In a world undergoing accelerating digital transformation, the value of modelling and simulation in tribology are becoming increasingly apparent [1]. One ‘holy grail’ of tribology is the ability to predict performance using a virtual copy of a laboratory experiment, machine component, or even engineering system. Digital twins are precise, virtual copies of machines or systems [2]. As well as the common issues with data acquisition and processing, tribological systems pose additional problems in terms of the complexity of the physicochemical processes occurring between sliding surfaces [1]. Thus, in order to construct physics-based digital twins in tribology, improvements in fundamental understanding of all the processes occurring in a system is first required.

The modelling tools now available to tribologists enable processes to be studied from the smallest to the largest scales. Over the last few decades, atomistic simulations, such as molecular dynamics (MD) and density functional theory (DFT), have given unique insights into the nanoscale structure and friction of lubricants and additives [3]. In many cases, processes that occur at these scales govern the macroscopic tribological response. However, continuum methods, such as finite element analysis (FEA) and continuum fluid dynamics (CFD) are required to model systems at the component and system-levels [1]. Thus, for a digital twin in tribology, multiscale/multiphysics modelling techniques are urgently required. Depending on the systems and conditions, either hierarchical or concurrent coupling approaches may be required [1].

At the same time, as in many other areas of science and engineering [4], machine learning is poised to play an increasingly important role in tribology. The potential to supplement or even replace physics-based models with statistical approaches is attractive given the extraordinary complexity of the physicochemical processes that occur in tribological systems [1]. This would require the deployment of extensive sensor arrays to gather sufficient data to accurately predict friction and wear.

In reality, a combined physics-statistical approach is likely required to yield the most effective digital twins in tribology. This meeting will bring together experts in both of these approaches from across the world in a virtual environment. The aim is to discuss the state of the art in various techniques needed to create a digital twin and gather insight in how to progress towards this aim. During the lectures, all participants will be able to learn about techniques they might not be familiar about; during the round-table discussion, they will brainstorm and reflect on the future challenges in constructing digital twins in tribology.