Some Algorithms for Evaluating Nuclear Data and Generating Uncertainty Covariance Matrices

Nuclear data is evaluated using a realistic covariance matrix of observable uncertainties in experimental data. This matrix is built on a model of the piecewise constant energy dependence of experimental uncertainties. Using this model least square problems can be solved without inversion of ill-conditioned matrices with a rank equal to the total number of experimental points and with inversion of a matrix with a rank equal to the number of energy zones of the function. Within the least squares approach the problem of the combined processing of already known evaluations and new experimental data is considered. The algorithm for construction of the covariance matrixes of the uncertainties of the different reactions multigroup cross sections, correlated because of the use of the same reference cross sections in the relative measurements, is described.