Machine Learning Models for the Prediction of the Compressive Strength of Self-Compacting Concrete Incorporating Incinerated Bio-Medical Waste Ash

Round 1

Reviewer 1 Report

In this study, SCC was prepared by partially substituting coarse aggregate with LECA, partially substituting cement with GGBS and partially substituting fine aggregate with IBMWA, and the compressive strength values were modeled by machine learning approach. There are some suggestions before this manuscript considered for publication.

1.In section 2.5. What instrument or method was used to obtain the chemical composition of IBMWA, please specify it. Please add the full name of the specification in the text.

2.In 3.2.1. Figures 5-8 can be merged and the size of the images can be standardized.

3.Please add references in 3.5.

4. In 4.1.1. What is the significance of the four test methods? Please give a recommended fresh property test method.  

5. In 4.1.2. Please explain the changes in SCC performance.

6. Suggest authors to merge the Figure 13 and Figure14.

7. In 4.2. Please give the weight of influence of the variable factors.

Author Response

Reviewer 1

In this study, SCC was prepared by partially substituting coarse aggregate with LECA, partially substituting cement with GGBS and partially substituting fine aggregate with IBMWA, and the compressive strength values were modeled by machine learning approach. There are some suggestions before this manuscript considered for publication.

Reply: The authors would like to thank the reviewer for their precious time and invaluable encouraging comments. The suggestions for improvement are considered and carefully addressed.

1. In section 2.5. What instrument or method was used to obtain the chemical composition of IBMWA, please specify it. Please add the full name of the specification in the text.

Reply: The authors wish to clarify that, the chemical composition of IBMWA presented in Table 7 was determined by X-ray fluorescence (PAN Analytical) conducted at an outsourced chemical laboratory following the guidelines of IS 12803 (1989): Methods of analysis of hydraulic cement by X-ray fluorescence spectrometer. The details are added in the text.

2. In 3.2.1. Figures 5-8 can be merged and the size of the images can be standardized.

Reply: The suggestion is taken into consideration and the figures are merged appropriately.

3. Please add references in 3.5.

Reply: The suggestion is taken into consideration and the equations are properly referenced with equation numbers appropriately.

4. In 4.1.1. What is the significance of the four test methods? Please give a recommended fresh property test method.

Reply:   The slump flow test is used assess the horizontal free flow of self-compacting concrete in the absence of obstructions. In J Ring test, the SCC sample is allowed to flow/spread in all directions while being restrained by a circular arrangement of reinforcing bars. The V funnel test measures the ease of flow of concrete, shorter flow time indicates greater flow ability. The L-Box test measures the filling (flowing) and passing capacity of self-compacting concrete to flow through tight openings, including gaps between reinforcing bars and other barriers without splitting or blocking. The following details are available in section 3.2.1. All the four methods are unique and hence all the four are recommended and hence included in the study.

5. In 4.1.2. Please explain the changes in SCC performance.

Reply: The suggestion is taken into consideration and the section 4.1.2. is further elaborated. Please check the 4.1.2 of revised version.

6. Suggest authors to merge Figure 13 and Figure14.

Reply: The suggestion is taken into consideration and the figures are merged appropriately.

7. In 4.2. Please give the weight of influence of the variable factors.

Reply: Since, all the SCC constituents were considered as input parameters for the model building process, the weightage of influence of the variable factors was not considered in the study. The input variables include SCC ingredients: LECA (%), GGBS (%), IBMWA (%), density of SCC (kg/m3), water to binder (W/B) ratio, superplasticizer (SP) (%) and the compressive strength (MPa) of SCC was considered as the output parameter. Knowing that GGBS, IBMWA, and LECA were replaced for cement, fine aggregate, and coarse aggregate, respectively, it is obvious that the cement (%) may be simply computed if the GGBS (%) is known, and vice versa. Hence, the SCC constituents, namely cement (%), fine aggregate (%) and coarse aggregate (%) were excluded as input parameters.

Reviewer 2 Report

 

Dear Authors,

 

 

Manuscript's strengths:

The article is well presented with an interesting contribution to the research area.

The introduction is well presented with good justificative.

The results were clearly presented and well interpreted by the authors.

Overall, I understand that the article is well written.

