Вы находитесь на странице: 1из 33

Introduction

Circular paper recording charts ("charts") have been used for years to record pressures and
temperatures from dry-flow orifice meters. This method of collecting the raw data necessary to
calculate fluid volumes has proven satisfactory in the past, but there are certain inaccuracies
inherent in the technology which have always been present. With changing times, the
importance (and consequently the price) of the fluids being measured has increased
exponentially, making these inaccuracies far more critical, and thus diminishing the value of
these charts as a data recording method.

The technology for recording and reading the data from these charts has remained basically
unchanged for decades and has not kept up with the advent of changing technologies which could
enhance the accuracy of the data. Recent advances in digital technology have resulted in tools
becoming available to overcome many of the inaccuracies that have, to this point, been accepted
as an inevitable consequence of limitations to available technology. In order to maintain the
relative validity of the data being recorded on such charts, the oil and gas industry has a
responsibility to be aware of these advances, and to take advantage of them when appropriate
(usually determined by the cost/benefit guidelines established by applicable regulatory
authorities from time to time).

While most people involved in gas measurement acknowledge that there is a degree of inaccuracy
involving paper charts, there is a popular misconception that the effect is minimal, and that in
any event the errors will balance themselves out eventually. An error that is equivalent to a
single pen width (1/32 of an inch) would appear to be minimal at first blush, but even though
the average chart is more than 11 inches across, the size of the grid range where measurement
actually occurs is only 3 and 5/8 inches, so an error of 1/16 of an inch amounts to a 0.862%
error.

Although there are a myriad of circumstances that can give rise to inaccuracy in the use of
charts, this summary deals only with those that have proven capable of some level of resolution
through the proper use of technology.

Meter Calibration Inaccuracies

Orifice meters used in the Oil & Gas industry have a twofold purpose - to measure flow and to
record it. Even if the meter's measurement is perfectly accurate, unless it is recorded with
comparable accuracy, the measurement precision will be lost. Once the flow has been
measured, it is only the recording of the measurement that is used for all purposes, so the
accuracy of the meter will only be as good as the accuracy of the recording. Since most paper
charts start out with inherent inaccuracies introduced in the original production of the chart,
the flow data that it records can be inaccurate.

In order for the data produced from different meters to be usable by flow calculation formulae,
and also for it to be comparable to the data generated by other physical meters, it must have a
common set of references. For example, static pressure pens on two different charts might be
mirror images of each other, but if one meter refers to "gauge" pressure and one refers to
"absolute" pressure, (i.e. one takes into account atmospheric pressure and the other doesn't)
the data could produce significantly different results, even though the charts look identical.
Similarly, physical meters must be calibrated so that not only does the data from the meter
accurately reflect the actual pressure of the meter, the data produced from any given meter will
also be comparable with data from any other meter. The way that charted orifice meters are
calibrated for recording to a chart is to close the meter's flow and to align the meter's pens with
the inner hub of the chart, representing a "no-flow" situation. If the meter's pen rests on the
inner hub line of the chart, it is imperative that such a measurement in fact represents "zero"
pressure. It is just as critical that when the same pen rests on the outer chart line, it in fact
represents "maximum" (100%) pressure. If all meter pens are calibrated in the same fashion,
the relative measurements of similar charts on any similar meter will be comparable.

The inaccuracy inherent in calibrating the meter in this manner stems from the fact that this
method relies on the printed chart gridlines to establish the "reference" measurements, and that
requires that every printed chart be identical. Since most printed charts are printed on large
offset presses, from a single carefully measured static photographic "plate", the gridlines from
each chart should be taken to be identical for all practical purposes. However, the meter is
constructed so that the arms holding the pens used in the measurement do not vary except
through pressure changes. They attach to the meter at a fixed point or fulcrum, which ensures
that once their pens are "calibrated" the relative measurements made by the machine will be
consistently comparable. The values applied in "calibrating" these instruments are established
relative to the distance their pens are from the exact centre of the chart.

If we concede for discussion purposes that the printed gridlines on the charts are identical, the
inaccuracy arises from the fact that in order to be comparable, these "calibration" points all
assume that the charts being used are perfectly circular, and that they do not move. Both of
these attributes are controlled by a spindle in the middle of the chart which holds the chart in
place. The hole that is cut in the centre of the chart is placed over the spindle snugly holding the
chart, but in order for the chart to be perfectly circular both the spindle, and therefor the
chart's centre hole must both be located in the absolute centre of the printed chart. Assuming
the spindle on the meter is physically located exactly in the centre of the chart range and does
not move relative to its pens (a reasonable assumption given current machining techniques)
then the only thing required to make the chart perfectly circular is for the centre hole on the
chart to be cut exactly in the centre of the printed chart - "and therein lies the rub".

It is extremely difficult to accurately cut a circular centre hole in the printed circular chart,
and in fact it almost never happens. In order to examine why that is, we have to start with the
actual production of the paper chart to see where inaccuracies can be introduced.

1. Paper Cutting
Raw paper stock supplied to offset printers from the mill are cut and shipped in "squares".
While not always 'square' as the name implies, the original stock is always rectangular. There
are several reasons for this, including the fact that almost all printed documents are
rectangular, so historically the raw stock has always matched the intended output. Rectangular
bales allow for tighter packing, reducing transport and storage space and thus cost. Raw stock
has to be delivered in varying sizes depending on its purpose, so in order to minimize waste, the
raw stock has to be cut, and rectangular paper allows for cutting with a straight edge.

More importantly than any of these however, is the requirement for "straight edge
justification". In order to cut paper with a straight edge for example, it is useful to be able to
align the cutting blade in a "justified" position. The raw rectangle is "justified" by aligning the
outer straight edge of the bale against a straight guide. Once this is done, the cutting blade only
has to be positioned a given distance parallel to the justified edge, and the resultant cut will
produce a matching straight edge on the cut side. While it will appear to be straight, the larger
the bale being cut with a single knife pass, the more inaccuracy will be introduced. To
illustrate this, cut a single sheet of paper with a pair of scissors along a straight line. The cut
should be reasonably accurate. Now try cutting 10 or 20 sheets together along the same line.
The bottom sheet will not be nearly as accurately cut as the top .

Similarly, in order to ensure that paper is fed into the press (or a printer, scanner, copier
etc.) justification guides are used to align the paper edge so that its top and bottom are
perpendicular to the plate (or camera as the case may be) so that the image that is reproduced
on the paper is aligned to the straight edges of the paper stock.

