Review Reports

Round 1

Reviewer 1 Report

1)This paper investigates the deep learning based estimation of muckpile fragmentation using cloud data. The contributions and motivations of this paper were insufficient, in which the system model considered have been discussed in many treatise.

2)How to understand the estimation methods using 2D and 3D images and apply deep learning to muckpile fragmentation?

3)The conventional estimation methods should be selected as the benchmarks for the purpose of comparison. 

4)All simulation parameters have to be justified with proper references in numerical results.

no comments

Author Response

Dear Reviewer,

Thank you for taking the time to review our manuscript and for providing valuable feedback. We truly appreciate your insights and have addressed your concerns as follows:

1. Contributions and Motivations: We recognize the importance of delineating our paper's unique contributions and motivations. We will revisit this section and provide a more detailed discussion, highlighting what sets our approach apart from existing methods and the reasons for our chosen methodology.

Response: "We understand the concerns raised about the unique contributions of our work. We will endeavor to better emphasize the novelty and significance of our approach in the revised manuscript, differentiating it from existing treatises on the subject."

2. Estimation Methods using 2D and 3D Images: To ensure clarity, we will incorporate a more comprehensive section discussing the methods of estimation using 2D and 3D images. This will provide readers with a better understanding of how deep learning is applied to muckpile fragmentation.

Response: "Your suggestion to elaborate on the estimation methods is well-taken. In our revised version, we will delve deeper into the methodologies of 2D and 3D imaging and their implications in deep learning for muckpile fragmentation."

3. Conventional Estimation Methods as Benchmarks: We agree with your assessment. To give a holistic view, we will incorporate conventional estimation methods and compare their results with our proposed deep learning-based technique. This will underscore the advantages and potential limitations of our approach.

Response: "We acknowledge the significance of benchmarking our approach against conventional methods. The revised manuscript will present a comparative study to better showcase the efficacy of our deep learning-based method."

4. Justification of Simulation Parameters with References: Your point is well-taken. In the revised manuscript, we will ensure that each simulation parameter is justified and adequately referenced, providing readers with a clearer understanding and verification of the parameters chosen.

Response: "Thank you for emphasizing the need to justify and reference our simulation parameters. We will make the necessary adjustments in the revised manuscript to address this concern."

Once again, we are grateful for your insights and constructive feedback. We believe that your suggestions will significantly improve the quality of our paper, and we are committed to making the necessary revisions.

Warm regards,

Author Response File: Author Response.pdf

Reviewer 2 Report

I have gone through the MS entitled Deep Learning-Based Estimation of Muckpile Fragmentation Using Simulated 3D Point Cloud Data, where authors tried to estimate the blast muck fragmentations using cloud-based 3D photogrammatry method (the 3D point clouds generated by 3D photogrammetry from multi-view images) including PointNet network (ANN based program). Please use references for application of ANN in rock blasting which may be-Application of Artificial Neural Network (ANN) for Prediction and Optimization of Blast-Induced Impacts.

Prediction or estimation of blast fragmentation is really tedious task in mining operations but it's very important to utilize heavy machinery and mines to mill down operations. Authors need to improve the literature review, there are some literature may be helpful to enhance the MS quality, it may be Numerical assessment of spacing–burden ratio to effective utilization of explosive energy 

The paragraph has been repeated from line 123 to 138, so please delete it.

Authors should also discuss or keep the example of digital image analysis software including with WipFrag, like Split desktop, Portametrics and Fragalyst etc. They are almost the same program generally used in 2D image analysis of muck piles. 

Cyber world to generate training data with known particle size distributions for deep learning, but question arise that how to calibrate the real muck pile data with CG based training data? I mean is there any calibration or reference defined so that CG rock fragments size correlated with real muck size at blasting field site. 

The MS have ample amount of analysis and just need minor revision before publish. 

 

 

Author Response

Dear Reviewer,

We greatly appreciate your time and effort in reviewing our manuscript titled "Deep Learning-Based Estimation of Muckpile Fragmentation Using Simulated 3D Point Cloud Data." Your feedback is invaluable, and we have carefully considered each point you raised.

References for ANN in Rock Blasting:
Thank you for pointing out the need for references on the application of Artificial Neural Networks (ANN) in rock blasting. We will incorporate the suggested reference – "Application of Artificial Neural Network (ANN) for Prediction and Optimization of Blast-Induced Impacts" – and explore more literature to substantiate our claims.

