Options
Nutzung von Methoden des Maschinellen Lernens für die Werkstückqualität- und Werkzeugverschleißüberwachung bei semi-automatischen Bohrprozessen
Publikationstyp
Conference Paper
Publikationsdatum
2023-11-10
Sprache
German
Institute
M-18
Citation
72. Deutscher Luft- und Raumfahrtkongress (DLRK 2023)
Contribution to Conference
Publisher DOI
Publisher
Deutsche Gesellschaft für Luft- und Raumfahrt
Riveting structural components in aircraft construction requires the production of a large number of rivet holes. Quality standards must be maintained for each well to avoid rework and waste. This results in high demands on process stability and quality monitoring during drilling. The production of rivet holes is increasingly being carried out with electric, semi-automatic drill feed units (BVE), replacing pneumatically operated machines. The electrical BVEs enable them to be equipped with internal sensors that enable recording of drive currents and vibrations of the machine structure. In this work, the process data is examined with regard to its potential for determining the currently prevailing tool wear condition and the resulting workpiece quality using machine learning (ML) methods. For this purpose, process data from drilling TiAl6V4 and CFRP were recorded in an experimental environment and examined with the help of data processing scripts in Matlab. The suitability of selected methods for dimension reduction for model training with regard to predicting tool wear and workpiece quality was determined. Relevant features could be identified primarily in the time domain and in the time-frequency domain of the wavelet transformation of the feed motor current. Prediction models could be successfully trained for both wear and quality based on the highlighted characteristics. The evaluation of the classification and regression showed good prediction of tool wear using classification trees as well as irregularized and regularized linear regression. Good quality predictions could be achieved for the exit burr height when drilling in TiAl6V4 and for the delamination on the exit side when drilling in CFRP. The results achieved form a basis for further development of methods using machine learning in quality control during drilling.
Schlagworte
Strukturmontage
Qualitätsüberwachung
Nietverbindungen
Bohrtechnologien
Maschinelles Lernen
DDC Class
620: Engineering
004: Computer Sciences