ML20133H481

Sanitized Insp Rept 70-1113/83-28 on 830920-23.Noncompliance Noted:Failure to Adequately Reevaluate Control Chart Limits at Least Once Per Matl Balance Period & Failure to Establish Written Procedure Re Evaluating Limits

ML20133H481
Person / Time
Site:	07001113
Issue date:	10/28/1983
From:	Mcalpine E, Richards B NRC OFFICE OF INSPECTION & ENFORCEMENT (IE REGION II)
To:
Shared Package
ML20133H069	List: IR 05000730/2008002 ML20133H065 ML20133H094 ML20133H271 ML20133H285 ML20133H304 ML20133H337 ML20133H382 ML20133H421 ML20133H456 ML20133H481 ML20133H506
References
FOIA-85-158 70-1113-83-28, NUDOCS 8508090363
Download: ML20133H481 (6)

Text

%

k, , 42 cag ,

'o, L .; LEAR REGULATORY COMMISSION

<g

_O o cE110N 11 3 E 101 M ARIETTA STREET, N.W.

k ATLANTA. GEORGIA 30303

%.....* OCT 311983 Report No. 70-1113/83-28 Docket No. 70-1113 License No. SNM-1097 Safeguards Group No. III Licensee: General Electric Company P. O. Box 780 Wilm,ington, NC 28402

' Date of Inspec' tion: September 20-23, 1983 Type of Inspection: Unannounced Material Control and Accountability Inspector': S.u/ '

[ ~

/3/sr/v4 p B. E. Rii: hards, Statistician Date S1(ned -

Approved by: E E. J. McA1)ine, Sectioh Chief, Material Control

/O[2M83 Date Signed and Accountability Section, Safeguards Branch Division of Emergency Preparedness and Materials Safety Programs Inspection Summary ,

Areas Inspected: Measurements and Statistical Controls.

The inspection involved 25 inspector-hours by one NRC inspector and was begun during the regular hours.

1esults: The licensee was found to be in compliance with NRC requi.ements in the one area examined during the inspection except for the following items:

1. Failure to adequately reevaluate control chart limits at least once each material balance period and failure to establish a written procedure to describe the methods used for evaluating control chart limits. ,

2. The licensee's procedure which described control chart limit calculations was inadequate in that it included inappropriate statistical techniques.

4 PDR FOIA '

RATNER 85-158 PDR

x 1

M l 4

REPORT DETAILS l l

Report No. 70-1113/83-28

1. Key Persons Contacted

• B. F. Bentley, Acting Manager, Fuel Manufacturing

• T. D. Brechtlein, Manager, Fuel Chemical Quality Control R. L. Bruce, Analyst, Licensing and Nuclear Materials Management

• R. H. Foleck, Acting Manager, Licensing and Nuclear Materials Management W. B. Haverty, Analyst, Licensing and Nuclear Materials Management

., *W. J. Kemp, Acting Manager,. Fuel Fabrication

• E. A.- Lees, Manager, Quality Assurance

• R. I. Parnell, Measurement Control Engineer

• L. A. Sheely, Manager, Fuel Quality

• W. B. Smalley, Acting Manager, Regulatory Compliance

• J. P. Stewart, Senior Engineer, Nuclear Safety Evaluation

• J. W. Summey, Acting Manager, Planning and Projects

• T. P. Winslow, Manager, Chemet Laboratory The inspector also interviewed several other licensee employees.

• Denotes those present at the exit interview

2. MC 852068 - Measure'ments and Statistical Controls

a. Control Chart Program The licensee's statistical program for establishing and maintaining the control charts used to monitor the quality of special nuclear material

- accountability measurements was inspected. The inspection consisted of a review of the licensee's applicable written procedures, the licensee's methods for detecting out-of-control situations (associated with standard measurements), the data base used to establish the control chart limits, and the statistical methods employed in the evaluation of the data base. The data base and statistical methods used to determine the control chart limit: which were in effect during the current inventory period were examined. The program data for the following measurement systems were inspected: standard control charts for uranium concentration measurements of uranium dioxide powder and clean scrap, enrichment measurements of uranium dioxide powder, and bulk measurements of uranium hexafluoride.

The licensee maintained control charts for the measurement of standards to monitor the performance of applicable measurement systems. Control charts have been established for individual measurements of standards.

The licensee appeared to have an adequate system in place for detecting out-of-control situations for standard measurement,s for those measure-ment systems reviewed during the inspection.

