After the sample plane has
been selected, perform reference traverses to determine
velocity and concentration maps. For these traverses, it
is preferable to use 6 x 6 to 8 x 8 arrays. Using traverse
points, compute line averages in each direction and plot
the results. Typical results are shown in Figure 14. The
direction which should be selected for row averaging is the
one which has the flatter profile (X direction in Figure 14).
This direction may be different for velocity measurement
and gas sampling.
3. Determination of Row Location - Two factors influence this
decision unless the Annubar is to be used as the flow sensor.
If the Annubar is to be used, it should be located midway
between the walls. The two factors are flatness of the row
average curve in the vicinity of the proposed row location
and deviation of the Row Average from the overall average.
The first deals with random errors in that the flatter the
local profile is, the lower will be the random error due
50
-------
Sample
Plane
Location
A. For Systems Using Annubar, D = 1.5W +'
B. For Systems Using Row
A/erage Method for
Velocity
D = .8W
+.40W
-.60W
Figure 13. Recommended sample plane locations
downstream of a rectangular elbow
Normalized
Row
Average
1.0
' Y Rows
X Rows
Row Average Data Point
Curve Fit
Length in % of Duct Dimension
Figure 14. Typical plot of normalized row averages
51
-------
to flow variations. The second 1s associated with systematic
errors. The first factor should usually take precedence
since systematic errors of this type can be eliminated through
in-place calibration. A desired row location is illustrated
in Figure 14.
4. Installation and In-Place Calibration - After the row loca-
tions have been selected (they will be presumably different
for velocity and gas sampling) the hardware can be selected
and installed, with a minimum of eight sample points along
each Row. The system should then be calibrated in place by
means of 16 point velocity and gas sample traverses. At
least three traverses should be performed at each flow con-
dition. In place calibration is essential to minimize
systematic errors.
The Row Average Method is recommended because it has demonstrated
acceptable accuracy in the laboratory and in the field, and because it
results in minimum cost and system complexity. Like the proposed circular
duct techniques, it makes use of the uncoupling of velocity and concentra-
tion in the sample plane.
4.2.3 General Sampling Considerations
Many of the comments in Section 4.1.3 on manual sampling apply as
well to continuous monitoring. In addition to those, the following should
also be given appropriate consideration to assure accurate, reliable
measurements:
• Selection of In-Stream Components - This will be covered
in more detail in Section 5. The emphasis should be placed
on purchasing or fabricating components whose performance
will not be degraded due to long term exposure to the flow
stream. For example, theeffect of particulate buildup on
aerodynamically important surfaces of velocity sensors is
not usually a problem for manual traverses because of the
short exposure times involved. For long term applications,
the effect of the buildup would tend to appear as a syste-
matic shift in output which may be overlooked if the sensor
52
-------
continues to function. In a case such as this, the system
should not be calibrated in place until a steady state
buildup is reached, so that the calibration will be appro-
priate for subsequent data.
Probe Locations and Alignment - The important point here
is to use proper hardware (e.g. fittings) and installation
procedures to insure that proper probe placement and align-
ment are maintained after installation. Process stream
ducting is usually subject to continuous vibrations which
could easily result in movement or rotation of in-stream
probes. In addition, once probes are installed the locations
should be permanently marked in such a manner that the probes
can be removed for maintenance and then replaced in exactly
the same location. This must be done to avoid compromising
calibration data previously obtained.
Leakage - Proper fittings should be used to avoid leakage
problems in pressure and gas sample lines. We have had the
best experience with Swagelok fittings, which are designed
to resist loosening due to vibration and can be repeatably
connected and disconnected without degradation. Pipe thread
fittings should be used as little as possible, and only with
a proper thread sealant such as Teflon tape. Leakage prob-
lems in gas sample lines can be minimized by maintaining as
much of the line as possible at higher than ambient pressure.
Data Reviews - Regular, methodical reviews of system output
data are a good way to discover system errors. For example,
a steady change in output over an extended period of time
when flow conditions are known to be constant indicates the
development of a systematic shift.
Accuracy Considerations - The proposed sampling techniques
are intended to give significantly better system accuracy
than the currently popular single point techniques, while
avoiding the cost and complexity of large arrays. The
possibility of having a system with a systematic error perhaps
"53
-------
as high as 10% due to the sampling configuration selected
is considered acceptable due to the fact that such an error
can be eliminated through 1n-place calibration. It is
large random errors which cannot be tolerated, since they
cannot be reduced. The body of data available for Row
Average and Log-Linear velocity measurements with 8 points,
and for the Annubar, suggests that acceptably small random
flow variations can be expected for a given installation
due to the stabilizing effects of the local geometry on
the flow distribution. It is our belief that for most
installations, the 2o random error due to flow and compo-
sition variations should not be greater than +6% and*>425,
respectively, which is roughly twice the error expected
for a properly performed 16 point traverse.
Basic continuous measurement system procedures are given in
Section 6.
54
-------
5. HARDWARE
This section is concerned with the characteristics of velocity
measurement hardware for manual and continuous applications, and with
devices used to obtain gas samples. It is not specifically concerned with
gas sample train components which are not in the stream, but several of
these are discussed briefly. The purpose of the section is to show the
acceptability or unacceptability of various pieces of currently available
hardware, to indicate where hardware development is needed for continuous
monitoring systems, and to show a prototype monitoring system which incor-
porates the proposed methodology and hardware.
5.1 VELOCITY MEASUREMENT
5.1.1 Pi tot Devices
Pi tot probes of various types are used to sense velocity by means of
the relation in Equation 3. An extended treatment of the subject is
given in Reference 12. Two probe characteristics are of primary importance
for stationary source monitoring — calibration factor when the probe is
aligned with the flow, and sensitivity to flow orientation angle. The two
most common types of pitot probes are the pitot-static probe, Figure 15a,
and the Stauscheibe or S probe, Figure 15b. The primary output of each
probe is the pressure difference between the high and low pressure ports.
It is common practice to associate a calibration factor K with the probe
such that
<38>
where
K = calibration factor, dimensionless
Ap = true free stream differential pressure, pQ-pm, torr
Ap = measured differential pressure from instrument
•J\
A typical calibration curve is shown for a commercially available S probe
in Figure 16.
" 55
-------
AXIS
IMPACT
(HIGH PRESSURE)
PORT
STATIC
(LOW PRESSURE)
PORTS
Figure 15a. P1tot-static probe
Figure 15b. S. probe
LOW
PRESSURE
PORT
HIGH
PRESSURE
PORT
For the pitot-static probe, orienta-
tion can be specified by one angle,
e, due to the axisymmetry of the
probe. For the S probe, orientation
is a function of two independent
angles: e^, the rotation angle
and Sp. the tilt angle.
