Total Measurement Uncertainty for NDA for SNM in Process Materials and Waste

---

View full version of this Technical Paper (PDF)

D.R. Davidson, M.J. Koskelo, B.M. Gillespie, R.D. M^cElroy
CANBERRA Industries
Meriden, Connecticut, USA

T. Nishida
Toyo Corporation
Tokyo, Japan

Abstract

Non-Destructive Assay for safeguards and waste applications requires accurate determination of the Pu and U content of the samples. NDA systems must be designed to handle a variety of sample sizes, chemical forms, isotopics and matrices, all of which complicate the analysis of the measurement data. Canberra has evaluated regulatory requirements and nuclear material types worldwide to identify a standard set of NDA instruments that address these diversified applications. As these systems are becoming more common, the need for verification and certification of their performance for various sample types increases. Performance data for most of Canberra's safeguards and waste have been published in recent years. Canberra has combined all of these data into a composite evaluation of the total measurement uncertainty (TMU). This paper includes a summary of these TMU and each of the components that produce this overall uncertainty for two standard NDA systems: Waste Drum Assay Systems and the Waste Assay Scanner. This approach for determining the TMU for an assay has been accepted by the US Department of Energy and the approach has been implemented in Canberra’s standard neutron and gamma NDA software. Also, the major contributors to the TMU have been identified and new assay techniques have been identified to detect and minimize the TMU.

1. Introduction

Uranium and plutonium bearing materials are routinely measured in facilities related to the production of uranium or MOX fuel for nuclear power plants. The measurements are made for a variety of reasons, including satisfying the materials balance reporting requirements for international safeguards, verification of uranium enrichment or plutonium isotopics in process materials, and determination of the presence or absence of uranium or plutonium in waste and scrap. While it is possible to do some of these measurements with destructive techniques, such as chemistry or mass spectrometry, Non-Destructive Assay (NDA) methods are becoming more and more common. NDA systems are very desirable because they typically provide a much faster result than the destructive methods. They also permit measuring the entire material stream so that no uncertainties are introduced by the sampling process, and do not require workers to open containers that contain potentially hazardous materials.

To handle all of these diverse needs, an NDA system must be designed in such a way that it can handle the samples presented to it in many different chemical forms, with different sample sizes, and with varying amounts of interfering material present, all of which complicates the analysis of the measurement data. In recent years, Canberra has embarked on a goal of developing a standard set of NDA instruments that address these needs[1]. These instruments range from portable uranium enrichment analyzers [2] to large waste measurement systems [3]. As these systems are becoming more and more common, it has become obvious that their performance for various sample types needs to be verified and certified. Without a performance verification, it is difficult for the regulatory agencies to accept the results from such instruments. Performance data for most of Canberra's safeguards and waste instruments under varying conditions has been published over the last few years [4-9]. We have recently combined all of this data into a composite evaluation of the total measurement uncertainty (TMU) that can be expected in large neutron and gamma NDA systems. This paper will include a summary of these TMU and each of the components that produce this overall uncertainty.

2. TMU Approach

TMU can be estimated using two basic approaches [10-11]. In the first approach, normally labeled Type A, the measurement technique uncertainties are estimated by a standard deviation of the sample measurements themselves. This works well if the samples are somewhat uniform relative to the quantity being measured, and where the measurement result can be expected to follow some predictable distribution. In the case of NDA measurements, this type of an approach is typically not feasible because it is not clear what is the overall distribution of the measured quantity whether it is total Pu content or alpha activity.

Therefore, we have chosen to use the Type B approach instead. In a type B evaluation, the measurement uncertainty is evaluated from a composite pool of information that includes previous measurements of test cases, experience and general knowledge of the behavior and properties of the materials in the NDA samples, calibration data and other uncertainties assigned to reference data taken from handbooks and data bases. For example, we have tested the effect of various attenuating materials on the net signal as a function of the placement of a small Pu source, and evaluated the overall expected uncertainty assuming a certain probability for the placement of such sources. This approach to TMU places bounds on the possible drum content which is consistent with the reported Pu mass of the drum. For example, if a 250 kg steel drum was reported to contain 1 gram Pu total (~300 nCi/g) the actual content is between 190 and 460 nCi/g at the 2s confidence level.

To illustrate the approach further, we will discuss TMU results for two standard NDA techniques that are routinely used in Japan: the Waste Drum Assay System (WDAS) and the Waste Assay Scanner (WAS). The WDAS is a passive coincidence counter which measures the spontaneous fission neutrons emitted by the even isotopes of plutonium to provide an accurate and reliable assay result of the plutonium mass when combined with known isotopics. The Figure 1. WDAS installed at PNC in Tokai.

