Clearing Unexploded Ordnance: Bayesian Methodology for Assessing Success

Anderson, K K.

Pacific Northwest National Laboratory (U.S.)

关键词: Risk;    Uncertainties;    Unexploded Ordnance;    Bayesian Modeling;   

DOI  :  10.2172/882983
RP-ID  :  PNNL-15420
RP-ID  :  AC05-76RL01830
RP-ID  :  882983

美国|英语

来源: UNT Digital Library

【 摘 要 】

The Department of Defense has many Formerly Used Defense Sites (FUDS) that are slated for transfer for public use. Some sites have unexploded ordnance (UXO) that must be cleared prior to any land transfers. Sites are characterized using geophysical sensing devices and locations are identified where possible UXO may be located. In practice, based on the analysis of the geophysical surveys, a dig list of N suspect locations is created for a site that is possibly contaminated with UXO. The suspect locations on the dig list are often assigned into K bins ranging from ``most likely to contain UXO" to ``least likely to be UXO" based on signal discrimination techniques and expert judgment. Usually all dig list locations are sampled to determine if UXO is present before the site is determined to be free of UXO. While this method is 100% certain to insure no UXO remains in the locations identified by the signal discrimination and expert judgment, it is very costly. This paper proposes a statistical Bayesian methodology that may result in digging less than 100% of the suspect locations to reach a pre-defined tolerable risk, where risk is defined in terms of a low probability that any UXO remains in the unsampled dig list locations. Two important features of a Bayesian approach are that it can account for uncertainties in model parameters and that it can handle data that becomes available in stages. The results from each stage of data can be used to direct the subsequent digs.