Abstrict A system and method for correcting a meter calibration curve in
a fuel dispenser. The meter may be a turbine flow meter. A meter
calibration curve is expressed in a finite number of data points.
The meter calibration curve is used to determine the volume and/or
flow rate of liquid flowing through the meter. Since the meter calibration
curve may contain errors, the meter calibration curve is corrected
by determining poles in the meter calibration curve for given known
flow rates and/or volumes of liquid passing through the meter, determining
an error correction at each of the poles in the meter calibration
curve, correcting the expanded meter calibration curve at each of
the poles, using linear interpolation to correct points in the meter
calibration curve adjacent to the poles.
Claims I claim:
1. A method of correcting a calibration curve of a meter used to
measure fuel in a fuel dispenser, comprising the steps of: (a) receiving
a meter calibration curve used to determine flow rate or volume
of liquid passing through the meter expressed in a finite number
of data points; (b) determining the location of a plurality of poles
on said meter calibration curve by determining an expected value
on said meter calibration curve for a plurality of flow rates or
volumes of liquid flowing through the meter; (c) computing error
in said meter calibration curve at the location of said plurality
of poles by determining the difference between the flow rate or
volume of liquid indicated by said meter calibration curve at the
location of said plurality of poles and said plurality of flow rates
or volumes for said plurality of poles; (d) computing an error correction
for said meter calibration curve at each of said plurality of poles
based on said computed error in said step (c); and (e) correcting
said meter calibration curve to form a corrected curve by applying
said error correction to said meter calibration curve at the location
of said plurality of poles and using linear interpolation to correct
said meter calibration curve at points adjacent said plurality of
poles.
2. The method of claim 1 further comprising expanding said meter
calibration curve using linear interpolation to create an expanded
curve having a greater number of data points than said meter calibration
curve before said steps (b) (e), wherein: said step (b) comprises
determining the location of a plurality of poles on said expanded
curve by determining an expected value on said expanded curve for
a plurality of flow rates of liquid flowing through the meter; said
step (c) comprises computing error in said expanded curve at the
location of said plurality of poles by determining the difference
between the flow rate or volume of liquid indicated by said expanded
curve at the location of said plurality of poles and said plurality
of flow rates or volumes for said plurality of poles; and said step
(e) comprises correcting said expanded curve to form a corrected
curve by applying said error correction to said expanded curve at
the location of said plurality of poles and using linear interpolation
to correct said expanded curve at points adjacent said plurality
of poles.
3. The method of claim 2 further comprising expressing said corrected
curve in a number of finite points equal to the number of finite
points of said meter calibration curve after said step (e).
4. The method of claim 1 wherein said meter calibration curve
is received from memory in the meter in said step of receiving.
5. The method of claim 1 further comprising loading said corrected
curve into memory in the meter.
6. The method of claim 1 wherein said meter calibration curve
is a Roshko-Strouhal curve.
7. The method of claim 1 wherein said step (d) comprises computing
an error correction coefficient for said meter calibration curve
at each of said plurality of poles based on said computed error
in said step (c).
8. The method of claim 7 wherein said step (e) is comprised of
multiplying said error correction coefficient with said meter calibration
curve at the location of said plurality of poles and using linear
interpolation to correct said meter calibration curve at points
adjacent said plurality of poles.
9. A system for correcting a calibration curve of a meter used
to measure fuel in a fuel dispenser, comprised of: a control system,
wherein said control system: (a) receives a meter calibration curve
used to determine flow rate or volume of a liquid flowing through
the meter expressed in a finite number of data points; (b) determines
a location of plurality of poles on said meter calibration curve
by determining an expected value on said meter calibration curve
for a plurality of flow rates or volumes of liquid flowing through
the meter; (c) computes error in said meter calibration curve at
the location of said plurality of poles by determining the difference
between the flow rate or volume of liquid indicated by said meter
calibration curve at the location of said plurality of poles and
said plurality of flow rates or volumes for said plurality of poles;
(d) computes an error correction for said meter calibration curve
at each of said plurality of poles based on said computed error;
and (e) corrects said meter calibration curve to form a corrected
curve by applying said error correction to said meter calibration
curve at the location of said plurality of poles and using linear
interpolation to correct said meter calibration curve at points
adjacent said plurality of poles.
10. The system of claim 9 wherein said control system expands
said meter calibration curve using linear interpolation to create
an expanded curve having a greater number of data points than said
meter calibration curve before performing said functions (b) (e),
wherein: said function (b) comprises said control system determining
the location of a plurality of poles on said expanded curve by determining
an expected value on said expanded curve for a plurality of flow
rates of liquid flowing through the meter; said function (c) comprises
said control system computing error in said expanded curve at the
location of said plurality of poles by determining the difference
between the flow rate or volume of liquid indicated by said expanded
curve at the location of said plurality of poles and said plurality
of flow rates or volumes for said plurality of poles; and said function
(e) comprises said control system correcting said expanded curve
to form a corrected curve by applying said error correction to said
expanded curve at the location of said plurality of poles and using
linear interpolation to correct said expanded curve at points adjacent
said plurality of poles.
