Triboinformatics: Machine Learning algorithms and Data Topology methods for surface roughness, friction, and wear

Abstract

Friction and wear are very common phenomena found virtually everywhere. However, it is very difficult to predict the tribological (i.e., related to friction and wear) structure-properties relationships from the fundamental physical principles. Consequently, tribology remains a data-driven, mostly empirical discipline. With the advent of new Machine Learning (ML) and Artificial Intelligence (AI) methods, it becomes possible to establish new correlations in tribological data to better predict and control the tribological behavior of novel materials. Hence the new area of triboinformatics has emerged combining tribology with Data Science. We review ML algorithms used to establish correlations between the structure of metallic alloys and composite materials, tribological test conditions, friction, and wear. We also discuss novel methods of surface roughness analysis involving the concept of data topology in multi-dimensional data space, as applied to the macro- and nanoscale roughness. Other triboinformatic approaches are considered as well.