About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
AI-Enabled Materials Processing: Integrating Accelerated Experimental Workflows and Processing-Aware Machine Learning
|
| Presentation Title |
Data-Driven Multi-Objective Optimization of Ultrashort-Pulse Laser Cutting of Magnesium Alloy Microtubes for Bioabsorbable Stent Manufacturing |
| Author(s) |
João Cruz, Alex Mainguy, Benedikt Bornschlegel, Simon Pöstges, Christian Rüdiger Hinke, Constantin Leon Häfner, Alexander Kopp |
| On-Site Speaker (Planned) |
João Cruz |
| Abstract Scope |
Ultrashort-pulse laser cutting of WE43MEO microtubes requires balancing competing quality responses across a constrained process space. A four-factor central composite design was used to model surface roughness, taper angle, burr height, recast-affected width, and surface darkening. Ordinary least-squares (OLS) response-surface models were benchmarked against nonlinear surrogate algorithms using leave-one-out cross-validation. OLS matched or outperformed the alternatives across all five responses while retaining interpretability of factor effects and interactions. Pareto analysis and weighted desirability-based multi-objective optimization identified a balanced condition within the experimentally feasible region. Experimental validation (n = 5) showed that all geometric and roughness responses lay within their 95% prediction intervals, while surface darkening was lower than predicted. When transferred to stent manufacturing, the optimized condition produced the highest radial strength, approximately 9% above the lowest-performing condition (Welch t-test, Holm-adjusted p = 0.010). This workflow links interpretable data-driven optimization with functional validation in bioabsorbable metallic devices of complex geometry. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Magnesium, Machine Learning, Biomaterials |