Measuring and evaluating noise
Axial piston pumps convert mechanical energy into hydraulic energy. They help to lift heavy loads on construction or agricultural machinery, and are used in industrial conveyor systems. Working with partners, scientists at the Fraunhofer Institute for Digital Media Technology IDMT have installed battery-powered sensors on axial piston pumps. These sensors are capable of recording the pump noise via the air, processing them, comparing them with reference audio data and wirelessly transmitting them to a digital evaluation unit. Not only does this enable the early detection of possible faults, it also provides information on the nature of the problem – play in bearings, perhaps, or hydraulic issues. There is then the possibility of intervening before the powertrain or hydraulics suffer more serious damage.
Using machine learning
"We trained the cognitive system using machine learning methods based on pump audio signals we had previously recorded," explains Danilo Hollosi, Manager of "Acoustic Event Recognition" at the Hearing, Speech and Audio Technology project group of the Fraunhofer IDMT, based in Oldenburg. A central infrastructure for data processing is not required. Therefore, while servers can gobble up tens of thousands of euros, the price per sensor remains in two figures. Moreover, because the signals are processed in situ, less data is needed for training. "Customers benefit from a secure technology platform, which is suitable for a great variety of audio scenarios, is easy to retrofit and can be scaled as desired. Sensors can be linked in a smart network via the internet for the purpose of remote maintenance," says Hollosi, summing up the advantages.
The technology is being funded by the German Federal Ministry of Education and Research (BMBF) in the ACME 4.0 project. In the meantime, the partners are into their third year of the project and have reached technology readiness level 8. The prototype will be put through field trials in 2018. At the same time, the scientists are working with Infineon on predictive maintenance for chip production.