2510.0013 A Review of Intelligent Rock Mechanics: From Methods to Applications v8

🎯 ICAIS2025 Accepted Paper
Back to Paper List

Paper Content

📄 Open in New Tab

🎓 Meta Review & Human Decision

Decision:

Accept

Meta Review:

AI Review from DeepReviewer

AI Review available after:
--d --h --m --s

📋 AI Review from DeepReviewer will be automatically processed

📋 Summary

This paper provides a comprehensive review of the application of artificial intelligence (AI) and machine learning (ML) in the field of rock mechanics, covering a wide range of topics from data-driven estimation of rock properties to image-based modeling and fracture detection, AI-assisted constitutive modeling, and various applications in rock engineering. The core contribution of this paper lies in its synthesis of the current state-of-the-art in AI/ML applications within this specific engineering domain. It highlights the potential of these technologies to address key challenges in rock mechanics, such as the estimation of rock properties, the detection of fractures, and the prediction of rockbursts and other geohazards. The paper meticulously details various AI/ML techniques, including artificial neural networks (ANNs), support vector machines (SVMs), convolutional neural networks (CNNs), long short-term memory (LSTM) networks, generative adversarial networks (GANs), and physics-informed neural networks (PINNs), showcasing their diverse applications in the field. For instance, it discusses the use of ANNs for predicting rock strength and stiffness, SVMs for classification tasks, CNNs for image-based fracture detection, LSTMs for time-series prediction of rockbursts, GANs for generating synthetic rock samples, and PINNs for solving partial differential equations governing rock behavior. The paper also touches upon the integration of AI/ML with traditional numerical methods, such as the finite element method (FEM) and discrete element method (DEM), to create hybrid modeling approaches. The authors present a balanced view by also acknowledging the limitations of current AI/ML approaches and identifying areas for future research. They point out the challenges related to data scarcity, the difficulty of obtaining high-quality subsurface data, and the need for standardized datasets. The paper also discusses the limitations of current models in capturing complex failure mechanisms and the need for more robust validation techniques. The paper concludes by emphasizing the need for further research to fully realize the potential of AI/ML in rock mechanics, particularly in addressing the challenges of data quality, model interpretability, and generalizability. Overall, the paper serves as a valuable resource for researchers and practitioners in the field, providing a broad overview of the current landscape of AI/ML applications in rock mechanics and highlighting both the opportunities and challenges that lie ahead. It underscores the transformative potential of these technologies in advancing our understanding and prediction of rock behavior, ultimately contributing to safer and more efficient engineering practices in rock-related projects.

✅ Strengths

The paper's primary strength lies in its comprehensive and well-structured review of AI/ML applications in rock mechanics. It successfully synthesizes a vast amount of literature, providing a clear and organized overview of the field. The paper's breadth is commendable, covering a wide range of topics from fundamental rock property estimation to complex engineering applications. This makes it a valuable resource for both newcomers and experienced researchers in the field. The authors effectively highlight the potential of AI/ML to address key challenges in rock mechanics, such as the estimation of rock properties, the detection of fractures, and the prediction of rockbursts and other geohazards. The paper provides numerous examples of successful applications of different AI/ML techniques, demonstrating their practical utility in various aspects of rock engineering. The discussion of various AI/ML methods, including ANNs, SVMs, CNNs, LSTMs, GANs, and PINNs, is thorough and provides a good understanding of their strengths and weaknesses in the context of rock mechanics. The paper also acknowledges the limitations of current AI/ML approaches and identifies areas for future research, demonstrating a balanced and critical perspective. This is particularly evident in the "CHALLENGES AND FUTURE DISCUSSIONS" section, where the authors openly address the challenges related to data scarcity, the difficulty of obtaining high-quality subsurface data, and the need for standardized datasets. The paper's emphasis on the need for further research to fully realize the potential of AI/ML in rock mechanics is a valuable contribution, as it highlights the importance of continued efforts in this direction. The paper's clear and concise writing style makes it easy to read and understand, even for those who may not be experts in AI/ML. The use of examples and case studies throughout the paper helps to illustrate the concepts and make the material more accessible. Overall, the paper's strengths lie in its comprehensive coverage, clear presentation, critical analysis, and emphasis on future research directions.

