Data-driven model for estimation of friction coefficient via informatics methods

As technologies progress, the development of new mechanical systems demands the rapid determination of friction coefficients of materials. Data mining and materials informatics methods are used here to generate a predictive model that enables efficient high-throughput screening of ceramic materials, some of which are candidate high-temperature, solid-state lubricants. Through the combination of principal component analysis and recursive partitioning using a small dataset comprised of intrinsic material properties, we develop a decision tree-based model comprised of if-then rules which estimates the friction coefficients of a wide range of materials. This datadriven model has a high degree of accuracy with an R ² value of 0.8904 and provides a range of possible friction coefficients that accounts for the possible variability of a material's actual friction coefficient.