 

 

Manuscript's weaknesses:

 

Despite the good discussion of the results, I understand that the authors need to include a comparison with results previously obtained in the literature. This part is even important to justify the research that was carried out.

 

 

A point-by-point list of your major recommendations for the improvement of the manuscript:

 

·      A little more in-depth analysis of the results including a comparative analysis of the results with similar works in the literature on the context of the research presented. If there are no similar works in the literature, the authors should make this clear in the work.

 

no comments

Author Response

Reviewer 2

Manuscript's strengths:

The article is well presented with an interesting contribution to the research area.

The introduction is well presented with good justificative.

The results were clearly presented and well interpreted by the authors.

Overall, I understand that the article is well written.

Reply: The authors thank the reviewer for appreciating our efforts and recommending our manuscript.

Manuscript's weaknesses:

Despite the good discussion of the results, I understand that the authors need to include a comparison with results previously obtained in the literature. This part is even important to justify the research that was carried out.

Reply: The suggestion is taken into consideration and the comparison of our results with previously obtained in the literature is included. Please refer to section 4.

A point-by-point list of your major recommendations for the improvement of the manuscript:

A little more in-depth analysis of the results including a comparative analysis of the results with similar works in the literature on the context of the research presented. If there are no similar works in the literature, the authors should make this clear in the work.

Reply: The suggestion is taken into consideration and the comparison of our results with previously obtained in the literature is included. Please refer to the text highlighted in red color in section 4. For the first time, the CS of SCC produced by blending GGBS, IBMWA and LECA is modeled utilizing the most recent ML techniques.

Reviewer 3 Report

1 The number of references is not rich enough and they are somewhat outdated. Please further increase the number of references and the proportion of publications in the past five years

2 In Chapter 3.2.1, the paper used four methods to measure the performance of concrete, but the purpose and significance of these measurement experiments were not clearly explained. Figures 6-8 are experimental operation images and cannot characterize the inherent performance of concrete.

3 The image format of the paper is too single and not aesthetically pleasing, such as too many bar charts and some experimental operation images taken too casually. Please optimize and adjust.

4 When introducing the principles of the algorithm model, there is a lack of necessary illustrations, formulas, etc. to explain, and only textual descriptions are clearly insufficient.

5 The optimization process of how the hyperparameters of the algorithm models are determined is not reflected in the paper.

6 The conclusion section of the paper is too simple and the content is somewhat thin. Please optimize it.

The authors may add more state-of-art application articles for the integrity of the manuscript. For ML algortihm applications, please refer to An experimental investigation and machine learning-based prediction for seismic performance of steel tubular column filled with recycled aggregate concrete; Reviews on Advanced Materials Science.

Author Response

Reviewer 3

1 The number of references is not rich enough and they are somewhat outdated. Please further increase the number of references and the proportion of publications in the past five years.

Reply: The suggestion is taken into consideration and recent literature related to SCC are included in the Introduction and Results section.

2 In Chapter 3.2.1, the paper used four methods to measure the performance of concrete, but the purpose and significance of these measurement experiments were not clearly explained. Figures 6-8 are experimental operation images and cannot characterize the inherent performance of concrete.

Reply: The following details are available in section 3.2.1. The slump flow test is used assess the horizontal free flow of self-compacting concrete in the absence of obstructions. In J Ring test, the SCC sample is allowed to flow/spread in all directions while being restrained by a circular arrangement of reinforcing bars. The V funnel test measures the ease of flow of concrete, shorter flow time indicates greater flow ability. The L-Box test measures the filling (flowing) and passing capacity of self-compacting concrete to flow through tight openings, including gaps between reinforcing bars and other barriers without splitting or blocking. All the four methods are unique and hence all the four are recommended and hence included in the study.

Figure 6-8 is replaced with a single image (Figure 5)

3 The image format of the paper is too single and not aesthetically pleasing, such as too many bar charts and some experimental operation images taken too casually. Please optimize and adjust.

 

Reply: The suggestion is taken into consideration and the figures are merged appropriately.

4 When introducing the principles of the algorithm model, there is a lack of necessary illustrations, formulas, etc. to explain, and only textual descriptions are clearly insufficient.

Reply: Since, the authors have used standard machine learning (ML) methods and no alterations or improvements have been done to standard ML methods, for detailed information related to the implemented algorithms, the authors redirect the readers to refer to standard literature mentioned in respective sub-sections. Reproducing same formulas and illustrations is not useful in the context of this article.

