Loss-pattern identification in near-real-time accounting systems

To maximize the benefits from an advanced safeguards technique such as near-real-time accounting (NRTA), sophisticated methods of analyzing sequential materials accounting data are necessary. The methods must be capable of controlling the overall false-alarm rate while assuring good power of detection against all possible diversion scenarios. A method drawn from the field of pattern recognition and related to the alarm-sequence chart appears to be promising. Power curves based on Monte Carlo calculations illustrate the improvements over more conventional methods. 3 figures, 2 tables.