Intelligent Signal Smoothing and Cumulative Sum Control Applied to Nuclear Source Search

Precise identification of activities associated with the use of nuclear materials – illicit or legal - is the main mission of nuclear security. Data analytics significantly contributes to that cause by recognizing patterns of interest and making inferences. In this thesis one new algorithm involving smoothing is introduced and another method which involves cumulative sum control (CUSUM) is used for abrupt change detection in gamma ray measurements. Such changes may designate activities pertaining to nuclear material, and thus, their fast and accurate detection is of paramount significance. The first method is based on the synergism of fuzzy number and data smoothing to label each datapoint in time series measurements as containing a potential threat or merely noise. The second method simply applies a derivative to a cumulative sum control. The methods were tested using five time series of real-world gamma measurements with infused peaks of varying magnitude and varying location. The ratio of the number of correctly identified peaks is recorded as a number between zero and one inclusive. On average the intelligent smoothing method identified 66.6% of the peaks correctly and the cumulative sum control method identified 83.4% of the peaks correctly.