Vanadium detectors offer accuracy and long life

Rhodium-based detectors have commonly been used in PWRs for in-core neutron flux measurement. In contrast, Canadian Candu reactors have used vanadium-based detectors for over 15 years. Vanadium offers a low depletion rate - about twenty times less than rhodium - and relatively simple cross-section characteristics. What is more, vanadium-based detectors can be more mechanically robust than rhodium detectors. But because vanadium has a longer half-life than rhodium, developers had to consider whether vanadium-based detectors would take longer to respond to a changing power distribution.

Westinghouse considered these aspects in developing the Parssel in-core instrument assembly, which is designed to replace the rhodium-based assemblies used in all ABB-CE and B&W PWR designs.

The Parssel ICI design has an overlapping detector arrangement, using long vanadium elements. Westinghouse says the Parssel provides signal levels comparable to rhodium ICIs, so the detector signals can be measured accurately throughout the PWR operating range using existing signal processing electronics. The Parssel ICIs were used successfully in a demonstration programme at the St Lucie 1 reactor in Florida. The plant was designed by ABB-CE and is operated by Florida Power and Light. It is currently operating its fifteenth cycle.

At St Lucie 1, three common causes of mechanical failure in rhodium detectors have been observed.

The first failure mode is the separation of the rhodium wire from the Inconel lead wire. This is related to the difficulty of joining the rhodium to the Inconel wire: their metallurgical properties are so different that a solid union is difficult, so a brazed connection is commonly used. This junction is less capable of withstanding shear stress than the other components of the detector. As the effects of temperature ageing and radiation damage increase the junction becomes even more fragile and prone to failure. The problem is less important in vanadium-based detectors. Since vanadium wire is more ductile than rhodium it can form a hard mechanical joint, which is not subject to the thermal ageing and radiation damage that affects the brazed wire junction used for rhodium.

The second failure mechanism is a direct result of the brittleness of the rhodium detector wire. Rhodium accepts very little work in an unirradiated state before crumbling to powder and radiation damage makes it still more brittle. As time goes by, the rhodium becomes more likely to develop a surface crack, and cracks become more likely to propagate, especially if the detector is subject to lateral stress. Vanadium's higher ductility makes it less prone to cracking. To provide extra protection, in its Parssel design Westinghouse used a solid Inconel-690 outer assembly sheathing, which is expected to virtually eliminate the formation of surface cracks during insertion through small-radius bends or arising from vibration-induced stresses.

The third failure mechanism is not directly related to the rhodium. Short circuits can be caused when stress corrosion allows small perforations to develop in the Inconel-600 sheath covering the detector wires. Using Inconel-690 sheaths should solve this problem.

In Candus, apart from a small number (0.3%) that fail on installation, vanadium detectors typically operate for over 15 years without mechanical failures or loss of signal. Some 10% fail after 10 years of continuous operation because of radiation-induced failure at the junction between the vanadium emitter and the Inconel signal wire. The Parssel design addresses this failure mechanism, using a different joining technique, and Westinghouse also believes the problem is less important in the Parssel assembly because the junctions in the Parssel element are near the end of the active fuel, where neutron flux is very low.

The Parssel ICI assembly used in the St Lucie demonstration project is composed of five vanadium detector elements and one K-type thermocouple. Each vanadium detector comprises a vanadium emitter element mechanically attached to an Inconel-600 signal lead, surrounded by a thin alumina insulation running along the length of the wire and sealed in a thin Inconel-690 sheath. Using vanadium makes an extra instrument segment available, because, unlike rhodium, it does not require a background wire. The ICIs were manufactured by Instrumentation and Sensing Technology of New York, under Westinghouse patents.

In operation, the output from the instrumentation is digitised by the plant computer and examined by Beacon System software. The Beacon System compares measured and predicted detector signals, to adjust its prediction of the reactor power distribution. The Beacon fixed in-core measurement methodology has been generalised to allow short-segment rhodium detectors and Parssel detectors to be used at the same time.

Two Parssel detector assemblies were installed in positions B5 and W16 in St Lucie's core.

The B5 and W16 assemblies both have symmetric partners that house rhodium detectors - there are three instrumented symmetric partners for B5 and seven for W16. This feature was very useful in obtaining accurate estimated neutron flux distributions in the Parssel detector assembly locations.

A series of data sets of SPD signal and core condition measurements was collected through the Beacon interface with the plant computer. The data sets were obtained at regular intervals from the contents of the plant computer and the corresponding 'current' core model files generated by the Beacon monitor function. After dealing with unrelated data processing problems, three sets of data were obtained from the detector elements to six-digit precision, and this was used to analyse the Parssel detector signal response characteristics. One package included data collected during a reactor power transient.

The vanadium detector response model was generated for St Lucie 1 using the same methodology as for the rhodium detector. The model is applicable for fuel to be used in future, including fuel enrichment up to 5wt%, U-235 burnup to 50GWd/tU, gadolinia loading to 8wt% and detector depletion to 25%.