', 2

b. Control Chart Limit Evaluations The licensee is required by Section 4.5.1 of his Fundamental Nuclear Material Control Plan to reevaluate existing control chart limits once each material balance period. The purpose of this evaluation is to determine whether the existing control chart limits are still applicable. j

- An examination of the control chart limits associated with the NBL97 i standard which is used to monitor uranium concentration measurements I of uranium dioxide powder disclosed that the limits were calculated in Novem6er 1981, from measurement data generated between April 1, 1981 The licensee did not have a program in place and ' November 12, 1981.

to compare, by means of formal statistical tests, current measurement data to the measurement data associated with the existing control chart

, limits. The licensee generated a report every three months which sum-marized standard measurements from the current three months along with key statistical parameters associated with the standard measurements.

The licensee compared the current three months of measurcment data to the measurement data from the preceding three months. However, the licensee's program did not include a provision for comparing current measurement data to the measurement data upon which the existing control chart limits were based - a necessary step in determining the applicability of the existing control chart limits. If the licensee had performed such a test, he would have found that the variance associated with standard measurments changed significantly on two

. separate occasions. The variance of the February 7 - May 8, 1983 period in part'icular decreased by a factor of five relative to the variance of the April 1, 1981 -

November 12, 1981 period. These variar.cas arc statistically significantly different at the .0001 level.

It should be noted that tne variance associated with standard measure-ments is a key parameter in determining control . chart limits. These independent tests performed by the inspector indicated that measurement performance had significantly changed, based on the statistical varia-tion of standard measurements from period to period, and that the control chart limits should have been updated. The variance of the measurements of standar.ds associated with the February 1 - May 2,1982 time period was also found to be significantly different (at the .05 <

level) from the variance of the measurements of standards associated with the data base from which the existing control chart limits were {

calculated. The licensee failed to reevaluate control chart limits associated with the gravimetric uranium measurement system during the August 10, 1982 through August 8, 1983 material balance period.

In addition, the licensee did not have a procedure that described the methods used for evaluating control chart limits. Aspects of the program such as frequency of evaluation, composition of the data base, and the statistical tests and criteria to be a' plied to the evaluation of the data base were not addressed in any procedure. The licensee's failure to establish a written procedure for this aspect of the measurement control program and the failure to reevaluate the control chart limits is a violation. (83-28-01)

\ 3

c. Control Chart Limit Calculations The statistical techniques used by the licensee in calculating control chart limits were examined and found to be appropriate except for two concerns. It was disclosed that the licensee was using an

-1< appropriate method to obtain the "t"-statistic. The "t"-statistic is j a key p~arameter used in the formula to calculate control chart limits. I For some measurement systems, the licensee analyzed the same standard a j numbe,r' of times when a batch of process material was analyzed. The

, lice,nsee chose to control a series of standard measurements at the .05 and .001 levels of significance. This resulted in the licensee controlling each individual standard measurement at levels other than

.05 and .001, as is required by 10 CFR 70.57(b)(11). In order for a j

, series of measurements to be in control, all the results in that series must be in control. A series of measurements is out-of-control if any one or more of the results are out-of-control.

The control chart limits associated with standard NBS 950B (which is f' used to monitor uranium concentration in clean scrap), for example,

( were calculated incorrectly because of the licensee's inappropriate statistical techniques. The licensee controlled a pair of standard measurements for this measurement system at the .05 and .001 levels of r significance. However, the licensee compared each standard measurement

}

result to the calculated control chart limits. The existing control chart limits went into effect July 13, 1983. The licensee established ,

the control chart limits from the measurement data associated with 83 individual measurements of the standard (NBS 9508) performed between '

Feoruary 7,1983 and May 8,1983. The degrees of freedom associated with the "t"-statistic used in calculating the control chart limits thus became 82 (n-1=83-1). To choose the correct "t"-stati stic for significance levels of .05 and .001, one should look up the row of

'- values in a "t"-table corresponding to the applicable degrees of j freedom, 82 in this case. Then the two values in that row corre- l sponding to alpha levels of .05 and .001 should be selected. The I inspector found these values (after interpolation) to be t (82; .05) = f I

1.99 and t (82; .001) = 3.424. The "t"-statistics that the licensee used in the formula to calculate the control chart limits for this system were 2.2781 and 3.6250. The corresponding significance levels associated with these "t"-statistics are .028 and .0007 (after interpolations are performed). Thus, in order for the licensee to control the pair of standard measurement results at the .05 and .001 significance levels, each individual standard measurement had to be controlled at the .028 and .0007 significance levels. The use of significance levels that were smaller than .05 and .001 resulted in values for the "t"-statistic that were too large, and therefore control chart limits that were too large. This type of anomaly resulted in inaccurate control chart limits for this measurement system, . as evidenced by the following table:

4 Significance "t" -

Control Chart Levels Statistic Limits GE .028 2.2781 84.65, 84.89 method .0007 3.6250 84.59, 84.95 1[5CFR70.57(b)(11) .05 1.990 84.67, 84.87 criteria method .001 3.424 84.60, 84.94

, The 11cen.see's applicable procedure for calculating control chart limits, COI #002, entitled " Chemet Laboratories Control Charting", was found to be inadequate. It included a provision for controlling individual standard

- measurements at significance levels other than .05 and .001. The proce-dure,was reviewed by the appropriate managers, but the review did not detect the provision for using an inappropriate "t"-statistic. Therefore, the licensee's review of the procedure (in accordance with 10 CFR 70.57(b)(5))

was not adequate. The most recent revision (6) of the procedure was issued May 6, 1983, after various licensee representatives reviewed it between April 15 and April 27, 1983. The inspector examined the process resulting in approval of tnis procedure. It appearea that sufficient technical or regulatory expertise was not obtained to properly evaluate the adequacy of this procedure. In this matter, it should be noted that the requirement for an adequate review differs from the requirement for establishing a procedure The failure to perform an adequate review of an already established proce-dure is equally significant and its proper function is required by 10 CFR 70.57(b)(5). The Ticensee's use of a procedure which was inadequately reviewed, in that it contained inappropriate statistical techniques, is a violation (83-28-02).

It should be noted that the licensee has an exemption from 10 CFR 70.57(b)(11)

(License Condition 4.3) for establishing control chart limits. The exemption was granted to allow the licensee to use least scale divisions in calculating control chart limits in lieu of exact .05 and .001 levels of significance for bulk measurements. At no time was the intent of the exemption to allow the licensee to control individual standard measurements at levels signifi-cantly different than 05 and .001.

The second issue concerns the licensee's method for calculating control chart limits associated with uranium hexafluoride weight measurements. The licensee representative who was responsible for these calculations was not on site during the inspection. Therefore, the licensee's statistical methods could not be fully evaluated in this area. This item is unresolved, pending further inspection in this area by the NRC. (83-28-03)

5

3. Unresolved Items Unresolved items are matters about which more information is required to determine whether they are acceptable or may involve violations or deviations. One new unresolved item identified during this inspection is discus. sed in Paragraph 2.

4. Exit Interview The inspedtion scope and findings were summarized on September 23, 1983, with thos'e persons indicated in Paragraph I above.

e e

e 4

o

\

df3 +35

. ~~ C .* 5 . ~

, .GENER AL h ELECTRIC WILMgGTONy ' RJNG DEPARTMENT GENERAL ELEUT}t!G'dd$NY SCUBOM .WILMINGTON, N{TgCLlyy, J (919) 343-5656 .

June 7, 1984 Mr. J. Philip Stohr, Director Division of Emergency Preparedness &

Materials Safety Programs '

f U. S. Nuclear Regulatory Commission, RII Post Office Box 2203 Atlanta, Georgia 30301

References:

(1) NRC License SNM-1097, Docket 70-1113 (2) NRC Inspection Report 70-1113/83-28, 10/31/83 (3) Letter, C. M. Vaughan to J. P. Stohr, 11/29/83 (4) Letter, J. P. Stohr to J. A. Long, 1/5/84 (5) Letter, C. M. Vaughan to J. P. Stohr, 2/3/84 (6) Letter, J. P. O'Reilly to J. A. Long, 5/11/84

s. .

Thank you very much for your letter reporting your review of General Electric's responses to the findings of the referenced inspection.

General Electric reolies pertaining to the items of apparent noncomoliance are alven in the attachment to this letter. We welcome further discussion with your staff on these items or to our related replies for further clarification.

Pursuant to 10 CFR 2.790(d), General Electric Company reguests

.that the attachment to this letter be withheld from public

_ disclosure since this attachment identifies details of General Electric's control and accounting procedures for safeguarding licensed special nuclear material.

Very truly yours, GENERAL ELECTRIC COMPANY h N. A ^ ' '

Charles M. Vaughan, Manager Licensing & Nuclear Materials Managemer.t M/C J26 CMV:cd SGD-I Attachment

.cLR ADOCK 07C01113 le h C. PDR

.