56
-------
Calibration
Factor, K
.90
.86
.84
.82
.80
J_
8
16 24 32
Velocity, m/sec
Figure 16, S probe calibration factor as a function
of velocity
There are S probes which have more constant calibration factors than the
one shown in Figure 16, but users must be cautioned not to assume a constant
calibration factor for a given probe - they should be calibrated individ-
ually to verify the true calibration.
There are several forms of pitot-static probes, the most desirable of
which for process streams use is the ellipsoidal nosed pitot-static probe.
The following statement regarding calibration accuracy is taken from
Reference 12, page 35, based on a calibration factor K = 1.000: " ....
it seems reasonable to assume that the maximum error on velocity will not
exceed 0.5 per cent over the entire speed range from 3 to 200 ft/sec."
The upper limit of the range is determined by compressibility effects, and
the lower end is proportional to the probe diameter. The statement applies
specifically to probes of diameter .79 cm (.31 inches), so that a probe of
diameter .64 cm (.25 inches), which is a commonly available size, would
have a lower limit of 1.1 m/sec.
In terms of calibration factor above, a properly designed S probe
with a constant calibration factor would be preferable to a pitot-static
57
-------
probe because of the higher differential pressures it produces. When
orientation sensitivity is also considered, the preference changes. An
orientation sensitivity plot for the same commercially manufactured S probe
is shown in Figure 17, which shows TRW test data taken from Reference 1.
Figure 17 shows that orientation sensitivity is a function of velocity as
well as angle. Comparison data for the S probe and ellipsoidal nosed
pi tot static probe are given in Table 6, where the pi tot-static probe
data are from Reference 12. The angles are defined in Figure 15.
The angles are defined in Figure 15.
TABLE 6. YAW CHARACTERISTICS OF PITOT-STATIC
AND S PROBES
9
Degrees
0
5
10
15
20
25
30
Error with Respect to u = U cos 9, %
Pi tot Static
Probe
0
+ .6
+2.0
+3.3
+2.7
+1.7
+ .3
S Probe
Rotation Angle, e-.
0
+3.3
+6.9
+7.4
+9.0
+11.9
+15.0
Tilt Angle, 9g
For 92 > 0, errors
are higher than
corresponding ef
errors; for 69< 0,
errors are of the
same order as 9,
errors.
The entries in Table 6 are exactly in the form of the error term
[(tan 9)a0] in the error analysis equation, Equation 24a, so that the
o
effect on system accuracy can be easily determined.
The accuracy savings offered by the higher pressure output of the
S probe are more than offset by the large errors due to alignment sensi-
tivity. Field measurements taken with S probes traditionally tend to
result in overly high estimates of average velocity, and orientation sensi-
tivity is believed to be the cause. The S probe is unquestionably a good
instrument for qualitative investigation work, since its bi-directionality
58
-------
u
1.06r
indicated
uo
= velocity at e = 0
1.04
1.02
1.00
.98
.96
.94
.92
.90
.88
.86
uo
6 m/sec
17 m/sec
30 m/sec
.84
i 1 i i i . . . \ . . . .
-50 -40 -30 -20
-10
0 10 20 30 40 50
Orientation Angle e1 , degrees
Figure 77. S pitot probe orientation sensitivity data
59 ~
-------
quickly reveals such things as recirculating regions in the flow. For
quantitative work, however, the pitot-static probe is much more desirable
because of its accuracy characteristics. The S probe is suitable for
quantitative work only when very good flow alignment with respect to the
duct axis can be verified, as in the case of a long, straight circular duct
with no swirl. TRW test teams have used pitot-static probes for manual
traverses in field applications involving high particulate grain loadings
and in very moist, droplet laden streams without encountering clogging
problems. Some care is required to avoid bending the probe head during
insertion and withdrawal, but this does not hamper operating efficiency.
EPA Method 2 recommends that a pitot-static probe be used to calibrate
S probes. After examining and working with both types of probes, we have
concluded that the S probe is unacceptable for accurate work under most
circumstances,and~the ''primary standard" - the pitot-static probe - should
be used instead.
5.1.2 Continuous Monitoring Velocity Instruments
In this section, two instruments which are known to be acceptable
are discussed. In addition, an unacceptable instrument is briefly dis-
cussed, and general criteria for instrument selection are presented.
Finally, some advanced types of instruments are mentioned to indicate
desirable directions for future development.
5.1.2.1 Ramapo Fluid Drag Meter
The Ramapo Mark VI Fluid Drag Meter is shown in Figure 18. The
drag force on a target disc is measured, and velocity is calculated
from knowledge of the drag coefficient of the disc. Velocity is propor-
tional to the square root of the output in the same way that velocity is
proportional to the square root of the differential pressure from a pitot-
static probe. The standard probe has a nominal accuracy of +_!% of the
measured velocity. This accuracy was confirmed through TRW testing as
described in Reference 1. The key to-the accuracy of the instrument is
the constant drag coefficient of the target disc, which is treated in
Reference 13. Instrument calibration can be checked and adjusted elec-
tronically through the use of a decade box, which helps to minimize
calibration problems. As with the S probe, orientation sensitivity varies
as a function of velocity. For the velocity range lOm/sec to 25 m/sec,
60
-------
TARGET
DISC
PROBE IS
BI-DIRECTIONAL
LEVER
ARM
STRAIN
GAGE
BRIDGE
Drag on the disc 1s
transmitted by the lever
arm to the strain gage
bridge which provides the
probe output voltage.
VOLTAGE
OUT
txJ
TRAVERSE
MECHANISM
IOLTAGE
OUT
For line average measurements, the probe
should be attached to an automatic
reciprocating traversal mechanism to
obtain measurements at various points
along the line. This would be less
expensive than a multiple sensor rake.
DUCT
Figure 18. Ramapo Fluid Drag Meter
61
-------
the maximum error for flow misalignments up to 30° would be about 3%; the
figure would be somewhat higher for velocities outside this range.
The instrument is acceptable for long term continuous monitoring
because of its desirable accuracy characteristics and because of its
ability to withstand hostile flow environments. The only important com-
ponent which may be susceptible to corrosion over a period of years is the
target disc itself, and that component is very inexpensive. The probe
requires only electricity — no purge gases or compressed air is used.
The Ramapo Fluid Drag Meter was found to be the best of the point
sensors tested in Reference 1, and is definitely recommended for use in con-
tinuous monitoring applications involving point velocity sensing techniques.
5.1.2.2 Ellison Annubar
The Annubar, mentioned briefly earlier in the report, is shown in
Figure 19. As a velocity sensor, the Annubar is somewhat unique because
it does not take measurements at a single point but along a line across
the duct. The instrument consists of a tube with four holes which face
directly upstream, and a single hole on the backside of the probe. The
tube is hollow so that the four upstream ports combine to provide a single
pressure output, which is then compared to the downstream hole pressure
to provide a single differential pressure output. In practice then, the
instrument has the same type of output as an S probe, but responds to
the flow velocity at five points in the stream instead of one.
