A study about the sensitivity of expert-system method for characterizing flaws in the NPP fuel

The experience from CANDU type NPP operation highlighted the fact that the fuel failures appear to have a very low rate. Even so, the evolution of the fuel failure state remains a current issue with direct implication for NPP operation safety. To monitor the fuel breakdown in NPPs there exist special techniques and methods of investigation, as for instance, RMHTS (radiation monitoring of heat transport system) and FFLS (failed fuel localization system). Moreover, an expert-system method was developed and applied at INR Pitesti as a complementary analysis method for failed fuel, consisting in determining at a given moment the number of defects in reactor core. This is done using the data from measurements of coolant activity levels from fission gases (Xe, Kr) or iodine isotopes. This method is based on the results of tests performed at CRNL with failed fuel, CANDU type NPP operational experience, theoretical studies on simulating failed fuel behaviour as well as on the authors' experience built up in this field. This work aims at improving the sensitivity of the expert system method. The quantities and the parameters used by this method belong to the following categories: 1. General characteristics of the reactor core, namely, reactor thermal power, linear power on the failed fuel element, information on degassing system, information concerning the equilibrium state; 2. Fuel element characteristics, i.e., burnup degree per failed fuel element, the number of fuel elements in reactor core, the length of pellet column in a fuel element and the fuel element weight; 3. Cooling system characteristics, i.e., the gamma ray activity of the coolant from noble gases (Xe, Kr) and iodine, coolant flowing rate, efficiency of the coolant purification system; 4. Parameters for handling, analysis and interpretation of data, i.e., the diffusion coefficient of the fission gases in failed fuel element, the coefficient of fission gases' released into the coolant, gamma activity values of the coolant. For each of these parameters studies were performed to identify those with major effects in final results evaluation. These are the linear power per failed fuel element, which is a quantity escaping an exact determination (small defects and large defects are classified according the values of the fission gas release rate); the coefficient of fission gas rate of release in the coolant as determined at CRNL (10-8 - 10-4) is too high in comparison with the computations and the analyses carried out restricted the range to (10-7 - 10-4); - the diffusion coefficient of the fission gases from the failed fuel element which is the most important parameter appearing in the simulation mathematical models for failed fuel behaviour. The dependence of the diffusion coefficient on the linear power (Pl) was improved by using new data fitting methods. Also important are the measurements of coolant activity as well as a set of parameters related to the NPP operation. By using this modifications presented above an improved experts system method resulted which was applied. A comparison between the Xe-133 activity measurements' data and correspondingly the number of defects calculated for a three years period from the failed fuel element discharge at the dates reported by the Cernavoda NPP's team shows a rather good agreement