A modular approach to online monitoring for laser-based 3D printing using deep MWIR features
- Journal of Intelligent manufacturing (submitted),
Adrian Pallas, Anton Garcia-Diaz, Veronica Panadeiro
We present ConvLMD, a novel approach to generic image-based real time monitoring of laser metal deposition using convolutional features from MWIR coaxial images. ConvLMD is a simple CNN regression model trained in a feasible large size dataset. It may be easily fine-tuned to work with additional target magnitudes. We demonstrate the capability of the model to measure online a variety of major process quality indicators, to our knowledge for the first time ever. These results bridge a key gap in the field of 3D printing of large metal parts. We also release the first large dataset of annotated images of laser metal deposition.