Lean data in manufacturing systems: Using artificial intelligence for decentralized data reduction and information extraction

In the course of digitization, a drastically increased amount of acquired data in production systems can be observed. Nevertheless, only a minor part of the acquired data is practically used for near real-time analysis and optimization within production systems. This paper introduces a concept for the realization of a decentralized data analysis integration. Therefore, an analysis system using artificial neural networks is conducted at the measurement point in the main supply of a production plant, to classify different operating states. The classification accuracy in all evaluation models is at least 99.82% and proves that it is capable to recognize the operating states of a production machinery reliably. The significantly, without loss of information, reduced amount of data is handed over to a superordinate instance of the production system for further use of data