Researchers use big data to keep complex production running smoothly

Industrial plants must function effectively. Remedying production downtimes and breakdowns is an expensive and time-consuming business. That is why companies collect data to evaluate how their facilities are doing.

Large amounts of data are produced when industrial companies monitor their facilities. Sensors check temperature, pressure, power or energy use data.

“If you’re scanning to the nearest second, it’s easy to rack up several terabytes of information in under a week,” says Dr Olaf Sauer from the automation business unit at the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation (IOSB). But often there is a lack of suitable methods to evaluate the information.

Industrial plants are complex structures, and anyone who wants to monitor them must monitor the data from a wide variety of process variables. (Image © Daimler AG).

“Today’s operators use only about 7% of this data for maintenance or protection from breakdowns,” adds Sauer. IOSB developers from the Fraunhofer Institute at Karlsruhe and Lemgo have developed systems that can process the collected data in a way that increases plant availability significantly while reducing energy consumption.

The scientists from Karlsruhe focus on continuous processes in the chemical or pharmaceutical industries.

“At the Fraunhofer Application Center Industrial Automation IOSB-INA in Lemgo, we have developed solutions we can use to easily monitor complex manufacturing technology and hybrid systems,” explains Prof Oliver Niggemann, deputy head of IOSB INA in Lemgo. Together, the methods are suitable for all kinds of capital-intensive industrial facilities that need to run for long periods of time.

“Even wind farms could be monitored this way,” says Sauer.

The researchers use smart data mining methods to calculate the optimum operational steps for each production process. This reference model is then compared to data from current operations in order to quickly identify and precisely locate any discrepancies before thoroughly eliminating them. There is no need for detailed separate modelling of complex plant structures. Special data storage systems record the information in real time and send it over the network to a database. Next, software normalises the data, makes it comparable and establishes relationships. The results are then presented in a clear way - for instance, in the form of a three-dimensional map. ‘Mountains’ and ‘valleys’ depict the individual process phases; any disruptions or anomalies can be quickly identified.

“Our tool has been used successfully in industry for some time now,” says Niggemann. “The toolbox proKNOWS for instance is currently being tested in two projects with industry partners.”

The scientists from Karlsruhe and Lemgo also use it to monitor and analyse facilities’ energy demand. This information makes it possible not only to judge the state of the facilities but also to reduce their power consumption by way of appropriate adjustments to the controls. “This is something we’ve already demonstrated with conveyor drive technology,” says Sauer.

Condition monitoring is the name production experts give this use of modern control systems to monitor industrial facilities so as to prevent breakdowns. Most industrial companies today use technology of this sort, but in many cases they monitor only individual components and not the entire facility - even though that is what they should be doing. This is especially true for continuous manufacturing processes, where creeping change can suddenly cause a breakdown unless operators have their eye on all the variables. One example could be a pipeline blockage as a result of a gradual build-up of material deposits on the pipe’s inner walls. It is also rare for people to work up a reference model directly from collected data in the way the IOSB tools do. An additional challenge is that today’s monitoring systems generally run on a standalone basis and are not part of the production control system.

“But industry has recognised that it needs to catch up, and there is clearly a trend toward integrating them into manufacturing execution systems (MES). We’re still a long way away from the vision of Industry 4.0, in which smart machines automatically report of their own accord when they need maintenance or spare parts. But our methods bring us one step closer to reaching that goal,” says Sauer.