Bertrand PEROT
SGN
(Société Générale pour les techniques Nouvelles)
1 rue des Hérons
78182 St Quentin en Yvelines CEDEX
FRANCE
Phone : (33) 1 39 48 61 34
Fax : (33) 1 39 48 78 55

Christian PASSARD & Alain MARIANI
CEA
(Commissariat à l’Energie Atomique)
Centre d’Etudes de Cadarache
13108 S^t Paul-lez-Durance CEDEX
FRANCE
Phone : (33) 4 42 25 24 66
Fax : (33) 4 42 25 23 67

Christophe HEULIN
COGEMA
(COmpagnie GEnérale des MAtières Nucléaires)
1 rue des Hérons
78182 St Quentin en Yvelines CEDEX
FRANCE
Phone : (33) 1 39 48 52 75
Fax : (33) 1 39 48 51 63

Jean-Luc FERRAGUTO
MELOX
(a COGEMA & FRAMATOME company)
BP 124
30203 Bagnols sur Cèze CEDEX
FRANCE
Phone : (33) 4 66 90 64 13
Fax (33) 4 66 90 64 69

Solid alpha bearing waste characterization can be performed using an active technique based on the use of a pulsed neutron generator. This method has been implemented with the PROMETHEE system for the measurement of technological wastes in the French MOX fuel fabrication plant MELOX. The paper first presents an experimental validation of neutron transport calculations performed with a three-dimensional Monte Carlo code. The second part deals with the application of these calculations to the qualification of the industrial measurement device PROMETHEE installed in MELOX.

Active measurement with a pulsed neutron generator is a very sensitive method for alpha bearing waste characterization. The generator produces 14,2 MeV neutron bursts which are slowed-down and thermalized in the cavity of measurement, then inducing fissions in the nuclear material. The method consists in detecting only prompt fission neutrons. In order to avoid the counting of interrogating neutrons delivered by the generator, a physical energy discrimination in the detection blocks is used (cadmium filters).

More precisely, the signal due to fast interrogating neutrons disappears in a few hundred microseconds, which correspond to their thermalization time. It then becomes possible to induce thermal fissions, and to detect at that time the fast neutrons emitted with the fission process. This signal has a longer decay time and lasts typically several milliseconds (see Fig. 1). It decreases with the diminution of the thermal interrogating neutron flux, due to leaks and neutron captures in the cavity materials. This difference of decay time between the initial background noise, due to interrogating fast neutrons, and the useful signal corresponds to the duration difference between thermalization and thermal scattering. It is the basis of the method. and explained its name: «Differential Dieaway Technique (DDT)».

Simulation calculations can be used either to define and develop measurement devices based on such methods, or to ease qualification during calibration procedures. In fact, these calculations allow the evaluation of physical phenomena which have a significant influence on the measurement, such as self-shielding, matrix (with the meaning of waste matter) effects and position of the fissile material inside the drum. They give a support for measurement interpretation, and are used to define appropriate corrections. The work presented in this paper is composed of two aspects :

• The first one concerns the validation of DDT calculations with the Monte Carlo modeling code. This work has been performed at the Nuclear Measurement Development Laboratory (CEA) using an R&D PROMETHEE cell/1,2/ (see Fig. 2).
• The second one concerns the application of Monte Carlo calculations to the pre start-up calibration tests of MELOX industrial measurement device PROMETHEE.

The first step of qualification does not take into account matrix effects. Fissile material samples have also been measured in an empty cavity (R&D PROMETHEE system). Such a measurement result is shown in Fig. 1.

