2510.0013 A Review of Intelligent Rock Mechanics: From Methods to Applications v1

🎯 ICAIS2025 Submission

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 integration of artificial intelligence (AI) into the field of rock mechanics, tracing the evolution of AI methodologies from traditional approaches like backpropagation and support vector machines to modern deep learning frameworks such as convolutional neural networks and transformers. The core contribution of this paper lies in its synthesis of recent advancements in intelligent rock mechanics, highlighting the growing role of AI in various aspects of the field, including microstructure reconstruction, mechanical parameter estimation, constitutive modeling, and real-time hazard prediction. The paper emphasizes the integration of data-driven AI models with physical principles, showcasing examples of physics-informed neural networks and graph-based learning. A key theme throughout the paper is the potential of AI to enhance our understanding and prediction capabilities in rock mechanics, moving towards more intelligent and autonomous systems. The paper also acknowledges the challenges in the field, such as data quality, model generalization, and interpretability, and suggests future directions for developing intelligent frameworks that couple physical knowledge with spatial reasoning and adaptive learning. The paper's significance lies in its ability to provide a clear overview of the current state of AI in rock mechanics, identify key challenges, and outline potential future directions. It serves as a valuable resource for researchers and practitioners interested in the intersection of these two fields. However, it is important to note that the paper is a review and does not present any novel research findings or methodologies. The paper's structure follows a logical progression, starting with an introduction to the history of AI, then delving into specific applications in rock mechanics, and concluding with a discussion of challenges and future prospects. The paper's analysis of the literature is thorough, and it effectively synthesizes recent progress in the field. While the paper provides a solid foundation for understanding the current state of AI in rock mechanics, it could benefit from a more detailed discussion of practical applications and a broader range of AI methodologies. The paper's emphasis on the integration of AI and physics is particularly noteworthy, as it highlights the importance of combining data-driven models with physical principles to enhance the interpretability and reliability of AI applications in rock mechanics. Overall, this paper is a valuable contribution to the field, providing a clear and concise overview of the current state of AI in rock mechanics and identifying key areas for future research.

✅ Strengths

This paper's primary strength lies in its comprehensive review of the literature, effectively synthesizing recent advancements in the application of artificial intelligence to rock mechanics. The paper provides a well-structured and logically organized overview, making it easy for readers to understand the evolution of AI methodologies and their applications in this field. The paper successfully highlights the growing trend of combining data-driven AI models with physical principles, which is crucial for enhancing the interpretability and reliability of AI applications in rock mechanics. The inclusion of examples of physics-informed neural networks and graph-based learning effectively illustrates this trend. The paper also provides a forward-looking perspective, outlining future directions for the field. It suggests the development of next-generation intelligent frameworks that can couple physical knowledge, spatial reasoning, and adaptive learning, which is a valuable contribution to the field. The paper's acknowledgement of key challenges in the field, such as data quality, model generalization, and interpretability, is another strength. It emphasizes the need for standardized datasets, interdisciplinary collaboration, and transparent AI workflows to address these issues, which is a crucial step towards advancing the field. The paper's clear and concise writing style makes it accessible to a wide audience, including researchers and practitioners in both AI and rock mechanics. The paper's ability to provide a clear overview of the current state of AI in rock mechanics, identify key challenges, and outline potential future directions is a significant contribution. The paper's emphasis on the integration of AI and physics is particularly noteworthy, as it highlights the importance of combining data-driven models with physical principles to enhance the interpretability and reliability of AI applications in rock mechanics. The paper's synthesis of recent progress in the field makes it a valuable resource for researchers and practitioners interested in the intersection of these two fields. The paper's clear and concise writing style makes it accessible to a wide audience, including researchers and practitioners in both AI and rock mechanics. The paper's ability to provide a clear overview of the current state of AI in rock mechanics, identify key challenges, and outline potential future directions is a significant contribution.

