Can unsupervised machine learning boost the on-site analysis of in situ synchrotron diffraction data?

Abstract

We explore the use of unsupervised machine learning to analyze in situ diffraction data of an additively manufactured Ti-6Al-4V alloy. The model is trained on a dataset consisting of four thermal cycles. The α/α’-β phase transformation results in a steep gradient of the reconstruction error, whose derivative is applicable to detect periods of fast phase transformation. Moreover, the latent space features of the autoencoder correlate well with the volume fractions of α/α’ and β. The methodology can be implemented to monitor phase transformation kinetics on-site during experiments at synchrotrons without the need of continuous training or manual data labeling.