5 The optimization process of how the hyperparameters of the algorithm models are determined is not reflected in the paper.

Reply: The hyperparameters of all the models tuned by trial and error approach during the model development process are presented in Table 9.

6 The conclusion section of the paper is too simple and the content is somewhat thin. Please optimize it.

Reply: The suggestion is taken into consideration and the conclusion section is revised. Please refer to manuscript.

The authors may add more state-of-art application articles for the integrity of the manuscript. For ML algortihm applications, please refer to An experimental investigation and machine learning-based prediction for seismic performance of steel tubular column filled with recycled aggregate concrete; Reviews on Advanced Materials Science.

Reply: The suggested article is appropriately cited and referenced.

Reviewer 4 Report

The manuscript describes the modeling of the compressive strength (CS) of SCC mixes using machine learning approaches, using the herein determined CS of experimental SCC containing partially replaced components by LECA, GGBS, and IBMWA, respectively.

ML development using own experimental data approach is appropriate and is expected to generate significant results.

The experimental part was executed well and the results are well presented. The only problem is that it was not clear what the n was for each test group, and the results are presented as single absolute values rather than mean (SD). Further, there lacks statistical analysis for verifying the statistical significance of each substitution’s influence on the SCC properties.

The models are of high quality, applicable and suitable for the study. Detailing each model and the formulas was necessary and thus was done properly.

A major revision is recommended .

Some suggestions:

1. Sentence in Line 18-19 means the opposite due to poor English usage. Rewrite. Or even better, should combine with the previous sentence and add the % into each replacement.

2. Introduction Line 41 and 45 are repetitive.  

3. “…aluminosilicates and calcium silicates that are acquired in a molten state…” could use an important citation doi 10.1002/advs.201902209

4. Section 2 Materials, the full product info of each procured materials should be given for reproducibility. The data Tables, were they determined by the authors, or by the producers? If it is the latter, the sources of Tables 1-7 should be given. Technically, if these were determined by the authors, the test details should be provided. Since these are not the main tests of the study, these can be provided in the supplementary info.

5. What is the control 0-0-0 mix design cement:coarse:fine aggregates?

6. What is the controlled environment of the curing tank?

7. There are many models, why did you choose these 3 ML models?

8. Please change N/mm^2 to MPa.

9. Does CBR predict well for control 0-0-0 as well as for SCC with partially-replaced components?

minor revision

Author Response

Reviewer 4

The manuscript describes the modeling of the compressive strength (CS) of SCC mixes using machine learning approaches, using the herein determined CS of experimental SCC containing partially replaced components by LECA, GGBS, and IBMWA, respectively.

ML development using own experimental data approach is appropriate and is expected to generate significant results.

Reply: The authors thank the reviewer for appreciating our efforts and recommending our manuscript.

The experimental part was executed well and the results are well presented. The only problem is that it was not clear what the n was for each test group, and the results are presented as single absolute values rather than mean (SD). Further, there lacks statistical analysis for verifying the statistical significance of each substitution’s influence on the SCC properties.

Reply: The reviewer concern couldn’t be understood clearly. What is ‘n’? The manuscript is available as preprint and the data used for modeling is published online along with preprint. The reviewer could refer to data presented in ANNEXURE.

The models are of high quality, applicable and suitable for the study. Detailing each model and the formulas was necessary and thus was done properly.

A major revision is recommended.

Reply: The authors thank the reviewer for appreciating our efforts and recommending our manuscript.

Some suggestions:

1. Sentence in Line 18-19 means the opposite due to poor English usage. Rewrite. Or even better, should combine with the previous sentence and add the % into each replacement.

Reply: The sentence is modified as given below.

“LECA, GGBS, and IBMWA were replaced with coarse aggregate, cement, and fine aggregate, respectively at different substitution levels of 10%, 20%, and 30%.”

2. Introduction Line 41 and 45 are repetitive.

Reply: The authors really don’t feel so. Please we request the reviewer to have a relook.

3. “…aluminosilicates and calcium silicates that are acquired in a molten state…” could use an important citation doi 10.1002/advs.201902209

Reply: The suggested article is appropriately cited and referenced.

