Machine Learning Techniques in Dosing Coagulants and Biopolymers for Treating Leachate Generated in Landfills
Round 1
Reviewer 1 Report (New Reviewer)
This paper presents a method to optimize the coagulation-flocculation process for treating landfill leachate using machine learning techniques. A model that combines artificial neural network (ANN), support vector machine (SVM) and K-nearest neighbor (KNN) algorithms is proposed to predict the optimal doses of coagulant and flocculant to achieve the lowest turbidity of leachate. The performance and accuracy of the model are validated by laboratory experiments and compared with other linear and nonlinear algorithms. It is found that the SVM algorithm has the best fit and error indicators. The reliability of the prediction results is confirmed by repeated experiments.
However, the paper does not provide a deep analysis and argumentation of the key issues, lacks a detailed description of the principles and parameter settings of the machine learning algorithms, and does not give the mathematical expression of the model, which will affect the understanding and reproduction of the model. The specific problems are as follows
1. The introduction section should introduce the research background, motivation, objective and contribution of the paper, and point out the shortcomings and innovations of the existing research in the literature review. The current introduction section lacks a literature review and an explanation of the innovation points. It is suggested to add a review of relevant literature, highlight the research significance and value of this paper, and compare and evaluate it with other methods.
2. In the materials and methods section, the principles and parameter settings of the machine learning algorithms should be supplemented, and the mathematical expression of the model should be given, so that readers can understand and reproduce the model.
3. The table format and font in the results and discussion section should be consistent, and some tables have text problems, such as two Initial turbidity in Table 2 on line 185, which should be Initial turbidity and final turbidity.
4. The results and discussion section also lacks some necessary graphs and data, such as curves of turbidity, pH, COD and other parameters of leachate during coagulation-flocculation process; error analysis graphs between predicted results and actual results of different machine learning algorithms; optimal coagulant and flocculant dosage range and corresponding leachate water quality index table. It is suggested to add these graphs and data, and explain and discuss them reasonably, highlighting the effects and contributions of this paper in leachate treatment, and pointing out the shortcomings and improvement space.
5. The current conclusion section is simple and vague, does not clearly answer the research questions raised by the paper, nor highlights the innovation points and contribution points of the paper. It is suggested to summarize the main content and conclusions of the paper in the conclusion section, emphasize the innovation and advantages of this paper in leachate treatment, discuss the limitations and applicability of the model appropriately, and propose future improvement directions and prospects.
The English language can be improved.
Author Response
Thank you for your comments; we have provided a letter responding to each of your points. In addition, the modified document with your comments
Author Response File: Author Response.pdf
Reviewer 2 Report (New Reviewer)
· While the study focuses on the application of machine learning techniques to optimize coagulation-flocculation processes for leachate treatment, the methodology lacks in-depth explanation. It would be beneficial to provide a more detailed description of the experimental setup, data collection methods, and the selection criteria for input variables.
· The effectiveness of machine learning heavily depends on the quality and quantity of training data. The study should elaborate on how the dataset was compiled, including potential biases or limitations in data collection that could impact the model's accuracy.
· While the Support Vector Machine (SVM) emerged as the optimal technique, the study doesn't sufficiently discuss why SVM was chosen over other machine learning algorithms. A comparative analysis of different algorithms and their suitability for this specific problem would enhance the credibility of the chosen approach.
· The study primarily focuses on a specific range of dosages (0.1 to 22 mL). It's crucial to address how well the optimized doses from the machine learning model can be generalized to a broader range of dosages and different leachate compositions. This would reflect the practical applicability of the model.
· The study mentions successful validation using experimental data generated by the algorithm. However, the validation process itself needs more clarification. How was the external validation performed? Were there any discrepancies between predicted and actual results, and if so, how were they addressed?
· Given the complexity of the problem and the potential high dimensionality of the data, overfitting is a concern. The study should discuss how the risk of overfitting was minimized, either through regularization techniques, cross-validation, or other methods.
· Machine learning models like Support Vector Machines can lack interpretability, making it challenging to understand the underlying reasons for the model's predictions. Incorporating feature importance analysis or other interpretability techniques would enhance the study's insights.
· The study should elaborate on the preprocessing steps applied to the raw data, as these steps can significantly impact the model's performance. Details on outlier handling, missing data treatment, and normalization methods should be provided.
· A sensitivity analysis of the model's performance with respect to changes in input variables would provide a better understanding of which variables have the most significant impact on dose prediction accuracy.
· While the study presents promising results in terms of dose prediction and turbidity reduction, it should conclude with potential future directions. Addressing challenges such as scaling the model to larger datasets, incorporating real-time data, and integrating the model into practical landfill operations would add value to the study's implications.
