Project Description
In today's globalized world, the continuous advancement of know-how in production and development is essential for technological leadership and prosperity in Bavaria and Europe. In order to produce competitively, the Overall Equipment Effectiveness (OEE), i.e. product quality, performance and availability of the machines must be maximized. Aiming at increasing the OEE, M@OK (Machine @ Online Knowledge) links data analysis approaches to expert knowledge about the machines of manufacturers and operators as well as process and alarm data and uses machine learning methods to uncover optimisation potentials and thus make them usable. The main goal of the research project is the reduction of machine downtimes. To this end, data mining will be used to find correlations in the sensor, actuator and message data of a machine that indicate wear and thus the need for maintenance in the near future. In order to enable a targeted, result-oriented data analysis right from the start, human knowledge about the machine and the processes running in it should be incorporated into the evaluation. This has the advantage that only interrelated sensors, actuators and messages are analyzed, which reduces the complexity of the models. The results of the data analysis will be used to indicate the wear of a machine. This requires an automatic analysis of the current sensor, actuator and message data during plant operation and includes an online interpretation of the analysis and the derivation of recommendations for action. The use of agent systems will therefore coordinate the data mining process. The learning of models as well as the testing of the process or the plant on the basis of these models is triggered and carried out by software agents. The subsequent implementation of the data mining results is controlled by further agents. Recommendations for action should be provided to support the operator. Integrated in plants and machines, the agent system and the developed models can increase plant availability.