The Annubar was developed for use in circular ducts, the impact hole
positions being located according to a circular duct mapping scheme.
Extensive TRW testing described in Reference 1 has also shown it to be
suitable for measurement in rectangular ducts. The primary advantages of
the Annubar are practical ones. It has been demonstrated that a single
Annubar can achieve accuracies comparable to arrays of up to eight point
sensors. Since accuracies are comparable, the advantages of using a
single probe with a single output rather than eight sensors with eight
outputs are immediately obvious. Testing in the laboratory has resulted in
the conclusion that 2o random errors for Annubars used in short runs of
circular and rectangular ducting should be on the order of 3% to 7%, which
is comparable to the accuracy of the Row Average Method using eight points.
62
-------
Oi
CO
ASME PAR UW-16 DRILL OR BURN
ONE 1-3/8" DIAMETER HOLE AND
PROVIDE 1/4" MINIMUM WELD BEAD
ALIGN TO PIPE AXIS (TYPICAL) -
ASTM SPEC. A181 GRADE No. 2
2/3000 LBS. 1" NPT FORGED
STEEL WELD COUPLING IS
SUPPLIED.
4- 3/8" DIAMETER SENSING PORTS
ARE LOCATED IN APPROPRIATE
CENTERS OF CONCENTRIC ANNUL).
FLANGE INCLUDED
TO CUSTOMER'S
SPECIFICATIONS
FACTORY TIG WELDED
UNLESS OTHERWISE
SPECIFIED
FLANGE MOUNT OPTION 1/8" NPTF CLEAN-OUT PLUGS-
MATE RJALj^ 303 STAINLESS. OTHER MATERIALS OPTIONAL.
A 3 TO 4 PIPE DIAMETERS IS RECOMMENDED FOR DOWNSTREAM SIDE.
X60R MORE PIPE DIAMETERS IS RECOMMENDED FOR UPSTREAM SIDE
V"J AFTER VALVES. ELBOWS AND ETC. SEE CALCULATION REPORT E - 79
^PERMANENT TAG SHOWING MIN.. NORM, ft MAX. DESIGNED FLOWS. METER
'"'READINGS FOR DESIGNED FLOWS. TAG NO.. LINE SIZE. SER. NO. » METERED FLUID.
ARMCO 17-4 PH
HARDENED STAINLESS
COMPRESSION FERRULE^
PERMANENT
METAL TAG
WITH 3" CHAIN.
STAINLESS 1/2". 1/4" or
1/8" NPT. or 1/4" TUBE
COMPRESSION. OR
BRASS SHUT-OFF VALVE
WITH 1/4"SAE FLARE
CONNECTION
SECTION
DETAIL
HALF SIZE
SP?ciFYPplPEZ0 D
lD
i.u.
"ELEMENT MADE TO THIS DIMENSION
UNLESS OTHERWISE SPECIFIED. PARTS
IN THIS REGION & GASKET NOT SUPPLIED
EXCEPT WELD COUPLING
fr
OISTBRICH STANDARD CORPORATION
• L I. I • O N INSTRUMENT DIVISION
ORAWSR M SOULOKR COLORADO •OJOS USA
751 TO 765 ANNUBAR FLOW ELEMENTS
Wy
5-19-70
//- 7-
REVISED
-------
Like the Row Average Method, the Annubar should be calibrated in place
to minimize systematic errors.
The Annubar rear orifice is subject to clogging, as described in
Reference 1, so an intermittent purge using compressed air is recommended
in particle laden streams, especially when the moisture content is high.
Also, buildup of particulate on the back side of the probe can affect
the reading of the rear orifice. This accuracy consideration was discussed
in Section 4.2.3. Because of this effect, the Annubar should not be cal-
ibrated until a steady state buildup has been reached. Purging will not
affect the bulk of the buildup, but will just keep the orifice itself clear.
In sources such as power plants, this steady state condition should be
reached within about 24 hours after installation. A nominal purge cycle
would be five minutes per hour.
5.1.2.3 General Considerations for Instrument Selection
Several of the characteristics to be mentioned below will ordinarily
be determined from manufacturer's data. The following situation is pre-
sented to show that such data must be examined with care to avoid being
misled about instrument performance.
A Hastings-Raydist Flare Gas Flow Probe was selected for evaluation
as a continuous monitoring probe. The following paragraph is quoted from
the factory manual section on accuracy:
"The initial calibration of the probe is +2% of the
full scale voltage; that is, ±.10 volt. Since the
velocity vs. voltage curve is non-linear, the vel-
ocity tolerance must be detemined for each segment
of the curve. For example, ±.10 volts represent a
velocity tolerance at 1000 fpm (2.00 volts) of +50
fpm, but at 2500 fpm (4.00 volts) it represents
+150 fpm."
After reading this, most people would be inclined to remember only the
+2% figure, and tend to think of the instrument as having a +2% accuracy,
despite the presence of the two qualifying sentences. The accuracy in
terms of velocity is shown in Figure 20.
64
-------
16
Accuracy of
Velocity
Measurement,
Percent
8
l i I I I I I I 1 _L I i I I
8
16 24
Velocity, m/sec
32
Figure 20. Velocity accuracy for Flare Gas Flow Probe
Accuracy in terms of velocity, which is the real parameter of interest, is not
on the order of 2%. The point of the illustration is that accuracy should
always be considered in terms of the main parameter, and not a secondary
parameter such as voltage.
Selection of a velocity sensing instrument for continuous monitoring
applications should involve consideration of the following:
1. Accuracy - This relates primarily to simple straight align-
ment (0=0) accuracy but also includes such features as
repeatability, drift, zero and full scale stability, and
hysteresis. All of the effects should be combined, using
the R.M.S. technique exemplified in Equation 23, to produce
a single curve of accuracy given as percent of reading, as
in Figure 20. This approach avoids possible misconceptions
concerning instrument accuracy. When the data are presented
in this manner, a determination can be made regarding the
suitability of a particular sensor for a specific application.
2. Orientation Sensitivity - This, parameter is being emphasized
throughout the manual because of its importance in process
stream measurements. So long as measurements must be made
in ducts of complicated geometry with many disturbances and
changes of direction, significant flow angularity will be
65
-------
encountered. Orientation sensitivity is not a parameter
for which manufacturers normally supply data, so the
work usually has to be performed by the customer. In any
event, this type of calibration may be very important
relative to overall system accuracy and should be treated
accordingly.
3. Survivability - It is essential that the sensor be able to
perform in a process stream environment for long periods of
time. Survivability data are usually best obtained from
people who have been using the instrument for a prolonged
period in application similar to or worse than the
intended one.