Quality of English Language is ok
Author Response
Thank you for your comments; we have provided a letter responding to each of your points. In addition, the modified document with your comments
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report (New Reviewer)
The paper can be accepted in the current form.
The quality of English language is fine.
Reviewer 2 Report (New Reviewer)
Author now address all queries raised by reviewer. The manuscipt is now may consider for pubication.
Quality of English Language is OK
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.
Round 1
Reviewer 1 Report
GENERAL IMPRESSION
The application of ML including SVM is an interesting and required method in water treatment. However, in this paper, scientific discussions with enough results are lacking.
GENERAL COMMENTS
1. The content of the discussion section is like reference study that is usually contained in the introduction section.
2. Grammar should be checked throughout the manuscript and revised (e.g., L43, 85,89–90, 108–109)
3. Isn’t there a copyright related to the satellite map (Fig.1)? The source of the map should be indicated.
4. In Table 1, some error ranges exceed the average (maybe) concentrations, which may mislead to negative concentration.
5. Past and present tenses were mixed in some paragraphs (e.g., L117–126). Results are recommended to be expressed in a past tense.
6. Why the parameters in Table 1 selected? According to the exceedance of criteria? It should be mentioned.
7. The numbers of actual experimental results seem to be lacking.
SPECIFIC COMMENTS
1. Title: the classification of the leachate should be specified in the title. Also, the period should be removed. Why was the sequence of biopolymer-coagulant applied in the title which was different from the sequence in Abstract?
2. L34–35: ‘leachate becomes more alkaline’ may indicate that leachate is initially alkaline. However, in many cases, leachate can also be acidic. Also, in ‘present a high concentration of dissolved solids product of’, use of ‘dissolved solids product of’ should be revised. For example, ‘high concentrations of cations such as….’
3. L43: ‘lower’ does not seem to be appropriate.
4. L82: Abbreviation (KNN) should be defined at the first appearance.
5. The name of the country does not appear in the section 2.1.
6. L204: R2 was not adequately explained.
Grammar should be checked throughout the manuscript and revised (e.g., L43, 85,89–90, 108–109)
Author Response
Thank you for your feedback. I have attached a response letter addressing each of your comments.
Reviewer 2 Report
This study attempts to propose a scheme for optimizing the coagulation treatment of landfill leachate through the use of machine learning techniques such as artificial neural networks. To this end, different concentrations of coagulants were added to simulate the leachate treatment process, and laboratory results were used as input variables for the algorithm used; Use computational models to estimate predicted data and obtain predictions for the optimal dose with high statistical adjustment indicators. Through this approach, the author believes that the experiment can be optimized to reduce operational costs without affecting sufficient dose results.
The other specific opinions:
1. The two conclusions obtained in “5 Conclusion” seem inconsistent with the research topic.
1)“it was proved that by mixing cassava starch (coadjutant agent) and ferric chloride (coagulating agent), a great p tential effect in coagulation-flocculation was achieved, which greatly helps in the treatment of leachates to clarify the liquid in question with the removal of almost 98.31% and thus also reduce its pollutant load”. However, this is not a coagulation process experiment, and the removal effect and optimal process are not your research goals.
`2) “The effectiveness of the ML will depend a lot on the definition of the input varia- 354bles that directly influence the prediction of the data……”, which is not the topic of research either.
2. In this study, turbidity, SS, and pH were selected as the parameters for evaluating the dosage of coagulants. Coagulation treatment has an effect on reducing the turbidity and suspended solids of wastewater, but in the treatment of leachate, turbidity and SS are the easiest parameters to meet, as ultrafiltration(UF) has been widely used.
If coagulation has some value, which may be reflected in reducing the concentration of organic matters, especially refractory ones, while improving the biochemical degradability of leachate. However, this study did not provide a reduction in COD as the main effect of coagulation, which is a limitation for the evaluation of coagulation.
L309-311:“In the study presented by Patel [30], it tests several coagulants, including ferric chloride for leachate treatment, showing significant decreases of several pollutants such as total organic carbon and COD.” This citation does not provide evidence for this study to illustrate the relationship between turbidity and COD, as the relationship between turbidity and COD varies among wastewater from different sources. This study also did not attempt to establish a connection between turbidity and COD. The lack of a connection between turbidity and organic pollutants weakens the value of research.
3. Other issues
1)L189,L173:“According to the table above”should be “According to Table 3”….
2) L204: root mean square error (RMSE) and root mean square error (Rsquared) were tested.
3) Table 6å’ŒTable 5 could be combined as one Table.
Author Response
Thank you for your feedback. I have attached a response letter addressing each of your comments.
Author Response File: Author Response.pdf