A probabilistic approach to data fusion for non-destructive structural integrity assessment
Abstract
Information obtained from the analysis of non-destructive testing [NDT] can be employed as an effective material characterization tool for conducting an integrity assessment of structural systems. Data fusion has proven to be an accurate method of synthesizing NDT data from various sources and developing a comprehensive assessment of the condition of a structural system. A B-52 test plate with artificially induced faults was developed by Boeing to conduct a benchmark study of different NDT techniques and their ability to detect different faults. Three different tests were conducted for this purpose including Pulse Echo, Eddy Current, and Magnetic Resonance test. The tests yielded a single value for each point in the Pulse Echo test and two components (real (X) and imaginary (Y)) for the Eddy Current and the Magnetic Resonance. A total of 12 images were developed from the test results where the components of Eddy Current and Magnetic Resonance were synthesized to produce a total of 7 grayscale bitmap images. The grayscale values were normalized and probability values were then assigned using the Gauss Kernel density estimator. Bayes Theorem was the probabilistic technique applied for data fusion. The theorem was applied to the images using a computer algorithm, which produced a final image that reflected the probability that damage exists at every point in the tested specimen. A comparison with other data fusion methods was made to check the reliability of the data fusion based in Bayes theorem.
Subject Area
Civil engineering
Recommended Citation
Carrasco Martinez, Gilda Lizeth, "A probabilistic approach to data fusion for non-destructive structural integrity assessment" (2004). ETD Collection for University of Texas, El Paso. AAIEP10528.
https://scholarworks.utep.edu/dissertations/AAIEP10528