4. Support and Interface Requirements - Does the instrument
require additional devices such as voltmeters or pressure
transducers, and if so, what is the impact of these addi-
tional devices on cost and system accuracy? It is also
important to know the extent of assembly required by the
customer after delivery. What type of power interface
is required, and does the instrument need compressed air
or water supplied to it? Some instruments are very sen-
sitive to changes in line voltage, so it may be necessary
to provide a voltage regulation capability to damp out
line fluctuations.
5. Maintenance and Calibration - What are the extent and
frequency of normal maintenance and calibration? Can
calibrations be performed in place, as with the Ramapo
Fluid Drag Meter, or will it be necessary to remove the
instrument and ship it back to the factory?
6. Cost - A nominal cost breakdown estimate should be per-
formed for a specified period of time. Items included
should be the purchase price of all identifiable system
components, an estimate of expected replacement costs
during the time period, and cost in manhours to assemble,
learn to use, install, calibrate, and maintain the
equipment.
66
-------
Any given instrument may have a particular characteristic in any
one of the above categories which would make it unsuitable for use, so each
category should be considered before making a purchase. It is recommended
that the Ramapo Fluid Drag Meter and Ellison Annubar be used as standards
for comparison for point sensors and line averaging sensors, respectively,
when evaluating velocity devices for a given application.
5.1.2.4 Advanced Concepts
The great majority of currently available instruments are point sensors.
Since point sensor arrays are expensive and single point measurements are
generally inaccurate, point sensors are not the most desirable type to use
for process stream measurements, especially for continuous monitoring.
The next major development step should logically be line averaging devices.
The Annubar is a good start in that direction. The near future should see
development of techniques such as laser anemometry for general industrial
applications. A line averaging device should be able to do just what the
name implies -measure the average velocity along the line of interest.
In a circular duct, such an average would be in error because of the area
considerations, but the error would be systematic in nature and could be
accounted for by a calibration factor. In a rectangular duct, a line
average is exactly what is desired for the Row Average Method. The
methodology for line average devices may be said to already exist, at least
in elementary form, which lends credance to their development.
5.1.3 Supplementary Velocity Instrumentation
This section deals with pressure and temperature measurements which
are usually required in support of velocity measurements. The parameters
of interest are differential pressure (nominal range 0-5 torr), absolute
static pressure (700-800 torr), and static temperature (0-400°C): Static
temperature and pressure are reasonably easy to measure with accuracies
on the order of 1% or better, using properly compensated thermocouples and
electronic pressure transducers or manometers. Recall from Section 3 that
static temperature should be measured at each point where velocity is
measured, while static pressure can be considered constant across the
duct. In continuous monitoring systems, it is usually not acceptable to
assume that static pressure remains constant over long periods of time
and use a single value in data reduction. Stream pressures usually change
~87"
-------
as a function of ambient pressure, which can change by 3% to 4% over a
period of days. Velocity sensors should ideally incorporate integral
thermocouples, and should be purchased in that configuration if possible.
Differential pressure measurement is traditionally a source of large
uncertainties because of the small pressures involved in gas flow measure-
ments. Electronic pressure transducers are capable of providing accurate
measurements in the range of interest, but are more expensive than liquid
manometers. In most cases where good accuracy is of interest, the expense
will certainly be justified. Nominal accuracies for a U-tube manometer,
Statham pressure transducer, and MKS Baratron pressure transducer are
shown in Figure 21. The cost for the Statham unit would be about $400,
and about $3000 for the Baratron. The Baratron cost is similar to that
for a typical continuous gas analyzer. For a continuous monitoring
system which employs a differential pressure device such as the Annubar,
the velocity measurement error is likely to be the largest single error
source, so use of an adequate pressure transducer is essential, just as it
is also essential for maximum accuracy of manual traverses.
5.2 GAS SAMPLING
This section is concerned with hardware for extracting gas samples,
that is, gas sample probes. Other sample train components are mentioned
as necessary in the discussion.
5.2.1 Single Point Sampling Probes
Gas sampling is not subject to the isokinetic requirements which
govern particulate sampling, so the shaped nozzle commonly used for par-
ticulate sampling is not required. In fact, its use can sometimes lead to
sampling errors. The objective in gas sample extraction is to locate the
sample probe at the point of interest and withdraw the gas sample without
causing any changes in composition or phase of the gaseous stream components,
Basically this means that changes in temperature and pressure along the
sampling line must be minimized in order to maintain the gas sample at the
local stream pressure and temperature. Pressure losses can be minimized
by using the largest reasonable size of tubing and low flow rates. The
extraction flow rate must of course be compatible with the method of
analysis to be used. Temperature control is best provided by having a
68~
-------
1.5
APPROXIMATE VELOCITY, m/sec
5 15
30
45
100
10
UJ
UJ
CL,
o
I
o
O
5
cc.
o
o
d .1
.01
U TUBE
MANOMETER
(REF. 12)
TRANSDUCER
(REF. 14)
BARATRON 145B
TRANSDUCER (REF. 15)
..I
i i i
.02 .1 1
DIFFERENTIAL PRESSURE, TORR
Figure 21. Single point velocity accuracy based on differential
pressure measurement accuracy versus differential
pressure for three pressure measurement methods
69
-------
thermocouple at the probe Inlet and maintaining as much of the sample line
as possible at the probe inlet temperature. Thermal control is most
critical for substances such as water and volatile hydrocarbons which may
be subject to significant evaporation or condensation in the sample lines.
Aerosol evaporation in the sample lines can best be avoided through
use of a sample probe whose inlet is normal to the flow or, preferably,
pointing downstream. This will minimize the collection of droplets which
could subsequently be vaporized. A particulate filter at the probe entrance
should be used to remove any entrained particles. If aerosols are kept out
of the sample lines, thermal control of the probe is made easier, since it
then becomes a matter of insuring that the line temperature does not fall
below the temperature at the probe entrance.