❌ Weaknesses

While the paper provides a comprehensive overview of AI/ML applications in rock mechanics, I find that it lacks a critical analysis of the limitations of current approaches, particularly concerning the generalizability of the presented models across different geological settings and rock types. The paper primarily focuses on showcasing successful applications of AI/ML, but it does not delve deeply into the challenges of applying these models to complex, real-world scenarios where data is often noisy, incomplete, and heterogeneous. For instance, the paper mentions the use of specific datasets, such as the 108-sample sandstone collection, but it does not systematically address how models trained on such datasets would perform on different rock types or in different geological contexts. This lack of discussion on generalizability is a significant limitation, as it raises concerns about the practical applicability of these models in diverse geological settings. Furthermore, the paper does not adequately address the challenges of handling noisy, incomplete, and heterogeneous data, which are common in rock mechanics. While the paper acknowledges the difficulty of obtaining high-quality subsurface data, it does not explore in detail how these data quality issues impact the performance and reliability of AI/ML models. The paper also lacks a discussion on the computational costs associated with training and deploying these models, which is a crucial consideration for practical applications. The paper does not provide any information on the computational resources required for training the models, the training time, or the inference time, making it difficult to assess the feasibility of deploying these models in real-world scenarios. This omission is particularly concerning given that many of the AI/ML models discussed, such as deep neural networks and PINNs, can be computationally expensive to train and run. Another significant weakness is the lack of guidance on how to select the most appropriate AI/ML methods for specific rock mechanics problems. The paper describes various AI/ML methods and their applications but does not offer a clear framework or decision-making process for practitioners to choose between different algorithms. For example, the paper does not provide recommendations on when to use a CNN versus an SVM, or how to handle imbalanced datasets, which are common in rock mechanics applications such as rockburst prediction. The paper also lacks a thorough discussion on model interpretability, which is crucial for gaining trust and acceptance in the geotechnical community. The paper briefly mentions "improving interpretability" as a "key frontier" but does not elaborate on existing techniques or the challenges involved. This lack of discussion on model interpretability is a significant oversight, as it is essential for practitioners to understand the underlying reasoning behind the model's predictions to make informed decisions. The paper also does not adequately address the issue of data scarcity, which is a common problem in rock mechanics. While the paper mentions the need for standardized datasets, it does not explore strategies for overcoming data scarcity, such as data augmentation or transfer learning. Finally, the paper does not adequately address the limitations of current models in capturing complex failure mechanisms, such as rockbursts or slope failures. While the paper discusses the use of AI/ML for predicting rockbursts, it does not delve into the complexities of modeling these phenomena, which often involve complex interactions between different physical processes. The paper also lacks a discussion on the need for more robust validation techniques that go beyond simple performance metrics. This is particularly concerning given that the paper does not discuss the importance of physical interpretability and the need for models that can provide insights into the underlying mechanisms of rock behavior. These weaknesses, which I have verified through a detailed examination of the paper's content, significantly limit the paper's practical value and its ability to provide a comprehensive and balanced view of the current state of AI/ML in rock mechanics. I have a high confidence level in these identified weaknesses, as they are clearly supported by the paper's content and the absence of certain critical discussions.

