Applying a Neural Network-Based Machine Learning to Laser-Welded Spark Plasma Sintered Steel: Predicting Vickers Micro-Hardness

Round 1

Reviewer 1 Report

Dear authors,

On the basis of which the welding parameters described in section 3.3 are selected. Literature references are required.

“As the literature indicates, the chemical composition…” What kind of literature? Confirm please.

The authors present various calculated and measured data, however, there is no discussion of these results, which is confirmed by the absence of a conclusion.

In addition, it is not clear how the calculations take into account the structure and composition of the material? What are the limits of such calculations? Are they applicable only to the material studied in the work?

The conclusion is more like an Introduction and does not show any conclusions on the work, but only describes the work done. The conclusion needs to be completely changed.

In addition, the abstract states that the model can be used to predict the mechanical properties of materials, however, apart from hardness, there is no other data in the work. This needs to be clarified.

Author Response

Point 1: On the basis of which the welding parameters described in section 3.3 are selected. Literature references are required.

 Response 1: Please provide your response for Point 1. (in red)

I have indicated how I arrived at the welding parameters used.

Point 2: “As the literature indicates, the chemical composition…” What kind of literature? Confirm, please.

Response 2: Please provide your response for Point 2. (in red)

As the literature indicates, the chemical composition of the WZ is significant for the mechanical integrity of welded metallic alloy. The microstructure of the WZ differs from the base metal (BM) because of the variations in the chemical composition and the thermal history of the WZ [29].

The literature consulted has been cited as reference 29

Point 3: The authors present various calculated and measured data, however, there is no discussion of these results, which is confirmed by the absence of a conclusion".

Response 3: Please provide your response for Point 5. (in red)

I have tried to mention and put some discussion in section 5 and concluding sections about R2 value for training and test data, including MSE and MAE values. I have also further added more flesh to the discussion of the data.   

I have highlighted in section 5.0, as explained below: Under Table 4, I explained that Table 5 and 6 shows the maximum and minimum error values possible for the prediction of hardness value to be 9.57% and 0.09% (Pg 13).

Also, I included the hardness value for figure 8, in chapter 4

Point 4: In addition, it is not clear how the calculations take into account the structure and composition of the material? What are the limits of such calculations? Are they applicable only to the material studied in the work?

Response 4: Please provide your response for Point 3. (in red)

The major focus is to predict the mechanical property(Hardness in this case) of the welded sintered material.

For machine learning prediction, the y(Hardness), which is the dependent variable is been predicted by the sintering processing parameters and welding processing parameters as the independent variables (xi+1). These parameters have effects on the overall hardness.

Before setting out on this research I consulted some research work, which guided the formulation of the machine learning prediction theory.

This theory can be applied to any material.

Part of the work consulted has been cited in the introduction part of the manuscript.

Application of artificial neural network to predict Vickers microhardness of AA6061 friction stir welded sheets

Point 5: The conclusion is more like an Introduction and does not show any conclusions on the work, but only describes the work done. The conclusion needs to be completely changed.

Response 5: Please provide your response for Point 4. (in red)

The conclusion has been reconstructed; the issue is with the sentence I started with. There are some findings indicated in the conclusion that is not in the introduction. The conclusion talks about the findings and proposes full integration of analysis and prediction into one framework.

I can say that the way the conclusion is written is different from the introduction

Point 6: In addition, the abstract states that the model can be used to predict the mechanical properties of materials, however, apart from hardness, there is no other data in the work. This needs to be clarified.

Response 6: Please provide your response for Point 5. (in red)

The second sentence of the Abstract clearly stated that the aim is to use the ANN algorithm to estimate the Vickers hardness at the weld Zone.

Note: I have also tried to improve the grammatical errors noticeable in the write-up.

Author Response File: Author Response.pdf

Reviewer 2 Report

1）When using the English abbreviation for the first time, the full English name should be indicated, such as” WZ”, “ADAM” abbreviation.

2) There are many mistakes in the text, please check and correct them carefully.

For example, a) the Table2 in subsection 3.3 should be Figure2, b) the two different tables on page 11 and page 8 are both Table2.

3） After the experimental equipment it should be the experimental design, write a separate subsection as the experimental design, specific powor, time, temp, and speed, and use the table to fill in the content of the experimental design, rather than the line graph in Figure 8

4) It is recommended to mention the parameters or roles of the activation function, the cost function, Back-propagation and Optimizer in the flowchart of Figure 4 in Chapter 4, and to beautify the flowchart.

5）In Page 11 " Meanwhile, different activation functions and optimizers available under Keras API were tested , but the one that can give the best optimized predicted value is ReLU and RMSProp ", while,  in the text, there is no comparative testing data to show.

Author Response

Point 1: When using the English abbreviation for the first time, the full English name should be indicated, such as” WZ”, “ADAM” abbreviation.

Response 1:

I have worked on this, the changes are highlighted in yellow

Point 2:  There are many mistakes in the text, please check and correct them carefully.

For example, a) Table2 in subsection 3.3 should be Figure2, b) the two different tables on page 11 and page 8 are both Table2.

Response 2:

I have sorted and corrected the Tables appropriately corresponding to where the tables were mentioned

Point 3  After the experimental equipment it should be the experimental design, write a separate subsection as the experimental design, specific powor, time, temp, and speed, and use the table to fill in the content of the experimental design, rather than the line graph in Figure 8

Response 3: Please provide your response for Point 3. (in red)

Section 4 Was not filled with graphs only.

The article dwells more on machine learning design than experimental design. Therefore the outline is as follows

• Introduction
• Significance of research
• Experimental Process
• ANN Algorithm
• Results and Discussion
• Conclusion

Point 4: It is recommended to mention the parameters or roles of the activation function, the cost function, Back-propagation and Optimizer in the flowchart of Figure 4 in Chapter 4, and to beautify the flowchart.

Response 4: Please provide your response for Point 4. (in red)

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.