Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > EDA/IP

Sensor fusion reveals machine status to prevent failure

Posted: 06 Apr 2015 ?? ?Print Version ?Bookmark and Share

Keywords:Saarland University? machine? sensor?

During the normal course of its operation, any large machine shows signs of wear and tear, before it actually clutches its chest and drop dead, so to speak. However, researchers from the Saarland University have developed innovative ways to precisely identify the parts that need to be replaced soon, by means of sensor fusion techniques.

The research team led by Andreas Schutze monitors continuously the status of the machinery. When bearings or other mechanic and electric components are starting to wear out, the characteristics of their operational parameters starts to change: frequencies, spectral composition of the vibrations as well as temperatures start to deviate from their values in normal operation. The system under development by the researchers from Saarbrucken interrelates the measurements from various sensors and thus can detect even minuscule variations.

Sensor fusion tech to predict machine failure

Saarland University team develops sensor fusion technology to predict machine failure.

"Our sensor system allows us to observe the current condition of a plant," noted Schutze. "We are working on getting the system to issue very early warnings at the first sign that the plant may fail or malfunction. By combing multiple sensors we are able to register even the minor deviations, changes that would simply not be detectable with a single sensor."

The team's approach involves attaching vibration sensors at numerous positions on the machine to provide a continuous stream of measurement data. The engineers also incorporate data from the process sensors that are now installed as standard on most of today's machines. "We are studying how we can correlate sensor signal patterns, such as vibrational frequencies, with typical damage and failure modes, such as reduced cooling performance or a drop in accumulator pressure," said Schutze. To do this, the researchers have been analysing large quantities of measurement data in order to identify those patterns in the data that can be assigned to particular changes in the machine's state.

From the mass of data acquired the researchers filter out a manageable quantity of relevant sensor data that is characteristic of certain machine damage scenarios. The aim is to reliably detect disturbances in the machine's operating cycle during the incipient damage phase and to devise mathematical models for different fault levels.

This information about the relationship between sensor signal patterns and incipient malfunction or damage is used by the engineers to "teach" the system to enable it to identify these states automatically. Through continuously monitoring the machine's condition, the system can also recommend when to carry out particular remedial measures, such as replacing a part. According to Schutze, this makes it easier to plan maintenance operations on large or difficult-to-access plant machinery. It also helps to avoid unnecessary maintenance. "As the system is also capable of analysing whether production machinery was operating properly during a manufacturing process, it can also be used for quality control purposes. There are a large number of potential applications of this system, particular in the smart manufacturing processes envisaged under Industry 4.0," the researcher stated.

The project is a collaborative enterprise between Schutze's team of engineers at Saarland University and the Centre for Mechatronics and Automation Technology (ZeMA) and researchers at the German Research Centre for Artificial Intelligence (DFKI) and the HYDAC group.

- Christoph Hammerschmidt
??EE Times Europe

Article Comments - Sensor fusion reveals machine status...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top