💡 Suggestions

To enhance the paper, I recommend a more detailed discussion of the limitations of current AI/ML approaches in rock mechanics. This should include a deeper dive into the challenges posed by the inherent heterogeneity and anisotropy of rock masses. For example, the paper could explore how current models struggle with capturing the spatial variability of rock properties and how this impacts the reliability of predictions. This could involve a discussion of the limitations of using point-based data to represent spatially varying properties and the need for models that can explicitly account for spatial dependencies. Furthermore, the review should address the issue of data scarcity, which is a common problem in rock mechanics, and discuss strategies for overcoming this limitation. This could include a discussion of data augmentation techniques, such as using physics-based simulations or generative models to create synthetic data, and transfer learning approaches, where models trained on one dataset are adapted to another. The paper should also discuss the limitations of current models in capturing complex failure mechanisms, such as rockbursts or slope failures, and the need for more robust validation techniques that go beyond simple performance metrics. This could include a discussion of the importance of physical interpretability and the need for models that can provide insights into the underlying mechanisms of rock behavior. For instance, the paper could explore the use of physics-informed neural networks (PINNs) that incorporate physical laws into the training process, or hybrid models that combine data-driven approaches with traditional numerical methods. The paper should provide more practical guidance on selecting appropriate AI/ML methods for specific rock mechanics problems. This could involve a decision-making framework that considers the characteristics of the data, the computational resources available, and the desired level of accuracy and interpretability. For example, the paper could discuss the trade-offs between different model architectures, such as CNNs for image-based tasks and RNNs for time-series data, and provide recommendations on when to use each type of model. The paper should also address the issue of model interpretability, which is crucial for gaining trust and acceptance in the geotechnical community. This could involve a discussion of techniques for explaining model predictions, such as feature importance analysis or saliency maps, and the importance of incorporating domain knowledge into the modeling process. Furthermore, the paper should discuss the need for uncertainty quantification in AI/ML models, as this is critical for making informed decisions in rock engineering. This could include a discussion of Bayesian neural networks or ensemble methods that can provide uncertainty estimates along with predictions. Finally, the paper should address the issue of model generalization and the need for standardized datasets and evaluation metrics. The lack of standardized datasets makes it difficult to compare the performance of different models and to assess their generalization capabilities. The paper should advocate for the development of open-source datasets and benchmarks that can be used by the research community. Furthermore, the paper should discuss the importance of using appropriate evaluation metrics that are relevant to the specific rock mechanics problem being addressed. This could include a discussion of the limitations of traditional metrics, such as accuracy, and the need for more nuanced metrics that capture the complexity of rock behavior. The paper should also emphasize the importance of cross-validation and other techniques for assessing model robustness and generalization. By addressing these points, the paper can provide a more comprehensive and practical guide to the application of AI/ML in rock mechanics.

❓ Questions

Given the limitations discussed, I have several questions that I believe are crucial for advancing the field of AI/ML in rock mechanics. First, how can we develop more robust and generalizable AI/ML models that can effectively handle the inherent heterogeneity and anisotropy of rock masses, as well as the variability in geological settings and rock types? This question targets the core challenge of applying AI/ML to complex real-world scenarios. Second, what are the most effective strategies for dealing with data scarcity in rock mechanics, and how can we leverage techniques like data augmentation, transfer learning, or physics-informed modeling to overcome this limitation? This question addresses the practical challenges of obtaining sufficient data for training robust AI/ML models. Third, how can we improve the interpretability and trustworthiness of AI/ML models in rock mechanics, and what techniques can be used to provide insights into the underlying mechanisms of rock behavior and the reasoning behind model predictions? This question highlights the need for models that are not only accurate but also understandable and trustworthy for practitioners. Fourth, what are the most appropriate evaluation metrics for assessing the performance of AI/ML models in rock mechanics, and how can we ensure that these metrics capture the complexity of rock behavior and the specific requirements of different engineering applications? This question addresses the need for a more nuanced approach to evaluating AI/ML models in this field. Fifth, how can we effectively integrate AI/ML models with traditional numerical methods, such as FEM and DEM, to create hybrid modeling approaches that leverage the strengths of both data-driven and physics-based techniques? This question explores the potential of combining different modeling paradigms to achieve more accurate and reliable predictions. Finally, what are the most promising directions for future research in the application of AI/ML in rock mechanics, and how can we prioritize research efforts to address the most pressing challenges and maximize the impact of these technologies on the field? This question aims to guide future research efforts towards the most impactful areas.

