Authors: S M Robinson E R Siciliano J E Schweppe
Publish Date: 2008/05/04
Volume: 276, Issue: 2, Pages: 447-453
Abstract
Developing and testing improved alarm algorithms is a priority of the Radiation Portal Monitor Project RPMP at PNNL Improved algorithms may reduce the potential impediments that radiation screening presents to the flow of commerce without affecting the detection sensitivity to sources of interest However assessing alarmalgorithm performance involves calculation of both detection probabilities and false alarm rates For statistical confidence this requires a large amount of data from drivethrough or “dynamic” scenarios both with and without sources of interest but this is usually not feasible Instead an “injectionstudy” procedure is used to approximate the profiles of drivethrough commercial data with sources of interest present This procedure adds netcounts from a predefined set of simulated sources to raw grosscount drivethrough data randomly selected from archived RPM data The accuracy of the procedure — particularly the spatial distribution of the injected counts — has not been fully examined This report describes the use of previously constructed and validated MCNP computer models for assessing the current injectionstudy procedure In particular this report focuses on the functions used to distribute the injected counts throughout the raw drivethrough spatial profiles and for suggesting a new class of injection spatial distributions that more closely resemble actual cargo scenarios
Keywords: