Optimization of 3D Printing Parameters on Deformation by BP Neural Network Algorithm

Round 1

Reviewer 1 Report

Dear authors,

 

This manuscript addresses a method to predict deformation of additively mnufactured polymer materials by using a machine learning method. Machine learning methods such as neural netowork are important technology to expand application of additive manufacturing technologies. However, the journal "metals" is not appropreate journal for polymer materials which this manuscript addresses, and then, it is better for you to submit this manuscript to another journal.

In addition, when you submit this manuscript to anot ther journal, this manuscript should be revised by considering following points.

1. You used the amount of deformation of the printed material as an inspection index, but you did not show how to measure the amout of deformation. The deformation is not scalar but vector.

 

2. You shoud show the amount of the deformation in the case of Figure 7b.

 

Sincerely yours.

 

 

 

Author Response

1. You used the amount of deformation of the printed material as an inspection index, but you did not show how to measure the amount of deformation. The deformation is not scalar but vector.

Response: Thanks so much for your professional and kind suggestion. The deformation is vector. In this paper, we consider the average value of the deformation in the X/Y/Z direction. Such as the X direction. The calculation process of the deformation amount is as follows: the set length is a1 (mm), the actual printing length is b1 (mm), and the deformation amount is A=｜b1-a1｜/a1. In the Y direction, the calculation process of the deformation amount is as follows: the set length is a2 (mm), the actual printing length is b2 (mm), and the deformation amount is B=｜b2-a2｜/a2. In the Z direction, the calculation process of the deformation amount is as follows: the set length is a3 (mm), the actual printing length is b3 (mm), and the deformation amount is B=｜b3-a3｜/a3. The average value of the deformation is S=(A+B+C)/3. Thank you very much for your support and suggestions for our work. We will explain this aspect in detail in the follow-up work, and I hope you can pay attention to our follow-up work.

2. You should show the amount of the deformation in the case of Figure 7b.

Response: Thanks so much for your professional and kind suggestion. We will design the content you suggested in future experiments. Thank you very much for providing ideas for our follow-up experimental verification.

Author Response File: Author Response.pdf

Reviewer 2 Report

Manuscript numbered “metals-1857620” has been reviewed:

The introduction needs some improvements.

Please note that NN is a predictive tool. You need to use optimization methods for optimization, the simultaneous use of NN and GA also is an effective method.

Results have been just reported, please compare your findings with other research.

Please explain the "BP" abbreviation and method used in the paper. This will be useful for researchers who study this for the first time.

Please improve the quality of the figure and increase the fonts in the figures.

The following papers are suggested for introduction:

Optimization of LB-PBF process parameters to achieve best relative density and surface roughness for Ti6Al4V samples: using NSGA-II algorithm

Multi‐objective optimization of part‐building orientation in stereolithography

Efficient design optimization of variable-density cellular structures for additive manufacturing: theory and experimental validation

Process parameters optimization for improving surface quality and manufacturing accuracy of binder jetting additive manufacturing process

 

 

 

Author Response

Reviewer #2

1.The introduction needs some improvements.

Response: Thanks so much for your professional and kind suggestion. We have revised the Introduction for better understanding by the reader.

 

2. Please note that NN is a predictive tool. You need to use optimization methods for optimization, the simultaneous use of NN and GA also is an effective method.

Response: Thanks so much for your professional and kind suggestion. In the subsequent experimental design, the relevant design algorithm (such as: NN and GA algorithm) is used for optimization.

 

3. Results have been just reported, please compare your findings with another research.

Response: Thanks so much for your professional and kind suggestion. We will make a comparative analysis in the follow-up work.

 

4. Please explain the "BP" abbreviation and method used in the paper. This will be useful for researchers who study this for the first time.

Response: Thanks so much for your professional and kind suggestion. We have added the abbreviation of BP, namely backpropagation (BP).

 

5. Please improve the quality of the figure and increase the fonts in the figures.

Response: Thanks so much for your professional and kind suggestion. We have improved the quality of the figure and increased the fonts in the figures in the revised manuscript.

