Augmented Data-Driven Approach towards 3D Printed Concrete Mix Prediction

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript presents research on the possibility of data augumentation and the use of various machine learning models in predicting the parameters of 3D printed concrete. The research used to train and test machine learning models was taken from available literature sources and is not the result of the authors' research. Additionally, in their analyses, the authors focus mainly on assessing the quality of prediction of selected parameters (slump, compresive strength, and anisotropy of compresive strength). Perhaps the presented models could be used to optimize concrete for 3D printing, however, in my opinion, such an issue is not analyzed in the manuscript, so I think that a more appropriate title would be "Augmented Data-Driven Approach Towards 3D Printed Concrete Mix Prediction".

Apart from the question of whether the title is consistent with the content, I rate the quality of the manuscript highly. The authors conducted an in-depth literature review. The research methodology was clearly described. The conclusions are consistent with the results obtained from the model analyses. I have a few minor comments, mainly related to text editing.

1) Fig. 1 - In my opinion, it is appropriate to show buildings that were actually made using 3D printing, so I propose replacing the habitat printed on Mars with another example.

2) Fig 4 - For examples b); c) and d) I propose to show the 3D printing directions - if they were supposed to be marked in colors on the cross of the coordinate system (Dir1, 2 and 3), they are shown incorrectly

3) The abbreviation HWRA was first used on page 17 (6th line from the bottom), I did not find its meaning

4) Fig. 7 – The bars contain four colors, I found a description of only three of them in the legend. What is the meaning of the fourth color (purple?)

5) Page 20, last sentence in the first paragraph - it seems to me that the wrong table was referenced in the hyperparameters reference - should it be Table 3?

6) Page 21, last sentence - should it be Table 4?

6) Page 23, last sentence - should it be Table 5?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The submitted “Article” with the Manuscript ID: applsci-3119357 and the title: “Augmented Data-Driven Approach Towards 3D Printed Concrete Mix Optimization” investigates the application of machine learning models trained on augmented datasets to predict properties of 3D printed concrete mixes, addressing challenges of limited data availability in the emerging field of 3D concrete printing. By enhancing the predictive capability of the models through data augmentation techniques like deep generative adversarial networks and bootstrap resampling, the research is trying to enable precise predictions of both fresh-state properties (printability) and hardened-state properties (mechanical strength). To conclude, this research streamlines the mix design processes, reduces resource-intensive iterations, and facilitates the formulation of printable concrete mixes suitable for various 3D printing machines, ultimately advancing the adoption of 3D concrete printing technology in the construction industry.

The manuscript presented in the paper is comprehensive, showcasing a well-organized composition. Nevertheless, certain areas require refinement to meet the necessary academic standards. The following comments and suggestions are provided for the authors' review:

1. While the introduction furnishes foundational context, the scope of the literature review seems constrained and somewhat outdated. To bolster the aims of the study, it is advisable to integrate the following pertinent subjects:

(a) Recent Advances in 3D Printed Concrete Mix Optimization: Discuss innovations in mixing techniques, materials used, and the implications for structural integrity and design flexibility.

(b) Integration of Smart Technology in Construction Materials: Address the integration of smart technologies in construction materials for monitoring structural health, stress responses, and environmental interaction. It is essential to address nondestructive evaluation (NDE) techniques that use big data to monitor the structural health of the material. Discuss the potential for future smart cementitious composites to provide real-time data and adapt to changing conditions. How could the experimental data from those NDEs provide valuable data for applying machine learning models to predict the properties of 3D-printed concrete mixes?

Implementing a comprehensive literature review aligned with the aforementioned suggestions would prove advantageous. As illustrative references, the subsequent articles may serve as exemplars, offering insight into the discussed issues:

-"Acoustic monitoring for the evaluation of concrete structures and materials," in Acoustic Emission and Related Nondestructive Evaluation Techniques in the Fracture Mechanics of Concrete (Second Edition), 2021.

-"Revealing the nature of metakaolin-based concrete materials using artificial intelligence techniques," Construction and Building Materials, 2022.

2. All abbreviations should be described at the beginning of usage.

3. A more detailed theoretical explanation of all the results presented is needed.

4. Throughout the manuscript, figures and tables frequently precede their corresponding mentions in the text. It is recommended that this arrangement be revised to ensure that all figures and tables are introduced in the text before their appearance.