Image 1.1

While the accuracy of the raw stock measurements may vary from mill to mill, the
measurement tolerance is usually warranted to only 1/64 of an inch - about half the width of a
dime, as illustrated by the yellow line in Image 1.0 (shown as actual size).

Add to that the fact that most printers will usually print charts on a 2-up or 4-up press
(meaning 2 or 4 identical images on a single wide piece of raw stock, which are then cut again
after printing.) The bale cutting inaccuracies mentioned above are reintroduced and the
accuracy tolerance quickly goes from 1/64 of an inch to 1/32 of an inch or more. While paper
inaccuracies of 1/32 of an inch may not seem like much, it amounts to a 0.862% difference in
most charts. Then take into account the fact that since the charts are supposedly circular, a
0.862% error at one place on a chart will be duplicated (albeit with reversed polarity) on the
opposite side of the chart, and the cumulative effect of these physical errors becomes more
evident.

Now, take the cut chart 'squares' and feed them into another press in order to cut out the centre
hole and then again to cut out the outer edge of the circular chart. The only measurement that
the press has to go on is the now inaccurate "justified" edge of the cut square, which can result
in the circular cutting die moving slightly from chart to chart.

There are other circumstances that can contribute to the inaccuracy of the chart's production.
For example, offset press rooms try to maintain a constant humidity, since paper can shrink or
expand with its moisture content, but this is sometimes difficult to do so paper size resulting
from humidity can vary, thus changing cutting measurements. This effect can also differ
depending on the direction of the paper's grain vis-a-vis the edge being cut.
2. Physical Transport
When you add to cutting inaccuracies, the fact that in order to run the charts through the
presses a physical transport mechanism is required which provides for some minor movement
of the paper as well. This can be the result of several factors such as slippage from rollers that
become covered in paper dust etc., or the fact that the speed of an offset press is such that flat
paper can act like a foil or an airplane's wing and lift into the air slightly, allowing for
movement as it's re-introduced to the roller mechanism. Now all of a sudden the chart that was
supposed to be perfectly circular, with a perfectly circular centre hole exactly in the middle of
the chart, is almost certainly an ellipse - the only question being how badly out of round the
final product is.

3. Pre-Press
Another place where inaccuracies can be introduced is in the preparation of the printing
"plates". If the images used to create plates for an offset press are simple raster (scanned)
images that are copied, there is not way to avoid any inaccuracies which are introduced by the
original scan used.

As we have indicated, it is extremely difficult to accurately cut a circular centre hole, and no
matter where the centre hole is cut, if it is not exactly in the centre of the printed chart, then
as it spins the resulting recorded data will appear to be elliptical rather than circular, and the
resulting measurements will be skewed.

For example, in Image 1.2 on the next page the red line was created by placing the centre hole
"A" over the meter's spindle and setting the red pen on the 40% line, then manually spinning
the chart for one complete revolution. The blue line was created by placing the centre hole "B"
over the meter's spindle and setting the blue pen on the 60% line, then manually spinning the
chart for one complete revolution.

In the case of the red pen, the meter was constantly recording a pressure of 40 in. H2O. Since
the centre hole was reasonably accurately punched, the red pen stays on the 40% line of the
chart all the way around, so when the chart is removed from the meter and used to analyze the
data from the meter, the information will be correct.

In the case of the blue pen, the meter was constantly recording a pressure of 60 psi (the
gridlines on the chart range between 0 and 100), but since the centre hole was not accurately
punched, the data doesn't read the pressure as a constant 60 psi, but ranges from a low of 48 psi
to a high of 82 psi. When the chart is linearized (see Image 1.3) it is clear to see that the effect
of the offset centre hole is to cause the pen to trace in a single cycle sine wave pattern, with the
amplitude of the sine wave being directly proportional to the amount of the offset. The value of
the sine wave's inflection point (the value that is equidistant from the highest and the lowest
points on the sine wave) will be the average value of the pen over one full revolution of the
chart, but this value will only be 60 psi if the operator happened by sheer chance to calibrate
the pen at exactly the inflection point (which of course can't be seen until after the chart has
completed its revolution).
Image 1.2

In the example of the chart in Image 1.2, the blue line shows how the offset centre hole can
affect the data recording. While it is extremely exaggerated in our example, we assumed the
meter to be reading a constant pressure of 60 psi. The blue pen was placed on the 60 psi line
where indicated, but the average value of the blue pen for the chart 64.8 psi. Since this chart is
the only record of the pressure data produced by the meter, there is now no way for the data to
be accurately reproduced, or analyzed. Even if the pen was perfectly traced or reproduced, it
would yield an average value of 64.8 psi, even though the meter was measuring pressure at a
constant 60%. The problem is that depending on where on the chart the meter was "zeroed" the
calibration will vary. The magnitude of the problem is exacerbated when the chart used is set
Image 1.3

to a higher (e.g. 1500 psi) spring range. In our example, instead of reading 900 psi (the
known pressure from the well), the chart would read 972 psi.

We have used this exaggerated example to illustrate the effect of the offset centre hole, when in
reality, it is so far off centre that most operators would discard the chart. There is a popular
misconception that the amount of error introduced by physical distortions such as these will not
have a significant effect on the data, but this is not always the case.

Most often, the amount of offset in a centre hole punch is small enough that it is impossible to
discern with the human eye, so it is impossible for a field operator to know how serious the
problem is.

The chart in Image 1.4 looks like a perfectly normal chart, but when properly analyzed from
the location of the centre hole as shown in Image 1.5 it is clear that there is actually substantial
distortion.
Image 1.4

The black concentric circles in Image 1.5 are mathematically created guide lines centred on the
actual centre hole of the chart. These guide lines converge with the chart's grid lines at
approximately the 11:00 o'clock and 5:00 o'clock positions, indicating most likely that this is
where the pen lines pass through the inflection point of the sine wave mentioned earlier, and
that the distortion is the result of the centre hole being offset.
Image 1.5

When the chart is "unwrapped" into a linear image, the sine wave of the 100% grid line around
the true straight computer generated guide line is plainly evident (Image 1.6), however, it is
imperative that in analyzing the pen lines on the chart, the sine wave not be corrected because it
was produced by the centre hole offset, which existed when the meter pens were zeroed, so if the
image is offset before being analyzed, the data will not reflect what was recorded by the meter.