11. The system of claim 10 wherein said control system expresses
said corrected curve in a number of finite points equal to the number
of finite points of said meter calibration curve.
12. The system of claim 9 wherein said control system receives
said calibration curve from memory in the meter.
13. The system of claim 9 wherein said control system loads said
corrected curve into memory in the meter.
14. The system of claim 9 wherein said meter calibration curve
is a Roshko-Strouhal curve.
15. The system of claim 9 wherein said control system computes
an error correction at each of said plurality of poles by computing
an error correction coefficient for said meter calibration curve
at each of said plurality of poles based on said computed error.
16. The system of claim 15 wherein said control system multiplies
said error correction coefficient with said meter calibration curve
at the location of said plurality of poles and uses linear interpolation
to correct said meter calibration curve at points adjacent said
plurality of poles.
17. A method of correcting a Roshko-Strouhal curve of a turbine
flow meter used to measure fuel in a fuel dispenser, comprising
the steps of: (a) receiving a Roshko-Strouhal curve used to determine
flow rate of liquid passing through the turbine meter expressed
in a finite number of data points; (b) expanding said Roshko-Strouhal
curve using linear interpolation to create an expanded Roshko-Strouhal
curve having a greater number of data points than said Roshko-Strouhal
curve; (c) determining a location of a plurality of poles on said
expanded Roshko-Strouhal curve by determining an expected value
on said expanded Roshko-Strouhal curve for a plurality of flow rates
of liquid flowing through the turbine meter; (d) computing error
in said expanded Roshko-Strouhal curve at the location of said plurality
of poles by determining the difference between the flow rate or
volume of liquid indicated by said expanded Roshko-Strouhal curve
at the location of said plurality of poles and said plurality of
flow rates or volumes for said plurality of poles; (e) computing
an error correction for said expanded Roshko-Strouhal curve at each
of said plurality of poles based on said computed error in step
(d); (f) correcting said expanded Roshko-Strouhal curve to form
a corrected expanded Roshko-Strouhal curve by applying said error
correction to said expanded Roshko-Strouhal curve at the location
of said plurality of poles and using linear interpolation to correct
said expanded Roshko-Strouhal curve at points adjacent said plurality
of poles; (g) expressing said corrected expanded Roshko-Strouhal
curve in a number of finite points equal to the number of finite
points of said Roshko-Strouhal curve to form a corrected Roshko-Strouhal
curve; and (h) storing said corrected Roshko-Strouhal curve in memory
of the turbine meter.
Description FIELD OF THE INVENTION
The present invention relates to a process and method to correct
error in a meter calibration curve in a meter used in a fuel dispenser
to measure fuel, and in particular to correct error in a Roshko-Strouhal
calibration curve for a dual rotor turbine flow meter.
BACKGROUND OF THE INVENTION
A turbine flow meter can be used to measure the flow rate of a
liquid. Flow rate information can be used to derive the volume of
the liquid to accomplish a volumetric meter. One example of a turbine
flow meter is described in U.S. Pat. No. 5689071 (hereinafter
the "'071 patent." In the '071 patent, two turbine rotors
are contained within the liquid flow path of a meter housing. The
turbine rotors rotate as liquid passes across the rotors. The liquid
passes through the first turbine rotor and is directed into the
second turbine rotor such that the second turbine rotor rotates
in a direction opposite from the first turbine rotor.
Another example of a dual rotor turbine flow meter is described
in U.S. Pat. No. 5831176 (hereinafter the "'176 patent").
In the '176 patent, a description is provided on how the flow rate
of liquid is derived from information concerning the rotation of
the turbine rotors. During calibration of the turbine flow meter,
a known volumetric flow rate of liquid is placed through the meter.
As the liquid flows through the meter, the liquid passes through
the rotors causing the rotors to rotate. The rotational frequency
of the turbine rotors is measured at various flow rates and frequencies
to arrive at a "Strouhal" number for each turbine rotor.
The Strouhal number is the frequency of the rotor divided by the
volumetric flow rate as follows: ##EQU00001##
The corresponding "Roshko" numbers for each of the Strouhal
numbers are determined for each of the turbine rotors by dividing
the frequency of the rotor by the viscosity of the liquid, as follows:
##EQU00002##
The Strouhal and corresponding Roshko numbers are plotted on a
Roshko-Strouhal (R-S) curve and/or are stored in an array of finite
points with the Strouhal numbers being plotted in one axis or an
array, typically in the y-axis, and the corresponding Roshko numbers
being stored in another axis or corresponding array, typically the
x-axis. An example of a R-S curve is illustrated in FIG. 2.
During operation, when the R-S curve is used to determine a Strouhal
number from a calculated Roshko number as discussed below, linear
interpolation is used to determine numbers that fall in between
the finite points of the R-S curve. However, since the R-S curve
is not a linear function, linear interpolation will introduce error.