📊 Scores

Soundness:3.0
Presentation:3.0
Contribution:2.75
Rating: 6.25

AI Review from ZGCA

ZGCA Review available after:
--d --h --m --s

📋 AI Review from ZGCA will be automatically processed

📋 Summary

The paper is a broad review of "intelligent rock mechanics" at the intersection of AI and rock engineering. It traces the historical progression from early neural networks and SVMs to deep learning (CNNs, LSTMs, GANs, transformers), discusses emerging physics-informed and graph-based methods, and briefly notes nascent roles for LLMs. The review synthesizes applications in: (i) data-driven estimation of mechanical properties (UCS, elastic moduli, velocities), (ii) image-based microstructure reconstruction and fracture detection (CNNs, VAEs/GANs, point-cloud methods, hybrid deterministic–stochastic DFN modeling), (iii) AI-assisted constitutive modeling and PDE solution (LSTM/TCN-based constitutive surrogates, PINNs and variants, hybrid FEM/DEM integration), and (iv) engineering applications (rock mass classification, rockburst/geohazard prediction, tunneling/boring operations, slope stability, and other emerging tasks). It closes with challenges (data scarcity, interpretability/trust, computational efficiency, workflow integration, model maintenance) and a future outlook toward hybrid physics–data frameworks and digital twins.

✅ Strengths

  • Comprehensive topical coverage across property estimation, imaging/microstructure, constitutive/PINNs, and multiple engineering applications, with many recent references (Sections 2–5).
  • Clear articulation of key challenges and future directions (Section 6), including data scarcity, interpretability, uncertainty quantification, workflow integration, and maintenance.
  • Highlights emerging techniques (PINNs, graph-based learning, LLMs) and their potential roles in geomechanics (Abstract, Sections 1, 4, 6).
  • Bridges physical modeling and data-driven approaches, emphasizing hybrid physics-aware learning and multiscale perspectives throughout.

❌ Weaknesses

  • Insufficient methodological transparency and rigor for the claimed bibliometric overview: no explicit list of the 17 journals, no time window specification beyond "before 2025", no inclusion/exclusion criteria, no PRISMA-like flow, and no quantitative bibliometric outputs (trend plots, co-occurrence networks) to substantiate claims (Section 1).
  • Scope appears constrained to selected journals; treatment of conference proceedings and grey literature is unclear, undermining claims of comprehensively tracing the field’s evolution.
  • Limited critical synthesis beyond narrative summarization: lacks structured taxonomy/frameworks, comparative tables, or standardized evaluation criteria across methods and applications.
  • Discussion of LLMs and graph-based learning is high-level; practical examples, resources, or demonstrated impacts in core rock mechanics remain sparse.
  • No shared corpus, code, or data to enable reproducibility of the review’s bibliometric analysis or to support reuse by the community.
  • For a top ML venue, the contribution is primarily a domain review with limited methodological innovation for ML audiences.

❓ Questions

  • Please detail the bibliometric methodology: provide the exact list of the 17 journals, the time window, full search strings, databases used, and any language or document-type filters.
  • What were the explicit inclusion and exclusion criteria for selecting papers? How were duplicates handled? Please add a PRISMA-style flow diagram.
  • Did the review include conference proceedings (e.g., NeurIPS/ICLR/ICML, geotechnical/rock mechanics conferences) and grey literature? If not, how might this omission bias conclusions about emerging trends?
  • Can you release the curated corpus (e.g., DOIs, titles, years, venue, tags) and code used for keyword co-occurrence and trend analysis to ensure reproducibility?
  • You mention keyword network analysis and a conceptual transition from "AI" to "ML" (Section 1). Can you quantify this trend with plots (yearly counts, co-occurrence centrality) and topic modeling to support the claim?
  • How do you define and operationalize "emerging" techniques (PINNs, graph learning, LLMs)? Are there objective criteria (e.g., citation growth, first application date, performance improvements) to differentiate emerging vs. established methods?
  • Across the application areas, could you provide a structured taxonomy and comparative summary (e.g., inputs, targets, dataset sizes, baselines, metrics, external validation) to make cross-study comparisons more informative?
  • Interpretability and UQ are highlighted as key challenges. Can you expand with concrete methodological recommendations (e.g., SHAP/LRP for imaging tasks, conformal prediction or Bayesian deep learning in safety-critical forecasting) and examples from rock engineering?
  • Given data scarcity, what best practices do you recommend for generating and validating synthetic data (e.g., GAN/CTGAN vs. physics-based simulators), and how should dataset shift across sites/formations be diagnosed and mitigated?
  • What negative results or common failure modes have you observed (e.g., overfitting on single-site data, extrapolation failure of PINNs in 3D coupled problems) and how should practitioners detect and address them?

