적층 제조공정 불확실성을 고려한 PLA시편의 심층 신경망 기반 피로수명 신뢰성 평가Deep Neural Network-Based Reliability Assessment on Fatigue Life of PLA Specimens Considering Uncertainty of Additive Manufacturing
- Other Titles
- Deep Neural Network-Based Reliability Assessment on Fatigue Life of PLA Specimens Considering Uncertainty of Additive Manufacturing
- 문현철; 노우승; 유현승; 도재혁
- Issue Date
- Polylactic Acid (PLA); Additive Manufacturing (AM); Fatigue Life; Deep Neural Network (DNN); Reliability Assessment
- 신뢰성 응용연구, v.22, no.1, pp.37 - 47
- Journal Title
- 신뢰성 응용연구
- Start Page
- End Page
- Purpose: This study aims to assess the reliability of fatigue life for additively manufactured polylactic acid (PLA) specimens by considering the physical uncertainty of additive manufacturing (AM) conditions based on a deep neural network (DNN).
Methods: The fatigue specimens were manufactured by fused deposition modeling (FDM) in accordance with AM conditions. The fatigue life test was then conducted using the rotational fatigue test equipment. Furthermore, to consider the physical uncertainties resulting from AM processes, the variation of AM parameters was assumed as a normal distribution. The DNN model was trained to predict the fatigue life based on experimental data in accordance with AM conditions. The reliability assessment was then performed with Monte Carlo simulation (MCs).
Results: The DNN-based reliability assessment is carried out by employing MCs. The result shows that the reliability is converged by increasing the number of sample data. In addition, the predicted reliability is verified by error evaluation under a 95% confidence level.
Conclusion: The framework of DNN-based reliability assessment is suggested to predict and assess the reliability of fatigue life in accordance with AM conditions. This framework can be used for the reliability assessment of fatigue behavior for various product designs manufactured with AM.
- Files in This Item
- There are no files associated with this item.
- Appears in
- 융합기술공과대학 > 기계소재융합공학부 > Journal Articles
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.