 

6. The following papers are suggested for introduction: Optimization of LB-PBF process parameters to achieve best relative density and surface roughness for Ti6Al4V samples: using NSGA-II algorithm; Multi‐objective optimization of part‐building orientation in stereolithography; Efficient design optimization of variable-density cellular structures for additive manufacturing: theory and experimental validation; Process parameters optimization for improving surface quality and manufacturing accuracy of binder jetting additive manufacturing process.

Response: Thanks so much for your professional and kind suggestion. These references have been added in the Introduction in the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper was prepared correctly. The topics discussed in the paper are among the interesting issues in the development of contemporary technologies. Optimisation of the 3D printing process is important to optimise the quality of the process. The use of an artificial neural network BP with the Lavenberg-Marquardt algorithm was created to train the network and predict orthogonal experimental data is justified from a scientific research point of view.
The proposed method fulfils the expectations of high reliability in learning
simulation tests and numerical prediction.

Author Response

Reviewer #3

The paper was prepared correctly. The topics discussed in the paper are among the interesting issues in the development of contemporary technologies. Optimization of the 3D printing process is important to optimize the quality of the process. The use of an artificial neural network BP with the Lavenberg-Marquardt algorithm was created to train the network and predict orthogonal experimental data is justified from a scientific research point of view.
The proposed method fulfils the expectations of high reliability in learning simulation tests and numerical prediction.

Response: Thank you for acknowledging our work.

Author Response File: Author Response.pdf

Reviewer 4 Report

The study sent for review deals with an important aspect of the selection and optimisation of incremental manufacturing parameters using the FDM technique. In the study, the authors presented a methodology for the selection of process parameters by an optimisation method using BP neural network construction.

The study is interesting but has some inaccuracies and needs to be supplemented.

Major comments:

1. the authors show excessive detail in some issues, e.g. the photo showing the filament spool. On the other hand, in important issues they omit important information that makes it difficult for the reader to analyse the results. An example is the way in which the value of Rj was determined and the values contained in Table 5 were calculated. I suggest providing a relation or example for the calculation of a given quantity.

(2) The presentation of the final result of the proposed method of optimising parameter selection as a printout of a figure is an unreliable result that does not provide the possibility of objectively assessing the validity of the proposed approach to process parameter selection. The study should be supplemented with an authoritative way to evaluate the final effect. I suggest to perform a 3D scan and, in a suitable program, make a comparison between the geometry of the CAD model and the geometry of the manufactured model, or to choose a different geometry and, through classical measurement techniques, demonstrate the least deformation of the manufactured part.

Details notes:

1.       3D printing technology is commonly attributed as part of the fourth industrial revolution not the third as stated by the authors - page 1 line 25.

2.       How to understand the term "manufacturing under extreme harsh conditions" - p.2 line 48 - 3D printing requires, like most manufacturing techniques, the maintenance of strict manufacturing conditions. I suggest developing this thought a little further or changing the wording.

3.       What does the parameter "printing density" mean - p.2 line 79 - does it refer to filling density? If so this should be added.

4.       Table 2 last row column A should be a value of 3 not 4.

Author Response

Reviewer #4

The study sent for review deals with an important aspect of the selection and optimisation of incremental manufacturing parameters using the FDM technique. In the study, the authors presented a methodology for the selection of process parameters by an optimisation method using BP neural network construction.

The study is interesting but has some inaccuracies and needs to be supplemented.

Major comments:

1. the authors show excessive detail in some issues, e.g. the photo showing the filament spool. On the other hand, in important issues they omit important information that makes it difficult for the reader to analyse the results. An example is the way in which the value of Rj was determined and the values contained in Table 5 were calculated. I suggest providing a relation or example for the calculation of a given quantity.

Response: Thanks so much for your professional and kind suggestion. We have replaced the pictures in the revised manuscript. The value of Rj is calculated by the following formula: Rj= Max {}-min{}. The main calculation process is referred to in reference 32.

32. Sun W, Tian M, Zhang P, et al. Optimization of plating processing, microstructure and properties of Ni–TiC coatings based on BP artificial neural networks. Transactions of the Indian Institute of Metals, 2016, 69: 1501-1511.