If the R-S curve could be converted into a perfect equation instead
of being stored as finite points such that interpolation was not
required for its use, error would not exist as a result of interpolation.
However, the R-S curve is a complex curve that cannot be easily
described by a linear or complex order equation. Further, some processes
involve the manipulation of raw Roshko and Strouhal coordinate points
calculated during calibration of a meter to remove extraneous points
that would cause linear interpolation to introduce even more error.
However, even with such manual manipulation, error will still exist
to some degree in the R-S curve.
During operation of the meter, the liquid flow rate passing through
the meter can be determined using the Roshko and Strouhal data from
the R-S curve. If there is any error in the R-S curve, this error
will be propagated to the volumetric flow rate calculation that
is made by the meter during operation. The volumetric flow rate
of liquid flowing through the turbine meter is determined as follows:
First, the rotational frequencies of the turbine rotor are measured.
As discussed in the '071 patent, pick-off coils or other sensing
devices, such as Hall-effect sensors for example, are employed in
the turbine meter to detect the rotation of the turbine rotors.
The detection device detects the movement of each blade on the turbine
rotor and can therefore determine the frequency of rotation as is
described in the '071 patent. Once the rotation frequencies of the
turbine rotors are measured, the Roshko number for each rotor can
be determined according the formula for the Roshko number shown
above. After the Roshko number is calculated, the corresponding
Strouhal number is determined by the R-S curve or equation. The
Strouhal number and the frequency of the turbine rotor are then
used to determine the volumetric flow rate according to the rearranged
Strouhal formula below. The Strouhal numbers of each rotor may be
combined to use as the Strouhal number in the equation below: ##EQU00003##
The volumetric flow rate calculation is repeated continuously in
periodic time increments so that the volumetric flow rate of liquid
flowing through the turbine meter is known at any given time. The
volume of the liquid can be derived from the volumetric flow rate
using time as is well known.
If the turbine flow meter is used in an application in which the
liquid flow is distributed aperiodically and frequently, such as
in a fuel dispenser where a customer can constantly change the flow
of fuel delivery by engaging and disengaging the fuel nozzle, it
is more difficult to accurately measure flow rate and volume. Disruptions
in the liquid flow result in disruptions in the rotational frequency
of the turbine rotors, which in turn affect the calculation of the
Roshko numbers. Further, since the R-S curve is derived data that
forms a curve where an equation cannot be formulated to perfectly
match the curve, approximations of the Strouhal number using the
calculated Roshko number will cause error as well. When the R-S
curve is determined during calibration, a line is fitted through
the raw Strouhal and corresponding Roshko data points, and the process
of line fitting introduces error since extraneous points may alter
the final R-S curve. The liquid flow rate variations cannot be controlled,
but derivation of the R-S curve is a controlled operation. If the
accuracy of the R-S curve can be improved to reduce inherent errors
present, the accuracy of the volumetric flow rate calculation will
also improve.
Therefore, there exists a need to find a technique and method to
more accurately approximate data points in an automated fashion
where a perfect equation cannot be derived to match the data points
and thus finite points of data are used. In the preferred embodiment,
the data points are the Strouhal numbers of the R-S curve since
error in the Strouhal numbers will cause error in flow rate calculations
of a turbine flow meter.
SUMMARY OF THE INVENTION
A system and method of correcting a calibration curve of a meter
used to measure fuel in a fuel dispenser. The meter may be a turbine
flow meter that measures the flow rate of a liquid passing through
the meter. A meter calibration curve is received. The meter calibration
curve can be used to determine the flow rate or volume of liquid
passing through the meter. The meter calibration curve is determined
beforehand through a calibration process. The meter calibration
curve is expressed in a finite number of data points. The meter
calibration curve may contain errors.
A plurality of poles of meter calibration curve are determined
by a control system by determining an expected value on the meter
calibration curve for a plurality of flow rates or volumes of liquid
flowing through the meter. Next, the error in the meter calibration
curve is computed by determining the difference between the meter
calibration curve and the plurality of poles. An error correction
at each of said plurality of poles is next calculated. The meter
calibration curve is corrected to form a corrected curve by applying
the error correction to the meter calibration curve at the plurality
of poles and using linear interpolation to correct the meter calibration
curve at points adjacent to the plurality of poles. The error correction
may be an error correction coefficient or an adder/subtracter.
In an alternative embodiment, the meter calibration curve is expanded
first using linear interpolation to create an expanded curve having
a greater number of data points than said meter calibration curve
before the error correction is performed. The corrected curve may
be reduced back to the same number of finite data points as present
in the original uncorrected meter calibration curve at the end of
the correction process.
The meter calibration curve may be stored in memory in the meter
and obtained from the memory before being corrected, or the meter
calibration curve may be obtained from a system in the calibration
process. The corrected meter calibration curve may be stored in
the memory of the meter when completed or in a control system in
communication with the meter to be used to perform flow rate and/or
volume calculations regarding liquid passing through the meter.
