Data Management in Database Networks with Metadata Models

Individualized productions, such as in the Photovoltaic Technology Evaluation Center (PV-TEC), place high demands on the flexibility of software systems for production monitoring and the collection, interlinking and analysis of the resulting data due to the wide variety of manufactured products and the constantly changing production capabilities and processes. Conventional production control systems only meet these requirements to a limited extent, partly because the database structures are optimized for highly standardized data flows, which makes the processing of partly structured and unstructured data from heterogeneous data sources more difficult and thus the implementation of Industry 4.0 and big data concepts.

We are therefore developing innovative modular digitalization concepts for planning and tracking production orders as well as collecting and analysing the resulting data, which meet the requirements of individualized productions through two disruptive approaches:

  1. the dynamic recording of all production data without structural specifications (data lake) and
  2. the use of a network of NoSQL databases for the separate storage of raw data and metadata with an integrated metadata layer for interlinking and pre-structuring the data (lakehouse) for queries and analyses.

Typically, we use graph databases to store metadata (context information), document databases to store production data and other databases and storage solutions specialized for specific data types, which are linked via reference IDs to form a database network.

Data management in database networks with metadata models: Integrated system approach
© AI generated
Integrated system approach with
(a) fail-safe server infrastructure,
(b) maintenance-friendly implementation,
(c) configurable data pipelines and
(d) expandable applications integrated into a micro-service architecture.
Automation pyramid: top-down approach of classic production control systems vs. bottom-up approach of our SmartCell system. Instead of setting fixed standards, all quality and process data can be recorded bottom-up and can be structured and assigned through the processing of suitable context information.
© Fraunhofer ISE
Automation pyramid: top-down approach of classic production control systems vs. bottom-up approach of our SmartCell system. Instead of setting fixed standards, all quality and process data can be recorded bottom-up and can be structured and assigned through the processing of suitable context information.
Metadata model of wafer-based PV production with different data classes: (blue) Planning orders / experiments, (red) execution of real processes.
© Fraunhofer ISE
Metadata model of a wafer-based PV production with different data classes:
(blue) planning of manufacturing orders / experiments,
(red) Execution of real processes.

Our R&D services include:

  • support with concept development for laboratory digitization (variants, possibilities, restrictions)
  • development of comprehensive data models tailored to the respective production by interlinking all available data sources
  • connection of data sources to the system with minimal requirements for the data interface of the tools
  • development of complex database queries to provide extensive data sets for reporting
  • development of a customized API for central data access of all application and analysis tools

Further Information on this Topic

Research Topic

Artificial Intelligence and Data Management

Business Area

Photovoltaics:​ Production Technology and Transfer