Big Data Toolbox aimed at promoting Industry 4.0 in Germany
The increasing use of automation in machine and plant construction has inevitably led to a large growth in the number of industrial production processes being recorded and monitored by sensors. If the vast quantities of data this generates could be centrally evaluated in real time it would be possible for the results to be used to optimize internal processes and drastically reduce production costs. Unfortunately, the most commonly used data analysis tools are simply not designed to handle such enormous amounts of real time data. The SAKE project has been set up to resolve this problem by developing a framework specifically designed to analyze these vast streams of data. By implementing prefabricated modules it will also be possible to use individual applications in a number of different roles. The modules will be evaluated in real-life industrial environments by the project's industrial partners.
Turning RDF format and machine learning into fast analytical results
The information generated by the sensors is represented within the framework in the RDF format – a standard developed by the World Wide Web Consortium (W3C) for describing logical statements. As the format is based on a simple and clearly defined structure, strongly heterogeneous data streams can be consolidated and subsequently analyzed using modern machine learning processes. Because the RDF data gathered by the sensors will be modularized, it is no longer necessary to process all of the data but only the aspects of it relevant to the purpose of the analysis. This leads to a massive reduction in the workload required of the system resources and makes it possible to process the size of data streams required in Industry 4.0.
Automatic language generation provides high degree of user friendliness
The typical end user has generally found it difficult to fully understand the results of traditional statistic processes but the SAKE framework makes it possible for analytical results and the causes of errors to be processed in natural language. This is done via a combination of modern learning methods and automatic language generation processes. At the end of the process the user is not only given information about potential operational errors, but also about the cause of these errors.
The 3 year project is funded by the German Federal Ministry of Economics and Technology (BMWi). In addition to the University Leipzig and the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS, industrial partners such as AviComp Controls GmbH, Heidelberger Druckmaschinen AG, Ontos GmbH and USU Software AG are also part of this project. It is through these partners that the framework modules can be practically evaluated in real-life industrial environments.
Fotos © Heidelberger Druckmaschinen AG