Cost Effective Mass Standards Calibration Intervals (U)

W S R

C-MS-95-0027

Cost Effective Mass Standards Calibration Intervals

(U)

by

A.

H.

Shull

Westinghouse Savannah River Company Savannah River Site

Aiken, South Carolina 29808 J. P. Clark

A document prepared for INSTITUTE OF NUCLEAR MATERIALS MANAGEMENT36TH ANNUAL MEETING at Palm Desert from 07/09/95

-

07/12/95.

DOE Contract No. DE-AC09-89SRi 5035

This paper was prepared in connection ydh work done under the above contract number with the U. S.

Depafiment of Energy. By acceptance c: this paper, the publisher and/or recipient acknowledges the U. S.

Government's right to retain a nonexclusive, royalty-free license in and to any copyright covering this paper, along with the right to reproduce and ta authorize others to reproduce all or part of the copyrighted paper.

..

.

DISCLAIMER

This report was prepared

as

an account of work sponsored by an agency of the United States

Government. Neither the United States Government nor any agency thereof, nor any of their

employees, makes any warranty, express

or

implied,

or

assumes any legal liability

or

responsibility for the accuracy, completeness,

or

usefulness

of

any information, apparatus,

.

product,

or

process disclosed,

or

represents that its use

would

not infringe privately owned rights.

Reference herein to any specific commercial product, process,

or service by trade name,

trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement,

recommendation, or favoring by the United States Government or any agency thereof. The

views

and opinions of authors expressed herein do not necessarily state or reflect those of the

United States Government or any agency thereof.

This

report has been reproduced

directly from the

best available copy.

Available to

DOE

and DOE contractors from

the

Office

of

Scientific and Technical Information,

P.O.

Box

62,

Oak

Ridge,

TN

37831;.prices available

from

(615) 576-8401.

Available to the public

from

the National Technical Information Service, U.S. Department of

Commercef5285

_Port

Royal Road, Springfield, VA 22161.

COST EFFECTIVE MASS STANDARD CALIBRATION INTERVALS

*

A. Harper Shull & John P.

C M

Analytical Laboratories

Westinghouse Savannah River Company Aiken, SC 29802 USA

803/9 52-3970

Abstract

National Institute of Standards and Technology (NIST) traceable standard weights serve as the foundation of mass measurement control programs. These standards are normally recalibrated annually at a cost of approximately $1 00 per weight. The Savannah River Site

(SRS)

has more than 4,000 standard weights. Most have recalibration intervals of 1 year. The cost effectiveness of the current practice was questioned. Are these mass standards being calibrated too often, and are all of these

standards needed for calibration and QC

activities? Statistical analyses of data from the

calibration histories were performed on a random sample of eight weight sets. The analyses indicated no time effects or significant trends in the weight masses for periods of from 5 to 8 years. In other words, calibration

checks were being performed too frequently. In addition, current electronic balance technology does not require a traditional set of standard weights that cover the entire weighing range of a balance. At the

most,

only 2 or 3 standards are needed for most weighing systems. Hence, by increasing weight set recalibration

frequencies from 1 t o 3 years, and

by

reducing the number standards calibrated by 80%, annual cost savings of over $400,000 are attainable at

SRS.

Details of the data analysis, technological advances, and cost savings are included in the paper.

*

The information contained in this article was developed during work under Contract

No.

DE-

AC09-89SR18035 with the U. S. Department of Energy.

Table 1: WEIGHT SET: WL-1017

Nominal Wt (gms) Jrm-88 Ap-89 oct-90 oct-91 NOV-92 Des93

100 100.001200 100.000760 100.001117 100.000880 100.ooo%4 100.000757 50 50.000230 50.000460 50.000552 50.000570 50.000498 50.OOO460 1-20 20.oooO70 19.999930 20.oooO11 19.999880 20.000001 19.999940 ' 2-20 20.000170 20.oooO50 20.oooO28 20.000000 20.000045 20.000006 10 10.000192 ' 10.000210 10.000207 10.O00140 10.000171 10.000183 5 4.999993 5.000560 5.000476 5.000520 5.000469 5.000473