Since the pens on the meter record measurements relative to the centre hole, they are not
wavering - it is the chart gridlines that are wavering because of the offset, so even though the
data on the chart moves in a relative sine wave, the data is constant (remember in our example
the meter flowed at 60% constantly, no matter where the pen marked the chart) and correct.
Image 1.6

For the most part, there is nothing that can be done to correct inaccuracies that get incorporated
into charts in the chart production process. Cutting and transport issues will always exist and
there are, and always will be limitations to the accuracy that can be produced in bulk production
of paper stock. The only part of that process that can be controlled is to ensure that the press
plates are produced from originally created vector images in a digital image editor, which is far
slower and far more costly than using scans of existing charts, but also far more accurate..

While most printers are responsible for the pre-press preparation of the charts, the service
provider who orders the charts from the printer should be actively involved in the pre-press
artwork. If all the vendor does is take a sample of an existing chart to the printer, it will
almost certainly be scanned into the printer's system as the base of the new chart's "plate" as a
simple raster image, with cosmetic changes such as name etc., being substituted over the
scanned template to create a new plate.

While this might provide an adequate new plate as far as the human eye is concerned, it won't
provide any control over the image to the end user, and as we have seen, many existing scanners
stretch the original image to begin with.

The pre-press artwork for new gas charts should be created from scratch in a graphic creation
tool like Adobe Illustrator, as a vector image that can be controlled. The physical parameters of
the chart dimensions can then be mathematically altered using accurate trigonometric
calculations, giving the most accurate measurements to the chart plate. It will also allow for
quick chart alterations (such as changing the differential pen arc line in the grid). This would
also prevent the introduction of errors already existing in the images being scanned, and give
the press plates the accuracy of mathematical precision.

However, all of the various reasons for the production inaccuracies (stock tolerance, transport,
cutting etc.) manifest themselves in a single issue - the inability to accurately punch the centre
hole directly in the middle of the printed chart grid. This issue leads to physical calibration
inaccuracies, so the solutions to chart production inaccuracies and physical calibration
inaccuracies can be dealt with simultaneously.

We have seen that most of the chart production inaccuracies result in improper measurement
opportunities from relying on the position in the image of the centre hole and the gridline. The
simple truth of the matter is that all of these errors can be corrected with proper chart
processing software. The key is having the software extract the chart data in relation to the
centre hole of the chart, and not any printed part of it. If the processor uses the charts inner
hub gridline as its reference point, the data will reproduce all of the errors that might have
been introduced in making the chart, whether in printing or in cutting. Once the position of the
centre hole is known, the position of the gridline and other chart markings becomes irrelevant
Processing software will calculate everything by pixel measurement in relation to the centre,
and in fact won't even use the chart grid for anything other than a visual reference to the
viewer.

Because the calibration of the field meter is done in relation to a chart, in order for the data
from that chart to be accurate, it is imperative that the eventual processing equipment for the
chart, whether an integrator or a computer imaging program, be calibrated to the same
chartline segment as the field meter. If it is, then the processing equipment will be able to
reproduce the data recorded by the meter, but if it is not, then a new calibration error could be
introduced that could have a compounding effect, or could have a counterbalancing effect on the
data. The problem is that the compounding error will be random, and impossible to account for.
(This is illustrated later under Integrator Inaccuracies.)

Since in the case of both field meters and integrators, the calibration inaccuracy resulted from
the physical instrument being calibrated against the chart gridline (which can move in relation
to the charts centre hole) rather than the charts centre hole itself (which obviously cannot),
the solution to either inaccuracy must be resolved by chart data being calculated in relation to
the centre hole of the chart (no matter where it is punched) and not the chart gridlines.

With an integrator, the only way the machine can be calibrated is against a chart gridline. If the
integrator operator knows exactly where on the chart the field operator calibrated the meter,
and the integrator is calibrated at exactly the same place, then the resultant data will be
accurate. The problem is that this would require the integrator to be re-calibrated for every
chart which would substantially increase the already lengthy processing time for the charts. In
fact, most integrators are calibrated against a separate calibration chart which can be marked
up (and whose centre hole is probably cut totally different from the chart being read), and are
usually only re-calibrated after every 30 - 50 charts.

There is simply no effective way to correct this error using an integrator.

With a scanned image of the chart to process, every pixel of the image is under the control of the
processing software (or should be), so the appropriate algorithm should be capable of
accurately finding the centre hole of the chart no matter where it is. Once the centre hole is
found, all of the pen data will have a constant measurement reference point, calculated by the
relative distance of that point to the centre of the chart. As we noted above, if the centre hole
was punched off-centre, when the chart rotates around it the pen lines will appear to move in an
elliptical pattern but in fact that is not what is happening. The pens are moving in a circular
motion and it is the chart under the pens that is moving in an elliptical motion. The gridlines
will fluctuate, but the pen will not. If that is truly what is happening, then if the scanned chart
image can be adjusted to straighten out the pen line, the linearized chart should appear in a sine
wave pattern to account for the centre hole offset. Image 1.7 is a copy of the chart in Image 1.3
in which the static pen has been straightened out, and in fact the chart does appear in a sine
wave pattern as expected.
Image 1.7

Nevertheless, whether the pen appears in the sine wave pattern or the chart does, the pen data
will still be a constant distance from the centre of the chart because as we saw, that is how the
pen data was created by the field meter. If we know where the field operator calibrated the
meter on the chart, the chart image can actually be adjusted to align the chart gridlines with the
calibrated pen data, but whether this is done or not, if the software uses the chart centre as its
reference point, the data as written (traced) on the chart will be correct.

Once we have the ability to accurately read the data on the chart, it remains to ensure that that
data in fact reflects the pressure being measured by the meter. In order for the chart data to
accurately reflect the pressure being measured, one of two circumstances must exist. Either
the chart must be printed so that the centre hole is in the exact centre of the printed gridlines,
in which case the chart will rotate in a perfectly circular motion and every pressure reading
will be the exact same distance from of the centre hole, or the exact point where the field meter
was calibrated must be known, and the chart must be realigned to that point by the software,
effectively ensuring that the calibration point of the meter matches the calibration point of the
chart.

Having satisfied ourselves that the meter calibration issue can be dealt with in this way, it begs
the question "Does it matter?" In other words, if the offset is so small that it can't be seen by
the naked eye, is the adjustment really going to have a significant effect on the volume being
read? In the sine wave pattern in Image 1.3 it is clear that have of the pen values are above the
inflection point, and half are below, which will have a "smoothing" or "averaging" effect,
reducing the magnitude of the change, but when the chart is adjusted to a calibration point,
every pixel in the bitmap moves in the same direction, magnifying the amount of change the
correction will have. In initial tests the range of error has been found to be anywhere from 0%
to more than 11%, but the tools necessary to properly evaluate this question have only recently
been developed and the testing is very preliminary. It is clearly however and issue that needs to
be examined further.
Integrator Inaccuracies

Apart from the meter "calibration" inaccuracies already referred to, there are several other
inaccuracies which are produced by integrators by their very nature. Many of these can be
overcome through a reduction in speed, but lack of speed is probably the biggest single drawback
of integrators in the first place, so the inaccuracies remain in order to get data in a timely
(relatively speaking!) fashion.

