A New Use Strategy of Artificial Intelligence Algorithms for Energy Optimization in Plastic Injection Molding
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
The paper investigates a method for optimizing energy consumption in the plastic injection moulding process using AI. The focus is on reducing energy usage during the production of plastic fruit containers by optimizing process parameters. The authors proposed an unsupervised autoencoder with the K-Means algorithm to analyze production data and identify factors that influence energy consumption. However, there are several critical issues that need to be addressed to enhance the quality and clarity of the paper.
1. The literature review section is not sufficiently comprehensive and lacks a systematic analysis of previous works. I recommend conducting a systematic review that compares existing methods for optimizing and classification the injection moulding process.
2. A summary table should be added to clearly present different approaches from past studies, their strengths, limitations, and how they relate to the current work.
3. Additionally, include a more detailed discussion of relevant articles in the lit review that are closely related to energy optimization in IM using AI techniques.
4. The introduction needs to be rewritten to better emphasize the significance of the research, the novelty of the proposed hybrid AI approach, and the research gap.
5. Adding a flowchart to the methodology section can improve the understanding of the proposed approach.
6. Lines 80 to 82, there are some spacing issues.
7. Ensure all abbreviations are defined the first time they appear in the text (such as MES )
8. The section explaining machine learning algorithms is too broad and generic. It is unnecessary to cover the entire topic of machine learning, which diverts from the main focus. Instead, the paper should concentrate on the specific algorithms used in this study.
9. The method used to generate the correlation matrix in Figure 3 is not sufficiently explained. More details are required on how the features were selected and processed before computing the correlation.
10. The axes in Figure 4 are unreadable. Consider increasing the font size
11. The term "specific energy" should be clearly defined in the methodology section. It is crucial to specify how specific energy is measured .
12. There is insufficient information about the experimental setup and process settings used for data collection. How the data has been collected? The paper should include details on critical parameters such as melt temperature, mould temperature, injection pressure, and cooling time during the experiments and data collection.
13. The results section lacks a thorough explanation of the figures and the data.
14. The parameters used in the analysis are not well-defined. Ensure that all relevant variables and their significance to energy optimization are clearly outlined in the methodology section.
15. The paper only used K-mean clustering for classification. It needs a comparison with the more complex techniques such as random forest, SVM, ANN etc.
16. The feature selection process appears to be limited. Other critical factors such as cooling time, injection pressure, and material properties can also influence energy consumption.
Author Response
Comments 1 : The literature review section is not sufficiently comprehensive and lacks a systematic analysis of previous works. I recommend conducting a systematic review that compares existing methods for optimizing and classification the injection moulding process.
Response 1 : We thank the reviewer for the suggestion. Majority of scientific papers found in literature is related to quality control of injection moulding process, as mentioned in the manuscript. We have expanded the analysis of studies with focus on energy optimisation of injection moulding process.
Comments 2 : A summary table should be added to clearly present different approaches from past studies, their strengths, limitations, and how they relate to the current work.
Response 2 : We are grateful to the reviewer for the hints. We have introduced a summary table in the manuscript.
Comments 3 : Additionally, include a more detailed discussion of relevant articles in the lit review that are closely related to energy optimization in IM using AI techniques.
Response 3 : We thank the reviewer for the suggestion. The revised manuscript now includes some discussions about the studies described in cited papers
Comments 4 : The introduction needs to be rewritten to better emphasize the significance of the research, the novelty of the proposed hybrid AI approach, and the research gap.
Response 4 : We are grateful to the reviewer for the suggestion. We have revised the introduction accordingly.
Comments 5 : Adding a flowchart to the methodology section can improve the understanding of the proposed approach.
Response 5 : We have added a flowchart describing the phases adopted for our data analysis
Comments 6 : Lines 80 to 82, there are some spacing issues.
Response 6 : Thank you, Text has been updated in manuscript
Comments 7 : Ensure all abbreviations are defined the first time they appear in the text (such as MES )
Response 7 : Thank you. Text has been updated in manuscript
Comments 8 : The section explaining machine learning algorithms is too broad and generic. It is unnecessary to cover the entire topic of machine learning, which diverts from the main focus. Instead, the paper should concentrate on the specific algorithms used in this study.
Response 8 : We thank the reviewer for the observation. We have reshaped the paper according to the classes of AI algorithm adopted for our analysis (ANN autoencoder for dimensionality reduction – deep learning category, K-Means for clustering – unsupervised category and Random Forest for features selection – supervised category).
