Direct link to service navigation Direct link to main navigation Direct link to content

ExDRa

Exploratory Data Science over Raw Data

Application industry: Production; industry-independent
Technology area: Data and service management; machine learning

ExDRa is facilitating the analysis of large amounts of heterogeneous raw data from different sources and helping to ensure more reliable monitoring and fail-safe operation of industrial plants.

The baseline
Typical processes for evaluating data, so-called data science processes, are exploratory in many companies. This means that data scientists first formulate hypotheses, compile the necessary data centrally, and then search for patterns or prediction models in various analyses. It is not known in advance whether the process will generate usable results, so the data is not usually systematically acquired and processed. The process is repeated each time in a time-consuming manner, which results in high costs. Furthermore, due to legal and export restrictions, it is often difficult to transfer data collected at different locations to a central system.

The project goal
In the ExDRa project, a demonstrator is being developed that supports the explorative data science process using heterogeneous and distributed raw data. This includes data, for example, that originates from different computer systems. This should simplify and accelerate the assessment of new data products. Data products in this context are, for instance, predictions or machine learning models obtained during data analysis. As raw data in ExDRa can also be stored and processed remotlely, ExDRa also guarantees the legally compliant processing of sensitive or export-restricted data. The solution will undergo practical testing in the process industry at Siemens AG, for example, in the remote monitoring of pumps. The performance of these pumps is monitored using machine learning methods. ExDRa uses the resulting data directly on-site without sending the raw data to a central location.

Application and practical benefit
ExDRa will eventually improve monitoring models, making plants more reliable and productive, and reducing costs. ExDRa is therefore particularly suitable for remote monitoring of distributed systems, such as those found in the chemical and pharmaceutical industries or oil and gas production. These systems especially require quick reactions to avoid disruptions and downtimes. Following the project, the establishment of a market platform for distributed data products would be conceivable.

Term: 1 June 2019 to 31 May 2022

Consortium: Siemens AG (lead member), Technische Universität Berlin, DFKI GmbH, Graz University of Technology

Contact:
Dr. Claus Neubauer
Siemens AG

E-Mail: claus.neubauer@siemens.com

Website: www.exdra.de