Pattern recognition of temperatures produced in fast reactor fuel elements

The use of pattern recognition in interpreting temperature noise from monitoring thermocouples in LMFBRs has been studied. Blockages and other malfunctions in fuel-coolant passages produce temperature noise through the random mixing in coolant fluid turbulence. Other, benign, sources of temperature noise can make it difficult to interpret the noise. Pattern recognition can effectively unscramble the mixed signal and give an accurate measure of the size of the original malfunction. The study uses temperatures calculated by a turbulence model, and shows how a pattern-recognition classifier can be built to estimate the size of blockage from measurable features in the noise signal. (author)