Exploratory Development of adhesive Bond Flaw Detection.
Abstract
This project sought to establish the feasibility of automatic detection and classification of bond line defects in multiple-layer adhesively bonded structures. Bond strength was not the issue in this work; rather, the detection and classification of a bond line flaw per layer into one of three classes- unbond, void, porosity-independent of part geometry was the main objective. Each part was scanned from one flat surface using a 0.5-inch transducer in the pulse-echo mode. A decision was rendered automatically regarding the existence and classification of defects in each layer for each pulse-echo return in an overall scan pattern. Results show that flaws in multiple-layer structures can be found with accuracies of 91% as to defect versus non-defect discrimination, and 92% as to the correct type of defect. In this work, a false-alarm rate of 19% (non-defects called defects) and a false dismissal rate of 2% (defects called non-defects) was obtained. It is believed that the false alarm rate could be greatly reduced by use of militple-point(contextual) information from contiguous scans. It has been shown that it is feasible and practical to use ultrasound to detect and classify defects and flaws in mulitple-layer adhesively bonded structures, regardless of the layers in which they Lie. The method is, by design, insensitive to moderate variations in transducer orientation and to construction geometry. The accuracies on a blind evaluation test compare very well with the data used to train the ALN classifier system.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1978
- Accession Number
- ADA072724
Entities
People
- Anthony N. Mucciardi
- George A. Alers
- James M. Fitzgerald
- Murray H. Loew
- Richard E. Elsley