1-2 2.oooO39 2.oooO33 2 . q 2 7 2.oooO10 2.000022 2.oooO18

2-2 2oooO66 2.oooO32 2.000029 2.oooO11 2.000024 2.oooO19

1 1.oooO62 1.000042 l.ooOo73 l.ooOo30 1.000061 1.000042

0.5 0.500043 0500044 0500031 0500022 0500039 NoData

1-2 0.200078 0200076 0200070 0200074 0.200072 NoData

2-2 0.200092 '0.200076 0.200075 0.200081 0200078 NoData

Study and Evaluation

Weight standard recalibration at

SRS

consists of comparing the mass of each weight t o the mass of a National Institute of Standards and Technology (NIST) traceable standards,

determining the current mass of the weight, and issuing a certificate that specifies the current true mass, the apparent mass at a density of 8 grams per cubic centimeter, and an estimate of the uncertainty. No mass adjustments are made

1

to the weights. The certificates, which are filed by the using organizations, comprise a history file for each weight set.

Data histories for eight weight sets containing a total of 102 individual weights were obtained from various laboratories at

SRS.

A typical set

of data is shown in Table 1. Five of the weight sets were stainless steel and the remainder

were brass. The weight data, which covers a

period of from 5 to 8 years, was evaluated by several statistical tests t o determine weight values were changing over time. if the

Table 2: WEIGHT SET: WL-1023

gobmind Calibration Dates: Jm-87 AF-89 S ~ 9 0 NOV-91 Oct-92 NOV-94

TrueMass(gm) 30.00034 30.00036 30.00037 30.00038 30.00028 30.00029 Average 30.00034 UpperLimit 30.00041 30.00041 30.00041 30.00041 30.00041 30.00041 StdDev LowerLimit 30.00026 30.00026 30.00026 30.00026 30.00026 30.00026 O.ooOo2 20 TrueMass(gm) 20.00028 20.00020. 20.00025 20.00031 20.ooo24 20.00023 Average UpperLiit 20.00039 20.00039 20.00039 20.00039 20.00039 20.00039 StdDev LowerLimit 20.00011 20.00011 20.00011 20.00011 . 20.00011 20.00011 O.ooOo5 10 TrueMass(gm) 10.00022 9.99988 9.99990 9.99990 9.99989 9.99988 Average Differences 0,00034 0.00002 0.00000 o.ooOo1 0.00000 9.99995 UpperLimit 10.00015 10.00015 10.00015 10.00015 10.00015 10.00015 StdDev LowerLimit 9.99974 9.99974 9.99974 9.99974 9.99974 9.99974 O.ooOo7 30

Me€- o.oooO2

o.oooo1

o.oooO1 0.00010 0.00000

outlier NO NO NO NO NO NO

Diffemms O.oooO8 O.ooOo5 .O.oooO6 O.ooOo7 0.00000 20.00025

outlier NO NO NO NO NO NO

outlier YES NO NO NO NO NO

Out I ie

r

Test

The first analysis was an outlier test t o

determine if any of the certified values for each weight differed from their historical average. The test was simply a

+/-

(t)x(s) confidence limit about the historical average where t is the appropriate 99.7% t-value. The test is

graphically illustrated in Figure 1.

The standard deviation, s

,

was estimated from the replicate values for each weight over time. The method of moving range or successive differences was used t o estimate s t o prevent any trends from unduly inflating the estimate. The method estimates s by dividing the average difference bv a factor of 1

.I

28. A Dortion of

is shown in Table 2. Failure of the outlier test indicated that a particular weight value was significantly different from its historical average. Any value that failed the outlier test was excluded from any subsequent analyses. Four weight values failed the oklier test Since the next calibration check indicated a return of the "failed" weight t o near its historical

average, it appears that the cause is due t o some perturbation in the measurement process for that specific value and not a real change in the weight value itself. The inadvertent switching of measured values for different weights with the same nominal value is another possible the outlier ialculations for a typical'weight set explanation.

