GAUGING DATA POSE QUESTION ON STABILITY OF REFERENCE GAUGE HEIGHTS

Frank J. Berto
Consulting Engineer
San Anselmo, Calif.

A major tank gauging calibration project at a 130,000-b/d California refinery revealed that any top-down (outage) tank level measurement has an uncertainty of 0.5 in.

A surprising possible conclusion is that this occurs because of random movement of the reference gauge height (RGH). The implications of this are formidable.

At current prices and world production rates, about $500 billion worth of crude oil and $1 trillion worth of product are sold each year. An average of six custody transfers occurs between the wellhead and the end user.

Roughly half of all crude oil is custody transferred by tank measurement, and the other half is metered. Most custody transfers of product are done by metering.

Metered custody transfer is quite accurate-in the range of 0.1%. But even a 0.1% error amounts to roughly $1 billion/year.

ACCURACY

Tank measurement is ideally suited for inventory control. However, custody transfer based on tank measurements is a different matter.

Literature on tank measurement accuracy is quite limited. A decade of research on the subject has not revealed anything definitive on the true accuracy of tank volume custody transfer.

The results of the extensive series of RGH field measurements at Tosco Refining Co.'s Avon refinery at Martinez, Calif., suggest, however, that tank gauging outage measurements are relatively inaccurate because RGH varies randomly by 0.5 in. (See accompanying box for a list of possible causes of tank measurement errors.)

The reference gauge height is the distance from the reference gauge point (a fixed point on the top of the tank) to the datum plate at the bottom of the tank, as shown in Fig. 1.

This is also called the reference height, the reference depth, or the gauging height.

The datum plate is a level metal plate located directly under the reference gauge point to provide a fixed contact surface for the innage bob. (See the accompanying list of definitions of terms used in this article.)

The random variation in RGH limits the accuracy of both manual outage gauging and automatic tank gauges that measure outage. The variation is not linear with oil level, oil temperature, or ambient temperature, so it cannot be compensated for in the level calculations.

Tank innage measurements are more accurate because they are independent of variations in RGH.

INNAGE VS. OUTAGE

To manually measure the oil level in a tank, one must decide between measuring from the bottom of the tank up to the oil surface (innage gauging), or from the top of the tank down to the oil surface (outage or ullage gauging). Fig. 1 depicts the differences between these two methods of RGH measurement.

Manual innage gauging is harder to perform than outage gauging because the gauger must prevent the bob from tilting when it contacts the datum plate. Innage gauging is also messier because the tape is coated with oil over the entire oil depth.

And finally, if the datum plate becomes fouled with debris, it is not possible to perform accurate innage gauging.

With outage gauging, the RGH is accurately measured and displayed on a nameplate mounted next to the gauging hatch. The measured outage is subtracted from the fixed RGH to obtain the oil depth.

If the RGH is measured before each outage gauging, as is the practice of some gaugers, the result is an innage gauge-performed in two steps.

REPEATABILITY OF OIL DEPTHS

If the RGH is constant, innage gauging and outage gauging give exactly the same oil depth. However, some measurement experts have nagging doubts about the accuracy of outage measurement.

What if the RGH changes between the opening and the closing gauges?

Because of this uncertainty, many companies use outage gauging for inventory control and innage gauging for custody transfer. Outage gauging is traditionally used for measuring heavy fuel oil and asphalt, to avoid tape cleanup.

RGH is a critical part of the installed accuracy of most automatic tank gauging (ATG) systems. Float-operated, servo-operated, and radar ATGs are all outage gauges, measuring from the top down.

Any change in the RGH affects the accuracy of the level measurement, regardless of the inherent accuracy of the ATG.

Hydrostatic tank gauges (HTGs) are mass gauges that are often used to measure standard volume or level.

HTGs measure by innage gauging, but they are subject to their own unique errors when used to measure level.

Literature on this subject has been thoroughly searched, and very little has been published about tank measurement accuracy or about RGH variation.

TOSCO SURVEY

It was learned in 1989 that an extensive tank measurement accuracy survey had taken place at Tosco's Avon refinery. The survey required a full year to complete and included more than 500 measurements of reference gauge height on more than 100 tanks.

In 1988, Tosco had needed to accurately monitor the oil inventory of the Avon refinery using Varec remote-reading ATGS. To satisfy this need, the readings of the Varec ATGs were compared with manual innage gauging measurements on a regular basis.

This was done to confirm that the oil inventory measured by the ATGs agreed with the oil inventory measured by manual gauging.

Tosco has permitted the use of the survey results in this article.

ATG calibration checks were performed by teams of three: an experienced gauger, a data recorder, and an accountant. They conducted the gauging in accordance with API Standard 2545.

Tosco purchased new gauging tapes and woodback thermometers, which were used only for the ATG calibration checks, The level and temperature measurements were typical of the quality that would be used for tank-gauged custody transfer witnessed by an independent inspector.