BRIEF DESCRIPTION OF THE DRAWINGS
The accompanying drawing figures incorporated in and forming a
part of this specification illustrate several aspects of the invention,
and together with the description serve to explain the principles
of the invention.
FIG. 1 illustrates a block diagram of a turbine flow meter for
measuring liquid in the prior art;
FIG. 2 illustrates a Roshko-Strouhal (R-S) calibration curve having
forty coordinate points that is used by a turbine flow meter to
measure flow rate of a liquid for given turbine meter rotor frequencies
and viscosity of the liquid;
FIGS. 3A and 3B is a flowchart that contains the operation of the
present invention to eliminate error in a R-S curve;
FIG. 4 is an illustration of an expanded R-S curve to provide 3901
coordinate points linearly interpolated from the R-S curve illustrated
in FIG. 2;
FIG. 5 is an illustration of linear interpolation between two segments
of the R-S curve illustrated in FIG. 2 to formulate the expanded
R-S curve illustrated in FIG. 4;
FIG. 6 is an illustration of an error correction array containing
correction poles at various Roshko numbers indicative of different
flow rates of liquid;
FIG. 7 is an illustration of a corrected expanded R-S curve derived
using error corrections illustrated in FIG. 6 as applied to the
expanded R-S curve illustrated in FIG. 4; and
FIG. 8 is an illustration of a reduced corrected R-S curve that
is used by the turbine meter to derive flow rate of liquid, and
is formed by reducing the corrected expanded R-S curve illustrated
in FIG. 7 from 3901 coordinate points to 40 coordinate points.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
The embodiments set forth below represent the necessary information
to enable those skilled in the art to practice the invention and
illustrate the best mode of practicing the invention. Upon reading
the following description in light of the accompanying drawing figures,
those skilled in the art will understand the concepts of the invention
and will recognize applications of these concepts not particularly
addressed herein. It should be understood that these concepts and
applications fall within the scope of the disclosure and the accompanying
claims.
Throughout the description of the present invention below, references
are made to source code contained in Appendix A to this application,
which forms part of the detailed description of the invention and
is incorporated herein.
The preferred embodiment of the present invention is a process
and technique to perform error correction on a Roshko-Strouhal (R-S)
curve or array that is used for measurement of flow rate of liquid
flowing through a turbine meter. The original R-S curve is a complex
curve that is derived empirically and is approximated using discrete
data points. The R-S curve is used to derive flow rate of liquid
by determining a Roshko number and then correlating the Roshko number
to a Strouhal number that can be used to calculate flow rate as
previously described in the "Background of the Invention"
section above. The R-S curve is determined during a calibration
step and may contain error since approximations are used to represent
the R-S curve.
In the present invention, the Strouhal numbers in the R-S curve
are verified against correct Strouhal numbers calculated in a separate
verification step where known volumes of liquid are passed through
the turbine meter at known flow rates. The correct Strouhal values
are compared to Strouhal numbers stored in the R-S curve to determine
an error correction factor for each Strouhal number data point.
Linear interpolation is used to form an error correction curve that
is multiplied against the original R-S curve to form a corrected
R-S curve. The error correction data point is applied to the original
R-S curve to form a corrected R-S curve. The points of the corrected
R-S curve are then used going forward for obtaining the Strouhal
numbers that are used to calculated flow rate. The details and aspects
of the present invention are described below.
FIG. 1 illustrates a block diagram of a turbine flow meter 1 that
is used to measure the flow rate of liquid. The turbine flow meter
1 may be the turbine flow meter described in U.S. Pat. No. 5689071
(hereinafter the "071 patent") incorporated herein by
reference in its entirety. The error correction aspects of the present
invention that are used by the turbine flow meter 1 are described
below starting with FIG. 3A. As can be seen from FIG. 1 liquid
flow enters the turbine meter 1 on one end and exits the turbine
flow meter 1 on the other end. As the liquid flow passes through
the meter 1 two turbine rotors (not shown), like those disclosed
in the '071 patent, rotate. Pick-off coils or other detection devices
detect the rotation of the turbine rotors and provide the frequency
of the rotation in the form of rotor data 4. The turbine flow meter
1 contains a calculator 5 which may be a microcontroller, microprocessor,
or other electronic circuitry, that receives the rotor data 4 and
in turn calculates the liquid flow rate and volume of liquid passing
through the meter 1.
The meter 1 also includes memory 2 that is accessed by the calculator
5. During calibration of the meter 1 the Roshko-Strouhal (R-S)
curve 10 for each rotor may be loaded into memory 2 via an interface
3. The R-S curve 10 for each rotor is determined using the technique
described in the "Background of the Invention" section
above. The R-S curve 10 is used by the calculator 5 to determine
the volumetric flow rate (Vr) of the liquid flowing through the
meter and in turn calculate the volume (V) of the liquid. The calculator
5 communicates the volume (V) of the liquid to another system. In
the example of a fuel dispenser (not shown), the meter 1 would communicate
the volume of liquid data to a control system or other electronic
circuitry to control the volume display on the dispenser indicating
the volume of fuel dispensed. An example of a fuel dispenser and
how meters are used to determine volume for display to a customer
is contained in U.S. Pat. Nos. 6721669; 6067476; 6065638;
5630528; 4938054; 4876653; and 6651517 all of which are
incorporated by reference herein in their entireties.
