| About this Abstract |
| Meeting |
2022 TMS Annual Meeting & Exhibition
|
| Symposium
|
Additive Manufacturing: Beyond the Beam III
|
| Presentation Title |
A Multi-step Data Driven Model for Reverse Shape Compensation for Binder Jet Parts |
| Author(s) |
Hao Deng, Basil J. Paudel, Albert To |
| On-Site Speaker (Planned) |
Basil J. Paudel |
| Abstract Scope |
The large shrinkage of binder jet parts happens during the sintering process, which is used to solidify the object. This sintering distortion may result in unacceptable parts with low geometric accuracy. This work proposes an approach to compensate input geometry files based on simulations using data-driven method. A multi-step neural network approach is proposed for the first time to learn the deformation pattern and compensate the sintering deformation of an initial part geometry. Several initial geometries with different reverse scaling factors are generated and simulated to generate training database. Once the training dataset obtained, a reduced order modeling technique is applied to effectively extract the features of training dataset. A multi-step neural network model is trained and used to predict the compensated part. Finally, the compensated geometry file is validated by building parts and comparing the change in the part distortion. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Modeling and Simulation, Machine Learning |