Next Article in Journal
Research on Optimization of Lifter of an SAG Mill Based on DEM Simulation and Orthogonal Tests and Applications
Previous Article in Journal
Geochemical Insights and Mineral Resource Potential of Rare Earth Elements (REE) in the Croatian Karst Bauxites
 
 
Article
Peer-Review Record

Prediction of Mine Waste Rock Drainage Quantity Using a Machine Learning Model with Physical Constraints

Minerals 2025, 15(2), 194; https://doi.org/10.3390/min15020194
by Can Zhang 1, Liang Ma 2,* and Wenying Liu 1,*
Reviewer 1:
Reviewer 2:
Minerals 2025, 15(2), 194; https://doi.org/10.3390/min15020194
Submission received: 18 December 2024 / Revised: 3 February 2025 / Accepted: 11 February 2025 / Published: 19 February 2025
(This article belongs to the Section Environmental Mineralogy and Biogeochemistry)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The mining industry has a large amount of waste rock, which is often stored in waste dumps. Drainage from these waste dumps can pose serious environmental emissions by polluting surface and groundwater. Effective management of waste dump drainage mechanisms requires reliable prediction of drainage volume. This method provides a new machine learning model developed to improve drainage prediction from waste dumps.

The results of the study showed that the inclusion of physical pressure in the machine learning model significantly improves the accuracy of drainage volume prediction compared to the previous model. Moreover, this approach ensures that the data models are consistent with physical laws, thereby maximizing their interpretability, reliability, and applicability to similar conditions. This makes the model a more valuable source of new drainage in the mining industry and reduction of negative environmental impacts. The proposed model can be used to predict drainage water volumes in different climates and for different types of waste dumps. The results of this study can help mining companies manage drainage more effectively, reduce environmental risks, and reduce water treatment costs. In the future, the scope of the model can be expanded to include other factors affecting waste dump drainage to further improve the accuracy and reliability of the predictions.

 

However, the work has the following comments:

 

1. The paper presents a differential review of three categories of drainage forecasting models: basic water balance models, models based on fundamental results, and reactive models. However, the mathematical methods used to transform primary weather characteristics into secondary ones are not described precisely enough. What specific equations or algorithms were used?

2. The description of the differentiation of additives is too general. How was the threshold temperature achieved and was the threshold changed?

3. It is necessary to explain in more detail what specific data on critical moisture retention in the ground and unsaturated hydraulic conductivity are needed to apply the Richards equation.

4. The paper should focus on the main drainage methods, namely, the organization of a drainage network at the base of the dump to ensure the reception of runoff from the dump body and sufficient capacity to maintain technological circulating water supply, as well as the device of formation drainage at the base of the outer dump, which allows increasing the value of the stability coefficient of the mining structure. It is also interesting to consider the method of creating drainage at the base of dumps and tailings storage facilities, according to which waste is backfilled (washed) onto a prepared base with ready-made drains. More attention should also be paid to the consideration of tailings storage facilities and methods of their disposal.

5. It is necessary to present a more complete analysis of the scheme for constructing the structure of the neural network shown in Figure 2.

6. The paper substantiates the choice of Gated Recurrent Units (GRU) for constructing neural networks, citing their efficiency in processing time series and speed, especially for small data sets. However, the advantages of using this design should be given. 7. There is insufficient information on why LeakyReLU is used instead of ReLU or other activation functions and how this function is used to constrain negative results.

8. Based on the data in Figures 4-8, it is necessary to present specific mathematical models that allow calculating and predicting the output parameters under consideration during the regression analysis.

9. It is necessary to explain what is meant by the "effective contribution" of impurities, snowmelt and drainage volume? It is necessary to dwell on the methods of its calculation in more detail.

10. It is not entirely clear from the presented work how exactly the bias criteria ensure a balance between model performance and its interpretability?

11. In Section "3.4. The impact of current weather conditions on future drainage quantities" the NSE metric is mentioned for assessing model performance, but there is no information on specific NSE values obtained on different delay days.

12. It is necessary to explain in more detail why the model performance decreases more slowly at station 1 and faster at station 2? What factors influence this?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript describes prediction of mine-water drainage from waste rock piles using a machine learning model with physical constraints. In the previous paper, the authors developed the fundamental model, and in this manuscript, they updated the model. They trained and tested many times to construct a qualified model. However, the following should be considered to easily understand it and consider the applicability of the model.

1. A location map of the waste rock piles, monitoring stations, and weather stations is required. In addition, the volume of the piles should be described. This is because readers consider the applicability of the constructed model.

2. Monitored temperature, rainfall, and snowfall data should be added in supplementary information. The ranges of the data are enough. In addition, the calculated second weather features should be added also in supplementary information.

 

The specific comments are below.

1. L. 229-230 The sentences are incorrect.

2. Once the authors define the abbreviations, such as NSE, they should use the abbreviations afterward.

3. L. 325 Station 1 and 2 -> Stations 1 and 2

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors answered the questions posed and revised the article.

Reviewer 2 Report

Comments and Suggestions for Authors

Well revised.

Back to TopTop