WDAS[12] shown in Figure 1 was jointly developed by Canberra and Los Alamos National Laboratory (LANL) and Authorized For Use (AFU) by the IAEA. The WDAS utilizes a 4p detector geometry consisting of ³He detectors embedded in a high density polyethylene moderator to maximize efficiency and provide a uniform response. The Add-A-Source (AAS) correction technique, which was developed by LANL, measures the matrix effects on the counting rate from a small ²⁵²Cf source on the outside of the drum to correct for the matrix effects on the inside of the drum.

Figure 2.
Waste Assay Scanner.

The WAS which is shown in Figure 2 is designed to provide full quantitative analysis and isotopic ratios for small cans to 320L drums containing transuranic material, and fission/activation products. The WAS is often referred to as a Segmented Gamma Scanner (SGS) because the HPGe detector scans along the axis of the drum and analyzes vertical segments to identify "hot spots" in the waste. The system is based on Canberra’s new modular design to accommodate site-specific operator measurement requirements.

Modules include:

• transmission source module for matrix corrections.
• horizontal detector movement for closer sample-to-detector coupling.
• optional Low Energy Germanium (LEGe) detector module for measuring the plutonium isotopic composition and the uranium enrichment. The Multi-Group Analysis Code (MGA)[13] developed by Ray Gunnink, analyzes the LEGe data for the plutonium isotopic composition and MGAU[14] used to determine the uranium enrichment.
• dose rate meter for automatic measurement of surface dose rate.
• manually loaded or automatic conveyor modes of operation.

The components of the TMU that were evaluated for the WDAS and WAS are listed in Table 1, and testing results are discussed in following sections.

Table 1. Components of the TMU for WDAS and WAS.
WDAS	WAS
Neutron Counting Statistics	Gamma Counting Statistics
Matrix Effects/Source Distribution	Matrix Correction/Source Distribution
Isotopics/Chemical Form/Multiplication Effects on the Neutron Assay	N/A
Calibration	Calibration
Background	Background

3. Results and Discussions

WDAS Results

Table 2 indicates the expected total mass errors assuming all error contributors can be represented by a Gaussian distribution. The minimum is the worst case underestimation based on source positioning and the maximum is the worst case overestimation based on source positioning.

Table 2. Contributions for TMU for standard WDAS.
Matrix Error Source	125 kg Steel	200 kg Graphite	50 kg Combustibles	250 kg Lead	250 kg Lead
Counting Statistics	3%	2%	2%	3%	1%
AAS Correction	2%	5%	10%	2%	2%
Min/Avg Ratio	0.80	0.93	0.80	0.80	0.80
Max/Avg Ratio	1.20	1.07	1.20	1.20	1.20
Nominal 1 s Matrix Error	7%	2%	7%	7%	7%
Multiplication Effects	---	---	---	---	1%
Background effects	5%	0%	0%	11%	1%
Isotopics Error	6%	6%	6%	6%	6%
Calibration Error	1%	1%	1%	1%	1%
Typical 1 sigma uncertainty for the primary waste matrix types.
Reported Mass	1.0 g	1.0 g	1.0 g	1.0 g	10.0 g
Nominal TMU (1s)	11%	9%	14%	15%	9%
Range of Sample Mass consistent with assay result at 2s Limits
Minimum Value	0.76 g	0.82 g	0.71 g	0.69 g	7.9 g
Maximum Value	1.30 g	1.18 g	1.35 g	1.36 g	12.8 g

Note: the values in the above table are based on measured or published data with the exception of the graphite drums. The graphite performance has been estimated based on the known interactions of neutrons in graphite. The worst case estimate for the graphite drums has been determined by MCNP simulation of a 320 kg graphite drum.

Counting Statistics: Contributions to the counting statistics include the neutron emission rate, the cosmic background, and AAS correction statistics. The contribution of the counting statistics to the TMU are normally insignificant for Pu masses less than 1 g. For the WDAS, the counting statistics is equivalent to the precision.

Matrix/Source Distribution Effects: To correct for matrix and source distributions, the WDAS performs an AAS matrix correction for each drum. As shown in Figure 3, the AAS correction is accurate to within +/- 2% for uniform source and matrix distributions with the exception of a few matrix types (e.g. highly moderated and poisoned samples or high Z metals). We have therefore concluded that the use of the AAS correction introduces an additional uncertainty of 2%.

Figure 3.
Add-A-Source Correction Results for the WM3100 WDAS. The white line is the uncorrected relative response of the WDAS and the black line is the AAS corrected relative response.