Comments 9 : The method used to generate the correlation matrix in Figure 3 is not sufficiently explained. More details are required on how the features were selected and processed before computing the correlation.
Response 9 : Interesting point; thank you! We updated text in manuscript with more details about the correlation matrix in Figure 3
Comments 10 : The axes in Figure 4 are unreadable. Consider increasing the font size
Response 10 : The figure 4 caption was put to recover the graphical problem ( tags for X and Y axis)
Comments 11 : The term "specific energy" should be clearly defined in the methodology section. It is crucial to specify how specific energy is measured .
Response 11 : Thank you for the observation. We updated gave in manuscript a better specification about the features used for analysis, including the “Specific_Energy_per_part” feature
Comments 12 : There is insufficient information about the experimental setup and process settings used for data collection. How the data has been collected? The paper should include details on critical parameters such as melt temperature, mould temperature, injection pressure, and cooling time during the experiments and data collection.
Response 12 : All data (melt temperatures, mould temperatures and injection pressures) has been collected by means of MES, as already specified in the paragraph 2.1 which we hope is clear enough.
Comments 13 : The results section lacks a thorough explanation of the figures and the data.
Response 13 : We thank the reviewer. The revised version of the manuscript now contains more details about the figures and the data, describing step by step the analysis we have carried out.
Comments 14 : The parameters used in the analysis are not well-defined. Ensure that all relevant variables and their significance to energy optimization are clearly outlined in the methodology section.
Response 14 : We thank the reviewer. The revised version of the manuscript now contains some further considerations about the energy optimization which can be useful during real-time control of manufacturing process.
Comments 15 : The paper only used K-mean clustering for classification. It needs a comparison with the more complex techniques such as random forest, SVM, ANN etc.
Response 15 : We thank the reviewer. Nevertheless, as argued in the paper, the dataset is based on “unlabelled” features. For this reason it is not possible to apply supervised algorithms. In our analysis we have considered an unsupervised technique based on ANN(autoencoder) plus K-means clustering.
Comments 16 : The feature selection process appears to be limited. Other critical factors such as cooling time, injection pressure, and material properties can also influence energy consumption.
Response 16 : We thank the reviewer again. We totally agree with the comment about the need to extend experimentation in the future using other critical factors. Our intention with this paper is to define a path of adoption of a better configuration of temperature sensors to obtain useful information for AI analysis, as indicated in paragraph 5. Conclusion
Reviewer 2 Report
Comments and Suggestions for AuthorsUse of machine learning algorithms is becoming more and more apparent in different production processes. Sensors providing real-time data allow gathering of enormous amounts of data that become very difficult and time consuming for humans. Thanks to different IIoT technologies managing that data becomes much easier. The topic of the paper is up-to-date and scientifically sound.
The paper itself is coherent and well written.
However, the title and discussion are somewhat misleading. You mention energy optimization in the title, but the paper presents results on energy consumption analysis. Then in discussion you write: "The new hybrid use strategy proposed combines an unsupervised autoencoder with the K-Means algorithm to analyse production data and identify key factors influencing energy consumption." No further information on energy optimization is provided. Analyzing injection moulding prosess data is well decribed in the paper, but nothing is said on energy optimization based on the obtained analysis. I suggest you either change the topic or present information on energy optimization. The obtained results should be further process e.g.: using AI or ANN to optimize the plastic injection moulding prosess and support real-time decision-making.
I suggest extending the discussion section and confront you results with other scientific papers.
Figures 2-7 lack units.
Figure 12 has no energy unit and although the values are different for the three groups placing the results at the same level makes data misleading. Moreover, graphs should begin at 0 energy. This would show the differences between groups and make the figure clear. Same applies to figures 13-16.
Comments on the Quality of English LanguageEnglish language needs editing.
Author Response
Use of machine learning algorithms is becoming more and more apparent in different production processes. Sensors providing real-time data allow gathering of enormous amounts of data that become very difficult and time consuming for humans. Thanks to different IIoT technologies managing that data becomes much easier. The topic of the paper is up-to-date and scientifically sound.
The paper itself is coherent and well written.
Comment 1: However, the title and discussion are somewhat misleading. You mention energy optimization in the title, but the paper presents results on energy consumption analysis. Then in discussion you write: "The new hybrid use strategy proposed combines an unsupervised autoencoder with the K-Means algorithm to analyse production data and identify key factors influencing energy consumption." No further information on energy optimization is provided. Analyzing injection moulding prosess data is well decribed in the paper, but nothing is said on energy optimization based on the obtained analysis. I suggest you either change the topic or present information on energy optimization. The obtained results should be further process e.g.: using AI or ANN to optimize the plastic injection moulding prosess and support real-time decision-making.