Whenever the ATG level differed from the manual innage level by more than 1 in., the ATG was checked and readjusted to match the manual innage level.

When the program began, Tosco provided extra instrument mechanics to perform the readjustments.

The company assumed that once all of the ATGs were calibrated, maintenance requirements would decrease. But after a year, ATG maintenance was still excessive. Many ATGs had been checked and adjusted two or three times.

At that time, Tosco hired the author to review the calibration and maintenance records for the first year of operation. It was apparent that the ATGs were primarily being readjusted because the RGHs were changing.

To reduce ATG maintenance, it was recommended that Tosco:

se outage, rather than innage, manual gauging to check the ATGS. (A float-operated ATG measures by outage gauging and it should be checked against a manual outage measurement.) This is in accordance with the new API Chapter 3.1B on automatic tank gauging.
Increase the allowable deviation between the float-operated ATG reading and the manual outage reading from 1 in. to 2 in.
Concentrate maintenance on the ATGs with the worst performance, and perform complete ATG overhauls rather than just recalibrations.

DATA ANALYSIS

The significant feature of the Tosco data was that each calibration check included an accurate, witnessed measurement of the RGH. This was the most extensive set of RGH measurements the author had ever seen.

It had been assumed that RGH would vary with oil temperature and would probably vary with oil depth. Preliminary analysis of the data did not support this assumption.

Statisticians at a major pipeline company, and at two major instrument companies, analyzed the Tosco data using their most powerful regression analysis programs.

These rigorous regression analyses yielded just one major conclusion-that RGHs vary unpredictably for both floating roof tanks with gauging wells and fixed roof tanks without gauging wells.

The changes in RGH do not correlate with either oil temperature, ambient temperature, or oil level. No correlations are discernible above the random background.

The sample size was large enough to produce a 95% confidence level in the following conclusion:

For both floating roof and fixed roof tanks, 95% of the RGH measurements will fall in a range of 0.9 in. and 75% of the RGH measurements will fall in a range of 0.5 in.

The Tosco results suggest that if you use manual outage gauging or an ATG that measures by outage gauging, you can expect that one level measurement in four will be in error by half an inch, and that one level measurement in twenty will be in error by almost an inch.

The size and the randomness of the variations were surprising.

In a 1989 article on tank measurement accuracy based on available literature, this author concluded that RGH varied predictably with oil level and oil temperature.1 This conclusion was wrong.

The distribution of the changes in RGH (_RGH) is plotted in Fig. 2 for floating roof tanks and in Fig. 3 for fixed roof tanks.

The best that can be said is that the level errors caused by the changes in RGH do not produce a bias towards either the buyer or the seller.

ANALYSIS METHODOLOGY

The original "raw" data base included 507 measurements of the reference gauge heights of 111 tanks. These data were transcribed onto a floppy disc in Lotus 123 format.

The raw data included seven tanks that had only been measured once. These were eliminated. This reduced the data base to 500 measurements of 104 tanks.

No ambient temperatures had been recorded. The ambient temperatures were added to the data base after researching San Francisco Bay area weather records.

The measured RGHs were corrected for the thermal expansion of the gauging tape. The submerged section of tape was assumed to have been at ambient temperature.

Gauging tapes are calibrated to read correctly at 72 F. The tape length readings were corrected for temperature variations from 72 F. to produce a temperature-corrected RGH.

For each tank, multiple measurements of RGH were averaged. Then, the individual measurements of RGH for each tank were algebraically added to the average RGH of that tank to give _RGHS.

Finally, the _RGHs were compared against oil depth, oil temperature, and ambient temperature, using regression analysis. There were no significant correlations.

In order to determine whether human errors were masking the expected correlation between RGH and oil temperature or oil depth, a number of subsets of data were analyzed.

This eliminated as "outliers" those measurements of RGH which differed from the average RGH by an assumed tolerance. This tolerance was varied from 1 to 4 in.

None of the subsets showed anything but random variation.

The conclusion stated in this article is based on the subset that eliminated all _RGHs of more than 2 in. This eliminated 121 measurements as outliers and reduced the data set to 196 measurements of 46 floating roof tanks and 183 measurements of 56 fixed roof tanks.

A smaller tolerance eliminates more outliers and reduces the RGH variation. Therefore, the size of the RGH variation is dependent on the assumption that the 121 outliers were caused by human errors.

These human errors include gauging improperly, misreading the tape, incorrectly recording the level on the ticket, and incorrectly transcribing the level from the ticket to the computer.

Two years after the survey was completed, an effort was made to check the outliers, but the original gauging tickets had been thrown out.

A similar problem exists in "round robin" testing of octane or Rvp. How do you handle the laboratory result that comes in one octane number high? Do you include it in the averages or do you eliminate it as an outlier?

The following subsets of the data were also analyzed:

Tanks with four or more RGH measurements
Tanks with six or more RGH measurements
RGHs of less than 1 in. (eliminating all larger deltas as outliers)
Tanks with the same diameter and height.

