About this Abstract |
| Meeting |
2025 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2025)
|
| Symposium
|
2025 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2025)
|
| Presentation Title |
Transfer Learning for Process–Microstructure–Property Mapping in Micro Selective Laser Sintering |
| Author(s) |
Farzana Tasnim, Joshua Grose, Nathan F Sheu, Remi Dingreville, Michael Cullinan |
| On-Site Speaker (Planned) |
Farzana Tasnim |
| Abstract Scope |
This study presents a transfer learning framework utilizing pre-trained convolutional neural networks (CNNs) to correlate process parameters with microstructure evolution in micro-Selective Laser Sintering (µ-SLS). The approach combines SEM image analysis with process data to predict microstructure-property relationships and sintering conditions. A dataset of µ-SLS SEM images is preprocessed via denoising, contrast enhancement (CLAHE), and edge-preserving sharpening to standardize microstructural features. High-level features extracted by pre-trained CNNs are fused with process parameters to train a dual-purpose model: a regression network predicts electrical resistivity, while a classification network identifies laser exposure time. The framework achieves robust performance across both tasks, demonstrating noise resilience and computational efficiency through transfer learning. Its generalizability to unseen data underscores its potential to bridge process-microstructure-property linkages in materials science. This scalable solution advances microstructure analysis for rapid process optimization, with applicability extending to image-based characterization in diverse scientific domains. |
| Proceedings Inclusion? |
Planned: Post-meeting proceedings |