Improvement in Literature Review:
We acknowledge your observation regarding the need for a more extensive literature review. We'll incorporate the suggested literature, "Numerical assessment of spacing–burden ratio to effective utilization of explosive energy," and will further delve into other pertinent studies to provide a more comprehensive review of the topic.

Repeated Paragraph:
We apologize for the oversight. The repeated paragraph from lines 123 to 138 will be deleted in our revised manuscript.

Discussion on Digital Image Analysis Software:
Your suggestion to discuss and provide examples of digital image analysis software is apt. In the revised manuscript, we will provide insights into software such as WipFrag, Split desktop, Portametrics, and Fragalyst, emphasizing their relevance and applications in 2D image analysis of muck piles.

Calibration with CG-Based Training Data:
Your question regarding the calibration of real muck pile data with computer-generated (CG) training data is crucial. We will include a section in our manuscript detailing the calibration methods employed and the benchmarks set to ensure the CG rock fragment sizes correlate accurately with real-world muck sizes at blasting field sites.

Overall Feedback:
We're grateful for your positive feedback on the depth of analysis in our manuscript. We are committed to implementing the suggested revisions to improve the overall quality and ensure it meets publication standards.

Once again, thank you for your constructive feedback. We believe these revisions will enhance the quality and clarity of our work.

Reviewer 3 Report

Dear authors,

Below you can find few comments suggested for better understanding and improvement of article quality:

Conclusion part:

Figure 19. Estimation Results of Particle Size Distribution should not be part of conclusion. Please remove it from conclusion under the results.

It seems to me that the model gives more of a linear prediction of the size of the fragmented material and the accuracy is not that good. For example, in sample 2, 95% of the material is smaller than 10 mm and the prediction gives results of less than 50%. Also, there are no particles larger than 20 mm, while the prediction says that more than 40% of the particles have a size larger than 20 mmmm20mm. 20 mm. Please look at the results and try to explain the linearity and the error.

In conclusion is stated:”The estimation results revealed that our approach could estimate particle size distributions that were generally close to the actual values, with an accuracy of 96.3% and a loss function value of 0.1237”. I believe this is related to the training process, accuracy of estimation is les than 96.3%. Please explain.

 

Figure 6. T-Net Functions - please check do you need this figure; size of letter should be the like letter on others figures.

Figure 8. Example of Particle- size Distribution- Please rearrange figure to have diagram in the same line.

Data Availability Statement: I believe the text should be removed.

Also, please check Acknowledgments and Conflicts of Interest.

 

Author Response

Dear Reviewer,

Thank you for your insightful comments and suggestions on improving our manuscript. Your feedback is invaluable, and we are taking the necessary steps to address each point. Please find our responses below:

1. Regarding Figure 19 in the Conclusion: We apologize for the oversight. As suggested, we have relocated "Figure 19. Estimation Results of Particle Size Distribution" from the conclusion and placed it under the results section where it rightly belongs.

2. On Model Predictions: Your observation about the linearity of the model and the associated errors in some of the samples is astute. We have revisited our results and are working on refining our analysis. We will provide a detailed discussion explaining the perceived linearity and the discrepancies in certain predictions, particularly for Sample 2.

3. Accuracy in Conclusion: Thank you for pointing out the discrepancy. The stated accuracy of 96.3% indeed pertains to the training phase. We will clarify this in the manuscript and provide the exact estimation accuracy for a clearer picture of the model's performance.

4. Regarding Figure 6 & Figure 8: We have reviewed "Figure 6. T-Net Functions" and believe it adds value in visualizing our approach. However, we have standardized the letter sizes to maintain consistency with other figures. For "Figure 8. Example of Particle- size Distribution", we've adjusted the layout as recommended to ensure the diagrams are aligned.

5. Data Availability Statement: After consideration, we have removed this section from the manuscript. We appreciate your feedback on this.

6. Acknowledgments & Conflicts of Interest: We've reviewed and made the necessary corrections in these sections to ensure clarity and accuracy.

Once again, we sincerely appreciate your detailed feedback and are confident that these modifications will enhance the quality of our work. We look forward to your further comments and suggestions.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The reviewer has no further comments.

Reviewer 3 Report

I have reviewed the manuscript, and I'm pleased to confirm that all of the suggested changes have been successfully incorporated into the paper. Thank you for your effort.