An example of a R-S curve 10 for one of the rotors of the turbine
flow meter 1 determined during calibration of the meter 1 is illustrated
in FIG. 2. This R-S curve 10 is used to determine the volumetric
flow rate of liquid passing through the meter 1 as discussed in
the "Background of the Invention" section above, and is
known in the prior art.
The R-S curve 10 is shown in a graphical representation of a line
in two-dimensions, where for a given Roshko number in the x-axis,
a corresponding Strouhal number in the y-axis can be ascertained.
Since the R-S curve 10 is determined for each meter 1 and its rotors
specifically, there are two R-S curves 10 unique to a particular
meter 1 one for each rotor.
The R-S curve 10 for each rotor is stored in memory 2 in the form
of an array having a finite amount of coordinate data points. In
one embodiment, the R-S curve 10 for one rotor is stored in two
separate one-dimensional arrays, with one array containing the Roshko
numbers and the other array containing the corresponding Strouhal
numbers. The same index into the arrays will provide correlating
Roshko and Strouhal numbers. In this example, there are four R-S
arrays, two for the first rotor, referred to as "rotor A,"
and two for the second rotor, referred to as "rotor B."
The R-S array size is forty points meaning that the R-S curve will
be stored in a two-dimensional array having a size of forty coordinate
points. Linear interpolation can be used to derive Roshko and Strouhal
numbers that fall in between coordinate data points as illustrated
in the R-S curve 10 in FIG. 2. Examples of the Roshko number array
and a corresponding Strouhal number array are contained in source
code in Appendix A and as shown below. Note that the Roshko numbers
for each rotor are the same, so one array, "rotor.sub.--X,"
is used to store the Roshko numbers for each rotor.
TABLE-US-00001 double rotor.sub.--X [40]; // Roshko numbers for
each rotor // double rotorA.sub.--Y [40]; // Strouhal numbers for
rotor A // double rotorB.sub.--Y [40]; // Strouhal numbers for rotor
B //
As illustrated in FIG. 2 there are forty data points on the R-S
curve 10. The first point, point 0 is referred to as point 12 is
comprised of the Roshko and Strouhal numbers in the "0"
index into the R-S arrays referenced above. In this embodiment,
the Roshko numbers for each rotor are the same, but the corresponding
Strouhal numbers may be different. If FIG. 2 were the R-S curve
10 for "rotor A", the "rotor.sub.--X" array
and "rotorA.sub.--Y" array are indexed to find a point
on the R-S curve 10. Point 18 is the fortieth or last point in the
R-S curve 10. Point 18 is referred to as "point 39" since
the index into the R-S arrays ranges from "0" to "39"
comprising a total of forty points. Note that the arrays that approximate
the R-S curve 10 could be comprised or more or less data points
according to design preference.
Examples of Roshko and Strouhal numbers that may be used to approximate
a R-S curve 10 for rotor A and rotor B are shown below and are contained
in the source code in Appendix A. Note that the "rotor.sub.--X"
array contains Roshko numbers for both rotor A and rotor B.
TABLE-US-00002 double rotor.sub.--X[ ] = // Roshko values, 40 pts
// { 1 1.1937766 1.4251027 1.7012543 2.0309176 2.424462 2.8942661
3.4551073 4.1246264 4.9238826 5.8780161 7.0170383 8.3767764
10 11.937766 14.251027 17.012543 20.309176 24.24462 28.942661
34.551073 41.246264 49.238826 58.780161 70.170383 83.767764
100 119.37766 142.51027 170.12543 203.09176 242.4462 289.42661
345.51073 412.46264 492.38826 587.80161 701.70383 837.67764
1000 }; double rotorA.sub.--Y[ ] = // rotor A Strouhal values, 40
pts // { 1405.6828 1460.3792 1513.2728 1563.6144 1610.9346
1654.8483 1694.9973 1731.0703 1762.8546 1790.2879 1813.4942
1832.7913 1848.6715 1861.7561 1872.7321 1882.2804 1891.0044
1899.3688 1907.6551 1915.9424 1924.1126 1931.8831 1938.8634
1944.6291 1948.8052 1951.1497 1951.6237 1950.4391 1948.0697
1945.2206 1942.7487 1941.5366 1942.329 1945.5498 1951.1312
1958.4024 1966.1008 1972.5915 1976.4061 1977.2361 }; double
rotorB.sub.--Y[ ] = // rotor B Strouhal values, 40 pts // { 1661.8836
1751.0332 1820.3686 1878.71 1929.7376 1974.4551 2012.7645
2044.3883 2069.3314 2088.0332 2101.3316 2110.3256 2116.2034
2120.0781 2122.8601 2125.1773 2127.3481 2129.4007 2131.1304
2132.1812 2132.1374 2130.6145 2127.3337 2122.1761 2115.2072
2106.6729 2096.968 2086.5839 2076.0427 2065.8307 2056.3406
2047.8352 2040.4398 2034.1672 2028.9725 2024.8241 2021.7645
2019.9189 2019.3894 2019.9474 };
The prior art stops at this point and uses the R-S curve 10 as
illustrated in FIG. 2 to determine the volumetric flow rate and
volume of liquid passing through the meter 1. If the R-S curve 10
contains errors, these errors will propagate to the final volumetric
flow rate calculation. Any R-S curve that is stored in the form
of finite coordinate points and not as an equation will always contains
some degree of error. Several factors cause the R-S curve to have
greater error than desired. For instance, interpolation of Roshko
and Strouhal numbers that fall in between data points in the array
of the R-S curve 10 introduce error. Extraneous data points from
the raw Roshko and Strouhal numbers may have been present during
calibration that cause linear interpolation to be even more erroneous.