5.2.2 Multiple Point Sampling Probes
Multiple point sampling probes are of interest primarily for continuous
monitoring applications, but also may be of value in situations where the
process cycle time is too short to allow a manual traverse. There is one
fairly common misconception about the operation of multiple point sampling
probes which should be disposed of before proceeding further. Consider
the hypothetical situation shown in Figure 22a. It is desired to obtain a
sample at two points in a stream, where the velocity at one of the points
is higher than at the other. A common technique is to use a single sample
tube with holes at the two points of interest, and make the hole area
proportional to the local velocity. To avoid particulate clogging problems,
the holes are made as large as possible. The intent is to draw a sample
through each hole'in proportion to the local velocity. What will actually
happen is quite different. With normal sampling rates on the order of a
liter per second and tube diameters of 2 to 3 cm, the pressure drop along
the sample tube is very small, and will usually be smaller than the varia-
tion in stagnation pressure in the free stream. When this happens, and
when the sampling holes are large so that there is very little pressure
drop across the holes, the flow pattern in the probe develops as shown in
Figure 22b. Because of the relative pressure drops, only flow from the
high velocity region enters the probe. Some of it is withdrawn and becomes
the sample, while the remainder goes back out of the probe through the
70
-------
SAMPLE
INTENDED
FLOW PATTERN
C*^**» ^ ^ J^""^" ^ ^ ^ •» ^ ^ •* •* T*
^ •* ^* I **• ^ ^ •* -«.-1
I
HOLE HOLE ,
1 2
U > u
A. INTENDED DESIGN: HOLES 1 AND 2 ARE SIZED IN PROPORTION TO u
AND u2 WITH THE INTENT OF OBTAINING A FLOW
PROPORTIONAL SAMPLE THROUGH THE HOLES
1
SAMPl F
OUT ^ =^s=a
/"
/
A..
K,
ACTUAL
PATTERN
/
> ^
Dl
FLOW
^
1\P2
B.
ACTUAL OPERATION: FLOW IS PRESSURE DRIVEN. SINCE p3 > p],
FLOW WILL GO OUT OF HOLE 1 RATHER THAN IN,
SAMPLE IS OBTAINED ONLY FROM HOLE 2.
Figure 22. Illustration of multiple point gas
sampling misconception
71
-------
other sampling hole Into the lower velocity region. The result is that
the extracted sample is not the sample which was intended.
In order for a probe of the type shown in Figure 22 to function as
intended, the pressure drop across each sampling hole must be large in
comparison to the free stream pressure variations. In most cases, a
pressure drop across each sampling orifice of about 20 torr will be adequate
to insure proper flow through each orifice. This will typically result in
effective orifice diameters of about one millimeter. Prototype sampling
probes designed on this basis are shown in Figure 23. The probe in Figure
23a was fabricated and successfully used in proof of principle field test
work as described in Reference 2. Its use was limited by the need to clean
or replace the in-line filter used to remove particulate from the stream.
There were no problems associated with clogging of the orifices themselves
despite their small size. More suitable designs are shown in Figures 23b
and c. It is highly recommended that such probes be developed as standard
commercial items. They are designed so that particulate filters act as
the stream interface and provide the pressure drop necessary to insure
proper sampling. If the probes face into the flow, the filter elements
will tend to be self cleaning, since the impacting stream will help to
prevent buildup of particulate on the filter surface. Occasional back
purging without removing the probes could also be used to prevent clogging.
The probes shown in Figure 23 represent a very practical way to per-
form multiple point sampling, especially in continuous monitoring applica-
tions. Since only a single outlet line is involved, they would be no
more difficult to use than single point probes. The validity of the
approach was verified by the work described in Section 4, which showed
that spatial gas sampling is just as accurate for most purposes as flow
proportional sampling. This means that the probes only need to be designed
to insure that the same amount of sample is drawn through each orifice,
which is easy to do. There is no requirement to account for velocity var-
iations along the line of interest. The probes in Figure 23 are all line
averaging devices, which makes them exactly the type of probe desired for
measurements in circular ducts using the tangential method, and in rec-
tangular ducts using the Row Average Method.
72
-------
SAMPLE
OUT
SAMPLE
OUT
SAMPLE
OUT
k
-2.
mm^
i -/
\%
FILTER STRIP
• ^•••••«
2.54
•
\ FILTER PLUG
NOMINAL DIA. 1.3 cm
cm NOM.
\
V
SAMPLING ORIFICE
NOMINAL DIA. 1 mm
// (CONCEPT
/ ONLY)
/, (FIELD
/ DEMON-
' / STRATED)
FILTER
A - PARTICULATE REMOVED BY IN-LINE FILTER
B,C - PARTICULATE REMOVED BY FILTER AT
PROBE INLET
NOMINAL PRESSURE DROP ACROSS INLET = 20 TORR
Figure 23. Multiple point gas sampling probes
73
-------
5.2.3 Sample Trains
Two types of gas sampling have been mentioned in this report: quanti-
tative sampling, where the desired output is a single number, such as m.,
and investigative sampling, where the desired output is a concentration
map. Sampling trains which use continuous gas analyzers can perform both
types of sampling, while wet chemistry trains are usually practical only
for quantitative sampling since evaluation of each individual sample would
be too tedious. Manual traverses using wet chemistry methods can be made
easier by using the spatial sampling approach. By using a constant sampling
flowrate and sampling time at each point, the need for continued flow rate
adjustment and a dry gas meter disappears.
One of the most common problems with continuous gas analyzers is
calibration shifts, which can be as high as several percent over a period
of a few hours. It is imperative that these instruments be used in strict
accordance with the manufacturer's specified procedures. For manual
traverse work, zero and span readings should be taken at the beginning and
end of each work shift, and systematic calibration changes should be
accounted for during data reduction. Calibration gases should be selected
and used in accordance with manufacturer's specifications to avoid errors.
For example, an SOp analyzer may have a very different response to a cali-
bration gas which consists of 1000 ppm of SOg in air than to a gas which
consists of 1000 ppm of SOp in nitrogen.
5.3 PROTOTYPE CONTINUOUS MONITORING SYSTEM
The complexity of continuous monitoring systems is clearly a function
of the required system performance. At a minimum, a given system should
provide continuous data for y~. for at least one gas, such as SOo. The
next step would be to provide a volumetric flow measurement capability,
• • • .
which would then add m, m., and V . For environmental purposes, m. will
usually be the most desirable parameter, since a measurement in mass
per unit time can be easily combined with other data to give information
such as grams of S02 per kilogram of fuel or grams of S02 per kilowatt
produced in a power plant. More sophisticated systems would measure outputs
of additional gases, such as 0« for control of excess air.
74
-------
A full monitoring system for use in a combustion stream is shown in
Figure 24, and nominal prices for system components are given in Table 7.
Of the items mentioned, only the multiple port gas sampling probe is not
presently commercially available. Specific manufacturers and model numbers
are given only to demonstrate that the equipment does exist. Their men-
tion does not constitute a recommendation except for the Ellison and Ramapo
probes, which TRW has specifically evaluated for EPA. It is clear from
Table 7 that the bulk of the cost for a continuous system is for the out-
of-stream components. Also, the out-of-stream components will be almost
identical regardless of whether single or multiple point techniques are
used. The cost of a multiple point measurement system may be perhaps 10%
higher than that of a single point system, but the accuracy of the multiple
point system may be better by as much as a factor of three.
