Dissertation
Data-driven design optimization for metal additive manufacturing
Doctor of Philosophy (Ph.D.), Drexel University
Sep 2023
DOI:
https://doi.org/10.17918/00001884
Abstract
Metal Additive Manufacturing (AM) has become a viable method of fabricating complex geometries that are difficult and even impossible to manufacture using some of the traditional techniques. Design for Additive Manufacturing (DfAM) approaches leverages the vast potential of AM approaches for getting optimal and novel part designs. However, the inherent cyclic thermal nature of metal AM processes creates thermal gradients in the printed part resulting in geometry challenges such as residual stress induced distortions which compromise the overall part performance. Therefore, the reliability and performance qualification of metal AM parts is critical for their safe and successful adoption in engineering applications. The thermal profile in metal AM parts is dependent on the process parameters used for printing, the geometry of the part, and the orientation used while printing. For these reasons, a 3D-printed part optimized using standard design optimization formulations that do not account for these manufacturing parameters may not provide the desired part performance. The work presented in this dissertation proposes a potential approach that may help address this critical issue. Design considerations for AM rely on thermomechanical process simulations; which although provide valuable insights on the role of manufacturing parameters and overall geometry on the print quality, do not offer direct ways to adapt the target part geometry into the specific AM process used and thus create a disconnect between the model and the produced part. To address this issue, this dissertation presents an approach to directly couple In-Situ Monitoring (ISM) data of the AM process with a gradient-based topology optimization approach. To accomplish this goal, a database of thermal data is generated by printing candidate parts for which in-situ monitoring data is collected. Neural network-based surrogate models are then used to predict temperature and temperature gradients for novel part geometries. A bi-objective topology optimization with mechanical and thermal constraints that queries the surrogate model for the thermal data and runs a static structural finite element solver for the mechanical objective evaluation is used. The proposed data-driven approach is much faster than current trial-and-error DfAM approaches. This approach is also expected to provide more optimal design outcomes than by previous DfAM approaches proposed in literature.
Metrics
27 File views/ downloads
63 Record Views
Details
- Title
- Data-driven design optimization for metal additive manufacturing
- Creators
- Vignesh I. Perumal
- Contributors
- Antonios Kontsos (Advisor)Ahmad Raeisi Najafi (Advisor)
- Awarding Institution
- Drexel University
- Degree Awarded
- Doctor of Philosophy (Ph.D.)
- Publisher
- Drexel University; Philadelphia, Pennsylvania
- Number of pages
- xx, 179 pages
- Resource Type
- Dissertation
- Language
- English
- Academic Unit
- College of Engineering (1970-2026); Mechanical Engineering (and Mechanics) (1970-2026); Drexel University
- Other Identifier
- 991021230305404721