Even if manual manipulation of the raw Roshko and Strouhal numbers
is performed to remove extraneous data points, such manual manipulation
cannot eliminate error caused by linear interpolation.
Therefore, because of the error in the R-S curve 10 or any curve
that is used to approximate a measurement, such as for a flow meter
in the present example, the present invention presents a technique
and method that may be used to reduce the error present in data
points, such as data points that approximate a R-S curve 10 when
linear interpolation is used to approximate a valve in the curve.
A flowchart of the methods and techniques employed by the present
invention is illustrated in FIGS. 3A and 3B. FIGS. 4 8 illustrate
R-S curves that result from the operations of the steps set forth
in FIGS. 3A and 3B. Therefore, as reference is made to FIGS. 3A
and 3B below, reference will also be made to FIGS. 4 8 to show the
results of the operations performed by the present invention. Note
that although the present invention is described with respect to
a R-S curve 10 used to determine flow rate of a liquid passing through
a turbine flow meter 1 the present invention may be employed for
any type of data points or curve that is used for determination
of values, such as measurement values for example.
As illustrated in FIG. 3A, the process starts (step 100), and the
R-S curve 10 for each rotor of the turbine rotor meter 1 is received
via the interface 3 and retrieved by the calculator 5 from memory
2 (step 102). The R-S curve 10 for each rotor has been previously
calculated in a calibration process of the meter 1 as described
above. The calculator 5 then uses the R-S curve 10 for each rotor
stored in memory 2 to perform the present invention. As previously
discussed and as illustrated in FIG. 2 the R-S curve 10 is comprised
of forty discrete coordinate points in this example.
Next, the R-S curve 10 for each rotor is expanded by a factor of
100 to a total of 3901 points using linear interpolation to form
an expanded R-S curve 20 as illustrated in FIG. 4 (step 104). In
one embodiment, step 102 is performed by the "ExpandRoskhoStrouhal
( );" C++ function in the source code in Appendix A and as
contained below, to produce three arrays to form an expanded R-S
curve 20--"rotor.sub.--X.sub.--expanded," "rotorA.sub.--Y.sub.--expanded,"
and rotorB.sub.--Y.sub.--expanded." Again, the "rotor.sub.--X.sub.--expanded"
array contains the expanded Roshko numbers for both rotors' expanded
R-S curves 20.
TABLE-US-00003 // ExpandRoshkoStrouhal(void) -- expand Roshko array
from 40 to 3901 points void TForm1::ExpandRoshkoStrouhal (void)
{ char buf[128]; int i, j; double y; for (i = 0; i < 39; i++)
{ for (j = 0; j < 100; j++) { InterpolateLinear((double) (i),
rotor.sub.--X[i], (double)(i + 1), rotor.sub.--X[i + 1], (double(i)
+ double(j) / 100.0), &y); rotor.sub.--X.sub.--expanded[(i *
100) + j] = y; InterpolateLinear((double) (i), rotorA.sub.--Y[i],
(double)(i + 1), rotorA.sub.--Y[i + 1], (double(i) + double(j) /
100.0), &y); rotorA.sub.--Y.sub.--expanded[(i * 100) + j] =
y; InterpolateLinear((double) (i), rotorB.sub.--Y[i], (double)(i
+ 1), rotorB.sub.--Y[i + 1], (double(i) + double(j) / 100.0), &y);
rotorB.sub.--Y.sub.--expanded[(i * 100) + j] = y; } } rotor.sub.--X.sub.--expanded[3900]
= rotor.sub.--X[39]; rotorA.sub.--Y.sub.--expanded[3900] = rotorA.sub.--Y[39];
rotorB.sub.--Y.sub.--expanded[3900] = rotorB.sub.--Y[39]; }
Since the original R-S curve 10 is comprised of 40 points, expansion
of the 40 points, point 0 through point 39 by a factor of 100 will
yield 3901 points, point 0 through point 3900 in the expanded R-S
curve 20. Point 0 (12) in the R-S curve 10 in FIG. 2 remains Point
0 (12) in the expanded R-S curve 20 in FIG. 4. Point 39 (18) in
the R-S curve 10 in FIG. 2 is point 3900 (18) in expanded R-S curve
20 in FIG. 4.