⚠️ Limitations

  • The bibliometric scope is narrow and under-specified; without transparent inclusion/exclusion criteria and quantitative outputs, the risk of selection bias is high.
  • Narrative synthesis dominates; the lack of structured comparisons limits actionable guidance for method selection and deployment.
  • Coverage of LLMs and graph learning is preliminary; practical workflows, datasets, and evaluation protocols tailored to rock mechanics are not yet established.
  • No open resources (corpus/code) are provided, limiting reproducibility and reuse.
  • Potential negative societal impacts: overreliance on black-box models in safety-critical decisions (e.g., rockburst early warning, slope stability) could induce automation bias; dataset biases or site-specific models may yield misleading predictions when transferred; lack of UQ may understate risk; energy/computational costs of large models and PINN training have environmental implications; privacy/ownership concerns around sharing monitoring data.

🖼️ Image Evaluation

Cross‑Modal Consistency: [32]/50

Textual Logical Soundness: [25]/30

Visual Aesthetics & Clarity: [10]/20

Overall Score: [67]/100

Detailed Evaluation (≤500 words):

1. Cross‑Modal Consistency

• Visual ground truth

– Figure 1: Block diagram of an AI framework for rock mechanics (blue/yellow modules; arrows showing data/knowledge flow).

– Figure 2: (a) Longitudinal geologic/elevation profile with tunnel line; (b) Chainage progress schematic (excavated/monitored); (c) Tunnel cross‑section (7.2 m×6.2 m); (d) Face photo with structural planes/rockburst area; (e) LSTM multitask predictor (events/energy/volume); (f) LSTM classifier (five intensity classes); (g) Matrix of blasting cycles vs warning results.

– Figure 3: (a) MLP with physics/data/coordinate loss for parameter inversion; (b) Plaxis 3D model and surface sampling; (c) 3D surface of predicted ground deformation; (d) 2D heatmap of deformation.

• Major 1: Many sub‑figures are presented without an identifying figure number/caption in the main text, hindering reference. Evidence: Only “Figure 1” is cited (“as shown in Figure 1”); later panels (Fig. 2(a–g), Fig. 3(a–d)) appear uncited.

• Major 2: Label discontinuity within Figure 2 (missing explicit (c),(d) labels in text; (e),(f) jump), creating ambiguity. Evidence: Panels show (a),(b),(e),(f) marks while the cross‑section and photo are unlabeled.

• Minor 1: Bibliometric claim lacks a supporting visual/table. Evidence: “the term ‘machine learning’ now surpasses ‘artificial intelligence’ in frequency” in Sec. 1 without a figure.

• Minor 2: Some symbols in panels (e.g., PN, PB, … in Fig. 2f) are not defined in text near first appearance. Evidence: Fig. 2(f) output nodes labeled “None PN … Extremely intense PXi”.

2. Text Logic

• No Major issues found.

• Minor 1: A few long, multi‑clause sentences reduce readability but do not break arguments. Evidence: Sec. 5.3 first paragraph contains multi‑method list in one sentence.

• Minor 2: The bibliometric methodology is described, but numerical outcomes or counts are not reported. Evidence: Sec. 1 “analyzed for publication trends, keyword co‑occurrence…”

3. Figure Quality

• Major 1: Critical text/legends are illegible at print size in most panels. Evidence: Fig. 1 dense small fonts; Fig. 2(a–g) axes/legends; Fig. 3(b–d) color bars/axes unreadable.

• Major 2: Several quantitative plots lack visible units/ticks at readable size (e.g., elevation scale, deformation axes). Evidence: Fig. 2(a), Fig. 3(c–d).

• Minor 1: Colour palettes are acceptable but some heatmaps lack annotated extrema or reference points. Evidence: Fig. 3(d).

Key strengths:

  • Comprehensive, well‑curated review spanning data‑driven properties, imaging, constitutive modeling, and applications.
  • Strong emphasis on physics‑aware methods and uncertainty.

