Modeling of neutron flux density anomalies in the automated system of «Shelter» Object’s nu-clear safety control.

Analysis of the dynamics of neutron flux density (NFD) from fuel-containing masses (FCM) in the «Shelter» Object shows the presence of values that exceed the average values for different observation periods. Identification of such values by the criterion of «anomaly / non-anomaly» will allow excluding uninformative events from the array of observations. Or, in the case of the anomaly confirmation, it will allow forming effective actions for decision-making in order to eliminate the consequences of such events. To solve the problem of detecting anomalous measurements, now there is utilized the theory of statistical solutions which is based on the use of parametric methods. The utilization of these methods requires a priori information about the nature of the distribution of the measured process and its parameters. In order to find an effec-tive solution to the problem of detecting and eliminating anomalous measurements, it is necessary to know the statistical characteristics of normal and anomalous components. This paper proposes statistical criteria for estimating anomalies in time series of NFD which have different approaches to the formation of observation intervals, power and reliability of anomaly detection. Depending on the type of distribution of the array of observations, a set of criteria is proposed. These criteria are most expedient to use when checking the anomaly of series levels for the exponential distribution, the Poisson and Weibull distribution. The capacity of the criteria has been evaluated depending on the sample size. The article defines the accuracy of determination of anomalies by criteria with the known values of the mean and dispersion in the studied sample. As the result, it is recommended to use the Grubbs test to study the anomaly of the sample levels with n>700, and the Dixon and Smolyak-Titarenko criteria for the samples with n<50. The utilization of optimal criteria depending on the characteristics of the studied samples will increase the mathematical significance of the obtained results and, as a result, will improve the quality of management decisions and nuclear safety on the «Shelter» Object as a whole.