News

Research develops methods for machine learning in metrology for sensor networks

Automation industry concept

EMPIR project makes open access data sets available

EMPIR project ‘Metrology for the factory of the future’ (17IND12, Met4FoF) is working to develop calibration methods for advanced digital-only industrial sensors. These sensors can measure dynamic, time-dependent qualities such as acceleration, force, and pressure, and pre-process collected data are quickly becoming the de-facto standard in the industrial Internet of Things. It is vital for manufacturing processes that such digital sensors measuring dynamic qualities are accurate and well calibrated.

The project will establish the infrastructure and software needed to account for measurement uncertainty and quality together with measurement data, and synchronise data flow in sensor networks. This will help facilities at National Measurement Institutes to become digital ready, and enable European factories of the future to be competitive with their global counterparts.

 

Open access data sets

The 3-year project is still in its first year, but three open access data sets produced on three very different testbeds are already available and easy to download. These data sets, which are seldom provided from such sensor networks for research purposes, will be of interest to machine learning researchers and other data scientists.

The three data sets are:

  • Condition monitoring data set of the ZeMA electromechanical cylinder test stand. Lifetime prognoses and end-of-line tests of electromechanical cylinders – sound, piston rod vibration, plain bearing vibration, ball bearing vibration, axial force, pneumatic pressure, velocity, active current and the three motor current phases.
  • Condition monitoring data set of the ZeMA hydraulic system – pressure, volume flow, temperature while the condition of four hydraulic components is varied.
  • Sensor data set radial forging at AFRC. Data from GFM SKK10/R radial forge that uses two pairs of hammers operating at 1200 strokes/min, and providing a maximum forging force per hammer of 150 tons.

The data sets will be reviewed regarding their suitability for machine learning and updated regularly by the consortium throughout the lifetime of the project. Further data sets will follow; in particular, data from the testbed at SPEA, Italy, which provides distributed measurement of temperature. Web-based tutorials for applying machine learning to the data sets will be published very soon.

The project consortium hopes that the early publication of these data sets will enable discussion and exchange with the scientific community. Users should comment on findings in the data sets, suggest further documentation details, and publish their results from applying machine learning. Sascha Eichstaedt, the Project Coordinator would be delighted for users to get in contact with him!

Sascha said ‘These data sets together with the method toolbox to be developed in the project mark the beginning of a long-term activity to support quality in production and stimulate metrology research. Collaboration with the project will help to guide these developments and to benefit from the outcomes as early as possible.’

This EMPIR project is co-funded by the European Union's Horizon 2020 research and innovation programme and the EMPIR Participating States.


Want to hear more about EURAMET?
Sign up for EURAMET newsletters and other information

Select your area of interest