Machine learning models of the transition from solid to liquid lubricated friction and wear in aluminum-graphite composites

Abstract We study wear and friction of dry and lubricated aluminum-graphite composites and the transition between lubrication regimes. Using Principal Component Analysis, we perform dimensionality reduction for the 14 material and tribological variables to find clusters in friction and wear data. Five standalone and one hybrid supervised regression models were developed to predict friction and wear of lubricated composites. ML analysis identifies lubrication condition and lubricant viscosity as the most important variables. Unlike dry, graphite content has a reduced impact on the tribological behavior with liquid lubricants. The incorporation of graphite in the matrix of aluminum alloys enables them to run under boundary lubrication and run for more extended periods with lower friction even after the lubricant is drained out.