HY-80 Intercritical Metallography
This repository of images was used for a thorough quantitative metallography investigation of the intercritical temperature range for HY-80. It also contains corresponding segmented images between ferrite and other phases (generally austenite/martensite and carbides) developed as a comparison of different methods for performing the investigation. Some images are higher quality than others; however, the results show that a simple, yet effective, machine-learning segmentation model outperformed simple image thresholding techniques. This machine-learning method is very accessible to other researchers without requiring special knowledge on developing and implementing machine-learning and computer vision algorithms that are often utilized for these types of investigations. The results of this investigation were also shown to be effective for adjusting CALPHAD models for predicting phase fractions following industrially relevant heat treatments. The CALPHAD phase fractions were modeled for HY-80 using Thermo-Calc 2022b with the TCFE12 database. DSC testing results are also included that were collected using a TA Instruments Q600 SDT. A sigmoidal baseline was fit and the phase fraction was calculated from the austenite transformation endotherm.
Steps to reproduce
HY-80 steel samples were solutionized and normalized followed by a water quench. Long heat treatments were performed at a variety of temperatures to explore the intercritical temperature range. The samples were then prepared for optical metallography using standard sample preparation techniques. Tint etchants used were 10 w/v% sodium metabisulfite (Na2S2O5) or a mixture of 10 w/v% sodium metabisulfite and 3 w/v% potassium metabisulfite (K2S2O5) (Beraha's 10-3). Quantitative metallography was performed using a variety of techniques to segment the tinted phase (ferrite) from the untinted phases (everything else). The results will vary somewhat due to inherent subjectivity using the techniques; however, other researchers should be able to approximately recreate my results. The included thresholded himages can be compared to those that others develop to compare accuracy.