None of these subsets showed a significant correlation of RGH with oil depth, oil temperature, or ambient temperature. RGH varies unpredictably.

Interestingly, S. Sivaraman and C. J. Hollaway reached a similar conclusion when they measured the RGH variation on four tanks using precision surveying equipment.2

They concluded that: "Instability of the reference height in vertical, cylindrical storage tanks can have a significant effect on tank-gauging accuracy."

REASONS FOR VARIATIONS

Following are possible reasons for RGH variations:

Movement of the reference gauge point caused by thermal expansion of the gauging well on floating roof tanks. This can be caused by either oil temperature or ambient temperature. (A 50-ft length of steel pipe expands 1/4 in. with a 60 F. temperature increase.)
The gauging well may not thermally expand because the top of the gauging well is not completely free to move vertically. Any binding between the floating roof guides and the gauging well, or between the guide on the gauging platform and the gauging well, can introduce vertical forces with effects that mask the thermal expansion.
Movement of the reference gauge point caused by thermal expansion of the tank shell on fixed roof tanks. This can be caused by either oil temperature or ambient temperature.
Movement of the reference gauge point caused by shortening of the tank shell or deflection of the tank roof, as the tank takes a barrel shape when filling. This affects fixed roof tanks or floating roof tanks with gauging wells rigidly secured to the top of the tank shell.
Movement of the reference gauge point caused by upward movement of the gauging well when the tank takes a barrel shape during filling. This affects floating roof tanks with gauging wells rigidly supported from the lower course of the tank shell. (All of Tosco's floating roof tanks have gauging wells supported from the tank bottom.)
Upward movement of the datum plate as the tank takes a barrel shape when filling. This affects fixed roof tanks where the datum plate is cantilevered from the lower course of the tank shell. (All of Tosco's datum plates are on the bottoms of the tanks.)
Thermal expansion of the gauging tape caused by oil temperature and ambient temperature. (This was corrected for in calculations).
Initial calibration of the gauging tape or the use of kinked gauging tapes. (Tosco used new tapes.)
Movement of the datum plate as the bottom settles during tank filling.
Use of the RGH on the nameplate, rather than actual measurement of the RGH.
Measurement of a false RGH because the datum plate is fouled by rags, notebooks, bottles, etc.
Human error in performing the manual gauging, recording the data, or transcribing the data to the Lotus 1-2-3 format.

The last two items are believed to be the main causes of the gross errors in measuring the RGH in Tosco's survey. Therefore, the clearly questionable readings were eliminated as outliers.

Editor's note: Given the unusual conclusion based on this survey, the editor invites those readers interested in obtaining the complete 500-measurement data base to send a blank, double sided, double density, 5 1/4-in. diskette formatted for IBM compatible computer in a stamped, self-addressed mailer to Refining/Petrochemical Editor, Oil & Gas Journal, P.O. Box 1941, Houston, Tex., 77251. Readers outside U.S. need not stamp mailer. Offer expires Sept. 1, 1991.

After analyzing the data, the author and the editor invite brief comment on the subject of unpredictable variation in reference gauge height.

REFERENCES

Berto, Frank J., "Methods for volume measurement using tank gauging devices can be error prone," OGJ, Mar. 13, 1989, p. 57.
"Hydrocarbon Measurement: An Investigation into the Tape Length Calibration of Portable Gauging Systems," BP Research Center Report No. 21079, June 29, 1987.

BIBLIOGRAPHY

Patterson, I.W.F., "Standardization for Installation of Automatic Tank Gauges," Journal of the Institute of Petroleum, July 1972.
Patterson, I.W.F., "Installation of Liquid Level Gauges and its Effect on Accuracy," Whessoe report, November 1979.
Neesbye-Hansen, O., "Accuracy of oil-custody transfers can be improved," OGJ, Jan. 3, 1983, p. 97.
Sivaraman, S., and Thorpe, W.A., "Measurement of tank-bottom deformation reduces volume errors," OGJ, Nov. 3, 1986, p. 69.
Sivaraman, S., and Holloway, C.J., "Method measures cylindrical storage-tank reference height variations," OGJ, Dec. 12, 1988, p. 50.
Mei, Kenneth W., "Automatic tank gauges can be used for custody transfer," OGJ, Nov. 13, 1989, p. 81.
Sivaraman, S., "Field tests prove radar gauge accuracy," OGJ, Apr. 23, 1990, p. 89.
Piccone, Roland P., "All You Ever Wanted to Know About Tank Gauging," Treatise by Sarasota Measurements and Controls Inc., April 1990.
Berio, Frank J., "Hydrostatic tank gauges accurately measure mass, volume, and level," OGJ, May 14, 1990, p. 57.
Sprigg, Marshall W. III, "Union Carbide Achieves 0.4 in. Accuracy With Servo Gauges," Control, February 1991.