9.99%

I

I 1 I I I

. Sep90 NOV-91 06-92 NOV-94

Jan47 m 4 9

Analysis

of

Variance Test

The next statistical test was an analysis of variance (ANOVA) on the data for an entire weight set. 1 The purpose of this analysis was to determine if there were any significant effects due t o the different calibration times. Intuitively, the ANOVA makes this

determination by comparing the weight averages at each calibration time to one another. The average is taken across all, weights in the weight set. Before performing the ANOVA, the data was transformed by

calculating the deviation from the historical average for each weight. This technique

removes the differences due t o different weight sizes. Data for each weight set was treated as a 2-factor ANOVA without replication with the test of significance set at 95% confidence. The

only weight set that indicated a significant effect due t o time was

brass

weight set WL-

101 3 (Table 3). This significant effect yas

due t o a gain trend which is discussed later.

Table 3: Weight Set WL-1013, Deviations from WeGht Averages ANOVA : Two-Factor Without Replication

sowce of

Vmiatwn

ss

df MS F P-value Fcritical

Weights 9.762&26 Calibration Times 4.7556E-06

Error 1.9869E-05 12 5 59 8.13E-27 2.42E-20 1.00 951E-07 282* 0.02 337E-07 1.92 237

I

Total 2.4625E-05 76

*

SignXcant with 95% confidence

I

Regression

Test

The final statistical analysis was a regression on each weight set t o determine if there were any significant gain or loss trends in weights over time. The number of elapsed months since the last calibration was chosen as the time variable. Again the analysis was performed on the transformed data, which were the deviations from the historical average for each weight. Figure 2 gives a graphical representation of the transformed data for a typical weight set. No

significant trends were indicated in any of the weight sets except set WL-1013, which had already exhibited a significant time effect in the ANOVA. The regression analysis for WL-1013,

which is shown in a Table 4, indicates an

average increase of 0.000008 grams per month or less than 0.1 milligrams per year. After

two

years the increase is less than 0.2 milligrams, which is well within the class S-1 tolerance. lgure 2

0.0004 0.0003

1

’ DEVIATIONS FROM WEIGHT AVERAGES

WEIGHT SET AG1020 LEGEND _{- 5 0} ai

.

20-2 v) 0.0002 0 10 . 5

-

2-1 2-2

E

0.0001

g

-0.0001 0 1

3

-0.0002 x 0.5 a 0 Y

-

x 0.2-1 -0.0003 i

+

02-2 -0.0004 I

Jan47 Feb-88 Mar49 Mar-90 Apr-91 May-92 Aug-93

-

0-3 I

Technological Advances

There has been a major paradigm shift is the world of weighing. Weighing technology has changed as dramatically as the 21 jewel Swiss watches that have been replaced with the simple single quartz crystal watches. The need for buying and maintaining a set of weights for each balance no longer exists for 95% or more of our weighing applications using electronic balances. Randal Schoonover, head of the NlST

mass group, has published a paper, "The Use of the Electronic Balance for Highly Accurate Direct Mass Measurements Without the Use of External Mass Standards"

3

that stresses the lack of need for external mass standards. Hence, the need t o have mass standards calibrated is unnecessary, except t o have

calibrated ASTM Class 1 mass standards that are required t o calibrate electronic balances.

Table 4: Weight Set WL-1013, Deviations from WeIght Averages

Regression Statistics MultipIe R 03871991 R Square 0.1499232 Adjusted R 0.1385888 Standard 0.0005283 Observations 77 !3quare Error

Analysis of Variance Signa@xmce

Regression df 1 SwnofSquares 3.69186E-06 3.69186E-06 13.227 0.0005036 Meansquare F ofF Residual 75 209332J2-05 2.79109E-07

Total 76 24625E-05

Cmffiients StMdardError tSt&ic P-value L.tn4er9SQ Upper9Sg

Interceut -0.000359 0.000115593 -3.10467234 0.0027 -0.0005892 -0.00012

-

Slope 8.054E06 221453E-06 3.636938131 0.0005 3.643E-06 1.25E-o

Jerry Everhart, metrologist formerly with EG&G Mound Technologies, Miamisburg, Ohio, recommends using QC check standards that approximate the process for which the balance is used, rather than weighing stainless steel mass standards. 4 Therefore, the practice of buying, maintaining, and calibrating entire sets

of

weights for each balance is obsolete, expensive and unnecessary for approximately 400 balances at SRS.