Qualification was performed in two steps:

• The first step consisted in looking for the temporal shape of the signal in a quantitative way. More precisely, we evaluated the exponential decay constant of the background noise due to the neutrons coming from the generator, and the one of the useful signal due to fission neutrons (see Fig. 1). This allowed us to be confident in our measurement cavity and detection block modelization (accurate geometrical design and material compositions). It appears that only a detailed description of the device can lead to a good agreement between experimental and calculated decay constants.
• The second step consisted in comparing, in a quantitative way, the experimental and calculated detection efficiency and active useful signal. This allowed us to investigate other parameters like the emission of neutron sources (generator, calibration sources), self-shielding in the fissile material (described below) or electronic set-up.

We processed several sensitivity studies in order to identify the most influent parameters of the modelization and to improve accuracy.

• As already mentioned above, the precision of the modelization, more particularly the geometrical design of the cavity, is crucial: small details can modify notably the results of the calculation, especially for the fission rate calculation. For example, the neutron emission point surroundings in the generator itself have a great influence on the slowing-down (presence of electrical insulation oil). Another example is the presence of stainless steel in the cavity (fixing plate of the drum, generator wrapping, screw of the structure...) which absorbs thermal neutrons and then modifies the interrogating flux.
• The precise knowledge of the generator neutron emission is of first importance. However, this parameter is very difficult to assess precisely because it depends on the functioning point of the apparatus (high voltage, target electric current, pressure in the sealed tube), the tritium target mechanical abrasion and its radioactive decay. This neutron emission has been evaluated with two different methods in a good agreement.
• Self-shielding : this phenomenon consists of the absorption of interrogating thermal neutrons near the surface of the fissile material, which leads to a loss of interrogation in the middle of the volume, and to an important decrease of the signal in case of thick masses.
• Electronic set-up : the threshold setting of electronic and gamma background noises leads, in particular, to the loss of useful signal. This one is more or less important according to the collecting mode which is used (charge or current).

The comparison between experimental and calculated results is summarized in Table I.

Taking into consideration the uncertainties described above, especially on the neutron emission of the generator (as well as the uncertainty of the calibration sources of ²⁵²Cf and ²³⁵U), the agreement between experimental and calculated values is very satisfactory.

This device is a part of a global measurement system, including also a gamma spectrometry cell, used to assess the technological wastes produced by the MOX fuel fabrication plant MELOX. The basic design has been performed by the CEA (French Atomic Commission for Nuclear Energy) and COGEMA’s engineering company SGN was the main contractor during realization.

PROMETHEE allows the measurement of very low quantities of plutonium (less than 10 milligrams) for a wide range of matrix materials and weight. This neutron measurement cell is constituted of a moderating graphite, lead and polyethylene cavity, a neutron generator (supplied by a French company: SODERN), 75 helium neutron detectors included in 15 detection blocks made of polyethylene surrounded by cadmium and bore filters.

MELOX industrial PROMETHEE cell has been qualified by pre start-up tests. They involved the evaluation of calibration coefficients required for any further measurement interpretation. This evaluation was performed by the assessment of standard well known drums and nuclear sources. Because of the importance of matrix effects, 10 reference 120 liter waste drums have been supplied in order to cover a large panel of the kind of wastes produced in the MELOX plant. Table II lists the main characteristics of these different drums.

For each matrix, a calibration curve has been assessed (similar to the one of Fig. 3) with a range of standard fissile masses covering the scope planed by the operators. These test sources where set in the center of the drum. The slope of the calibration curve, corrected for self-shielding effect, is called the Calibration Coefficient (CC). It characterizes the sensitivity of the device in terms of c/s/mg of ²³⁹Pu for a punctual central position in the drum. If one wants to calculate the sensitivity for a homogeneous distribution of the fissile material, a Homogeneous Localization Factor (HLF) must be used. This coefficient is defined by the ratio of the signal obtained with several masses evenly distributed in the drum and the signal obtained in the center of the drum. The Calibration Coefficient for an Homogeneous localization (CCH) is the product of CC and HLF.

To illustrate the importance of matrix effects and fissile material localization with the DDT method, a factor of 3 can be observed between the CCH of drums n°5 and n°6. These drums have indeed a different chlorine content, this element having a significant neutron capture cross section. For drum n°6, a factor of 3 is observed between the signal in the center of the matrix and the homogeneous answer (HLF = 3).