6. A comprehensive notation list delineating all variables referenced within the equations is required.

7. In Figure 6, a correction in the captions is needed. The (a) is repeated twice.

8. It is recommended that the tables be adjusted in numbering to ensure alignment when they are mentioned in the text.

9. A more comprehensive theoretical explanation is required to understand why DGAN-trained SVM, ANN, and XGBoost models achieved higher R2 scores for cast compressive strength during training than BR-trained models. Similarly, these trends were observed consistently in printed compressive strength across directions 1, 2, and 3. In contrast, BR-trained models consistently outperformed DGAN-trained models in slump flow metrics.

Comments on the Quality of English Language

The English needs polishing since the phraseology seems cumbersome in a few places.

Further, extensive sentences could be divided into two or three smaller ones to improve the reading of the manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

In this paper, the dataset used to formulate the mix design of printable concrete was extracted from the literature. This dataset included mix constituents as input features, while the output features were slump flow, cast compressive strength, and anisotropic compressive strength. The research explores the usage of data augmentation techniques like deep generative adversarial network and bootstrap resampling to increase the size of the available data to train three machine learning models namely support vector machine, artificial neural network, and extreme gradient boosting regression.

This research is interesting and has certain engineering application. However, some aspects of the manuscript need to be improved.

1. The abstract of a paper should summarize the core content of the research, including the purpose, methods, results and conclusions. The summary of the manuscript does not explicitly state the conclusions of the study, and it is recommended to supplement.

2. Any abbreviation that appears for the first time in the abstract shall be preceded by its full name.

3. In the introduction, page 2: "Research has proven that this way of printing structures can reduce construction costs by 33% to 60%". Is this cost reduction clear-cut? It is recommended to list comparative data.

4. Manuscript page 2, “This way of construction holds immense potential as shown in Figure 1 and it is believed that a robust additive  construction system can be the effective solution for future infrastructure development in both lunar and Martian  environments ". Please confirm the relevance of Figure 1 to the statement in this paragraph.

5. In the introduction, 1.1. Research Significance can be used as the last paragraph of the introduction, and it does not need to be listed as a separate section, and the content is repeated.

6. In the manuscript, Figure 4. (a) Concrete printing through the extruder, (b) Cast compressive strength, (c) Printed compressive strength (direction 1), (d) Printed compressive strength (direction 2), (e) Printed compressive strength (direction 30. The expression "direction 30" is wrong and needs to be corrected.

7. Page 12 of the manuscript, "... printed compressive strength in direction 3 (MPa) (as shown in Figure 3), slump flow (mm) ", this part of the content is not consistent with Figure 3.

8. Table 1 shows Statistics of the real dataset features. The data source of the table should be clearly marked and complete.

9. The conclusions of the manuscript should be further condensed, with appropriate additions to the outlook for the future.

Comments on the Quality of English Language

Minor editing of English language required

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The revised version of the submitted “Article” with the Manuscript ID: applsci-3119357-v2 and the title: “Augmented Data-Driven Approach Towards 3D Printed Concrete Mix Optimization” has been improved. However, the following comments of the previous review round have not been considered adequately:

Comment 1. The scope of the literature review still seems constrained and somewhat outdated. Innovations in mixing techniques of 3D printed concrete and the implications for structural integrity should be discussed further. Report of the integration of smart technology in construction materials is still shallow and requires further explanations

Comment 3. A more detailed theoretical explanation of all the results presented is still required.

Comment 6. A comprehensive notation list delineating all variables referenced within the equations is required.

Comment 9. A more comprehensive theoretical explanation is required to understand why DGAN-trained SVM, ANN, and XGBoost models achieved higher R2 scores for cast compressive strength during training than BR-trained models.

All the other comments are well-addressed. No revision is required.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Comments and Suggestions for Authors

The new revised version of the submitted “Article” with the Manuscript ID: applsci-3119357-v3 and the title: “Augmented Data-Driven Approach Towards 3D Printed Concrete Mix Prediction” has been improved substantially. The efforts performed by the Authors to consider all the recommendations and to respond to the questions, criticisms, and comments of the previous review are greatly appreciated. All of them have been considered sincerely and have been replied to accordingly. Hence, the revised version of this paper is suggested to be accepted for publication as it is.