Establishment and Solution Test of Wear Prediction Model Based on Particle Swarm Optimization Least Squares Support Vector Machine
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper proposes a tool wear status identification model based on Particle Swarm Optimization and Least Squares Support Vector Machine. The model integrates data from multiple sensors, such as vibration and force signals, to extract key features reflecting tool wear, optimizes these features using the above methods for accurate tool wear prediction. Several areas could be improved to enhance clarity, methodology, and overall quality. Below are specific suggestions:
- The abstract is too long and should be condensed. Additionally, it is recommended to include specific results, such as accuracy assessment. Furthermore, the key contribution should be clearly stated.
- The introduction is well-written, providing an overview of existing research. However, the authors should clearly state the gap between existing methods and their approach. At the end of the introduction - review, a new paragraph should be added to summarize this gap and provide a brief overview of the paper's structure.
- The second section is very well written and contains all the necessary elements. However, the authors are advised to avoid repetition. For example, the section repeatedly emphasizes the importance of cutting force, vibration and power signals for monitoring tool wear. While these signals are indeed crucial, the same points are mentioned several times without any new insights being gained.
- The authors mention the condition for the PSO algorithm is mentioned, but it would be helpful to explain why 50 iterations were chosen and whether this number was determined empirically or theoretically.
- The results are presented well, but the authors could discuss potential limitations of the model and scenarios where it might not perform as well. In other words, can some limits be set in which the model works with acceptable errors.
- The figures (e.g., Figure 13, Figure 14) should be placed in a higher resolution, i.e. better quality.
- In the conclusion, many phrases are repeated from the abstract. The conclusion should focus on summarizing the key results. And not a repetition of the methodology similar to the abstract.
Author Response
Comments 1:The abstract is too long and should be condensed. Additionally, it is recommended to include specific results, such as accuracy assessment. Furthermore, the key contribution should be clearly stated.
Response 1:Agreed, thank you for your suggestion. The abstract has been edited and compressed, and specific results and key contributions have been highlighted at the end. Please refer to the abstract section of the manuscript.
Comments 2:The introduction is well-written, providing an overview of existing research. However, the authors should clearly state the gap between existing methods and their approach. At the end of the introduction - review, a new paragraph should be added to summarize this gap and provide a brief overview of the paper's structure.
Response 2:Agreed, thank you for your suggestion. A new paragraph has been added at the end of the introduction to summarize this gap and provide a brief overview of the paper structure.
Comments 3:The second section is very well written and contains all the necessary elements. However, the authors are advised to avoid repetition. For example, the section repeatedly emphasizes the importance of cutting force, vibration and power signals for monitoring tool wear. While these signals are indeed crucial, the same points are mentioned several times without any new insights being gained.
Response 3:Agreed, thank you for your suggestion. Modified, please refer to section 2.5.
Comments 4:The authors mention the condition for the PSO algorithm is mentioned, but it would be helpful to explain why 50 iterations were chosen and whether this number was determined empirically or theoretically.
Response 4:Thank you for your question. Based on experience, 50 times can fully optimize the search without excessive computation. View page 11 of (7)
Comments 5:The results are presented well, but the authors could discuss potential limitations of the model and scenarios where it might not perform as well. In other words, can some limits be set in which the model works with acceptable errors.
Response 5:Thank you for your comment on the limitations of our PSO - LS - SVM tool wear state identification model. We acknowledge that in extreme operating conditions, like extreme cutting speeds, feed rates, or temperatures outside the experimental range, the tool wear mechanisms change, affecting the model's performance. High - noise data, due to sensor issues or electromagnetic interference, can also reduce the model's accuracy.​
However, the current manuscript focuses on introducing and validating the novel combination of PSO and LS - SVM for tool wear prediction in multi - sensor data fusion. We concentrated on normal and moderately variable conditions, demonstrating key feature extraction, PCA - based feature vector optimization, and parameter tuning with the particle swarm algorithm. Adding a detailed limitations analysis would have extended the manuscript and diverted focus from our core objective of showing the model's superiority over traditional methods.​
In future research, we plan to explore more robust algorithms to enhance the model's performance under challenging conditions, conducting comprehensive studies on extreme conditions and high - noise data scenarios. This way, we maintain the integrity of this study while showing our commitment to future improvements.​
Comments 6:The figures (e.g., Figure 13, Figure 14) should be placed in a higher resolution, i.e. better quality.
