Detect Anomalies early and avoid Damage with Machine Learning

CAE News |

scaiLernverfahrensmall

For this, however, the system first requires a stable learning phase in which it gets to know all possible normal states. In wind turbines or bridges, this is only possible to a very limited extent, as they are exposed to, among other things, strongly fluctuating weather conditions. In addition, there is usually little data available on anomalous events. This prevents the system from categorizing the exception states. However, this would be important in order to recognize how dangerous the respective standard deviations are. These two problems are to be addressed in the project "Machine Learning Methods for Stochastic-Deterministic Multi-Sensor Signals" (MADESI). With the help of numerical simulations all imaginable scenarios can be approximated. For example, it is possible to simulate what happens when strong storm gusts hit a wind turbine. The monitoring system could then be trained with the data generated in these simulations and then independently recognize and interpret anomalies.

More Information: scai.fraunhofer.de/presse

返回