| About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
Efficient Microstructure Prediction in Additive Manufacturing Using a Novel Dimension Reduction Method |
| Author(s) |
Arulmurugan Senthilnathan, Paromita nath, Sankaran Mahadevan |
| On-Site Speaker (Planned) |
Arulmurugan Senthilnathan |
| Abstract Scope |
Variability in the additive manufacturing process and material properties affect the microstructure which influences the macro-scale mechanical properties. Uncertainty quantification and propagation, and process design under uncertainty, require numerous process-structure-property (P-S-P) simulations. However, the high computational cost of the P-S simulation (thermal and phase-field models), which relates the microstructure to the process parameters, motivates the construction of inexpensive surrogate models. Moreover, the high-dimensional microstructure image generated by P-S simulation presents a challenge in constructing a surrogate model. Therefore, a novel dimension reduction strategy is first introduced to reduce the microstructure image dimension. A surrogate model is then constructed to predict the reduced dimensional features given process parameters. The predicted features are then mapped to the original dimension to obtain the microstructure image that is verified against the phase-field predicted microstructure, using moment invariants. The propagated uncertainties in the microstructure image are quantified using second order statistical moments. |