Lessons learned using machine learning to link third body particles morphology to interface rheology

Abstract This paper reports a preliminary investigation on the ability of Machine Learning algorithms to relate the morphology of third body particles to the rheology of the contact interface that created them. A testing campaign is performed on a pin-on-disc tribometer, followed by a comprehensive observation of the worn surfaces. Several Machine Learning algorithms are then used to establish and quantify the logical relations between the rheological and a morphological databases built from this campaign. Success rates and thorough analysis of their predictions are used to validate the general approach and to propose possible improvements. It appears that Machine Learning presents an interesting potential in quantitative tribological analysis if the morphological and rheological databases are properly enriched.