Accelerating Analysis of In-Situ X-Ray Videos of Alloy Solidification using Machine Learning

The role that gravity plays on the solidification behaviour of equiaxed dendrites in metallic alloys has long been an area of research interest. Experimental data has previously been successfully obtained in pursuit of an answer to this question, most notably via the acquisition of in-situ X-ray solidification videos under different gravitational conditions, including in microgravity through the use of sounding rockets. However, analysis of these videos has, until now, proven to be a manually-intensive, time-consuming and costly exercise, with the image characteristics of these videos demanding pixel-by-pixel and frame-by-frame measurement. This has made attempting to gain insights from already completed experiments very expensive. To rectify this situation, an automated analysis system has been developed, using the result of the MASER-13 sounding rocket flight for primary experimental inputs. Through the use of machine learning and a number of bespoke image analysis and dendritic behavioural estimation techniques, two in-situ, X-ray videos of Al-Cu were successfully analysed: one under terrestrial gravitational conditions and the other under microgravity. The obtained and discussed dendritic measurements highlight that terrestrial dendrites exhibit greater out-of-plane orientations and an increased prominence of secondary arms, despite solidifying at an identical rate to their microgravity equivalents. These results demonstrate the effectiveness of the developed solution while the 72 hour timescale utilised provides a significant reduction in the time and resources required to obtain dendritic measurements, when compared to the manual alternative.