Speaker
Description
Orano Mining is evaluating the potential of the CeBr3 spectrometric gamma ray logging probe developed by Advanced Logic Technology (ALT) for estimating uranium concentration in roll-front deposits, where decay chain disequilibrium disrupts its relationship with gamma total count rate. The Nuclear Measurement Laboratory of CEA IRESNE, in Cadarache, France, is working on automatic prediction algorithms capable of fully exploiting the shape of the recorded gamma spectra. However, the current number of logged wells is insufficient to properly train and evaluate such algorithms. This calls for the creation of a database of simulated gamma logs, using the Monte Carlo N-Particle (MCNP) transport code. Until now, we have carried out analog simulations of single CeBr3 spectroscopic logging probe measurements, with a detector response provided by MCNP pulse-height (F8) tally. However, several requirements makes building the database with this approach impractical. First, each individual simulation is time-consuming since a good statistics is needed in every energy bin. Secondly, actual borehole measurements are conducted every 10 centimeters over several meters, making the number of probe vertical locations to simulate considerable. Thirdly, we need to model a wide range of stratified geological profiles to maximize the diversity of the training samples. Given this explosive computational cost, designing a more efficient yet sufficiently accurate simulation procedure is crucial. We propose a two-step method, where the flux reaching the probe surface and the detector response to this flux are simulated separately. The bulk of simulation time reduction is achieved in the first step. Two options are available to perform fast probe surface flux estimation. The first hinges on the point-detector (F5) tally, which provides a semi-deterministic estimate of the flux at a point. The variance reduction brought by this tally accelerates the completion of each individual simulation. Furthermore, its pointwise nature allows leveraging the fact that we model measurements taken at close successive vertical locations. Several point detector tallies can effortlessly be arranged along the vertical path of the probe, thereby providing flux estimation at multiple locations simultaneously. Modeling the probe surface as a point might be too much of a simplification however, raising concern about potentially significant bias in the point-detector tally. This motivates the investigation of a variant of the first step based on the surface flux (F2) tally. Although using this tally does not address the first driver of high simulation time (achieving good statistics), it still allows taking advantage of the spatial contiguity of probe locations during logging. Since probe surfaces overlap along its vertical trajectory, it is redundant to estimate the flux on each of them independently. By stacking 10-centimeter high cylindrical surface flux tallies all along this path and later merging their output appropriately, it is possible to get the probe surface flux at multiple locations in one shot. We assess the accuracy of the two variants of our two-step method by comparing the resulting spectra with reference spectra yielded by the original analog approach. The comparison is performed on an initial dataset of simulated single-point measurements in homogenous wells, with varying geological and drilling parameters to ensure the validity of the comparison across a broad range of situations. We use the results of the comparison to develop a post-simulation correction method that models bias as a function of energy, for each variant of the two-step approach. Finally, we evaluate the corrected two-step approach on an actual gamma ray log acquired in a uranium in-situ recovery (ISR) mine.
