Start of the DFG project to develop an adaptive control for oil-lubricated systems at the model level (ENDURE) in October 2025
As a consequence of the development of energy-efficient and sustainable technical products, reducing friction losses and preventing wear in lubricated contacts of machine elements has become extremely important. Oil is often used as a lubricant to separate the contact surfaces in these tribological systems. The height of the lubricating film that forms during operation of the system is of central importance. To prevent wear as far as possible, a stable lubricating film must be established. There is also a lubricating film height at which friction is minimal. If the lubricating film height rises above this critical value, friction increases along with the additional fluid resistance. Accordingly, it is desirable to measure and adjust the lubricating film height as accurately as possible in order to achieve a compromise between low friction and low wear. The lubrication condition can be influenced easily and directly via operating parameters such as rotational speed, load, or oil temperature. In contrast, real-time measurement of the lubricating film height has so far been associated with very high apparative effort and is hardly feasible in real applications. One approach is to indirectly calculate the lubricating film height based on directly measurable operating parameters. Empirical approximation equations allow for real-time calculation, but these are prone to error. Thermo-elastohydrodynamic (TEHD) simulations, on the other hand, enable detailed and accurate calculation. However, these TEHD-simulations involve a high level of numerical effort. The aim of the DFG project is therefore to develop a greybox approach that forms the core of an adaptive control system for oil-lubricated systems. The error-prone calculation using empirical lubricating film height equations is corrected by predicting correction factors. To this end, a combination of TEHD simulations and tests on a two-disc tribometer with machine learning methods is used. In addition to implementing the control system, the transferability of the control mechanism from the model level to real systems in industrial applications is also ensured.