Cost Savings Analysis

Hence, a two-fold program is underway at the

SRS

to be more cost effective in the realm of mass standards use and calibration. The program calls for reducing the number of standard weights calibrated annually by the Savannah River Standards Laboratory (SRSL)

and extending weight calibration frequencies t o 3 years. In the analytical laboratories alone, the annual expense of approximately $63 K for calibrating 530 weight standards can be

reduced t o less than $5 K by reducing the number of weights by 80% and extending the calibration intervals to 3 years. Detailed

calculations are shown in Table 5. Since the 530 weights in analytical laboratories

represent only one-eight of the approximately 4000 weights at SRS, site wide implementation will result in annual savings of 8 x $58,655 or $469,240.

Table5 CostDetails %

Existing Cost Baseline $63,070 New Method

Implementation Cost -$4,415

Annual Cost Savings $sa,sss Total Two Year Savings $1 17,310

Recommendations

Statistical analyses

of

the historical data for each weight set indicate no time effects or significant gain or

loss

trends in the weights for periods of 5 t o 8 years with the exception

of

one brass weight set. I t s time trend, though statistically significant, was less than 0.1

milligrams per year, which is well within the

also a few significant "excursions" from the historical average for some individual weights. However, since the weight would return to near the expected value at the next calibration check,

it is believed that these deviations are related to

error in the calibration process rather than an actual change in the weight value. On the basis of the technological advances, and the statistical and cost savings analyses, the following actions are recommended.

1 Extend the calibration check frequency to three years for stainless steel weights, unless the history for a specific set indicates otherwise.

2. Eliminate the use of brass weights. 3. If brass weights are not eliminated,

extend their calibration check frequency t o two years.

4. Eliminate all weight calibration except for weights that are used for qualify control ahd for electronic balance calibration.

Summary

Over 4000 weight standards are in use a t the Savannah River Site (SRS). To validate their accuracy, most are recalibrated annually by the Savannah River Standards Laboratory. The need for an annual recalibration was questioned by the authors. Histories on eight weight sets from Analytical Laboratories, DWPF, and SRTC were selected at random t o determine if the certified weight values were changing significantly over time. Since statistical examination of the histories indicated little or no change in the certified weight values over time, a decrease in the frequency of calibration can be affected without any adverse effect on quality. Therefore, a three year calibration . frequency is recommended. Technological improvements in balances have reduced the need for a large number of calibrated weights. If the recommendations are implemented, cost for weight standard calibration can be reduced by approximately 90%.

5

References

1. Walpole and Myers, Probabi1.b and

_{. .}

_{. .}

atistics for Fnatneers

AuGaa&&

MacMillan Publishing Company, 1972. 2. National Institute of Standards and '

Technology, Handbook 44, SgecifimtiorlS, Jolerances. And Other Technical

Requiremen ts For Wetahina And Measuring Devices, 1990.

. .

3. Schoonover, R. M.

NlST

and F. E. Jones,

NlST

Retired, The Use of the Electronic Balance for

hlv Accurate Direct Mass Measurements Without the Use of Fxtemal Mass Standads,

Proceeding of the National Conference of Standard Laboratories Work shop and Symposium, Washing ton, DC, July 1994. 4. Jerry L Everhart, Mound Laboratories,

Mass Measurement Process Error

Determmtlon and Control

,

Proceedings of 7 994 Measurement Science Conference,

Pasadena, CAY January 1994

. .

5. A. H. Shull & J. P. Clark, Achieving Calibration Cost Savinas Throuah Data Analvsis, Proceedings of the INMM 34h Annual Meeting, Phoenix, AZ, July 1993.