(2) The presentation of the final result of the proposed method of optimising parameter selection as a printout of a figure is an unreliable result that does not provide the possibility of objectively assessing the validity of the proposed approach to process parameter selection. The study should be supplemented with an authoritative way to evaluate the final effect. I suggest to perform a 3D scan and, in a suitable program, make a comparison between the geometry of the CAD model and the geometry of the manufactured model, or to choose a different geometry and, through classical measurement techniques, demonstrate the least deformation of the manufactured part.

Response: Thanks so much for your professional and kind suggestion. We will design the content you suggested in future experiments. Thank you very much for providing ideas for our follow-up experimental verification.

Details notes:

1. 3D printing technology is commonly attributed as part of the fourth industrial revolution not the third as stated by the authors - page 1 line 25.

Response: Thanks so much for your professional and kind suggestion. We have revised this mistake in the revised manuscript.

2. How to understand the term "manufacturing under extreme harsh conditions" - p.2 line 48 - 3D printing requires, like most manufacturing techniques, the maintenance of strict manufacturing conditions. I suggest developing this thought a little further or changing the wording.

Response: Thanks so much for your professional and kind suggestion. We have replaced "manufacturing under extreme harsh conditions" with "precision manufacturing processes".

3. What does the parameter "printing density" mean - p.2 line 79 - does it refer to filling density? If so this should be added.

Response: Thanks so much for your professional and kind suggestion. We changed the printing density to filling density in the revised manuscript.

4. Table 2 last row column A should be a value of 3 not 4.

Response: Thanks so much for your professional and kind suggestion. We have revised this mistake in the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 5 Report

1. Please clarify in the Abstract and Introduction the aim and novelty of the work.

2. Line 51-52: "there are more than one printing parameters" - Please clarify the range of the number of 3D printing parameters (e.g. 60-80) and from where they can be obtained automatically (from the printer software, from the slicer if it is a separate software, from CAD, etc.). The aforementioned clarification is important to clarify the complexity of the computational problem for AI.

3. Figures 4, 5, 6: enlarging the descriptions would improve the readability of the graphs.

4. Conclusion is very weak. It should be divided into two parts:

Discussion: compartment with results of the previous similar studies, limitations of the own results (low number of materials, samples printers and ANNs structures studied), and directions for further research,

Conclusions: short sum-up of the results taking into consideration aim of the study.

Author Response

Reviewer #5

1. Please clarify in the Abstract and Introduction the aim and novelty of the work.

Response: Thanks so much for your professional and kind suggestion. We have revised the Abstract and Introduction in the revised manuscript.

 

2. Line 51-52: "there are more than one printing parameters" - Please clarify the range of the number of 3D printing parameters (e.g. 60-80) and from where they can be obtained automatically (from the printer software, from the slicer if it is a separate software, from CAD, etc.). The aforementioned clarification is important to clarify the complexity of the computational problem for AI.

Response: Thanks so much for your professional and kind suggestion. There are more than one printing parameters, such as: printing speed (10～80 mm/s), printing temperature (100～300 ℃), filling density (10～70 %), scanning speed (10～70 mm/s) and initial layer thickness (0.1～0.5 mm), etc. These parameters are set in the setting system of the 3D printer.

 

3. Figures 4, 5, 6: enlarging the descriptions would improve the readability of the graphs.

Response: Thanks so much for your professional and kind suggestion. We have enlarged the graphs as you suggested.

 

4. Conclusion is very weak. It should be divided into two parts: Discussion: compartment with results of the previous similar studies, limitations of the own results (low number of materials, samples printers and ANNs structures studied), and directions for further research; Conclusions: short sum-up of the results taking into consideration aim of the study.

Response: Thanks so much for your professional and kind suggestion. In this manuscript, 3D printing technology, BP neural network and orthogonal experiment are combined to build a model that can be used to predict the deformation of PLA materials in 3D printing technology. Compared with the traditional processing technology, this technology has obvious advantages. Through BP model construction, orthogonal data training and print parameter optimization, the deformation is verified and predicted. Further, baked up by the principle of orthogonal experiment, the main factors that affect deformation are obtained as follows: T (℃)> A (mm/s) > W (%) >D (mm). The constructed BP neural network structure delivers good reliability in learning training, simulation test, and numerical prediction. The optimal process parameters are A=45 mm/s, T=220℃, W=30 % and D=0.3 mm, and the smallest amount of deformation is 0.0906 mm/mm. The predicted value obtained by the BP model is in good agreement with the experimental value curve, with small relative error (the maximum value does not exceed 0.1%) and a correlation coefficient of 0.99985.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Dear authors,

 

I think that the description about experimental methods has been revised.

