Deep Learning for In-Situ Layer Quality Monitoring during Laser-Based Directed Energy Deposition (LB-DED) Additive Manufacturing Process
, Ehsan Dehghan-Niri 1,*, Michael Juhasz 2, Kevin Luo 2 and Harold S. Halliday 3Round 1
Reviewer 1 Report
The manuscript entitled “applsci-1884133” dealing with Powder bed fusion has been reviewed. The paper has been nicely written but needs significant improvement. Please follow my comments.
1. Why author has a separate literature review? Please justy it.
2. The explanation in Figure 1 needs more improvement.
3. Please add a scale bar to Figure 4.
4. The wordings in Figure 3 “Design Diagram” is too small. Please correct it.
5. The text has some typos. Please check them.
6. The introduction needs to be updated by reading and adding the three following papers.
· Thermal modeling of directed energy deposition additive manufacturing using graph theory
· A novel methodology to manufacture complex metallic sudden overhangs in weld-deposition based additive manufacturing
· Development and experimental study of an automated laser-foil-printing additive manufacturing system
7. Additive manufacturing has many advantages over the conventional manufacturing method which can be highlighted in your paper. Please read the following article and add to the introduction to show the experimental application of additive manufacturing and the advantage of this process over conventional manufacturing like machining.
Additive manufacturing a powerful tool for the aerospace industry.
Author Response
Please refer to the attached file
Author Response File: 
Author Response.docx
Reviewer 2 Report
Overall, it is an interesting study with a high interest in applied science community. I would like to point out a few points for authors to clarify:
1. Feature selection plays a significant importance for the model performance. However the number of features are limited due to the selected in-situ monitoring techniques. Can authors add a discussion about what would be the benefits of adding another NDT during monitoring?
2. The generalization of the problem (other AM pieces etc.) and limitations of the proposed methodology can be explained in more details.
3. Authors can clarify what was the data structure that was used for training, testing and validation in the article with more care. It is not clear the data size used fr training the model - which can lead the overfitting.
4. In a thought experiment, could you explain how can an other lab performing in-situ monitoring to during AM can utilize the build model?
Author Response
please refer to the attached file
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The paper is ready to publish.
