THE BACK PAGE by Editor Andy Pye
COGNITIVE SYSTEM FOR PREDICTIVE ACOUSTIC MAINTENANCE
maintenance of production facilities. In the new system, intelligent battery-powered
A
acoustic sensors process audio signals from machines and systems on the spot. From the information that is forwarded wirelessly to an evaluation unit, it is possible to draw conclusions about the condition of the production facilities and to avoid possible damage.
SENSOR NODES Axial piston pumps convert mechanical into hydraulic energy. On construction or agricultural machinery, they help to lift heavy loads or are part of industrial conveyor technology. “So far, these systems have not had permanently installed acoustic condition monitoring,” says Danilo Hollosi, head of Acoustic Event Recognition at the Oldenburg Project Group for Hearing, Speech and Audio Technology at IDMT. “Cognitive systems can be very powerful in this regard.” The scientists have mounted battery-operated
sensors on axial piston pumps that are able to record the noise of the pump via the air, to process it, to compare it with reference audio data and to send the information wirelessly to a digital evaluation unit.
48 /// Environmental Engineering /// June 2018
t the Hannover Messe (23-27 April 2018), the Fraunhofer Institute for Digital Media Technology (IDMT) demonstrated the prototype of a cognitive system for predictive
Not only can conclusions about possible
undesirable developments be identified at an early stage, statements about the nature of the problems can also be made; for example, if there are problems concerning bearing clearance or hydraulics. This provides the opportunity to intervene before serious damage to powertrains or hydraulics occurs.
USE OF MACHINE LEARNING METHODS “We have trained the cognitive system with machine learning based on previously acquired pump audio signals,” Hollosi says. A central infrastructure for data processing is not necessary. This saves costs: while servers can cost five-figure sums, the price per sensor remains in the double-digits. Signal processing on site also requires less data for training. “This data-secure technology platform is suitable
for a wide variety of audio scenarios and can be easily retrofitted and scaled to any size. Networking of sensors via the Internet for remote maintenance is also possible,” adds Hollosi. “Our colleagues are experts in technologically recreating the capabilities of the human ear. They teach the systems to adhere to given parameters when evaluating audio data, to take into account environmental noise patterns and to exclude background noise.” In 2018, the system will be field-tested. At the
same time, the scientists are working with Infineon on predictive maintenance for chip production. EE
❱❱ Configured wireless sensor nodes (in the foreground) send status messages of the axial piston pump (left) to a tablet
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52