One of the attractive characteristics of the vanadium detector is its low depletion rate - estimated at 0.65% per equivalent full power year (EFPY), compared to 15% EFPY depletion for rhodium detectors. Rhodium detectors must be discharged before the depletion of the string reaches 60-70% — after four or five EFPY. The Parssel vanadium detector takes about 100 EFPY - twice the plant's lifetime - to reach this level of depletion.

The Beacon system depletes the vanadium detectors at the same frequency as the rhodium detectors, so the small impact of detector depletion will be properly included in the Beacon methodology.

The best-estimate neutron fluxes at the assembly locations are obtained using the full design detail ANC nodal model maintained by the Becon system at current core conditions. Beacon computes the instrument thimble flux using the nodal model based on the current plant conditions including fuel burnup, reactor power and power history, CEA position and the resulting xenon and other fission product poison distributions.

The power distribution measurement accuracy depends on:

• The accuracy of the relationship between the observed instrument thimble reaction rate and the local power density as measured by the detector.

• The number of detector strings and elements per string.

• The interpolation technique that infers the power distribution across the reactor from the measurements of power density.

The axial geometry of the rhodium detector assemblies is different to that of the Parssel assemblies. A three-dimensional interpolation methodology was therefore developed so Beacon can use both configurations simultaneously. It will allow failed or highly depleted rhodium assemblies to be replaced by vanadium assemblies as necessary, minimising the capital cost.

To evaluate the effect of the detector numbers and the interpolation technique, a methodology was developed to quantify power distribution measurement accuracy. It operates as follows:

• Two reaction rate distributions are prepared. One represents the 'true' distribution from flux trace measurements obtained via a moveable detector system used in Westinghouse PWRs. The second is a predicted distribution.

• The detector configuration - length, position, number and type of elements - is defined.

• For each configuration, the 'simulated' currents at the detector axial and radial locations are obtained from the 'true' distribution. Predicted currents at the detector axial locations are developed from the predicted power distribution.

• Applying the relationship of measured versus predicted currents - obtained from a 3D spline interpolation methodology across the entire reactor domain - at each segment produces the inferred power distribution.

• The accuracy of the inferred distribution can be judged by comparing it with the 'true' distribution.

Four combinations of instrumentation were evaluated. They were:

• Pattern 1: 12cm x seven detectors per string (as in B&W plants).

• Pattern 2: 40cm x four detectors per string (the rhodium detector configuration).

• Pattern 3: 73cm x five detectors per string (a full Parssel configuration).

• Pattern 4: similar to pattern 2 in odd-numbered thimbles and pattern 3 in even-numbered thimbles (in transition towards a Parssel configuration).

Beacon interpolation logic was used based on a 3D algorithm, instead of less-rigorous axial-radial treatment. This allows flexible and accurate handling of different axial detector layouts.

The accuracy of the inferred reaction rate distribution was evaluated for each pattern. The most accurate was pattern 1. Pattern 3 was ranked second, followed by the Parssel transition, pattern 4. The rhodium detectors, pattern 2, were found to be the least accurate.

The analysis did not consider the effect of detector length on the results - a short detector has a higher measurement variability, due to slight inaccuracies in the axial length and detector mass per length uniformity assumptions. The authors therefore warned that it should not be assumed that pattern 1 will be more accurate than pattern 3. But they say the analysis demonstrates that the Parssel detector array will provide better accuracy than the rhodium detectors in use at St Lucie 1.

Neutron interactions in vanadium and rhodium produce an electrical signal by collecting the charge produced by a beta-decay-generated electron. In rhodium, two isotopes contribute to the beta decay, giving an effective half life of 56s. Vanadium only has contributions from one isotope and its 225.6s half life means that it takes longer to respond to local power changes. That might have affected the accuracy of axial power distribution measurements made while reactor power is changing or during initial startup. However, as the above analysis at St Lucie 1 indicated, the Parssel detector in fact provided a more accurate measurement of the axial power shape than the one currently provided by the rhodium assembly, and allowed the shape annealing factor to be measured to the same accuracy .

During the demonstration, the average reaction rate error over all nodes associated with the Parssel configuration was much lower than the corresponding error associated with the four-element rhodium detector configuration. This was especially true for highly burned fuel assemblies, typically loaded on the core periphery, with 'double-humped' axial power shape cases.

The demonstration showed that the measurement uncertainty of the Parssel assembly was as low as a third of that provided by current rhodium-based assemblies. The variation between measured and predicted vanadium detector currents was much lower than that for rhodium detectors. Since the vanadium response characteristics depend on a single isotope, a simple time-response acceleration algorithm can be used to provide a much quicker detector response to power distribution and level changes. A signal response algorithm for the Parssel detectors was developed during the St Lucie demonstration, and in tests it did allow the vanadium detectors to respond more quickly and accurately than the rhodium detectors to these power changes. The possibility of using the algorithms to achieve an essentially prompt response is being evaluated.