Response 1 : We thank the reviewer. In the revised manuscript we describe in more detail the strategy for using the results of this study for energy optimisation (see paragraph 3.2 pag.15)
Comment 2 : I suggest extending the discussion section and confront you results with other scientific papers.
Response 2 : We thank again the reviewer for the suggestions. Majority of scientific papers is related to the use of AI to quality control of injection molding process, as mentioned in the manuscript. Then they are not so significant for the focus of energy optimisation of injection molding process.
Comment 3 : Figures 2-7 lack units.
Response 3 : The figures already contain legenda with the tag of each axis. It is not possible to identify units because the features have been normalized.
Comment 4 : Figure 12 has no energy unit and although the values are different for the three groups placing the results at the same level makes data misleading. Moreover, graphs should begin at 0 energy. This would show the differences between groups and make the figure clear. Same applies to figures 13-16.
Response 4 : We agree with this comment: thank you. The revised manuscript contains the mentioned figures following your indications
Round 2
Reviewer 1 Report
Comments and Suggestions for Authors
1. Including a tracked changes version of the document would help reviewers and editors understand the revisions made from the previous submission
2. Replace "IA" with "AI” in Line 94 and Table 1.
3. Cite Table 1 in the main text where its content is relevant.
4. Why the supervised learning is explained when they have not used it in the paper.
5. Line 258 it should be figure 3.
6. The font size in Figure 4 is too small, making it unreadable. Increase the font size for axis labels, legends, and other details for clarity.
7. The criteria and threshold for determining the number of clusters in K-Means (e.g., the elbow method, silhouette score) are not mentioned.
8. Increase the size of Figure 11 to improve readability and ensure the information is clearly visible.
9. Address the mismatch between clusters and temporal groups, explaining its significance or irrelevance.
1. Line 432, explain why the threshold is 0.001 .
Author Response
Comments 1 : Including a tracked changes version of the document would help reviewers and editors understand the revisions made from the previous submission
Response 1 : We thank the reviewer for the suggestion. We have used tracked changes version in the new revision of manuscript.
Comments 2 : Replace "IA" with "AI” in Line 94 and Table 1.
Response 2 : Thanks, we have replaced “IA” with “AI” as suggested.
Comments 3 : Cite Table 1 in the main text where its content is relevant.
Response 3 : We thank the reviewer for the suggestion. The Table 1 is now cited in the main text of revised manuscript as suggested
Comments 4 : Why the supervised learning is explained when they have not used it in the paper.
Response 4 : We are grateful to the reviewer for the comment. We have used the random forest algorithm for features selection. This is the reason to explain the supervised learning approach.
Comments 5 : Line 258 it should be figure 3.
Response 5 : Thanks for the suggestion. We have modified the text in the manuscript.
Comments 6 : The font size in Figure 4 is too small, making it unreadable. Increase the font size for axis labels, legends, and other details for clarity.
Response 6 : Thank you, we have increased the font size of text for figures 4,5,6,7 in the revised manuscript
Comments 7 : The criteria and threshold for determining the number of clusters in K-Means (e.g., the elbow method, silhouette score) are not mentioned.
Response 7 : We are grateful to the reviewer for the comment. We have used a fixed number k=3 because our intention was to identify 3 phases of the entire production process (beginning part, intermediate part and final part) .(see Lines 280-282)
Comments 8 : Increase the size of Figure 11 to improve readability and ensure the information is clearly visible.
Response 8 : Thanks, we have increased the size of Figure 11 as suggested.
Comments 9 : Address the mismatch between clusters and temporal groups, explaining its significance or irrelevance.
Response 9 : We thank the reviewer for the comment. We have described with more details the aspect related to the apparent mismatch between clusters and temporal groups in the revised manuscript.(see Lines 487-491)
Comments 10 : Line 432, explain why the threshold is 0.001 .
Response 10 : Thanks for the suggestion. We have explained in the revised manuscript the reasons to adopt the threshold = 0.001 (see Lines 434-439)
Reviewer 2 Report
Comments and Suggestions for AuthorsDear Authors,
thank you for considering my comments and providing the revised version of the paper that includes my suggestions.
Author Response
Thanks to the reviewer for your support