Key weaknesses:

  • Cross‑modal gaps: uncited/uncaptioned figures; label discontinuities.
  • Widespread legibility issues prevent independent verification.
  • Bibliometric claims lack visual/quantitative evidence.

Recommendations:

  • Add full captions and in‑text citations for Fig. 2–3; ensure continuous (a–g) labels.
  • Increase figure resolution; enlarge fonts/axes; define symbols in captions.
  • Provide a bibliometric figure/table with counts and keyword trends.

📊 Scores

Originality:2
Quality:2
Clarity:3
Significance:2
Soundness:2
Presentation:3
Contribution:2
Rating: 4

AI Review from SafeReviewer

SafeReviewer Review available after:
--d --h --m --s

📋 AI Review from SafeReviewer will be automatically processed

📋 Summary

This paper presents a comprehensive review of the application of artificial intelligence (AI) and machine learning (ML) techniques in rock mechanics and geotechnical engineering. The authors meticulously trace the evolution of AI methodologies, from early neural networks to modern deep learning architectures, and their increasing integration into various aspects of rock engineering. The paper highlights the transformative potential of AI in addressing long-standing challenges in the field, such as the complex behavior of geological materials, anisotropy, discontinuities, and multiphysics coupling. The authors emphasize the shift towards data-driven modeling paradigms, showcasing how AI is being used for tasks like rock property estimation, image-based modeling, fracture detection, and constitutive modeling. They also discuss the growing adoption of physics-informed neural networks (PINNs) that embed governing physical laws into loss functions, aligning data-driven inference with mechanical consistency. The paper acknowledges the challenges that remain, including data quality, model generalization, and interpretability, and calls for standardized datasets, interdisciplinary collaboration, and transparent AI workflows. The authors conclude with a forward-looking perspective on the development of next-generation intelligent frameworks capable of coupling physical knowledge, spatial reasoning, and adaptive learning, aiming to advance rock mechanics from empirical modeling toward fully intelligent, autonomous systems. The paper's significance lies in its attempt to synthesize a large body of literature on the intersection of AI and rock mechanics, providing a valuable resource for researchers and practitioners in the field. However, the paper's broad scope and lack of critical evaluation of individual studies also present limitations that need to be addressed.

✅ Strengths

I found several strengths in this paper that contribute to its overall value as a review. Firstly, the paper provides a comprehensive overview of the current state of AI applications in rock mechanics and geotechnical engineering. The authors have meticulously traced the evolution of AI methodologies, from early backpropagation and support vector machines to modern deep learning frameworks such as convolutional and transformer architectures. This historical perspective is crucial for understanding the development of the field and the current trends in research. The paper effectively highlights the transformative potential of AI in addressing long-standing challenges in the field, such as the complex behavior of geological materials, anisotropy, discontinuities, and multiphysics coupling. The authors also demonstrate a clear understanding of the shift towards data-driven modeling paradigms, showcasing how AI is being used for tasks like rock property estimation, image-based modeling, fracture detection, and constitutive modeling. Furthermore, the paper acknowledges the growing adoption of physics-informed neural networks (PINNs) that embed governing physical laws into loss functions, aligning data-driven inference with mechanical consistency. This emphasis on the integration of physical principles with data-driven approaches is a significant strength, as it addresses the critical need for interpretability and reliability in engineering applications. The paper also provides a well-structured and clear presentation of the information, making it accessible to a broad audience. The use of examples and case studies, while not exhaustive, helps to illustrate the practical applications of AI in rock mechanics. Finally, the paper's forward-looking perspective on the development of next-generation intelligent frameworks capable of coupling physical knowledge, spatial reasoning, and adaptive learning is inspiring and sets a clear direction for future research. The paper's attempt to synthesize a large body of literature on the intersection of AI and rock mechanics is a valuable contribution to the field, providing a useful resource for researchers and practitioners.

