Acoustic condition monitoring of milling tools using the Rayleigh-Ritz autoencoder

The condition of the milling tool and its cutting
edges is crucial for the surface quality of high-performance
components. In this paper, we propose a non-intrusive approach to condition
monitoring by analyzing the acoustic emissions generated
during the milling process, enabling tool condition
monitoring without interfering with the milling operation.
The acoustic data is recorded using a MEMS microphone
and analyzed employing a hybrid machine learning framework.
In the first step, the raw acoustic data is transformed
to a phase space representation, where the time-series data
of each lane is mapped to a rotational angle. Subsequently,
the Rayleigh-Ritz autoencoder is applied to the phase
space data. To incorporate process-specific knowledge,
constraints are defined using trigonometric functions.
This approach has demonstrated its effectiveness in detecting
progressive tool wear using only acoustic data
providing a reliable and non-invasive solution for tool
condition monitoring.