Probabilistic models and decision rules in residual material clearance or recycling

The European chemometrics standard says: 'The placing of the analytical sensor in these situations (process analysis within industrial production, and in situ measurements on contaminated land) constitutes the taking of a sample'. Therefore, each clearance measurement result must be considered, as a singular data pertaining to a statistical sample then, the result of each sampling plan is a set of data which treatment is the support of the decision making to select between the available options: waste management, unconditional clearance or, if any, additional material treatment (decontamination). Once all data are collected, they are statistically summarized and graphically analyzed. However, those analyses usually do not include the data fitting to probabilistic models. This paper will analyze the usual probabilistic models: normal distribution, lognormal distribution and Weibull distribution. The optimization of the clearance process depends strongly on the segregation and decontamination activities when they are necessary. However the lowering of clearance levels is less important when the probabilistic approaches were consistent. The simulations of those activities permit us to understand the appropriate probabilistic models in this process and the non parametric test appears robust and more consistent with the applied probabilistic models. Finally, the dose calculations permit us to understand the necessity to determine consistent mass factors avoiding unnecessary efforts and facilitating the waste volume and worker dose savings.(authors)