Detecting anomalies in aerospace additive manufacturing: A Convolutional Long Short-term Memory approach
(2024) DABN01 20241Department of Economics
Department of Statistics
- Abstract
- As additive manufacturing technology within the aerospace sector advances rapidly, optimizing process monitoring methods has become increasingly critical, given that existing techniques remain costly and time-consuming. This study presents a novel approach to monitoring the laser metal deposition (LMD) process, emphasizing real-time anomaly detection through the analysis and prediction of frames from melt pool videos captured by top-view in-situ cameras. By training the Convolutional Long Short-Term Memory autoencoder on anomaly-free videos, the study has developed a model capable of forecasting future anomaly-free frames and identifying deviations from expected patterns. The model’s performance was evaluated using both stable and unstable... (More)
- As additive manufacturing technology within the aerospace sector advances rapidly, optimizing process monitoring methods has become increasingly critical, given that existing techniques remain costly and time-consuming. This study presents a novel approach to monitoring the laser metal deposition (LMD) process, emphasizing real-time anomaly detection through the analysis and prediction of frames from melt pool videos captured by top-view in-situ cameras. By training the Convolutional Long Short-Term Memory autoencoder on anomaly-free videos, the study has developed a model capable of forecasting future anomaly-free frames and identifying deviations from expected patterns. The model’s performance was evaluated using both stable and unstable process datasets, demonstrating its effectiveness in detecting a range of anomalies through pixel-wise difference comparison between predicted and actual frames. The results underscore the value of integrating advanced deep learning techniques into the LMD monitoring system, contributing to the development of real-time quality control, preventive maintenance, post-production analysis, and process optimization. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9173373
- author
- Nguyen Dam Khanh, Linh LU
- supervisor
- organization
- course
- DABN01 20241
- year
- 2024
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- anomaly detection, autoencoder, convolutional long short-term memory
- language
- English
- id
- 9173373
- date added to LUP
- 2024-09-24 08:38:32
- date last changed
- 2024-09-24 08:38:32
@misc{9173373, abstract = {{As additive manufacturing technology within the aerospace sector advances rapidly, optimizing process monitoring methods has become increasingly critical, given that existing techniques remain costly and time-consuming. This study presents a novel approach to monitoring the laser metal deposition (LMD) process, emphasizing real-time anomaly detection through the analysis and prediction of frames from melt pool videos captured by top-view in-situ cameras. By training the Convolutional Long Short-Term Memory autoencoder on anomaly-free videos, the study has developed a model capable of forecasting future anomaly-free frames and identifying deviations from expected patterns. The model’s performance was evaluated using both stable and unstable process datasets, demonstrating its effectiveness in detecting a range of anomalies through pixel-wise difference comparison between predicted and actual frames. The results underscore the value of integrating advanced deep learning techniques into the LMD monitoring system, contributing to the development of real-time quality control, preventive maintenance, post-production analysis, and process optimization.}}, author = {{Nguyen Dam Khanh, Linh}}, language = {{eng}}, note = {{Student Paper}}, title = {{Detecting anomalies in aerospace additive manufacturing: A Convolutional Long Short-term Memory approach}}, year = {{2024}}, }