Recent improvements in small computers offer a precise calculation
capability not previously achievable. The most common current practice
is to record instrument outputs directly, as for y. or y~.. Proper calcu-
lation of m, m., or V , as given in equation 20 in Table 1, cannot be done
without processing the outputs of several instruments. The MITS Altair
data processor mentioned in Table 7 is inexpensive enough to be cost effec-
tive in a continuous monitoring system, and has the capability to perform
the calculations required by equation 20. This type of data processing
capability is required to maintain system accuracy. It would be impractical
to record data and process it by hand or even by computer at a later time.
Approximate labor requirements to acquire, install, and use a contin-
uous monitoring system are shown in Table 8. The ranges are necessarily
very broad due to the large number of variables involved, and so the numbers
should only be taken as a very general guide. The specific tasks called
out will usually apply for any given system, and the distribution of hours
among the individual tasks is considered representative. Since the daily
time requirements for normal system operation become the largest manpower
cost factor over a long period of time, it is essential to properly design
the system initially so that the maintenance time requirement will be
minimized. Proper system utilization also requires that a carefully
written operational procedure be produced to optimize the maintenance
schedule, minimize component degradation, and preserve system accuracy.
75
-------
SAMPLE PROBE
r- ANNUBAR
FLOW
^-THERMOCOUPLE
^ DUCT
HEAT TRACED
SAMPLE LINE
SAMPLE OR
PRESSURE LINE
ELECTRIC LINE
(OUTPUT VOLTAGEf
ABSOLUTE
PRESSURE
TRANSDUCES
Ap
TRANSDUCER!
PRINTER
OR STRIP
CHART
DATA
PROCESSOR
AND CLOCK
MONITOR
HYDROCARBON
ANALYZER
MULTIPLEXER
GAS
CONDITIONING
SYSTEM
MONITOR
co2
MONITOR
MONITOR
so2
MONITOR
CO
MONITOR
FLOW
Figure 24. Schematic of a complete continuous monitoring
system for a combustion process stream
76
-------
Table 7. ESTIMATED CONTINUOUS MEASUREMENT SYSTEM HARDWARE COSTS
COMPONENT -
VELOCITY MEASUREMENT
Probes
Ellison Annubar (5m length)
Ramapo Fluid Drag Meter (5m length)
Pitot-Static Probe for reference traverses
Support Equipment
Thermocouple and Cold Junction (Omega-CJ)
Differential Pressure Transducer
Baratron 145B(+10 torr, high accuracy)
Statham PM5TC(+? torr, medium accuracy)
Absolute Pressure Transducer
Statham P822(0-800 torr)
Traversing Mechanism for Ramapo probe
GAS COMPOSITION MEASUREMENT
Probes ,
Multiple point probe (5m length)
Single point probe for manual traverses
Continuous Gas Conditioning System (Beckman)
Gas Analyzers
02 Taylor OA 138
C02 Beckman 865
CO I
NO
S02 ,
H20 Beckman 865
Hydrocarbon Beckman 402
DATA HANDLING, REDUCTION, AND PRINTOUT
MITS Altair 8800 plus support equipment
ESTIMATED
cnsT
$1000
2500
300
150
2800
400
400
^1500
^1500
200
6000
1000
3300
3300
3300
3300
3300
6500
2000
1
cost a direct function of probe length
77
-------
Table 8. ESTIMATED CONTINUOUS MEASUREMENT SYSTEM LABOR REQUIREMENTS
*
ACTIVITY
ACQUISITION (Total)
Formulation of requirements
Site selection
Preliminary measurements for
instrument scaling
System design and component
selection
Equipment purchase
Instrument checkout
INSTALLATION (Total)
Manual traverses for in-stream
component location
System integration, installation,
and checkout
Manual traverses for in-place
calibration
Writing of operational procedure
USE (Total )
System checkout(daily)
Data review(weekly)
Calibration check by manual
traverse (monthly)
General refurbishing(annual)
ESTIMATED LABOR*, HOURS
ENGINEERING
(108-284)
16-40
8-16
4-8
40-120
20-40
20-60
(88-336)
8-16
20-120
20-80
40-120
(1 08-440) /yr
0
1-4
4-16
8-40
TECHNICIAN
(b2-l24J
0
0
8-16
16-32
8-16
20-60
(104-344)
16-40
40-200
*
40-80
8-24
<533-1498)/yr
1-2
1-4
8-40
20-80
Assumes that system measures total volumetric flow and at least
one gas species concentration
78
-------
As was shown in Table 3, a reasonable target for accuracy of a con-
• .
tlnuous system 1s a 2a uncertainty of +9% for m, m., and Vc, and +5% for
—• i s —
y.|. These uncertainties apply to Individual measurements. Over a period
of time, it is expected that the average values would have a smaller error.
For example, for measurements of V on a continuous basis at a coal fired
power plant as described in Reference 1, agreement between the monitoring
system value arid the value calculated from plant operating conditions
(monitored in the control room) was on the order of +3% over a 24 hour
period. When the system uncertainties are truly random, they are expected
to oscillate about the true mean value, so that system accuracy over a
period of time such as a day or a week would be expected to be better than
the single measurement uncertainty.
79
-------
6. PROTOTYPE CONTINUOUS MONITORING PROCEDURES
This section presents recommended general procedures for installation
of continuous monitoring systems. The recommendations are based upon the
methodology and hardware discussed in Sections 4 and 5, and the equations
derived in Section 2. The procedures are a self-contained unit, and may
be used apart from the remainder of the manual. The purpose of the pro-
cedures is to show the kind and extent of activities which are required to
install and calibrate a continuous monitoring system.
80
-------
PROCEDURE 1
FLOW AND GAS COMPOSITION MEASUREMENTS IN CIRCULAR DUCTS
1. Site Selection. When possible, select the sampling plane so that 80%
of the local straight run is upstream of the sampling plane. Select two
orthogonal diameters in the sampling plane. If the upstream flow disturb-
ance involves a change in direction, such as an elbow, one of the diameters
should be in the plane of the disturbance.
2. Survey Traverses. Perform survey traverses for velocity and gas com-
position at as many plant operating conditions as is feasible. Operating
conditions must remain constant during each traverse. Forty point traverses
should be performed (10 points per radius) using the Log Linear Method for
velocity and the tangential method for gas sampling, as given in Table 1-1.
TABLE 1-1
POINT LOCATIONS FOR SURVEY TRAVERSES, IN
FROM INSIDE WALL TO TRAVERSE
Point
Log Linear
Tangential
1 2 34 5
.82 4.4 6.5 9.9 12.7
1.3 3.9 6.7 9.7 12.9
6
16.8
16.5
PERCENT OF DUCT DIAMETER
POINT
7
20.1
20.4
8
25.4
25.0
9
29.9
30.6
10
40.2
38.8
A standard pitot-static probe should be used for the velocity traverse.
Reduce the velocity data using equation 1.1.1 from the appendix and plot
velocity and composition data as a function of distance. Compute gas
stratification levels.