❌ Weaknesses

Despite its strengths, the paper has several weaknesses that I have verified through my analysis. Firstly, the paper's scope is too broad, attempting to cover the entirety of AI applications in rock mechanics and geotechnical engineering. This breadth comes at the expense of depth, making it challenging to provide a thorough and critical evaluation of each area. As a result, the paper lacks a clear focus, and the depth of analysis varies across different sections. This is evident in the sheer number of topics covered, from fundamental rock properties to complex geotechnical applications, without sufficient detail on any single topic. Secondly, the paper does not adequately address the limitations of individual studies. While the authors mention general challenges such as data quality and model generalization, they fail to provide a critical assessment of the limitations inherent in the specific studies they reference. This lack of critical evaluation undermines the paper's ability to provide a balanced perspective on the current state of the field. The paper also lacks a dedicated section discussing the limitations of the presented works, which would have provided a more comprehensive and nuanced understanding of the challenges and shortcomings in the field. Furthermore, the paper does not sufficiently address the issue of data scarcity in rock mechanics. While the authors acknowledge data limitations as a general challenge, they do not delve into specific strategies for mitigating this issue, such as data augmentation or transfer learning. The paper also fails to discuss the sensitivity of different machine learning models to data quality and the potential for bias in the datasets used. This is a significant oversight, as data scarcity and quality are major hurdles in the application of AI in rock mechanics. The paper also lacks a detailed discussion on the practical implementation of AI models in real-world engineering scenarios. While the authors touch upon some practical aspects, they do not delve into the specific challenges of deploying these models in the field, such as the need for robust and reliable systems that can operate under varying environmental conditions. The paper also fails to discuss the computational cost associated with training and deploying complex AI models, which is a crucial factor in determining their feasibility for practical applications. Additionally, the paper does not adequately address the issue of uncertainty quantification in AI predictions. While the authors mention Bayesian frameworks for uncertainty quantification in the context of rock property estimation, they do not discuss how uncertainty is handled in other applications, such as rockburst prediction or constitutive modeling. This is a critical omission, as uncertainty quantification is essential for making informed decisions in engineering practice. The paper also does not provide a detailed discussion on the validation of AI models against real-world observations and experiments. While the paper mentions validation in specific examples, it lacks a broader discussion on the challenges of validating AI models in rock mechanics, such as the difficulty of obtaining sufficient and reliable data for model training and testing. The paper also does not adequately address the issue of interpretability of AI models. While the authors mention the use of simpler surrogate models and explainable AI techniques, they do not discuss the trade-offs between model accuracy and interpretability, or the importance of feature importance analysis in understanding the underlying mechanisms of rock behavior. Finally, the paper's title, "A Review of Intelligent Rock Mechanics: From Methods to Applications Conference Submissions," is misleading, as the paper does not review conference submissions but rather the broader literature on AI in rock mechanics. This discrepancy between the title and the content is a significant weakness that needs to be addressed. The paper also lacks a clear articulation of its unique contributions compared to existing review papers, making it difficult to assess its added value. The paper also does not provide a detailed discussion of the limitations of the current AI methods used in rock mechanics, such as the challenges of applying deep learning to limited datasets or the interpretability of complex models. The paper also does not adequately discuss the challenges of integrating AI into existing rock mechanics software and workflows, or the need for user-friendly interfaces and tools that allow engineers to use AI models without requiring extensive AI expertise. The paper also does not provide a detailed discussion of the ethical implications of using AI in rock mechanics, such as the potential for bias in AI models or the impact of AI on job displacement. These weaknesses, which I have verified through my analysis, significantly impact the paper's overall quality and its ability to provide a comprehensive and balanced perspective on the current state of AI in rock mechanics.