1. Calibration Compounding
As already discussed, there is an inherent calibration error built into the field meter as a result
of inaccuracy in the chart's production (especially in the punching of the centre hole), but
depending on where on a chart the field meter was zeroed, a similar calibration error on the
integrator can have the effect of compounding, neutralizing or randomizing the error from the
field meter.

An integrator operates on the exact same principle as the meter, with the focal point of the
measurement being the centre spindle, and the pens being anchored to a point that causes them to
produce data based on the distance of the pens from the centre hole of the chart. As a result, all
of the physical calibration inaccuracies described previously that pertain to a meter, also
pertain to an integrator - but an integrator presents an even bigger problem. By the time a
chart gets from the meter to an integrator, as we have seen, there may already be an element of
data distortion built in as a result of meter calibration against the chart gridlines which may
not be centred.

Now the same off-centred chart is again placed on a spindle (the integrator's) so the chart pens
can be traced, but the integrator must be calibrated as well, for the same reasons the meter was.
The problem is that the integrator operator may not know where on the chart the field operator
calibrated the meter, so if the integrator operator uses the actual chart to calibrate the
integrator, it is very possible that the integrator will have used a different reference point for
calibration than the meter, which will introduce further error into the charts data by reading
values other than what the meter recorded.

As a practical matter, integrators are not calibrated for every chart, and when they are, a
calibration chart (one that is used only for the purpose of calibrating the integrator) is used,
and in all probability that chart will have a totally different centre hole offset than the chart
being integrated. Even if the centre hole offset was exactly the same as the chart being
integrated, there would be no way for the integrator operator to know at which point on the
chart the calibration should take place. As a result, the integrator now introduces a completely
different error element on top of the one that already exists from the meter.

As an example, the chart that we saw in Image 1.2 already shows blue pen data that is 64.8%
(when we know the meter was flowing at 60%). This error is wildly exaggerated for
illustrative purposes, but let's assume that the centre hole offset error was a more realistic
1/16 of an inch. Now let's assume that the integrator was also calibrated with a chart that was
offset by 1/16 of an inch, and that at the point of calibration the offset was in exactly the same
direction as the meter calibration offset. We have now compounded what was a 1.75% data
error into a potential 3.5% error for all of the pen data from the chart, and all before we even
begin to integrate the chart.
If we understand why the sine wave effect occurs, and accept that the data relative to the centre
hole is nevertheless accurate, there remains one other problem to be dealt with. Even if we
trace the meter pen accurately, we will get proper relative values from the chart, but in order
to ensure that those relative values are the true values being measured by the meter, we have to
know exactly where the meter pen was zeroed, so that the relative values on the chart equal the
actual values being measured by the meter.

Image 2.1

For example, if the centre hole of the chart in Image 1.5 was offset by 1/16 of an inch, when
the meter is zeroed (at point "C" on Image 2.1), the actual pen value on the chart's inner hub
could be 1.75% rather than 0%, so a 60% line on the chart would actually be 61.75%. If the
chart used a 1500 psi static pen range, then the meter that actually measuring 900 psi would
show on the chart as 926.25 psi. Now when the chart is integrated, if the chart used to
calibrate the integrator is offset by the same amount in the exact same direction (again at point
"C" on Image 2.1), the 60% line on the chart would actually be 63.5%, and in the case of our
1500 psi static pen range, then the meter that actually measuring 900 psi would show on the
chart as 952.5 psi. By the same token, if the chart used to calibrate the integrator was offset
by the same amount, in the exact opposite direction, it would cancel the effect of the meter offset
(much like sound waves out of phase) and the meter that was measuring 900 psi would actually
be read as 900 psi.

If the meter was zeroed at a point on the chart that happened to be at 90 in either direction
from point "B" on Image 2.1 (Point "A" or "D") then the integrator will add no further
distortion, but calibration at any other point will introduce further offset distortion on a
different axis than the meter offset distortion, and will become very difficult to correct.

Regardless of what system is used to process the chart, it is imperative that the operator know
where the field operator zeroed the meter on the chart. Since the chart gridline can move in
relation to the centre hole, we may not know where the true centre hub line on a chart should
be, but if we know where the field operator thought it was (and where he zeroed the meter pen)
then we know where it is. In fact, the true zero line on the chart is wherever the operator
calibrated the meter to say it is. Even if the operator calibrated the meter to show the "zero"
line 1/8 of an inch below the hub line, that is where he will have set the meter to read zero, so
in fact that is the real "zero" line, no matter where the chart gridlines might be.
In similar fashion, once true "zero" line on the chart is known, the processing software should
be able to adjust the bitmap to align all of the chart data in relation to that zero line, and from
that point, all of the data points can be measured accurately against the corresponding values of
the meter that created the chart.

Image 2.2 is the same chart that is shown in Image 2.1, but the software has moved the centre
hole so that the chart gridlines are aligned with the computer generated gridlines and are now
equidistant to where the centre hole has been adjusted to (the yellow crosshairs in the circle).
What it has done effectively is to move the centre hole of the chart to where is should have been
cut on the original. This will straighten the sine wave pattern of the pens (but in the process
change the chart to a sine wave to compensate), which now shows that as we saw in Image 1.7
the static pen will read as a constant 60%, and is in fact straight.

Now that the we are accurately reading the data on the chart, it remains to ensure that the chart
data reflects the actual meter pressure. In order to do this, referring to the example in Image
2.1, we have to ensure that the calibration point on the chart where the field meter was zeroed
is in fact the inner hub of the chart. To word it differently, we have to ensure that the
calibration is made exactly at the inflection point of the sine wave.

Most regulatory authorities require that the meter's static pen be zeroed on each new chart. If
it is, the calibration point will be easy to find, in that it will be the point on the inner hub
where the static pen starts to float to pressure. Once we know the point where the meter was
zeroed, we have to ensure that the computer generated inner hub (which marks the true inner
hub in relation to the centre hole) matches the inner hub gridline printed on the chart. If it
does not, which is usually the case, then the software should have a mechanism by which the
whole bitmap image can be realigned so that it does. Once this is done, the physical calibration
error from the meter will be corrected, and the possibility of a calibration compounding error
from an integrator will be eliminated.