Deviations from homogeneity introduce additional errors. For purely moderating matrices, samples with a relative hydrogen content of less than 0.02 g/cc have source location error of less than +/- 10%. For higher moderator drums the effects increase dramatically. If mote than a single point source exists in the sample, the source distribution begins to look like a homogeneous distribution, therefore, the error contribution decreases relative to a single point source. This effect is characterized by the min/avg and max/avg ratios. For the 125 kg steel matrix, the lowest result is 20% lower than the result from a uniform distribution. Hence the min/avg ratio of 0.80 in the table. The maximum result is 20% high. Assuming that the worst case extremes represent a 3s case, the nominal 1s matrix error is 7%.

Isotopics/Chemical Form/Multiplication Effects: If the isotopics or chemical form are unknown, a multiplication correction is impractical with the WDAS. An upper bounding value was determined for nonmetallic Pu based on data collected at LANL and PNC. If the isotopic abundance has been adequately determined (via MGA) the error contribution will be less than 5% due to the isotopic data.

For samples containing significant quantities of Pu fluorides or nitrates, multiplication effects can be significant for masses as low as 50 grams. However, when these compounds are present the chemical form tends to vary making the standard multiplication correction difficult. The presence of nitrates or fluorides is indicated by small real to totals ratios.

Background Effects: In addition to counting statistics, the cosmic ray background is matrix dependent. As the metal content of the drum increases, the cosmic ray interaction rate increases resulting in a background count rate dependent on the sample mass. As shown in Table 2, the contributions to the TMU from background effects are 11% for 1 g Pu in a 250 kg lead matrix compared to 1% for 10 g Pu in the same 250 kg lead matrix. Light metals such as aluminum or graphite have little impact on the background levels, however, drums containing significant quantities of steel and lead require an adjustment to the background subtraction.

Calibration Errors: The base mass calibration was determined by using known ²⁵²Cf sources and confirmed using Pu standards at the Los Alamos National Laboratory. Error contribution due to the calibration for a benign matrix is less than 1%.

WAS TMU Results

Table 3 lists the TMU for the WAS/SGS obtained by summing all of the errors in quadrature for the same matrices and densities based on a 1s error propagation. For the purposes of counting statistics the following table assumes an assay of approximately 5 grams of Pu.

Table 3. Contributions to TMU for standard WAS.
SGS Error Source	Combustibles (25 kg)	Graphite (50 kg)	Steel (100 kg)	Lead (200 kg)
Counting Statistics	5%	8%	10%	15%
Matrix Correction	2%	4%	6%	10%
Source Distribution	12%	14%	23%	28%
Calibration	5%	5%	5%	5%
Nominal TMU (1s)	14%	17%	26%	34%

Counting Statistics: The counting statistics for the measurements are directly propagated through the assay using the standard spectroscopy techniques following normal operating procedures. For very low gram quantities of Pu or U. the counting statistics can become a significant part of the TMU, at higher gram ranges these errors become insignificant.

Matrix Correction: Transmission correction is used to measure the average density of a segment. This directly accounts for both the density and Z of the material. The only matrix related effect which is not accurately covered by the matrix correction is accounting for non uniform matrix densities in a segment. It is assumed that extreme matrix non uniformities such as lead shields or other high Z materials would be identified in a non destructive examination (NDE) of the drum. Therefore matrix non uniformities would take the form of voids caused by non uniform filling or settling of materials with significant structures or higher density materials in combination with lower density materials. If this error is kept separate from other potentially synergistic measurement errors (particularly source non uniformities) then the total measurement uncertainty associated with matrix correction will cause a maximum error of less than 10%.

Calibration Errors: There are two primary factors which contribute to the calibration error. The first is the error on the calibration source. Typically these sources have a total uncertainty of 4 - 5%. Obviously it is impossible to have a calibration accuracy which is better than the calibration sources. The second factor is the fit of the software calibration routine to the calibration data. If good calibrations are not performed this can become a significant error, however for carefully performed calibrations using good standards in appropriate configurations the overall calibration error is usually in the 5% range.

Source Distribution: This measurement error is small for low density matrices and increases to become the most significant measurement error for high density matrices. The statistical aspect of this uncertainty is based on the probability of a source being in any particular point in the drum. Experimental data shows that for equal volume elements of a drum the measurement error tends to change smoothly from the worst case low to the worst case high value. Therefore the worst case source positioning errors can appropriately be considered as 3s errors. Table 4 lists the expected worst case errors for several different material types and densities.