Response 6:Agreed, the image has been replaced.
Comments 7:In the conclusion, many phrases are repeated from the abstract. The conclusion should focus on summarizing the key results. And not a repetition of the methodology similar to the abstract.
Response 7:Agreed, thank you for your suggestion. The conclusion has been rewritten, please refer to the conclusion section.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe article is scientifically sound and suitable for publication. Below are some remarks and suggestions to improve its completeness:
The experimental conditions are not described at all ( workpiece material, cutting parameters: speeds and feeds, depth of cut) which is the biggest flaw in this article. There is no mention of how the model might perform under varying conditions( for example roughing vs finishing passes).
The article focuses only on PSO-LS-SVM without comparing it against other tool wear prediction methods (neural networks, other SVM variants, etc.), therefore its hard to see its relative advantages.
The study lacks discussion on computational efficiency (for example training and prediction times).
Author Response
Comments 1:The article is scientifically sound and suitable for publication. Below are some remarks and suggestions to improve its completeness:
The experimental conditions are not described at all ( workpiece material, cutting parameters: speeds and feeds, depth of cut) which is the biggest flaw in this article. There is no mention of how the model might perform under varying conditions( for example roughing vs finishing passes).
The article focuses only on PSO-LS-SVM without comparing it against other tool wear prediction methods (neural networks, other SVM variants, etc.), therefore its hard to see its relative advantages.
The study lacks discussion on computational efficiency (for example training and prediction times).
Response 1:
Hello! Thank you very much for your careful review of our paper and valuable feedback. In response to your point about the lack of research and description in the comparison section with other methods in the article, we provide the following explanations and clarifications, and kindly request you to reconsider.
Indeed, the paper did not provide a detailed comparative study and description of the proposed method compared to other methods. This is not due to our neglect of this part of the content, but rather a comprehensive consideration of the overall structure and research focus of the paper. This study is highly innovative, successfully combining PSO (Particle Swarm Optimization) and LS-SVM (Least Squares Support Vector Machine) for the first time in the field of tool wear prediction in multi-sensor data fusion. This innovative combination has never been explored in previous research. The fusion mechanism of its core algorithm, parameter optimization process, and unique advantages in tool wear prediction tasks all require extensive elaboration and analysis to ensure the completeness and reliability of the research.
At the current stage of research, due to the focus on proposing, constructing, and validating new methods for tool wear prediction, our resources and efforts are mainly focused on in-depth exploration of the PSO-LS-SVM fusion model. The construction process of this model involves many complex links, such as how the PSO algorithm accurately optimizes the parameters of LS - SVM, and how multi-sensor data can be efficiently fused under the fusion model and accurately predict tool wear status. These contents themselves have already made the length of the paper considerable. If a large number of comparative studies with other methods are added, it will not only lead to an imbalance in the structure of the paper, but also possibly distract readers' attention from the core innovation points.
In addition, there have been numerous comparative studies on different methods in the field of tool wear prediction in the past, but none of these studies have involved our proposed new model combining PSO and LS-SVM. Our research aims to open up a new research path and inject new vitality into this field. In future research, we plan to further expand the comprehensive comparative analysis of this method with other mainstream methods based on this study. Through rigorous experimental design and under various working conditions, systematically compare the differences between the PSO-LS-SVM fusion method and other classical and emerging methods in key indicators such as accuracy, stability, and computational efficiency of tool wear prediction, providing more solid theoretical support and data basis for the practical application of this method.
In summary, although the current paper does not include a detailed comparison with other methods, this is based on a reasonable arrangement of research focus and resource allocation. We firmly believe that the method of combining PSO and LS-SVM proposed for the first time in this study has outstanding innovation and application potential, and we will actively carry out related comparative research in the future. Thank you again for your understanding and support. We look forward to your further guidance on our paper.