However,  you have not shown the evidence to prove the deformation is small in the Figure 7. This is the manuscript submitted to the scientific journal, and then, you need to discuss based on the results.

 

Sincerely yours.

Author Response

Reviewer#1

I think that the description about experimental methods has been revised.

However, you have not shown the evidence to prove the deformation is small in the Figure 7. This is the manuscript submitted to the scientific journal, and then, you need to discuss based on the results.

Response: Thanks so much for your professional and kind suggestion. We have added a Table 7 in the manuscript to discuss the results. Moreover, for further verifying the results of deformation, the nine deformation values in the ears, head and limbs of the model were measured, as shown in Table 7. Through the analysis of the set and measured values of the nine parts (A～I, as marked in yellow solid line in Fig.7(b)), it can be seen that the deformation error value of the nine parts is less than 5%. It can be seen that it is feasible and reasonable to guide 3D printing entity construction by using BP neural network to predict parameters.

Table 7 Comparison of the set values and measured values of printed model 

location	The set value(mm)	Measured value(mm)		Error/´10-2
A	5	4.91	1.8
B	5	4.97	0.6
C	5	5.03	0.6
D	5	5.11	2.2
E	4	4.12	3.0
F	4.5	4.52	0.4
G	4.5	4.57	1.6
H	4	4.10	2.5
I	4	4.12	3.0

Author Response File: Author Response.pdf

Reviewer 4 Report

Response: Thanks so much for your professional and kind suggestion. We have replaced the pictures in the revised manuscript. 

I do not see the change of the picture in the revised manuscript you sent. My comment was that there was no point in inserting a photo of the filament, which as a material in the 3D printing process is known to everyone and such a photo does not contribute anything.

Response: Thanks so much for your professional and kind suggestion. We will design the content you suggested in future experiments. Thank you very much for providing ideas for our follow-up experimental verification.

Unfortunately, I cannot accept such a response to the comment sent.

The final verification should be completed with a meaningful evaluation method. The current presentation of the final result cannot be accepted.

Author Response

Response: Thanks so much for your professional and kind suggestion. We have replaced the pictures in the revised manuscript.

I do not see the change of the picture in the revised manuscript you sent. My comment was that there was no point in inserting a photo of the filament, which as a material in the 3D printing process is known to everyone and such a photo does not contribute anything.

Response: Thanks so much for your professional and kind suggestion. We will design the content you suggested in future experiments. Thank you very much for providing ideas for our follow-up experimental verification.

Unfortunately, I cannot accept such a response to the comment sent.

The final verification should be completed with a meaningful evaluation method. The current presentation of the final result cannot be accepted.

Response: Thanks so much for your professional and kind suggestion. We have deleted the Fig.1(b) in the revised manuscript. We apologize for not understanding your question accurately.

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

Dear authors,

 

The discussion about Fig. 7 has been adequately revised, but the conclusion is not supported the results in Table 7. Although the error for the real object (Fig. 7) is 3.0%, authors write that the maximum value does not exceed 0.1%. This may be ,isleading for the readers. After revising the conclusion, This manuscript can be accepted.

 

Sincerely yours.

Author Response

Reviewer#1-3：The discussion about Fig. 7 has been adequately revised, but the conclusion is not supported the results in Table 7. Although the error for the real object (Fig. 7) is 3.0%, authors write that the maximum value does not exceed 0.1%. This may be misleading for the readers. After revising the conclusion, this manuscript can be accepted.

Sincerely yours.

Response: Thanks so much for your professional and kind suggestion. We have revised the mistake in the manuscript for better understanding by the reader.

Reviewer 4 Report

The extension sent showing the validity of the parameter selection methodology adopted is hardly professional.

However, it provides a basis for assessing the validity of the adopted parameter selection. I accept the study in its revised form.

 

Author Response

Thank you very much for your recognition of our work. Your suggestion will certainly encourage us to continuously improve the quality of our work in the future.