SEDL / STP / STP1165-EB / STP25106S

Artificial Intelligence for Twin Identification

Friel, JJ
Technical director and product manager, Princeton Gamma-Tech, Princeton, NJ

Prestridge, EB
Technical director and product manager, Princeton Gamma-Tech, Princeton, NJ

Pages: 11    Published: Jan 1993

Download this paper for $25 PDF (420K)          View License Agreement
Abstract

The presence of twin bands in some alloys makes grain sizing by automatic image analysis difficult. Twin bands appear similar to grain boundaries; therefore, the measured grain size is too small. Although an experienced metallographer can ignore twins while measuring grain size manually, it is often desirable to use automatic image analysis to save time and increase reproducibility. Because twin bands are so similar to grain boundaries, traditional image processing techniques cannot distinguish them. However, the techniques of artificial intelligence (AI) can be used to impart knowledge to a computer that enables it to distinguish among microstructural features. This method was tested on standard grain size charts and on real specimens. The results showed that the use of artificial intelligence for twin identification and removal works sufficiently well to lead to an accurate measure of grain size.