Monte Carlo Error Estimation Applied to Safeguards Problems

Monte Carlo randomization of nuclear counting data into N replicate sets is the basis of a simple and effective method for estimating error propagation through complex analysis algorithms such as those using neural networks, constrained response-function fitting or tomographic image reconstructions. The error distributions of properly simulated replicate data sets mimic those of actual replicate measurements and can be used to estimate the standard deviation for an assay along with other statistical quantities. We have used this technique successfully to estimate the standard deviation in the radionuclide masses determined using the tomographic gamma scanner (TGS) and combined thermal/epithermal neutron (CTEN) methods, and to estimate the error in the isotopic composition determined using the gross count material basis set (GC-MBS) method. The effectiveness of this approach is demonstrated by comparison of our Monte Carlo error estimates with error distributions in actual replicate measurements. The main drawback of the Monte Carlo approach is that N additional analyses of the data are required, which may be prohibitively time-consuming with slow analysis algorithms