Table 4. Worst case errors for the WAS for various matrices.
	Combustibles (25 kg)	Graphite (50 kg)	Steel Scrap (100 kg)	Lead Scrap (200 kg)
Min/Avg Ratio	0.75	0.6	0.4	0.25
Max/Avg Ratio	1.3	1.5	2	3.5

Conclusions

This approach for determining the TMU for an assay has been accepted by the US Department of Energy after review by an expert panel. The TMU method has also been implemented in Canberra’s standard neutron and gamma NDA software. As a result of these measurements, the major contributors to the TMU have been identified and new assay techniques have been identified to detect and minimize some of these sources of error, especially for gamma measurements. The first new technique is a non uniformity detection software technique (NUDS) which provides indices to the level of both source and matrix non uniformities. To date this technique is only being used as information to be considered during data review of assay results. Review is underway to add this detection technique to neutron-based NDA measurements. The second technique is a differential peak absorption correction technique which is used to identify and provide some correction for significant source non uniformities or significant source densities. This technique can also be used as an indicator of a significant measurement problem to limit the acceptance of the assay results and in this way limit the total measurement.

References

[1] M. Zebarth, D. Davidson, M. Koskelo and G. Gardner, "Instrument Development for Safeguards Measurements", Presented at the 17th Annual Meeting of the INMM Japan Chapter, Tokyo, Oct 15-16, 1996.

[2] G. H. Gardner, M. Koskelo, R. L. Mayer II, B. R. McGinnis, M. Möslinger and B. Wishard, "The IMCA: A Field Instrument for Uranium Enrichment Measurements", presented at the 37th Annual INMM Meeting, Naples, Florida, July 28-31, 1996,

[3] D. Davidson, R. McElroy, D. Brochu and M. Villani, "A New 208 Liter Drum Neutron Coincidence Counter with Add-A-Source Matrix Correction", p. 960, Proceedings of the 36th Annual Meeting of the INMM, Palm Desert, CA, July 9-12, 1995.

[4] M.J. Koskelo and G. Gardner, "An Improved U-Pu System for Field Pu Measurements", p. 441 of the Proceedings of the 19th ESARDA Annual Symposium on Safeguards and Nuclear Material Management, Montpellier, France, May 13-15, 1997.

[5] B. Gillespie and M. Zebarth, "Comparison of a Variety of Gamma Attenuation Correction Techniques for Different Waste Matrices", Proceedings of the 14th ESARDA Annual Symposium on Safeguards and Nuclear Material Management, Salamanca, Spain, May 1992.

[5] B.M. Gillespie, "Detection and Correction of Inhomogeneities in Drum Waste Assay Systems", Proceedings of Waste Management '94, Feb 27 - Mar 3, 1994, Tucson, AZ.

[6] B. Gillespie, D. Davidson and J. Verplancke, "A Very Low Level Waste System for Fission Products and Transuranics", Proceedings of the 36th Annual Meeting of the INMM, Palm Desert, CA, July 9-12, 1995.

[7] D.R. Davidson, R.D. M^cElroy, D.B. Brochu and I.J. Koskelo, " Validated NDA Systems for Accurate Characterization of Transuranic and Low Level Waste", Presented at the ICEM'97 Conference, Singapore, Oct 12-16, 1997.

[8] D.L. Haggard, D. Davidson and C.J. Lemons, "Performance Tests on PNL's Transportable Neutron/Gamma Waste System", Presented at the 4th NDA and NDE Waste Characterization Conference, Salt Lake City, Utah, Oct 24-26, 1995.

[9] J. Verplancke, D. Davidson, B. Gillespie, M. Koskelo, F. Bronson, R. Gunnink, J.-L. Ma, J. Romeyer-Dherbey, S. Abousahl and M. Bickel, "Applying MGA for Waste Characterization", Presented at the WM'95 Conference, Tucson, AZ, Feb 26 - Mar 2, 1995

[10] "Guide to the Expression of Uncertainty in Measurement" ISO/TAG 4/WG 3 Report, 1992.

[11] Barry Taylor and Chris Kuyatt, "Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results", NIST Technical Note 1297, 1994.

[12] D. Davidson, R. M^cElroy, D. Brochu and M. Villani, "A New 208 Liter Drum Neutron Coincidence Counter with Add-A-Source Matrix Correction", Proceedings of the 36th Annual Meeting on Nuclear Material Management, July 1995, Palm Desert, California, p. 960.

[13] R. Gunnink, "MGA: A Gamma-Ray Spectrum Analysis Code for Determining Plutonium Isotopic Abundances". Lawrence Livermore National Laboratory Report UCRL-LR-03220, April 1990.

[14] R. Gunnink, W. Ruhter, P. Miller, J. Goerten, M. Swinhoe, H. Wagner, J. Verplancke, M. Bickel and S. Abousahl, "MGAU: A New Analysis Code for Measuring U-235 Enrichments in Arbitrary Samples". Presented at the IAEA Symposium on International Safeguards, Vienna, Austria, Mar. 8-14, 1994.

View full version of this Technical Paper (PDF)

---

Top of Page