Distributed evaluation of high-frequency measurement data from industrial production for quality optimization and condition monitoring

Publiziert November 2019.

Downtimes of machines and equipment in manufacturing are expensive and should be avoided as far as possible. A frequent cause of machine downtime is damage to bearings due to wear, corrosion or overload. These causes of damage were detected using high-frequency acceleration and structure-borne sound sensors. This information can be further used for preventive maintenance. In addition to early damage detection, data-driven models can be used to optimize high-dimensional plant parameters. In industrial use cases, there is usually an organizational separation between the plant manufacturer and the plant operator. The plant manufacturer has a strong interest in aggregating data from multiple customers in order to generalize and transfer results of the analysis. However, it is usually not possible to transfer the data of high-frequency measurements from the plant operator to the plant manufacturer. In addition to technical challenges, organizational hurdles, such as data confidentiality, must also be considered. The core idea in this project is the distributed evaluation of high-frequency data streams in a multi-stage approach. By pre-evaluating data streams, the amount of data is reduced and fewer conclusions can be drawn about any trade secrets of the plant operator. The challenge is to be able to assemble the partial models from different plants into an overall model, taking into account the differences between the plants.

The contribution of this project is the distributed evaluation of the data, so that only pre-evaluated models are transmitted by the plant operator. [Jayaraman17] present a similar approach. However, this one does not make reference to high-frequency data analysis. The very active community around high-frequency data analysis for condition monitoring [Tandon99] has so far not addressed hierarchical data analysis with an aggregation of individual models. The explosive nature and topicality of this issue is also underlined by the fact that in the Industrie 4.0 platform the condition monitoring use case in a multi-stage plant operator/system integrator/component manufacturer scenario is to be particularly investigated in order to validate the concept of the Industrie 4.0 management shell.

Through test series on real machines and plants at the unfunded project partner Dieffenbacher, a data set of relevant size will be created (initially on a laboratory scale). Due to the sensitivity of the data, the release of an anonymized data set is only possible by the unfunded project partner Dieffenbacher.

The exploitation of the scientific results is carried out by Fraunhofer IOSB. Methods and algorithms will be used in further research and industrial projects. Given the expected completion in mid-2020, a publication for CASE 2021 (IEEE International Conference on Automation Science and Engineering) is planned. Exploitation for the use case will initially be carried out by the unsubsidized project partner Dieffenbacher. Dieffenbacher is active in the field of mechanical and plant engineering and manufactures press systems, as well as complete production lines for the wood, automotive, aerospace and recycling industries. Dieffenbacher employs over 1,600 people and operates at 16 production, service and sales locations worldwide.

Project period

01.11.2019 – 30.06.2020


Dr.-Ing. Julius Pfrommer (julius.pfrommer@iosb.fraunhofer.de), Fraunhofer IOSB-ILT

Dr.-Ing. Thomas Usländer (thomas.uslaender@iosb.fraunhofer.de) , Fraunhofer IOSB-ILT

Jürgen Woll (juergen.woll@dieffenbacher.de), Dieffenbacher GMBH Maschinen- und Anlagenbau