After going through this whole process, the obvious question becomes "Does it matter?". If the
degree of correction to the charts volume calculation is not significant enough, it may not be
worth bothering. The truth is that the magnitude of correction gained in straightening of the
sine wave by moving the centre hole will vary by the amount by which the hole is off centre
(the amplitude of the sine wave) and where the meter was originally calibrated in relation to
the inflection point of the sine wave. Nevertheless, the raw data will still go both above and
below the inflection point, and this eventually tends to have a smoothing effect on the magnitude
of the correction. However, the moving of the bitmap's inner hub is unidirectional, and affects
every measurement of every pen, and the magnitude of this correction can be significant.

Image 2.2
2. Multiple Pen/Radically Fluctuating Pen Tracing
Since each of the pens being traced by an integrator operator are calculating data which is used
only in relation to the corresponding values of the other pen, integrators are designed to have
the operator trace the differential and static pressure lines simultaneously. This is an
incredibly difficult task, even for an experienced operator, but the alternative is to trace each
pen line separately, which would require double the processing time for an already painfully
slow process.

Similarly, many of the charts contain radical fluctuations in the pen values, which make them
virtually impossible to trace accurately, no matter how slowly the integrator turntable moves.
Image 2.3 is a good example of a chart that would be impossible to trace accurately on an
integrator.

Image 2.3
As a compromise result, the integrator is designed to run at a reasonable speed, with both pens
being traced at the same time, but the tracing is only a rough estimation of the different pens,
using a sine wave pattern when necessary to estimate the actual readings. This produces what is
at best a "smoothed" or "averaged" value over the duration of the chart rather than an actual
value. It can also fail to pick up the chart "spiking" effects of the pens.

Often when there is a spike or trough in static pressure, it results directly in a corresponding
variance in the differential pressure, and reading the corresponding values from both pens at
the same time with a sine wave estimation tends to smooth over the peaks and troughs, affecting
the resulting data.

Once reduced to a paletted, fixed size bitmap, a computer can process the pen ink data far more
accurately and far more quickly than an integrator operator tracing 2 lines at the same time
(See Image 2.4). Because of the speed at which modern computer processors work, the
computer program should be able to follow a pen line pixel by pixel. There will be multiple
factors which affect the accuracy of the pen tracing algorithm, and these factors will determine
the quality of the data extracted, and must be assessed and compared in order to determine the
best software available.

Image 2.4

3. Single Point Measurement


The physical properties of chart integrators are such that each integrator pen arm results in a
single pen tip being used to trace the chart data. In situations where the chart meter pen has
produced a "painted" band of ink, the only way in which the integrator pen can extract a
reasonable value is to create a "sine wave estimate" of the banded area, running roughly between
the upper and lower edges of the band. This produces an average of the actual values represented
by the ink on the chart.

In some circumstances, this average value created by a sine wave estimate can be as useful as a
pixel by pixel evaluation of the ink data. If a single average volume over the chart duration is
all that is required, the values generated by the sine wave estimate should be reasonably close to
the average values generated by the more accurate pixel by pixel evaluation. For example, if a
chart rotates for 6 days starting at no flow and increasing in a linear fashion to a point at the
end of the fourth day where flow is 100,000 m3, the average flow for the chart is 50,000 m3,
and that is the value that will be generated by either method. However, if the sine wave estimate
is used for anything other than a full given duration average, the data can become quite
meaningless. For example, a chart might no flow for 3 days then flow at 100,000 m 3 for 3
more days. The average over the whole duration is 50,000 m3, but as soon as you try to apply
the whole duration average to any part of the duration less than the whole, the data becomes
grossly distorted. During each of the first 3 days, the average states that flow is 50,000 m3
more than it actually is, and for each of the last 3 days, the average states that flow is 50,000
m3 less than it actually is. The sine wave estimate generates valid data only for the whole
duration of the estimate, and is really not accurate for anything else.

If a meter pen is placed on a chart and the pen does not move during the chart rotation, the pen
will create a single line with an intended single value as in Image 2.5.

Image 2.5

However, when that single line is scanned at 150 dpi and examined closely, it is clear that the
computer sees a line that is broken down into between 5 and 8 pixels.

Image 2.6
Where this can become a problem is in reading chart pen lines that have some degree of banding
inherent in them from pulsation etc., which includes the majority of charts. Because the
tracing pens on an integrator use only a single point value, they cannot be used to establish a
"range" value for a pen line. If, for example, a static pen is fluctuating rapidly between 49 and
25 psig as a result of line pulsations, in order to account for the Square Root Error ("SRE") the
1992 AGA calculation method recommends that the proper method of calculating the pens value
is to take the square of the roots average, which requires the measurement of both the highest
and the lowest value for the pen's range. (See "EFFECT AND CONTROL OF PULSATION IN
PIPELINE MEASUREMENT", Class # 1090 by Michael Royce Miller, of PGI International, and
"INSTALLATION AND OPERATION ERRORS IN GAS MEASUREMENT", Class # 1190 by Johnnie
Jarred, in the Proceedings of the 79th International School of Hydrocarbon Measurement, at
Oklahoma City, OK., May 17 - 20, 2004.)

The integrator's single point measurement precludes this type of adjustment, but since digital
imaging allows the chart front to be divided into pixels on a bitmap, multiple pixels can be
attributed to the same point in time, and therefor calculations to provide for SRE can be
accommodated.

4. Static/Temperature Pen Offsets


In order to allow meter and integrator pens to pass each other in recording the full range of the
chart, the pens must be offset, so the corresponding static or temperature pen measurement for
any given differential pen measurement will not be aligned on the chart and must be adjusted to
take into account this offset. Since the fulcrum of the pen arms on both the meter and the
integrator are fixed, this offset will always be identical. It would follow then that the offset on
any chart should also always be identical, but for a variety of reasons this is almost never the
case. Pen arms can bend; pen nibs can be angled differently depending on how they are fastened
to the arm of the meter or the integrator; one or more pens can be dragged on the chart surface
in either putting the chart on or taking it off, making it difficult to determine where the actual
measurement reading begins or ends. If a pen is screwed into the arm either more loosely or
more tightly than the last time it was used, the distance between the pens will vary. Also, arms
as very light and pliable and subject to bending.

Given that the offset distances can vary, it is important for digital processing software to have
the ability to be constantly reset so that the correct values for any given pen are properly
associated with those of each of the other pens. Once reduced to a visual image, it is easy to see
where the different pens start, so it should be a simple matter for the software to readjust the
starting pixel, thus allowing each subsequent pixel in the image to correspond with the
appropriate other pen values.

