Compaction of Lattice Data : Improved Efficiency in Nuclear Core Calculation

Abstract: Westinghouse Electric Sweden AB’s three-dimensional reactor core simulation program POLCA uses a large number of tables containing various fuel dependent data, such as cross sections, pin power maps with power distribution etc. POLCA uses quadratic and linear interpolation to extract the values needed for the simulation. However, finding the correct values to interpolate between takes time. This master thesis describes a method of compacting the tables by removing values, in order to shorten the needed simulation time. This is done so that no significant accuracy is lost in the interpolations. The method also finds deviant values and replaces these with new, interpolated values. The thesis shows that approximately 90 % of all values can be removed without losing significant accuracy. These results are however heavily dependent on the choice of accuracy loss criterion; a lower allowance for accuracy loss lowers the amount of values which can be removed sharply.