Application of artificial neural networks for feature recognition in image registration.

Image registration is a process of aligning two or more images taken at different times or using different sensors by transforming the same area into one coordinate system. Imaging conditions, image and area deteriorations from repeated sectioning, are serious impediments to successful image registration. The application of artificial neural networks for feature recognition is introduced to the field of metallurgy to assist in an automated approach to image registration of metallurgical microstructures. Low susceptibility to feature deterioration, often occurring during serial sectioning, is demonstrated and assessed. The process of image registration using an artificial neural network to aid in feature segmentation is performed using computer generated shapes and a metallurgical microstructure.

A comparison of EDS microanalysis in FIB-prepared and electropolished TEM thin foils.

This paper reports the results of a fine-probe EDS microanalytical study of cellular precipitation in a Cu-Ti binary alloy. Compositional profiles across the solute depleted Cu-rich FCC lamellae and the Cu4Ti lamellae within isothermally formed cellular colonies were measured in a FEG-TEM from thin-foil specimens prepared by conventional electropolishing and by a technique using a Ga+ focused ion-beam (FIB). The Cliff-Lorimer ratio method, with an absorption correction, was employed to quantify the compositions. Two FIB samples were prepared with different orientations of the lamellae with respect to the ion-milling direction. The compositional profiles across the Cu-rich FCC lamellae and the Cu4Ti compound lamellae in both the FIB-prepared samples and the electropolished sample were, within experimental error, numerically equivalent. The composition of the Cu4Ti compound phase lamellae was very close to the ideal stoichiometric composition of 20 at % Ti. It is concluded that for this system, and for the specimen preparation procedures used in this study, the Ga+ ion-milling process has had no detectable effect on the chemistry changes across the interlamellar interface at the scale studied. These results indicate that the possible sources of chemical artifacts which include redeposition, preferential sputtering and ion-induced atomic migration can be minimized if several precautions are taken during milling in the FIB. Consistent with previous investigators, it was also found that the ion-milling process does introduce significant structural artifacts (e.g., dislocations) into the softer FCC Cu-rich phase compared with a specimen produced by conventional electropolishing.

Synthesis and properties of metallic glasses that contain aluminum.

The synthesis and properties of a class of metallic glasses containing up to 90 atomic percent aluminum are reported. The unusual formability of the glasses and their structural features are pointed out. Mechanical properties including tensile fracture strength and Young's modulus are reported along with crystallization temperatures. The unusually high strengths of the aluminum glasses can be of significant importance in obtaining high-strength low-density materials.