Reviewer 3 Report
Comments and Suggestions for AuthorsManuscript title
Establishment and solution test of wear prediction model based on PSO-LS-SVM
Authors
Xiao Huang, Yuhui Mao and Yongguo Wang
Reviewer’s remarks
- The manuscript describes the development of an innovative method for accurate prediction of CNC tools wear using particle swarm optimization and least squares support vector machine, which give the PSO-LS-SVM model.
- After the current state of CNC tools wear identification methods when they discuss about cutting force signal, vibration signal and power signal, the manuscript addresses the following sections: analysis of tool wear status, selection of monitoring signals, force signal acquisition, vibration signal acquisition, tool wear and surface roughness collection, original signal analysis, wear stage identification model and parameter optimization, least squares support vector machine classification theory, LS-SVM parameter optimization based on particle swarm algorithm, experimental analysis, feature extraction and optimization, analysis of test results, and conclusions.
- The research is well structured, the methodology is clearly defined, and the analysis stages are described by logical sequence of the PSO-LS-SVM algorithm.
- All the stages of the PSO-LS-SVM method are properly discussed so that, in the end, an optimal solution is obtained for the CNC tools wear identification. Furthermore, the PSO-LS-SVM model optimizes model parameters through the two-dimensional coordinates of the particle swarm algorithm to adapt to the given training sample set, proving its effectiveness.
- At the end of the manuscript, the authors consistently discuss the results obtained and list the conclusions of the study undertaken.
- I would recommend carefully checking the text, as I have identified drafting errors, for example, the graph in Figure 10 is identical to the graph in Figure 11.
- Finally, I appreciate that the manuscript can be considered for publication after a minor revision.
Comments for author File: Comments.pdf
Author Response
Comments 1:
Reviewer’s remarks
1.The manuscript describes the development of an innovative method for accurate prediction of CNC tools wear using particle swarm optimization and least squares support vector machine, which give the PSO-LS-SVM model.
2.After the current state of CNC tools wear identification methods when they discuss about cutting force signal, vibration signal and power signal, the manuscript addresses the following sections: analysis of tool wear status, selection of monitoring signals, force signal acquisition, vibration signal acquisition, tool wear and surface roughness collection, original signal analysis, wear stage identification model and parameter optimization, least squares support vector machine classification theory, LS-SVM parameter optimization based on particle swarm algorithm, experimental analysis, feature extraction and optimization, analysis of test results, and conclusions.
3.The research is well structured, the methodology is clearly defined, and the analysis stages are described by logical sequence of the PSO-LS-SVM algorithm.
4.All the stages of the PSO-LS-SVM method are properly discussed so that, in the end, an optimal solution is obtained for the CNC tools wear identification. Furthermore, the PSO-LS-SVM model optimizes model parameters through the two-dimensional coordinates of the particle swarm algorithm to adapt to the given training sample set, proving its effectiveness.
5.At the end of the manuscript, the authors consistently discuss the results obtained and list the conclusions of the study undertaken.
6.I would recommend carefully checking the text, as I have identified drafting errors, for example, the graph in Figure 10 is identical to the graph in Figure 11.
7.Finally, I appreciate that the manuscript can be considered for publication after a minor revision.
Response 1:
Thank you very much for pointing out this potential issue, which is truly valuable for improving the quality of our work. Figures 10 and 11 are indeed designed to present different data aspects. Figure 10 specifically depicts the mean square frequency, while Figure 11 showcases the frequency variance. Although at a glance, they might seem similar due to the nature of the data they represent.
Upon a more in - depth analysis, it becomes clear that there is no straightforward linear or other easily - discernible trend in these two sets of changes. In the initial stages of the wear process, the values of both the mean square frequency and frequency variance show relatively stable states. However, as the wear progresses into the middle and late stages, they start to fluctuate significantly. These fluctuations can be attributed to various factors within the experimental environment, such as the gradual degradation of the materials under study, changes in operating conditions, or the influence of external forces that become more pronounced over time. We apologize for any confusion that the appearance of the graphs might have caused and hope this explanation clarifies the situation.