3. Equipment Selection and Installation. Select and acquire hardware for
the continuous monitoring system. If the gas stratification level is too
high to permit single point sampling, use a multi-port probe or equivalent.
If a point sampling velocity technique is selected, points along both
diameters should be selected in accordance with the Log Linear 4 technique
shown in Table 1-2. For gas sampling, eight points should be selected
along one diameter according to the tangential method in Table 1-3. The
diameter selected for gas sampling should be the one along which the
greatest variation in composition is measured during the survey traverses.
81
-------
TABLE 1-2
LOG-LINEAR 4 METHOD FOR VELOCITY MEASUREMENT, LOCATION IN
PERCENT OF DUCT DIAMETER FROM INSIDE WALL TO
SAMPLE POINT
Point
Location
1234
4.3 29.0 71.0 95.7
TABLE 1-3
TANGENTIAL METHOD FOR GAS SAMPLING, LOCATION IN PERCENT
OF DUCT DIAMETER FROM INSIDE WALL TO SAMPLE POINT
Point
Location
12 3.4 5 6 7 8
3.3 10.5 19.4 32.3 67.7 80.6 89.5 96.7
If an Annubar is selected as the flow sensor, it should be installed along
the diameter which showed the greatest variation during the survey traverse.
If the Annubar and gas sample probe are to be installed along the same
diameter, the gas sample probe should be located at least 15 cm downstream
of the Annubar and 3 cm to one side. If they are on different diameters,
the gas sample probe should be at least 15 cm downstream of the Annubar.
The system should be installed and made operational.
4. In Place Calibration. The continuous monitoring system should be cali-
brated in place using 16 point Log-Linear velocity traverses, given in
Table 1-4, and 16 point gas sample traverses using the tangential method
in Table 1-3.
TABLE 1-4
LOG LINEAR 8 METHOD FOR VELOCITY TRAVERSES, LOCATION IN
PERCENT OF DUCT DIAMETER FROM INSIDE WALL TO SAMPLE POINT
Point
Location
1 2345678
2.1 11.7 18.4 34.5 65.5 81.6 88.3 97.9
Calculate volumetric flow according to Appendix equation 1.1.2 and species
emission from Appendix equation 1.2.4. Using monitoring system outputs
obtained during or before and after each traverse, solve Appendix equations
2.1,and 2.3 for the calibration factors. Program the calibration factors
into the data processor, and the system is ready for use.
82
-------
PROCEDURE 2
FLOW AND GAS COMPOSITION MEASUREMENTS IN RECTANGULAR DUCTS
1. Site Selection. The preferred sampling location is downstream of a
mitered elbow. If this type of location is not available, select the
sampling plane so that 80% of the local straight run is upstream of the
sampling plane. For the elbow case, the sample plane should be selected
in accordance with Figure 2-1.
Sample
Plane
.Location'
T
W
I
Flow
A. For Systems Using Annubar, D = 1.5W
.25W
B. For Systems Using Row
Average Method for
Velocity
D = .8W
+ .40W
-.60W
Figure 2-1. Recommended sample plane locations downstream
of a rectangular elbow
2. Survey Traverses. Perform survey traverses for velocity and gas com-
position at as many plant operating conditions as is feasible. Operating
conditions must remain constant during each traverse. Traverse should be
performed using an n x n matrix with 6
-------
Stde
B
Figur
Row
e 2-
i I i
* ' , \
-*•-!• 1-
•• ! • ] -K.
1 _l_ I
r t 1 — — — •
1 1
' 1 ' 1 '
! . '
1 * ' ' 1
,^_. Rnw Rl
Area Segment
Sample Point
Side A
Al
2. Illustration of Centroid of Equal Areas Technique
for Rectangular Ducts
A standard pi tot-static probe should be used for the velocity traverses.
Reduce the velocity data using Appendix equation 1.1.1. Compute the
volumetric flow rates from Equation 1.1.2 and the average mole fractions
from 1.2.1 or 1.2.2, and the gas stratification levels.
3. Equipment Selection and Installation. Select the instruments to be
used. If single point gas sampling is not acceptable and/or the Row
Average Method or Annubar is to be used, the survey traverse data must be
examined to determine the required probe length (equal to the length of
side A or B in Figure 2-2). Compute the average velocity and/or concen-
tration as required along each row, as identified in Figure 2-2, for each
traverse and normalize the row average with respect to the overall average.
Plot the results versus distance as shown in Figure 2-3.
Normalized
Row
Average
i.o 4^=*-^=^:=^-
,/Row Average Data Point
Curve Fit
I
100
Length in % of Duct Dimension
Figure 2-3. Typical plot of normalized row averages
"84
-------
When possible, a row for actual sensor Installation should be selected
from the smoother of the curves (the A curve in Figure 2-3) since more
repeatable data will be obtained in that direction. For the case in
Figure 2-3, the preferable row would be an A row, meaning that the probe
length would be equal to that of side B. In the case of velocity measure-
ments downstream of an elbow, the desired row direction will be in the
plane of the elbow.
Once the row direction has been chosen, the actual row location
must be selected. The Annubar should always be placed in the middle of
the duct. For the Row Average Methods, if four or more survey traverses
have been performed, select the row location at which the best normalized
row average repeatability was achieved. For three or less traverses,
select the row location at which the normalized row average was closest
to unity.
Obtain and install system hardware and activate the system. A multi-
port gas sample probe, if used, should have 8 equally spaced ports.
4. In Place Calibration. The continuous monitoring system should be cali-
brated in place using 4x4 traverses as illustrated in Figure 2-2.
Calculate volumetric flow according to Appendix equation 1.1.2 and species
emission from Appendix equation 1.2.4. Using monitoring system outputs
obtained during or before and after each traverse, solve Appendix equations
2.1 and 2.3 for the calibration factors. Program the calibration factors
into the data processor, and the system is ready for use.