FIG. 5 illustrates linear interpolation performed between point
10 (14) and point 11 (16) in the R-S curve 10 to expand this region
of the R-S curve to contain 100 points. Although FIG. 5 does not
show 100 points formed between points 10 (14) and 11 (16) due to
limitations on space, the present invention contemplates linearly
interpolating 100 points between points 10 (14) and 11 (16) in the
present example. The linear interpolation consists of determining
100 evenly spread points in between each pair of adjacent points
in the R-S curve 10 to form an interpolated line between the two
adjacent points.
Expansion of the R-S curve 10 to provide more points in between
adjacent points will allow the calculated Roshko numbers to more
closely match a point in the expanded R-S curve 20 to perform error
correction of the corresponding Strouhal number as will be described
later below in step 110. If the R-S curve 10 were not expanded before
error correction is performed, the error correction would be performed
on one of the forty points in the R-S curve 10 when the error correction
actually applied to a data point in between that is not closely
adjacent to a R-S coordinate point in the R-S curve 10 thereby
introducing error. However, this step of expanding the R-S curve
10 is an optional step and may not need to be performed, especially
if the R-S curve 10 has a high enough resolution for the desired
tolerance of error that is allowable for a particular application.
The next step is to determine correction poles along the expanded
R-S curve 20 (step 106). Correction poles are used to determine
the true Strouhal number for a given Roshko number to then compare
against the corresponding Strouhal number stored in the expanded
R-S curve 20 to determine the difference or error present in the
R-S curve 10. In this manner, the error can be eliminated or nullified
in the expanded R-S curve 20 by determining an error correction
that makes the Strouhal value in the expanded R-S curve 20 equal
to the correct Strouhal number determined in a verification step
of the meter 1.
An example of how the poles are determined is found in the C++
function "ComputeRoshkos ( );" contained in the source
code in Appendix A. Each pole is determined by providing known flow
rates of liquid through the turbine meter 1 in a verification step,
after calibration where the original R-S curve 10 is determined,
and then determining the corresponding Roshko number for each rotor
at these known flow rates. The computed Strouhal numbers for each
Roshko number are then compared against the stored Strouhal numbers
in the expanded R-S curve 20 and any difference is error. In one
embodiment, five flow rates are used to verify the meter 1 thereby
producing five Roshko numbers or five poles.
Examples of the poles that are determined in a verification step
of the meter 1 are illustrated in FIG. 6. Note that five Roshko
numbers were determined by providing five different flow rates through
the meter 1. The Roshko numbers are converted into indexes into
the expanded R-S curve 20 by finding the Roshko number in the expanded
R-S curve 20 that contains a Roshko number closest to the Roshko
number determined in verification and then storing that index into
the array of the expanded R-S curve 20. In this example, the Roshko
numbers are labeled in FIG. 6 as points 32 34 36 38 and 40
having index values of 2448 2671 2956 3144 and 3349 into the
arrays--"rotorA.sub.--Y.sub.--expanded" and "rotorB.sub.--Y.sub.--expanded"
depending on the rotor being analyzed.
After each of the poles are determined and their corresponding
Roshko indexes into the expanded R-S curve 20 are determined in
the verification of the meter 1 the next step is to compute the
error correction coefficients for the entire expanded R-S curve
20 (step 108). One method to perform this function is contained
in the C++ function "ComputeCorrection ( ); contained in the
source code in Appendix A.
At the poles, points 42 44 46 48 and 50 indicate the error
correction coefficient necessary to be multiplied against the Strouhal
number in the expanded R-S curve 20 to make the Strouhal number
match the calculated, correct Strouhal number during verification
of the meter 1. For example, at the first pole 42 the correction
factor is 0.998258524 meaning that the Strouhal number in the expanded
R-S curve 20 for Roshko index 2448 should be multiplied by 0.998258524
to be correct and match the Strouhal number determined for Roshko
index 2448 during verification of the meter 1. The error correction
coefficient is obtained by dividing the correct Strouhal numbers
obtained during verification for a given Roshko index by the Strouhal
number in the expanded R-S curve 20 for the same given Roshko index.
At this point, only five error correction coefficients have been
calculated based on five actual calculated Strouhal numbers during
verification of the meter 1. Since it is not feasible to perform
verification for all 3901 points in the expanded R-S curve 20 linear
interpolation is used to determine the error correction line illustrated
in FIG. 5 that will be used to perform a correction of all the Strouhal
numbers in the expanded R-S curve 20 in a later step. Note that
more than five poles may be used. In order to perform this linear
interpolation, the end points, point 0 (12) and point 3900 (18)
are set at an error correction coefficient of 1.0 meaning no error
since the error for these indexes was not determined during verification
of the meter 1. The resulting error correction line 10 is stored
in the form of an error correction array 30 as illustrated in FIG.