At first, the calculation qualification method used for the R&D PROMETHEE equipment has been applied to MELOX industrial cell. We started also by a comparison of experimental and calculated results in the empty cavity:

• In a qualitative way with the temporal shape of the signal,
• In a quantitative way with the detection efficiency and active useful signal.

This first step allowed us to check the good modelization of the detailed design of the industrial cell. Particularly, we took into account important differences (with regard to measurement) between the laboratory cell and the MELOX cell: for instance, massive stainless steel antiseismic protections and drum handling devices. This phase allowed also the validation of self-shielding calculations (see Fig. 3).

Regarding the calibration coefficient, a good agreement has been recorded between calculation and experimental results thanks to the experience of accuracy improvement described above.

Fission rate calculations in the center of the matrix may require a long calculation time because the Monte Carlo method is less efficient when the concerned region is of a small dimension (e.g. punctual drum center). Therefore we used variance reduction techniques, such as geometry splitting or ‘‘Russian Roulette’’, to obtain satisfactory statistical precision and reliable results. For the absorbing matrixes (boron, stainless steel, chlorine in great amount), the calculation time required is quite long whatever the bias, because of the small number of particle reaching the matrix center. Calculation duration is much smaller with homogeneous fissile material distribution, even for very absorbing matrixes.

The relative differences between experimental and calculated results for the detection efficiency (e ), the Calibration Coefficient for Homogeneous distribution (CCH) and the Homogeneous Localization Factor (HLF) are presented in Table III.

The general agreement is excellent considering the large range of matrix materials and weights. The main causes of uncertainty are due to the precision of matrix modelization. It turns out that calculations are very sensitive to the geometrical description, principally for very heterogeneous matrixes, and to the exact composition of the materials, especially those absorbing neutrons.

Simulation is widely used for the design of measurement devices. Our work proved that a qualified calculation tool can also bring up many advantages for industrial calibration tests:

• it allows to evaluate, before the tests, the amplitude of the phenomena mentioned above, giving help for the interpretation of the measurements and the definition of corrections;
• it enables to reduce the number of tests, particularly thanks to self-shielding calculations. It is indeed possible to extrapolate the whole calibration curve from only one or two measurement points;
• it enables to assess the impact on the measurement performances due to last time modifications which always occur in an industrial context.

Monte Carlo calculation have been successfully applied to the simulation of active neutron measurements for alpha bearing waste characterization. The modelization has been qualified at first with the R&D equipment PROMETHEE. The main sources of uncertainties have been identified and we finally obtained an excellent accuracy.

Afterwards, calculations have been applied to the qualification of an industrial PROMETHEE device installed in MELOX for the characterization of technological wastes. This qualification was performed successfully, proving a sensibility below 10 mg of plutonium for a large scope of matrix materials and weights. The calculations allowed us to estimate self-shielding, matrix effects and localization of the nuclear material before the calibration tests. A further comparison with experimental results showed globally a good agreement and allowed us to check matrix modelization.

We finally proved the adequacy of PROMETHEE system for low level alpha bearing waste characterization in an industrial scale. We also showed the interest of the simulation tool for design as well as for qualification of active neutron measurement devices. These calculations are widely applied for the definition and the calibration of such equipments in COGEMA or CEA plants.

1. J. L. Ma & al, "Development of advanced device for low level waste assay based on 14 MeV neutron interrogation", Proceedings of Institute of Nuclear Materials Management, 35^th annual meeting, Naples, Florida (1994).
2. L. Martin-Deidier & al, "Trends and R&D in France to improve the performances of activity measurements systems for the reprocessing low level wastes", GLOBAL 95, International Conference on Evaluation of Emerging Nuclear Fuel Cycle Systems, Versailles, France (1995).

BACK