Gear in production plants need to be changed regularly. This is because the gear oils used lose their performance over time, be it due to ageing of the base oils, the degradation of the additives they contain or external influences such as the introduction of foreign substances. If the oil is not replaced in time, there is a risk of machine failure. Up to now, external laboratory analyses have been carried out at regular intervals to provide a reliable statement on the condition of the gear oil. In the process, a large number of different influencing variables are determined directly or indirectly by chemical analysis. However, this procedure is time-consuming and cost-intensive.
Together with gear manufacturers Klüber Lubrication, Josef Bernbacher & Sohn GmbH and Munich University of Applied Sciences, a Fraunhofer EMFT research team is now pursuing a new, efficient approach: by combining sensor technology and machine learning methods, the aim is to establish predictive maintenance. The development partners are using existing approaches to online condition monitoring, in which the current condition of a system, or in this case the lubricant, is monitored using sensors and a secure IT infrastructure. If a previously defined limit value is exceeded, the production management receives a warning with a recommendation for action. However, while this procedure only looks at current and past values, the new system is intended to provide a forecast of when the next maintenance will be due.