5 whereby an error correction coefficient is contained for each
index into the expanded R-S curve 20. Examples of the error corrections
arrays for each rotor, rotor A and rotor B, are contained in the
source code in Appendix A as "correctionA" and "correctionB."
Note that the error correction array could contain adders or subtracters
rather than coefficient multipliers.
An example of error correction coefficients for select Roshko indexes
for the error correction array 30 according to the example provided
above is shown below.
TABLE-US-00004 Index Error Correction Coefficient 0 1.0000000 1
0.9999992 2 0.9999985 3 0.9999977 . . . 3143 0.9991913 3144 0.9991934
3145 0.9991956 . . . 3347 0.9996287 3348 0.9996308 3349 0.9996330
3350 0.9996351 3351 0.9996373 . . . 3369 0.9996758 3370 0.9996780
3371 0.9996801 3372 0.9996807 3373 0.9996813 . . . 3897 0.9999982
3898 0.9999988 3899 0.9999994 3900 1.0000000
Next, in step 110 the error correction array 30 is used to compute
new Strouhal numbers in the expanded R-S curve 20. The correction
coefficients contained in the error correction array 30 are multiplied
by their corresponding Strouhal numbers in the expanded R-S curve
20 to perform error correction on the Strouhal numbers on the expanded
R-S curve 20. An example of this function is contained in the C++
function "ComputeNewStrouhals ( );" in the source code
contained in Appendix A. The code steps through each index and multiplies
the error correction coefficient at the index in the error correction
array 30 times the Strouhal number at the same index in the expanded
R-S curve 20 to form an expanded corrected R-S curve 40 as illustrated
in FIG. 7 and as shown in the code below.
TABLE-US-00005 for(i = 0; i < 3901; i++) { new.sub.--rotorA.sub.--Y.sub.--expanded[i]
= rotorA.sub.--Y.sub.--expan- ded[i] * correctionA[i]; new.sub.--rotorB.sub.--Y.sub.--expanded[i]
= rotorB.sub.--Y.sub.--expan- ded[i] * correctionB[i]; }
The expanded corrected R-S curve 40 is stored in the form of arrays
containing the corrected Strouhal numbers called "new.sub.--rotorA.sub.---
Y.sub.--expanded" and "new.sub.--rotorB.sub.--Y.sub.--expanded"
as contained in the source code in Appendix A. Note that the Roshko
numbers stored in the array rotor.sub.--X.sub.--expanded" are
still used for the x-axis into the expanded corrected R-S curve
40 for each rotor.
Lastly, in step 112 now that the expanded R-S curve 20 has been
corrected to form the expanded corrected R-S curve 40 the expanded
corrected R-S curve 40 is reduced from the expanded 3901 points
back down to the original 40 points to form a corrected R-S curve
that is stored in the memory 2 of the meter 1. An example of this
function is contained in the C++ function "ComputeNewStrouhals
( );" contained in the source code in Appendix A and is repeated
below where every 100.sup.th index point into the corrected expanded
R-S curve 40 is used as a data point in the corrected R-S curve
50.
TABLE-US-00006 // extract the 40-pt subsets from the expanded arrays
// for(i = 0; i < 40; i++) { new.sub.--rotorA.sub.--Y[i] = new.sub.--rotorA.sub.--Y.sub.--expanded[i
* 100]; new.sub.--rotorB.sub.--Y[i] = new.sub.--rotorB.sub.--Y.sub.--expanded[i
* 100]; }
An example of the corrected R-S curve 50 is illustrated in FIG.
8. The corrected R-S curve 50 is stored in the form of a 40-point
array, with the new corrected Strouhal numbers for each rotor being
stored in the arrays "new.sub.--rotorA.sub.--Y" and "new.sub.--rotorB.sub.--Y."
The Roshko points that make up the x-axis of the corrected R-S curve
50 are taken from the array "rotor.sub.--X" described
above.
The last step in the process (step 114) loads the corrected R-S
curve 50 for each rotor into memory 2 of the meter 1 in place of
the original R-S curve 10. If the original R-S curve 10 was never
loaded into memory 2 and the original R-S curve 10 was obtained
from another source, the corrected R-S curve 50 will not overwrite
a previously stored curve. The calculator 5 now uses the corrected
R-S curve 50 to determine the Strouhal number using the calculated
Roshko number during operation of the meter 1.
Note that the present invention can be used with any measurement
data or curve, and may be used with a turbine meter 1 having only
one rotor. Also note that another calculator or processor not contained
within the turbine flow meter 1 could be employed to perform the
present invention with the final corrected R-S curve 50 being loaded
into the memory 2 at the end of the process (step 114 in FIG. 3).
Those skilled in the art will recognize improvements and modifications
to the preferred embodiments of the present invention. All such
improvements and modifications are considered within the scope of
the concepts disclosed herein and the claims that follow. |