Monte Carlo Engine for EMI Survey Analysis
Abstract
This project produced and demonstrated proof-of-principle for a Monte Carlo tool that can calculate performance measures under any given set of survey conditions and analysis methods. The existing Monte Carlo tool at AETC was improved by incorporating more realistic noise models, inherent variability of UXO items, and the ability to utilize different discrimination algorithms. The tool was used to show the potential improvement of a hybrid approach to discrimination analysis over an unconstrained or weighted unconstrained approach. It was also used to investigate sensitivities to field survey conditions and noises and, even in the limited proof-of-principle runs, clear guidance on the strong effect of at least one system parameter (timing error) was obtained. Signal-to-noise ratio is a critical parameter for successful UXO discrimination, and accurate noise models are a key part of any Monte Carlo analysis. In this project we improved existing noise models by incorporating correlation scales observed from field data. These field data, however, typically contain only aggregate information, which makes it difficult to discover the magnitude of the various components involved. ESTCP project MM-0508 "Quantification of Noise Sources in EMI Surveys" is aimed at producing the data which will make these component determinations possible, and in fact we have used some preliminary data from that project in this work. At the start of each iteration in the Monte Carlo code, target response values (beta values) were randomly drawn from a library and synthetic data was produced using the dipole model. These synthetic data did not, therefore, exhibit non-dipolar effects, something which could be incorporated in future work using a more sophisticated forward model. Beta values were drawn from a list of 98 possible targets, representing four UXO types: 20mm, 60mm mortar, 81mm mortar, and 3 inch Stokes mortar.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2006
- Accession Number
- ADA476424
Entities
People
- Jonathan Miller