Regardless of the cause of the offset variance, integrators are incapable of adjusting for it,
whereas proper digital processing software can instantly adjust the pen offsets from a chart to
be accurately read.

In Image 2.7, the processing software used provides for a "drag 'n drop" function that allows the
static pen start pixel to be moved for alignment with the ink on the chart, using a different pixel
from the bitmap to represent the same time, so that in the top image the start time of 8:45 a.m.
is represented by the pixel value for pixel 1079, and in the bottom image the same start time is
represented by the pixel value for pixel 1115. All subsequent pixels are instantly re-aligned
to correspond with the correct differential pixel value.
Image 2.7

5. Slow/Fast Meter Clock Speeds


It is often the case that meter clock speeds do not correspond directly with the chart duration
and either run slow or fast. Integrators have no accurate way of adjusting the length of the pen
data to accurately reflect the chart duration in the event of a slow or fast clock. If the meter
clock is slow, the pens on the chart will fall short of the length that would properly reflect the
amount of time that the chart was actually on the meter. An integrator could only record the
correct time by tracing over a blank chart surface after the pen had stopped, until the correct
time was recorded, but sing the same process displayed in Image 2.7, the end pixel can be
dragged to align with the end of the chart's pen, and then every pixel in the duration can be
interpolated to reflect the correct time. By stretching every pixel by the same amount the
whole pen line on the chart is accurately stretched (or shrunk) to match the correct chart
duration, without distorting the recorded pressure values.
Since most orifice meters run on manual spring clock mechanisms, they tend to wear out, get
stuck, wind down, speed up or stop. Any of these situations will result in the length of the pen
lines showing as something other than the duration that they actually rotated on a meter. In
order to properly process the pressure values over the time that they were actually rotating on
a meter, processing software must have the ability to adjust the pen values, to represent the
actual duration of the chart's rotation.

Integrators have no physical ability to accurately adjust the clock speed to match the pen length,
although the software that processes the integrator data could be configured to deal with these
variations, if sufficient data was input, and processing time was allowed.

Once again, the flexibility of digital images is such that the interpretation of the pen data can
easily be extrapolated or interpolated to readjust the values for the pen data to "fit" the actual
duration of the chart's rotation. It is easy to visualize where pens start and stop, so by being
able to identify the start and stop values on the image, the length of the pen lines can be
mathematically adjusted so that each pixel making up the line's length can be re-evaluated to
match the desired length of the pen. It is important for any processing software to have the
ability to quickly and accurately adjust the actual pen length to match the actual duration of the
chart's rotation.

6. Out-of-Arc Pens

Given the nature of field meters as we have seen, they are prone to some inaccuracies in their
setup and calibration, and one of the most difficult to correct is pen arms that do not accurately
match the arc of the chart grid. Meter pen arms are attached at a fixed point fulcrum, which
necessitates that the sweep of the arm results in an arc. If the fulcrum of the integrator's pen
arms are comparable, the arcs will be the same and the pen tracing can be accurate, but as the
pen arms get seriously out of arc, it will become more and more difficult to trace the lines and
data will become more and more inaccurate.

Duration of a field meter, and integrator or a digital image is represented by a uni-directional


rotation of the chart. Neither meter pens nor integrator pens can go backwards, since this
would represent a reversal of time, and time only moves forward. If the meter pen is out of arc,
there is a possibility that the processing method will follow the pen line to a point where it
would have to move in reverse to accurately reflect the pen data, and since this is not possible
(without overwriting already existing data) the best a process can do is to roughly approximate
the pen.

The most obvious example of this is in vertically oriented pen sweeps such as in a plunger lift
well. In order for an integrator operator to properly attain the true height of the chart pen,
his/her focus has to be the highest point of the arc. In order to reach that point as the
integrator turntable continues to rotate, there is a natural tendency to start the upward sweep
of the tracing pen a bit before the chart pen actually starts its vertical rise, and in order to
maintain processing speed, there is a tendency to bring the trace pen back down to its lower
position a bit slower than the chart pen drops, resulting in the trace pen spending more time
representing flowing values, both before and after the "spike" than the meter pen did,
overstating the flow every spike.
If the chart pen arm was out of arc, the sweep of the integrator pen will not match that of the
chart. If, for example, the chart pen leans too far to the right, the only way for the integrator
operator to trace the pen would be to estimate the amount of the difference in the arc of the chart
and the arc of the integrator, and to preceded the chart pen by the amount of such difference
rather than trying to use the chart ink as a guide, and touch the peak of the chart pen with the
integrator pen. This is asking a lot of an integrator operator, who more than likely will simply
match the height of the tracing pen to the height of the chart pen. Once this is done, the
integrator pen will already have over-rotated and since it can't move in reverse, the best it can
do is to drop straight down, but that will lead to separation of the upsweep line and the
downsweep line, overstating the pen values over the duration of the bi-directional sweep.

Since the digital image processing can be accomplished instantaneously, the trace lines can more
accurately reflect the pen lines, provided the pen is in arc, so the geometry of the tracing
process matches that of the meter pen. Digital processing software has the ability to readjust
the digital image representing the chart data and re-draw it so that the pen lines can be adjusted
to be in arc before being processed.

7. Divergent Overlaps

It is not uncommon for charts to be over rotated for any number of reasons, given the
inaccessibility of some meters during the winter etc.. In any processing software, it is
important to be able to read the overlap period separately from the initial rotation of the chart.
This is especially true if the pens diverge during the overlap period. If a chart maintains
relatively constant pen values, the overlap period will have similar flow to the initial rotation.
If, however, the pen values during the overlap period are different than during the initial
rotation, the volume will have to be calculated separately to be accurate, so the processing
software will require the ability read multiple values from a chart, from the same chart
location.

Image 2.8
Scanning Inaccuracies

Since most of the inaccuracies inherent in the processing of charts to date have generally been
the result of the physical properties of the meters and integrators, and the partially analog
technologies used in extracting the data from them. In order to make use of computers to assist
in processing these charts, the charts themselves must be converted to a digital format, and this
is most efficiently done by scanning the paper charts, but scanning itself can introduce new
inaccuracies which must be recognized and dealt with.

Although scanning itself has been in widespread use for more than a decade, its primary use has
been the reproduction of pictures, and scanning for data extraction is in its relative infancy. By
its very nature the use of scanners to produce images has been focused on how the human eye
perceives the scanned image, and in order to reproduce images accurately, processes have been
introduced which skew the images and thus the data they contain.

