A Surface Design Strategy Based on Bayesian Optimization and the Boundary Element Method

  • Ferreira Rocha, Paulo Ricardo (Faculty of Engineering of University of Porto)
  • Couto Carneiro, António Manuel (INEGI)
  • Andrade Pires, Francisco Manuel (Faculty of Engineering of University of Porto)

Please login to view abstract download link

In a world driven by performance and efficiency, optimal designs are sought after in various fields. In the field of contact mechanics, the optimal properties of the interaction between two predefined bodies can be optimised by modifying their surface topography. However, manufacturing surfaces with highly controlled surface roughness and performing experimental testing are both difficult and costly. Therefore, numerical methods capable of designing and testing optimal surfaces emerge as a cost-effective alternative. Recently, patterned surfaces have emerged as a transformative approach to tunable contact, driven not only by growing application demand, validated proofs of concept, and ease of manufacturing. The control of friction through surface design is also being actively investigated, targeting the tuning of friction coefficients, bio-inspired designs, or the imposition of a friction law [1]. These studies open new avenues for the development of systematic design frameworks in this emerging research field. In this context, the present work introduces a computational framework for automatic design of patterned surfaces, integrating the open-source boundary element code Tamaas with the in-house optimization tool, Piglot [2,3]. In this context, the focus is on designing deterministic surface patterns parametrized by a small number of variables. The objective is to enforce a prescribed contact area evolution for a given applied pressure. One of the investigated patterns consists of a grid of spherical asperities, similar to that used in [1]. The proposed strategy is evaluated using a range of objective functions, from single-value targets to full functional prescriptions. The results obtained in this first study lay the groundwork for more advanced applications, involving adhesive and frictional contact, as well as stochastic surface patterns. REFERENCES [1] A. Aymard, E. Delplanque, D. Dalmas, and J. Scheibert, Designing metainterfaces with specified friction laws, Science, 383:200–204, 2024. [2] L. Frérot, G. Anciaux, V. Rey, S. Pham-Ba, and J.-F. Molinari, Tamaas: a library for elastic-plastic contact of periodic rough surfaces, Journal of Open Source Software, 5(51):2121, Jul. 2020 [3] R. P. Cardoso Coelho, A. F. Carvalho Alves, T. M. Nogueira Pires, and F. M. Andrade Pires, Piglot: An open-source package for derivative-free optimisation of numerical responses, Journal of Open Source Software, 9:6652, Jul. 2024.