Process Monitoring of One-Shot Drilling of Al/CFRP Aeronautical Stacks Using the 1DCAE-GMM Framework DOI Open Access
Giulio Mattera, Maria Grazia Marchesano, Alessandra Caggiano

et al.

Electronics, Journal Year: 2025, Volume and Issue: 14(9), P. 1777 - 1777

Published: April 27, 2025

This study explores advanced process monitoring for one-shot drilling of aeronautical stacks made aluminium 2024 and carbon fibre-reinforced polymer (CFRP) laminates using a 4.8 mm diameter tool unsupervised machine learning techniques. An experimental campaign is conducted to collect thrust force torque signals at 10 kHz sampling rate during the process. These are employed real-time monitoring, focusing on material change detection anomaly identification, where anomalies defined as holes that fail meet predefined quality criteria. innovative approach based proposed enable automatic signal segmentation, feature extraction, hole assessment. Specifically, semi-supervised Gaussian Mixture Model (GMM) 1D Convolutional AutoEncoder (1D-CAE) detect deviations from normal conditions. The method benchmarked against state-of-the-art supervised techniques, including logistic regression (LR) Support Vector Machines (SVMs). Results show these traditional models struggle with class imbalance, leading overfitting limited generalisation, reflected by F1 scores 0.78 0.75 LR SVM, respectively. In contrast, improves detection, achieving an score 0.87 more effectively identifying poor-quality holes. demonstrates potential deep learning-based methods intelligent enabling adaptive control in hybrid detecting anomalous While handles small imbalanced datasets, further research into application generative AI could enhance performance, aiming above 0.90, thereby supporting adaptation real industrial environments high performance.

Language: Английский

Process Monitoring of One-Shot Drilling of Al/CFRP Aeronautical Stacks Using the 1DCAE-GMM Framework DOI Open Access
Giulio Mattera, Maria Grazia Marchesano, Alessandra Caggiano

et al.

Electronics, Journal Year: 2025, Volume and Issue: 14(9), P. 1777 - 1777

Published: April 27, 2025

This study explores advanced process monitoring for one-shot drilling of aeronautical stacks made aluminium 2024 and carbon fibre-reinforced polymer (CFRP) laminates using a 4.8 mm diameter tool unsupervised machine learning techniques. An experimental campaign is conducted to collect thrust force torque signals at 10 kHz sampling rate during the process. These are employed real-time monitoring, focusing on material change detection anomaly identification, where anomalies defined as holes that fail meet predefined quality criteria. innovative approach based proposed enable automatic signal segmentation, feature extraction, hole assessment. Specifically, semi-supervised Gaussian Mixture Model (GMM) 1D Convolutional AutoEncoder (1D-CAE) detect deviations from normal conditions. The method benchmarked against state-of-the-art supervised techniques, including logistic regression (LR) Support Vector Machines (SVMs). Results show these traditional models struggle with class imbalance, leading overfitting limited generalisation, reflected by F1 scores 0.78 0.75 LR SVM, respectively. In contrast, improves detection, achieving an score 0.87 more effectively identifying poor-quality holes. demonstrates potential deep learning-based methods intelligent enabling adaptive control in hybrid detecting anomalous While handles small imbalanced datasets, further research into application generative AI could enhance performance, aiming above 0.90, thereby supporting adaptation real industrial environments high performance.

Language: Английский

Citations

0