4. Section 2 Materials, the full product info of each procured materials should be given for reproducibility. The data Tables, were they determined by the authors, or by the producers? If it is the latter, the sources of Tables 1-7 should be given. Technically, if these were determined by the authors, the test details should be provided. Since these are not the main tests of the study, these can be provided in the supplementary info.

Reply: The data presented in Table 1 is obtained from the cement manufacturer (Hardcopy). The data is not published online to cite the source. The data presented in Tables 2-7, is from the experimental study conducted by authors. The details related to tests and the specifications/guidelines used is mentioned in the respective sections. Clarification of Table 7 is provided owing to the comments from another reviewer.

“The authors wish to clarify that, the chemical composition of IBMWA presented in Table 7 was determined by X-ray fluorescence (PAN Analytical) conducted at an outsourced chemical laboratory following the guidelines of IS 12803 (1989): Methods of analysis of hydraulic cement by X-ray fluorescence spectrometer. The details are added in the text.”

5. What is the control 0-0-0 mix design cement:coarse:fine aggregates?

Reply: Mix design codes, such as SCC 0-0-0 (which stands for self-compacting concrete with 0% LECA, 0% GGBS, and 0% IBMWA), have been coined for ease of identification.

6. What is the controlled environment of the curing tank?

Reply: Ponding method of curing was considered for curing of test specimens. Water was filled up to a height of 25 mm to 50 mm inside the rectangular pond at least two times a day and maintained at room temperature.

7. There are many models, why did you choose these 3 ML models?

Reply: The CatBoost regressor is a relatively new paradigm and hence it was chosen to test its applicability to model Compressive strength of SCC mixes. ANN and GTB models were applied to compare their performance with that of CBR approach.

8. Please change N/mm^2 to MPa.

Reply: The suggestion is taken into consideration and changed accordingly

9. Does CBR predict well for control 0-0-0 as well as for SCC with partially-replaced components?

Reply: Yes, the CBR model provided generalized performance for all the test cases.

Round 2

Reviewer 3 Report

accept

Author Response

We thank the reviewer for accepting our article for publication.

Reviewer 4 Report

The majority of the issues have been addressed.

The following two remain unresolved:

1. SCC 0-0-0 stands for self-compacting concrete with 0% LECA, 0% GGBS, and 0% IBMWA, yes, this is true. But what water:cement:coarse:fine aggregates ratio did you use? This is for reproducibility.

 

2. n is sample size of each group, usually minimium 3 is required for statistical certainty, but it depends on experimental design, and the values are normally reported as mean (standard deviation). And for inter-group comparison, statistical analysis are used to determine if the differences are statistically significant or not.

a quick search found this post, both this reviewer and the authors may refresh our knowledge https://www.iwh.on.ca/what-researchers-mean-by/sample-size-and-power

Author Response

The majority of the issues have been addressed.

Reply: Thank you for accepting our revision.

The following two remain unresolved:

1. SCC 0-0-0 stands for self-compacting concrete with 0% LECA, 0% GGBS, and 0% IBMWA, yes, this is true. But what water:cement:coarse:fine aggregates ratio did you use? This is for reproducibility.

Reply: I sincerely hope the reviewer misunderstood the preceding response. We know that LECA is replaced with Coarse aggregates, so if the LECA replacement % is 20%, its obvious that 80% is the remaining coarse aggregate content. Hence, we didn't consider coarse aggregate separately as a input variable. Similarly if we know W/B ratio, water content is easily accounted there, hence water content is not considered separately. I hope the reviewer now understands the model input parameters considered.

 

2. n is sample size of each group, usually minimium 3 is required for statistical certainty, but it depends on experimental design, and the values are normally reported as mean (standard deviation). And for inter-group comparison, statistical analysis are used to determine if the differences are statistically significant or not.

a quick search found this post, both this reviewer and the authors may refresh our knowledge https://www.iwh.on.ca/what-researchers-mean-by/sample-size-and-power

Reply: The question of reviewer is somewhat fuzzy. However, we wish to clarify that during experimentation, triplet cubes were casted and tested for their compressive strength and the data of all triplet samples were considered for modeling purpose. We welcome the suggestion of the reviewer and consider it during our future works.

We thank the reviewer for his insightful comments and suggestion.