Automatic feature extraction from micrographs of forged superalloys

Abstract

The manual determination of metallurgical parameters of forged superalloys can be dramatically improved by automatic, image-processing-based feature extraction. With the proposed methods, the typical errors during grain size estimation for Inconel 718 and Allvac 718Plus ™ , caused by twins and other artifacts like scratches, can be eliminated. Different processing strategies for grain size estimation allow the application of a wide range of ASTM grain size numbers from G3 to G12 with the typical variations in the manifestation of metallurgical details and the magnification-related limitations of image quality. Intercept counting strategies show advantages for samples with pronounced anisotropy and can produce detailed statistics on grain orientation. In addition to a single grain size number, grain size histograms offer a more precise description of the material properties.