Solution monitoring: quantitative benefits to safeguards

This paper investigates how SM (essentially continuous monitoring of solution level, density, and temperature in all key process tanks) can improve loss detection in a formal statistical sense. We use a simulation code developed at Los Alamos (FACSIM) to simulate data from the Rokkasho Reprocessing Plant (RRP), which is now under construction in Japan. We show the possibility of reducing the effect of systematic errors by using solution monitoring data to calculate bias corrections. We divide the effort into 4 activities for a 3-tank system and for a more realistic 15-tank system. We say a tank is in wait mode if its level change is due only to measurement error. Otherwise, a tank is in transfer mode. The 4 activities are (1) monitor each tank for volume loss during each wait mode, (2) monitor each tank for mass loss during each wait mode, (3) monitor each tank for volume loss during each transfer mode, and (4) monitor each tank for mass loss during each transfer mode. The success of this effort will depend largely on the performance of our proposed bias corrections to the volume measurements. To apply bias corrections, we require that some reasonable number of transfers (for example, 20) are known to have no true loss. The effectiveness of the technique will depend on the relative sizes of the random and systematic errors involved because the main outcome is a reduction of the systematic error variances. If there are large variations in true (legitimate) temporary losses such as pipe holdup that would add to the random error variance in our model, the effectiveness will be reduced. Even in such cases we will show there can be improved protracted loss detection and will definitely be improved abrupt loss detection. (author)