Regensburg 2022 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MM: Fachverband Metall- und Materialphysik

MM 22: Data Driven Materials Science: Experimental Data Treatment and Machine Learning

MM 22.7: Vortrag

Mittwoch, 7. September 2022, 12:30–12:45, H46

Analysis of acoustic emission spectra for structural health monitoring — •Klaus Lutter, Viktor Fairuschin, and Thorsten Uphues — Institute for Sensor and Actuator Technology, Coburg, Germany

Today, the analysis of vibration and acoustic emission spectra is routinely used for health monitoring of gears in industrial production. Recent developments of extended IIoT networks provide even fleet comparison and optimization of required field service.

Here, we present an extended approach to utilize acoustic emission spectra to monitor the structural health of machining tools like mills or drills to extract degradation and lifetime information from the acoustic emission. A successful application of a spectral analysis will provide a huge impact on production quality as well as tool quality according to different production parameters which are transferred into related spectral properties. Furthermore we follow an experimental approach using contact microphones. Our diagnostic approach is a detailed analysis of the corresponding frequency spectra and in particular the existing harmonic frequencies during milling processes.

We demonstrate the classification of different process parameter sets according to different dominant frequencies via classification algorithms. The retrieved classes of spectra are used for a classical regression model assisted by neural networks to analyse characteristics changes over time. From an industrial perspective this type of analysis is a non-invasive and versatile approach and easily implementable, even in existing production machinery.