Data science techniques applied to in-situ¬ XRD measurements of copper under tribological load
Friction is inherently a multi-scale phenomenon: the mechanics that govern macroscale sliding originate at the nanoscale, and often leave permanent microstructural alterations even if no wear occurs at the surfaces of the contacting bodies. In metals, frictional energy is primarily dissipated through the material underneath the contact. X-ray diffraction (XRD) analysis is a well-established non-destructive technique, which provides information about a material’s local microstrain accumulation, grain-to-grain interaction or grain size evolution. Thus, combining tribological experiments with the continuous tracking of diffraction patterns, acquired at the same location, is a promising avenue for gaining insights about the connection between surface stresses and resulting microstructural alterations. We used KIT’s ANKA synchrotron source to collect XRD patterns of copper, which was slid against a sapphire sphere between each X-ray exposure. The tribological experiments we conducted spanned normal loads of 1 to 4 Newtons, which correspond to maximum Hertzian contact pressure between 420 MPa and 670 MPa, and sliding speeds of 0.5 to 8 mm/s.
However, analyzing these patterns in a traditional manner can often be a daunting task due to their sheer number: >13,000 in our case. Furthermore, finding meaningful correlations with the tribological data which was acquired in parallel, makes this effort extremely time consuming, if at all possible. To solve this issue, popular data science techniques were adapted and applied. Tribological patterns, previously hard to identify, were extracted and correlated with changes in the material’s structure.