For example, shapes of almost everything are made up of curves and textures with a huge
variety of colours, not right angle squares of a single colour, but since computer images are
made up from 'picture elements' (pixels) which are exactly that (small squares of a single
colour) subjective interpretation of the image affects every aspect of its digitization. This
subjective interpretation is twofold: It can be controlled by the scanner manufacturer, either
through the physical properties it builds into the scanner, through the scanner's driver
software, the compression algorithm used to compress the image etc., or through the chart
processing vendor in how it manipulates the image with its software. However, before dealing
with the software, there are physical properties of different scanners that the end user should
be aware of.

1. Anti-Aliasing
Everything in computer graphics starts with the simple fact that a computer screen is composed
of tiny square points of information called pixels. On a PC, the number of pixels on the screen is
determined by the display mode or screen resolution of the user's computer system. Because the
screen resolution of a PC display is fixed, the number of pixels doesn't change, no matter how
large a monitor screen is used. On larger screens, the dots just get larger. (On a Macintosh, the
number of pixels per inch is fixed, so a larger monitor displays more pixels and actually lets
you display more information.)

Computer screens don't have enough dots available to create small, complex shapes like letters.
This can be especially noticeable in large fonts where diagonal lines are created with a stair-
stepped appearance, sometimes called "the jaggies" or "pixelation." This effect occurs any time
a line on a computer screen is not perfectly vertical or perfectly horizontal. Now think of a
normal circular gas chart that traces a pen on an arc so that every line is a curve.

The problem is called "aliasing" and it is overcome by using the technique known as anti-
aliasing. Anti-aliasing causes the pixels along a diagonal border to be filled with a colour that is
in between the colour of the letters and the colour of the background (see below). The goal is to
create an optical trick that fools the eye into seeing a smoother shape, especially at a distance.
Notice the examples in Image 3.1.
Image 3.1

The purpose of digitizing the chart in the first place is to enable the user to apply sophisticated
graphic editing tools that are available on the computer to "read" the blue and red pen lines off a
chart, so determining exactly what is a red pixel and what is a blue pixel is critical. For most
pixels this is not a problem, but those pixels that are anti-aliased on the outer edges of the pen
can become difficult to characterize as they lighten up to blend in with the background chart
colour. This is especially true if the pen on the chart is light to begin with, or running out of
ink. The problem can be further exacerbated when dealing with vertically oriented charts such
as plunger lifts, where differential pen line move slightly in a horizontal direction with the
turning chart, but the pen is almost completely vertical. Dropping too many red pixels on each
side of a line for example can cause individual segments to be misread
Image 3.2

Anti-aliasing can also cause problems for charts that have substantial spiking with the spikes
being close to each other. The charts background colour are altered to blend with the red pixels,
which causes them to turn a light shade of red themselves. For a computer there is no middle
ground. A pixel has to be treated as either red or not red. Anti-aliasing can have the effect of
filling in the chart background with red shading, so that the computer treats the whole area as
red.

Notice the space between the vertical spikes in Image 3.2. Anti-aliasing has turned what would
otherwise have been a white or grey colour from the blank chart, into various shades of red and
pink. Since the spikes on the chart are so close together, it creates a whole area between the
spikes that is a lighter shade of red, and whether the analysis software being used reads this as
red or not, and how it distinguishes when it should and when it shouldn't, can have an effect on
the volume being calculated.

2. Scanning Resolution
Since aliasing is caused by trying to represent too many patterns of an image with too few
pixels, it stands to reason that the resolution (number of pixels contained in an a square inch)
of scanned images is critical in minimizing the problem. If a chart is scanned at a high enough
resolution, it will likely have sufficient information about the image pattern to accurately
reproduce it, but there are limitations that prevent us from simply taking scans at a higher and
higher resolution.

The most important is file size. Computer images are simply files just like word processing
files, but they contain substantially more information in order to reproduce colours accurately.
Since every pixel contains colour information unique to itself, the more pixels there are in an
inch (the higher the resolution), the larger the file size is going to be when it is saved to disc.
It is clear that if resolution is set too high, the file size of the chart images will take up too
much disc space and will take too long to deal with (open, copy, move, edit etc.) which would
negate much of the benefit of making them digital in the first place.
However, the converse is also true. If charts are scanned at too low a resolution, the images can
become seriously pixelated and can distort the pen values as well. In the following images, the
line marking the trace pens in two different chart processing software programs are compared.
In the chart on the right in Image 3.3 the pen line covers an area of more than 2.0% of the
chart. In the chart on the left the trace line covers an area of 0.3% of the chart. That
discrepancy alone determines a possible data range for the pen value that is 6.67 times more
accurate than the first. That discrepancy will be further defined by how accurately the program
calculates the actual data point from those pixels. The second program actually creates a value
that is 1/1000 of the image, but it is unknown what level of accuracy the first program uses.
If it does nothing more to define the pen range from the image, the second program will be 20
times more accurate.

Image 3.3

When you compare a magnified view, it is clear that the image on the left is far more accurate
than the image on the right, and this will enable the data extraction algorithms used to produce
more accurate data, but suppose that these images are being scanned remotely, and there are
100 images that have to be electronically forwarded to a processing point. It is clear that the
larger image size would be unworkable, so some level of compression has to be introduced - and
the tradeoff begins! There will always be tradeoffs in software, between processing speed,
accuracy, automation, file size, transmission speed etc., and it is the assessment of these
tradeoffs between the available chart processing software programs that will determine the
program best suited to an individual client's needs.

For example, as we have already noted, the higher the resolution that charts are scanned at and
the less compression used by the digital file format being used, the more accurate a graphic
representation of the chart will be. Images 3.4 and 3.5 appear to be identical to the naked eye,
but Image 2.5 is 2,936,013 bytes and Image 2.6 is 479,232 bytes.
Image 3.4 Image 3.5

One way in which scanning resolution can affect the accuracy of a chart is in plunger lift or
"spiking" differential readings. At 150 dpi resolution, the width of a normal meter pen is
between 5 and 8 pixels, and even a very fine meter pen on a rapid spike will usually be at least
3 pixels wide. The difficulty is in reading the pen spike, when a single pixel on a 7 day chart
accounts for 3.5 minutes of duration. If a pen spike is 3 pixels wide, and the spike is read to the
top, the programme will be reading at maximum flow for 10.5 minutes (3 pixels wide at 3.5
minutes/pixel) when in fact the spike might well have lasted only 30 seconds.

