Constructing the continuum-physics core of a digital twin for tribological contacts under boundary and mixed lubrication conditions

Project description

Friction and lubrication are crucial for the efficiency and durability of machines. As many components are increasingly designed to operate at their absolute performance limits, boundary friction is becoming more common, with lubricating films only a few atomic layers thick or even being completely squeezed out of the contacts. Traditional models of thermo-elasto-hydrodynamic lubrication reach their limits when lubricants are only present as very thin films, which is particularly the case under extreme loads in components such as roller bearings and gear pairs. The challenge is to make experimentally inaccessible friction-causing processes in technical systems that occur at the atomic level, such as viscosity changes of lubricants in nanoscale friction gaps and the sliding of the resulting solidified lubricant over material surfaces, calculable.

This is where the ERC-funded LubeTwin project comes in: it aims to better understand and optimize lubrication in machines and technical systems that rely on highly stressed friction contacts. To this end, a digital twin is being developed that maps all lubrication regimes – from dry friction to hydrodynamic lubrication. Using advanced molecular dynamics simulations and machine learning, LubeTwin will link the behavior of molecules at the nanoscale with macroscopic friction in technical components. Simulation methods and computational models that describe the mechanisms on different scales will be combined in a single tool in such a way that the friction-causing characteristics can be recorded and the behavior of the friction system predicted. The project will use automated workflows for high-throughput molecular calculations of friction contacts under a variety of load parameters. In addition, the latest generation of machine-learned interatomic potentials (MLIPs), which offer quantum mechanical accuracy at a fraction of the computational cost, will be used.

Once the cause-and-effect relationships between the atomic processes and the energy-consuming friction for the respective technical system have been described mathematically, the system can be optimized. The proposed approach will improve the understanding of lubrication under high loads and help predict optimal conditions for extremely low friction and minimal wear. 

Work packages:

1. Molecular dynamics simulations: Development and execution of simulations that map atomic processes in lubricants under extreme loads (e.g., in rolling bearings and gear pairs). This includes precise calculations of lubricant rheology and tribochemical reactions in nanometer-sized lubricating films.
2. Physical material models: Development of constitutive equations for the rheology and tribochemical reactions of lubricants under gigapascal pressure and their integration into continuum equations.
3. Automated high-throughput simulations: Implementation of a workflow for the automated calculation of friction contacts under various load parameters and use of the latest machine learning methods (MLIPs) to reduce computing costs.