85
-------
APPENDIX
FLOW AND GAS SAMPLE CALCULATIONS
1. MANUAL TRAVERSES
1.1 FLOW TRAVERSE USING PITOT-STATIC PROBE
1.1.1 Point Velocity at Standard Conditions
"s = ^ ^f"49'74 VS1" •«*
where
m
u = velocity at standard conditions,
w O C w
T = standard temperature, 760 torr
Ps = standard pressure, 293.16°K
.. 2
R = universal gas constant, 8314.32 —gm m 2
mole sec °K
p = static pressure, torr
oo
Ap = differential pressure from pi tot-static probe, torr
M = average molecular weight(wet),
T^ = static temperature, °K
1.1.2 Volumetric flow at standard conditions
or
where
V = total volumetric flow at st«
2
o
V = total volumetric flow at standard conditions, m /sec
A = duct cross-sectional area, m
N = total number of traverse points
86
-------
1.1.3 Total gaseous Mass Flow Rate
N
5 n=l
or
m = 2067.8 f|- ^MP^-V yillL , gm/sec
n=l T»n
where
m = total gaseous mass flow rate, gm/sec
1.2 GAS SAMPLE TRAVERSE
1.2.1 Average Species Mole Fraction, Dry
R
1
\
where
y. = point mole fraction, dry, moles i
D mole of dry mixture
P. = average mole fraction, dry, moles i
D mole of dry mixture
1.2.2 Average Species Mole Fraction, Wet
N
n=l
or
Tw "2
where
y. = point mole fraction, wet, moles i
w mole of wet mix
y. = average mole fraction, wet, moles i
w mole of wet mix
87
-------
1.2.3 Average Molecular Weight in a Combustion Stream
M = (1-7H 0) I":32i70 + 44.0l7CQ + .04 + 28.01/,99-y- -7CQ "
2 L 2D 2D \ 2D 2D
+ 18'01*H20
1.2.4 Gaseous Species Average Mass Flow Rate
. N
M-
m. = m — ]T. =
1 M nw
or
AM
i, • 2067.8 -j. ^ I^JZ" 1^, ^,/sec
\n=l ' • n /
where
mi = species mass flow rate, gm/sec
M. = molecular weight of species i, gm/mole
2. CONTINUOUS MONITORING
2.1 VOLUMETRIC FLOW AT STANDARD CONDITIONS
2.1.1 Annubar as Flow Sensor
VP APfl o
-=-^ , nr/sec
MTA
where
= Annubar calibration factor (nominal .65)
= Annubar differential pressure, torr
T. = Representative stream temperature, °K
2.1.2 Point Sensor Array or Automatic Traversing Point Sensor; Ramapo
Fluid Drag Meter Used as Example
^ , m3/sec
n=l,
88
-------
where
Ramapo calibration factor
V^ = Ramapo voltage ratio, dimension!ess
N = total number of points used (8 nominal)
2.2 Total Gaseous Mass Flow Rate
• •
m = 41.57 M V , gm/sec
2.3 Species Mass Flow Rate
m. = m (—) £. K , gm/sec
> M / \ wi
where
K = continuous system species calibration factor; ideally
i
89
-------
REFERENCES
1. "Continuous Measurement of Total Gas Flowrate from Stationary
Sources," EPA-650/2-75-020, Brooks, et al.. February, 1975.
2. "Continuous Measurement of Gas Composition from Stationary Sources,"
EPA-600/2-75-012, Brooks, et al., July, 1975.
3. "Guidelines for Stationary Source Continuous Gas Monitoring Systems,"
E. F. Brooks, TRW Report 24916-6027-RU-OO, November, 1975.
4. Foundation of Aerodynamics, A. M. Kuethe and J. D. Schetzer;
John Wiley & Sons, Inc., 1959.
5. An Introduction to Thermodynamics, the Kinetic Theory of Gases, and
Statistical Mechanics, F. W. Sears; Addison-Wesley Publishing
Company, Inc., 1953.
6. "Standard of Performance for New Stationary Sources", Environmental
Protection Agency; Federal Register, Vol. 36, No. 159, Part II,
9-17-71.
7. The Analysis of Physical Measurements, E. Pugh and G. Winslow;
Addison-Wesley, 1966.
8. Flue Gas Monitoring Techniques, J. N. Driscoll; Ann Arbor Science
Publishers Inc., 1974.
9. "Procedures for Measurement in Stratified Gases," EPA-650/2-74-086-a
and -b, Zakak, et al., September, 1974.
10. "A Simplified Integration Technique for Pipe Flow Measurement,"
F. A. L. Winternitz and C. F. Fischl; Water Power. Vol. 9, page 225,
1957.
11. Industrial Source Sampling; D. L. Brenchley, C. D. Turley, and
R. F. Yarmac; Ann Arbor Science Publishers Inc., 1973.
12. The Measurement of Air Flow. E. Ower and R. C. Pankhurst; Pergamon
Press, 1966.
13. Fluid-Dynamic Drag, S. F. Hoerner; published by the author, 1958.
14. Statham Instruments, Inc. Product Bulletin AM117, December 1973.
15. MKS Instruments, Inc. Bulletin 145B, September 1974.
90
-------
GLOSSARY
SYMBOL
A
C
D
f
G
H
~ I
K
r
M
m
N
->•
n
PO
Poo
AP
R
U
if
u
vs
X
x
y
e
yi
P
a
USAGE
area
constant
diameter
functional relationship
general parameter
general variable
total number of gas species
calibration factor
length
molecular weight
mass flow rate
total number of area segments
unit normal vector
stagnation pressure
static pressure
differential pressure
universal gas constant
static temperature
magnitude of velocity vector
velocity vector
axial velocity component
total volumetric flowrate at
standard conditions
general parameter
length
length
flow orientation angle
mole fraction
gas density
standard deviation
DIMENSIONS
m2
dimension! ess
m
dimensionless
m
gm/mole
gm/sec
dimensionless
dimensionless
torr
torr
torr
2
gm m
2
mole sec °K
°K
m/sec
m/sec
m/sec
3
m/sec
m
m
radians
moles of i/mole
gm/m
91
-------
GLOSSARY (Cont'd)
SYMBOL USAGE DIMENSIONS
OSE total system uncertainty percent
o_pE single point measurement percent
uncertainty
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/2-76-203
3. RECIPIENT'S ACCESSION NO.
4. TITLE. AND SUBTITLE
Flow and Gas Sampling Manual
5. REPORT DATE
July 1976
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
E. F. Brooks and R. L. Williams
8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORGANIZATION NAME AND ADDRESS
TRW Systems Group
One Space Park
Redondo Beach, CA 90278
10. PROGRAM ELEMENT NO.
1AB013; ROAP 21ACX-AE
11. CONTRACT/GRANT NO.
68-02-1412, Task 13
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Task Final: 3-11/75
14. SPONSORING AGENCY CODE
EPA-ORD
^.SUPPLEMENTARY NOTES project officer for this report is W.
Ext 2557.
B. Kuykendal, Mail Drop 62,
16 ABSTRACTThe manual summarizes work done on a project to measure volumetric
flowrate and composition in gas-phase process streams. It is intended for use by
measurement professionals to apply the developed techniques for continuous
measurements of volumetric flowrate. Also presented are techniques for extracting
a representative gas sample from a process stream. Information is given for
selecting the optimum number of sampling points to achieve reasonable accuracy in
a cost effective manner.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATl Field/Group
Air Pollution
Measurement
Gas Sampling
Flowrate
Gas Analysis
Industrial Processes
Air Pollution Control
Stationary Sources
Process Streams
13B
14B
20D
13H
8. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
100
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
93
------- |