This is not so much an error in the programme, as a physical limitation of the chart meter, but
it can lead to serious mis-readings of the spikes. In a 7 day chart with compressed spiking (e.g.
one spike per hour), the result can lead to 168 spikes being over-read by as much as 10 hours,
with the whole flow being at the charts absolute maximum rate. The programme should be able
to apply a correction factor to the chart based on a sample fast clock analysis of the actual length
of the spike.

In a fast clock analysis, the meter is set to a faster rotation speed - for example, what might
normally be a 7 day chart is run on a meter set to run on a 24 hour clock, so that the spiking is
stretched or spread out. This gives a little more detail of the actual spikes' duration with a view
to making a more accurate determination of how long the spike actually lasts. There are other
ways to determine the actual characteristics of the meter, such as with a fast/slow chart drive,
which runs a part of a chart on a fast speed, and the balance on a normal (slower) speed, or
better still, the field operator's input as to what the actual meter characteristics are.

Once the meter characteristics are known, they can be databased as input parameters that can be
applied to the volume analysis of the chart so that the output is correctly factored into anomalies
such as spikes so the physical limitations of the meter or integrator can be adjusted.
Another example of inaccuracy that high resolution can present is providing too much
information about the pen lines on a chart image. At a resolution of 150 dpi, the average pen
line from a chart will be from 5 to 8 pixels wide. That means that there are 5 - 8 pixels
representing what was a single point of the chart pen, and the question becomes which one of the
5 - 8 pixel values should be used for that single point.

An example of this "too much information" being a problem on chart images is where there is a
chart from a no-flow meter, with one of the pens running on the inner hub line (the "zero"
line) of the chart. The pen width can be wider than the chart gridline and even if it is
accurately positioned on the gridline, it will overlap on each side of the line, leaving ink traces
within the chart range which will result in minimal flow being calculated, even though the
chart is probably a no-flow chart. While the amount of flow will probably be negligible, the
number of flowing hours could be substantially wrong. A dilemma presents itself in such a
situation, in that if certain 'threshold' pixel values around the inner hub are ignored, there is a
very real possibility that valid flow data could be lost.

For example, the chart in Image 3.6 belonged to a company that dropped out any data less than
0.5% above the inner hub to avoid spurious time data from the situation described. When
processed while dropping out all pixels with less than 0.5% value, flow (the black line) was
calculated as 82 103m3. The spiking results from the fact that in the software being used, for
every pixel that shows no flow, the mapping line drops down to the hub, so every time there is a
spike, it represents single pixels that show flow (i.e. greater than 0.5%), followed by one that
doesn't.

Image 3.6

When this 0.5% threshold was dropped so that every pixel with a value greater than 0 is
calculated (Image 3.7), the flow jumped to 8,364 103m3. Because of the "over-accuracy" of
the programme, the dropout threshold was incorporated to attempt to control spurious flowing
time data, but it had the effect of causing the programme to drop out enough pixels that more
than 8,282 103m3 would have been lost.
Image 3.7

3. Transport Braking

Most scanners use a paper transport mechanism that consists of several rows of servo driven
wheels, discs or bars covered by a rubber sleeve that rotate as the chart is fed into the scanner.
Usually there is a similar offsetting roller for each that is "inert" (not servo driven) that sits
on top of the driven roller and is pressed against it by means of a spring, to hold the chart down
against the driven roller so it will push the chart forward. There is no easy, accurate way to
measure the speed of the inert rollers, but it is probable that they have a distorting effect on the
image as a result of the physics of their use.

The inert roller is spring loaded to allow for varying thicknesses of paper, and it is this
requirement for flexibility that prevents it from also being servo driven. While the driven
roller rotates at a constant speed, the inert roller rotates at the same speed and in the same
direction. The problem with this mechanism is that when the leading edge of a chart is
introduced between each set of rollers, the inert one is lifted off the driven one so its inertia is
momentarily broken, acting as a brake until it picks up the speed of the passing chart. Each
time this happens (at every row of rollers) the braking effect interrupts the constant speed of
transport, causing the chart to pass over the camera (which is stationary) at varying speeds,
stretching the chart image along the vertical axis. The resulting image is not longer a circle,
but rather most likely an ellipse which will distort the measurement of the pens if not adjusted
for. The truth is that the various physical properties of scanners can have different effects on
an image, both stretching or compressing it in all directions, sometimes doing both on the same
image.

Image 3.8 is an example of a chart that was scanned on a scanner which is subject to transport
braking. As in the case of Image 1.5, the black concentric circles in Image 3.8 are
mathematically created guide lines centred on the actual centre hole of the chart. The effect of
the scanner stretching is that the top half of the chart does not align with the proper gridlines,
but this effect diminishes from the top of the chart to approximately mid-chart, by which point
the effect has ceased and the bottom half of the chart appears to be properly aligned. This is
assumed to be because by the time the mid point of the chart is over the camera, the front
(leading) edge of the chart has crossed the last transport roller, so the breaking effect stops.

Image 3.8

This effect will cause the data between Saturday and Monday to be skewed unless the effect can be
corrected. The degree of distortion on the chart in Image 3.8 is more than 2%.
It is clear from the unwrapped image 3.9 that the stretching effect is severe.
Image 3.9

The proper software process should be able to identify the area over which the stretch occurs...

Image 3.10
Image 3.11
and either compress (or expand if necessary) the scan lines of the image to reconstruct it in a
perfect circle.

Image 3.12
Image 3.13

Summary

The biggest benefit in the advance of technology as it applies to the processing of gas charts is
the ability to digitize the chart image. The physical hardware used in processing gas charts (the
meters and integrators) has always been finely machined and extremely accurate, but much of
the benefit of that accuracy has been lost because of the necessity of calibrating the hardware
against the printed charts which are not consistently accurate or uniform. While the physical
constraints of recording data on the paper charts (such as the necessity of calibrating the meter
against the chart, or the difficulty in punching a perfectly accurate centre hole in the chart)
continue to exist, creating a digital image allows the management of the data to be brought under
the control of the computer, and once that has been accomplished, the powerful tools of current
computers can be used to detect and correct errors, which can provide for more accurate output.

Although chart scanning and digital processing of dry-flow orifice meter charts can lead to much
more accurate data, and can do so with significantly less processing time and at a fraction of the
cost of manual integrators, it is important to remember that whether chart data is processed on
an integrator or the most sophisticated digital imaging and processing software, different
processing equipment is only a tool to be used, and while having the ability to edit and adjust
values as the situation requires will allow for much more accurate data, it will only be as useful
as the experience and diligence of the operator who uses it will allow.

Вам также может понравиться