Neural Network Technology for the Rapid Identification of Corrosion Damage in Aging Aircraft.
Abstract
Corrosion damage in aging aircraft is an increasingly critical concern for the U. S. Air Force. Effective, but inexpensive, techniques are needed to identify and evaluate corrosion damage to aircraft structures. The extent of material loss due to corrosion can be reliably measured from x-ray data, but x-ray measurements are costly, slow, and usually require significant aircraft disassembly. Corrosion by-products, which typically occupy more volume than uncorroded material, often causes slight aircraft surface deformations, or 'pillowing.' Pillowing can be measured with various inexpensive, rapid, and nondestructive optical imaging techniques. However, the relationship between percent material loss and 'pillowing' surface deformation is complex, and conventional methods for quantifying the relationship typically lead to unacceptably low correct-detection rates or unacceptably high false-alarm rates. Neural net technology offers a potentially more accurate approach for establishing this relationship. This paper describes the results of a study applying neural net technology to evaluate the percent of material loss and pillowing surface deformation measured with optical imaging techniques.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1997
- Accession Number
- ADA328369
Entities
People
- Paul M. Hartke
- Shing P. Chu
- Steven C. Gustafson
- Theresa A. Tuthill
Organizations
- University of Dayton