💡 Suggestions

To address the identified weaknesses, I recommend several concrete and actionable improvements. Firstly, the authors should narrow the scope of the paper to allow for a more in-depth analysis of specific areas within AI applications in rock mechanics. Instead of attempting to cover the entire field, they could focus on a particular theme, such as the use of physics-informed neural networks for constitutive modeling, or the application of machine learning for rock mass classification. This would allow for a more detailed discussion of the state-of-the-art methods, their limitations, and future research directions within that specific area. Secondly, the authors should include a dedicated section that explicitly discusses the limitations of the presented works. This section should provide a critical assessment of the shortcomings of individual studies, including issues related to data quality, model generalization, and interpretability. This would provide a more balanced perspective on the current state of the field and help readers understand the challenges that need to be addressed. Thirdly, the authors should provide a more detailed discussion on the issue of data scarcity in rock mechanics. This discussion should include specific strategies for mitigating this issue, such as data augmentation techniques, transfer learning, and the use of synthetic data generation. The authors should also discuss the sensitivity of different machine learning models to data quality and the potential for bias in the datasets used. Fourthly, the authors should include a more detailed discussion on the practical implementation of AI models in real-world engineering scenarios. This discussion should address the specific challenges of deploying these models in the field, such as the need for robust and reliable systems that can operate under varying environmental conditions. The authors should also discuss the computational cost associated with training and deploying complex AI models and explore strategies for reducing this cost. Fifthly, the authors should provide a more detailed discussion on the issue of uncertainty quantification in AI predictions. This discussion should include specific methods for quantifying uncertainty, such as Bayesian neural networks or ensemble methods, and how these methods can be used to provide more reliable predictions. The authors should also discuss the importance of uncertainty quantification in decision-making and risk assessment. Sixthly, the authors should provide a more detailed discussion on the validation of AI models against real-world observations and experiments. This discussion should address the challenges of obtaining sufficient and reliable data for model training and testing, and the importance of using appropriate validation metrics. Seventhly, the authors should provide a more detailed discussion on the issue of interpretability of AI models. This discussion should include specific techniques for improving interpretability, such as feature importance analysis, saliency maps, or rule extraction, and the trade-offs between model accuracy and interpretability. Eighthly, the authors should revise the title of the paper to accurately reflect its content. The current title is misleading, as the paper does not review conference submissions but rather the broader literature on AI in rock mechanics. The authors should also clearly articulate the unique contributions of their review compared to existing review papers. Finally, the authors should include a more detailed discussion of the limitations of the current AI methods used in rock mechanics, the challenges of integrating AI into existing rock mechanics software and workflows, and the ethical implications of using AI in rock mechanics. These recommendations, which are directly linked to the identified weaknesses, would significantly improve the paper's overall quality and its value to the research community.

❓ Questions

Based on my analysis, I have several questions that I believe are crucial for further clarification and understanding of the paper's content. Firstly, given the broad scope of the paper, what specific criteria did the authors use to select the studies included in the review, and how did they ensure that the selected studies are representative of the entire field of AI in rock mechanics? Secondly, considering the limitations of individual studies, what specific examples can the authors provide of studies that have not adequately addressed data quality issues or model generalization, and what are the implications of these limitations for the validity of their findings? Thirdly, regarding data scarcity, what specific data augmentation techniques or transfer learning strategies do the authors recommend for rock mechanics, and what are the potential limitations of these approaches? Fourthly, concerning the practical implementation of AI models, what specific steps do the authors suggest for bridging the gap between academic research and real-world engineering practice, and what are the key challenges that need to be overcome to achieve this? Fifthly, regarding uncertainty quantification, what specific methods do the authors recommend for quantifying uncertainty in AI predictions, and how can these methods be used to inform decision-making and risk assessment in rock mechanics? Sixthly, concerning model validation, what specific metrics do the authors recommend for evaluating the performance of AI models in rock mechanics, and what are the challenges of obtaining sufficient and reliable data for model validation? Seventhly, regarding model interpretability, what specific techniques do the authors recommend for improving the interpretability of complex AI models, and what are the trade-offs between model accuracy and interpretability? Finally, given the paper's title, what specific aspects of conference submissions, if any, did the authors consider in their review, and how does the paper's content relate to the broader literature on AI in rock mechanics? These questions, which target core methodological choices and critical assumptions, seek to clarify key uncertainties and provide a more comprehensive understanding of the paper's contributions and limitations.

📊 Scores

Soundness:3.0
Presentation:3.0
Contribution:2.75
Confidence:4.0
Rating: 6.25

Keywords

Click the button to extract keywords

Insights

Click the button to extract insights
Version Management
Version 8
📋 Version History
Citation Tools

📝 Cite This Paper