Creating configurations for testing radiation portal algorithms using factor covering combinatorial designs

Most ports in the United States now inspect a large number of goods for radioactive cargo to address the potential smuggling of illicit nuclear material for terrorism. The U.S. Department of Homeland Security (DHS) has sponsored research to develop systems for detection of illicit nuclear materials at the ports. We present a systematic approach for testing the efficacy of different algorithms used in portals to identify Special Nuclear Materials (SNM) in presence of a number of other confounding factors including background radiation and naturally occurring radiation material (NORM). Algorithms are sensitive to several factors, and for realistic situations, the number of factors is large and exhaustive testing using full factorial designs is impractical. Our method, instead, extends factor covering combinatorial design to create a smaller number of configurations to cover all pairwise or 3-way factor combinations. Further, our method leverages the inherent hierarchy amongst the factors to iteratively develop a more nuanced model using a notion called seed designs. This paper presents representative factors and their values which impacts the performance of the detection algorithms, provides details of the new testing method, and describes how the approach can be used to generate various configurations for evaluating the performance of various detection algorithms.