A deep learning approach to the texture optimization problem for friction control in lubricated contacts

Alexandre Silva,Veniero Lenzi,Sergey Pyrlin,Sandra Carvalho,Albano Cavaleiro,Luís Marques

DOI: https://doi.org/10.48550/arXiv.2212.06523

2022-12-13

Computational Physics

Abstract:The possibility to control friction through surface micro texturing could offer invaluable advantages in many fields, from wear and pollution reduction in the transportation industry to improved adhesion and grip. Unfortunately, the texture optimization problem is very hard to solve using traditional experimental and numerical methods, due to the complexity of the texture configuration space. In this work, we apply machine learning techniques to perform the texture optimization, by training a deep neural network to predict, with extremely high accuracy and speed, the Stribeck curve of a textured surface in lubricated contact. The deep neural network was used to completely resolve the mapping between textures and Stribeck curves, enabling a simple method to solve the texture optimization problem. This work demonstrates the potential of machine learning techniques in texture optimization for friction control in lubricated contacts.