Big Data Analysis & Predictive Maintenance

The Cloud Research Lab of our Partner university Furtwangen organized a workshop on "Big Data Analysis for SMEs in Industry 4.0" at the Chamber of Industry and Commerce in Strasbourg as part of the "Project HALFBACK".

Halfback is a project that deals with the Development of highly available production processes busy. With the help of Predictions should Defects on machines, Loss of quality or missing Material availability can be specifically prevented. This is done by Maintenance plannedexchange or Replacement of components and materials handled more efficiently or even by a Intelligent relocation of production to another production facility.
The project is funded by the EU Regional Development Fund and thus has an international mandate in cooperation with INSA Strasbourg and the University of Strasbourg.
Other partners from industry complement the project with their experience.

Practical example: Predictive maintenance for wind turbine gearboxes

Furtwangen University has chosen us on the basis of our experience from Predictive maintenance and big data topics. This enabled me to share our expertise and develop a specific project - a the intelligent wind turbine gearbox - in a little more detail.
The report on experiences and tips on best practices was of particular interest, as the wealth of experience here is still rare. Although Process models are available - but some of them have yet to be tested.
The audience was quite astonished when the figures for Data volumes were discussed. This large amount of data results from the Sampling rate and operating time of a wind turbine gearbox and the Number of sensors. This means that per turbine and per operating day over 120,000 data recordswhich then have to be processed.

Best practices for predictive maintenance projects

Best practice tip no. 1: Find ways to check and validate incoming data immediately!

The basis for meaningful predictions are valid data. These must therefore be continuously checked for quality and it must be ensured that there are mechanisms that only allow further processing if the Quality is right.
Best practice tip no. 2: Use data from the field as quickly as possible!

This applies to all IoT and predictive maintenance projects - regardless of the size of the project or company. Only the Real data supply chain with real data allows you to quickly add further unforeseen sources of error to identify. From the reliability of the sensors to the compilation of the data via the IoT device and the transmission of the data, there are many potential disruptors that need to be taken into account. Simulations simply not considered and thus have a strong influence on the project.