Statistical Analysis of Eddy Current Data from Fastener Site Inspections (Preprint)
Abstract
Recent work on reliably detecting and characterizing cracks in multi-layer airframe structures has used modeling and simulation to extract features from raw eddy current data, and to assist in the evaluation of probability of detection (POD). This paper focuses on the statistical analysis of the data from these studies. Hit/miss, linear, and physics-inspired methods are employed to evaluate POD. The Box-Cox transformation is used as a remedy for violations of homoscedasticity. In addition, a bootstrapping method is introduced for confidence bound calculation on a 2nd order linear model. The objective of this work is to provide on insight how different models and assumptions impact POD evaluation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 2012
- Accession Number
- ADA564933
Entities
People
- Inseok Park
- Jeremy S. Knopp
- John C. Aldrin
- Ramana V. Grandhi
Organizations
- Air Force Research Laboratory