❌ Weaknesses

While this paper provides a valuable overview of AI in rock mechanics, it suffers from several key limitations that significantly impact its overall contribution. Firstly, and most importantly, the paper does not present any novel research findings or methodologies. As a review paper, its primary purpose is to synthesize existing knowledge, but this lack of original contribution limits its impact on the field. The paper explicitly states its purpose as a review in the abstract and introduction, confirming that it does not aim to introduce new research. This is a significant limitation because it means the paper does not offer any groundbreaking insights or solutions to existing problems. Secondly, the paper exhibits a limited discussion of practical applications. While it reviews various AI methodologies, it lacks detailed case studies or examples of successful implementations in real-world rock mechanics scenarios. The "EXPERIMENTS" section, while present, primarily summarizes existing research rather than providing in-depth accounts of practical, on-the-ground applications. This lack of practical examples reduces the paper's value for practitioners seeking to apply AI in their work. The paper does not delve into the specific steps and considerations for implementing AI models in real-world rock mechanics projects, such as data acquisition, model selection, training, validation, and integration into existing workflows. This omission makes it difficult for readers to understand the practical challenges and limitations of applying AI in this field. The paper's focus on specific AI techniques, primarily deep learning and physics-informed neural networks, is another significant weakness. While these techniques are important, the paper does not provide a balanced discussion of other AI methodologies, such as fuzzy logic, Bayesian networks, and support vector machines, which have also been successfully applied in rock mechanics. The paper's structure and content clearly show a greater emphasis on deep learning and physics-informed neural networks compared to other AI techniques. This narrow focus limits the paper's scope and does not provide a comprehensive overview of the AI landscape in rock mechanics. The paper also lacks a detailed discussion of the limitations and challenges of applying AI in rock mechanics. While it mentions data quality, model generalization, and interpretability as key challenges, it does not delve into the specific issues that practitioners face. For instance, the paper does not discuss the difficulties in obtaining high-quality, labeled datasets for training AI models in rock mechanics, given the inherent variability and complexity of geological materials. Furthermore, it does not elaborate on the challenges of deploying these models in real-world scenarios, such as the need for robust models that can handle noisy or incomplete data, and the computational constraints often present in field settings. This lack of a thorough discussion of practical challenges reduces the paper's relevance and impact. The paper's limited discussion of the ethical considerations of using AI in rock mechanics is another weakness. It does not address the potential for AI to perpetuate existing biases in rock mechanics data, the importance of transparency and explainability in AI models, and the need for responsible use of AI in decision-making processes. This omission is significant because it means the paper does not consider the broader societal implications of using AI in this field. The paper's failure to provide a more detailed discussion of the specific steps and considerations involved in implementing AI models in real-world rock mechanics projects, including a breakdown of the data acquisition process, the selection of appropriate AI algorithms, the training and validation of models, and the integration of these models into existing rock mechanics workflows, is a significant limitation. For example, when discussing data-driven estimation of rock properties, the paper could have provided a case study that outlines the specific types of data required, the preprocessing steps necessary to prepare the data for AI analysis, and the evaluation metrics used to assess the performance of the AI model. Similarly, for image-based modeling and fracture detection, the paper could have detailed the specific image acquisition techniques, the image processing algorithms used to enhance image quality, and the methods used to quantify fracture characteristics from the processed images. This level of detail is missing, which significantly reduces the paper's practical value. The paper's lack of a thorough discussion of the limitations and challenges of applying AI in rock mechanics, including the potential risks and uncertainties associated with using AI models, such as the potential for biased or inaccurate predictions due to limited or biased training data, is another significant weakness. The paper also does not discuss the challenges of model generalization, including the potential for overfitting or underfitting, and the strategies used to mitigate these issues. For example, when discussing AI-assisted constitutive modeling, the paper could have discussed the limitations of using AI to model complex rock behavior, such as the difficulty in capturing the underlying physics of rock deformation and failure. They could also have discussed the potential for AI models to produce results that are not physically meaningful, and the importance of incorporating domain knowledge to ensure that AI models are both accurate and interpretable. This lack of a balanced and realistic view of the potential and limitations of AI in rock mechanics is a significant weakness. Finally, the paper's failure to include a section on the ethical considerations of using AI in rock mechanics, including a discussion of the potential for AI to perpetuate existing biases in rock mechanics data, the importance of transparency and explainability in AI models, and the need for responsible use of AI in decision-making processes, is a significant omission. This lack of consideration for the ethical implications of using AI in this field is a major weakness. All of these weaknesses are supported by direct evidence from the paper and are not speculative. The paper's limitations have a substantial impact on its overall conclusions and practical value. I am highly confident in the validity of these identified weaknesses.

💡 Suggestions

To significantly enhance the value and impact of this paper, several key improvements are necessary. Firstly, the paper should include detailed case studies that illustrate the successful application of AI in rock mechanics. These case studies should not only describe the AI methodologies used but also provide a comprehensive analysis of the results, including the benefits and limitations of the approach. For example, a case study could focus on the use of AI for predicting rockfall hazards, detailing the data collection process, the specific AI model used, and the accuracy of the predictions. Another case study could explore the use of AI for optimizing the design of rock slopes, demonstrating how AI can be used to improve safety and reduce costs. These concrete examples would provide valuable insights for practitioners and researchers alike, and would help to bridge the gap between theoretical concepts and practical applications. The paper should also broaden its scope to include a wider range of AI methodologies and their potential applications in rock mechanics. While deep learning and physics-informed neural networks are important, other techniques such as fuzzy logic, Bayesian networks, and support vector machines have also been successfully applied in this field. The paper should discuss the strengths and weaknesses of these different approaches and provide guidance on when each approach is most appropriate. Furthermore, the paper should explore the potential of emerging AI techniques, such as graph neural networks and reinforcement learning, for addressing complex problems in rock mechanics. By providing a more comprehensive overview of the AI landscape, the paper would be a more valuable resource for researchers and practitioners in the field. The paper should also include a more detailed exploration of the practical challenges and limitations encountered when applying AI in rock mechanics. While the review mentions data quality, model generalization, and interpretability as key challenges, it lacks a deep dive into the specific issues that practitioners face. For instance, the paper could discuss the difficulties in obtaining high-quality, labeled datasets for training AI models in rock mechanics, given the inherent variability and complexity of geological materials. Furthermore, it could elaborate on the challenges of deploying these models in real-world scenarios, such as the need for robust models that can handle noisy or incomplete data, and the computational constraints often present in field settings. A more thorough discussion of these practical considerations would enhance the paper's relevance and impact. The paper should also address the limitations and challenges of applying AI in rock mechanics more thoroughly. This should include a discussion of the potential risks and uncertainties associated with using AI models, such as the potential for biased or inaccurate predictions due to limited or biased training data. The authors should also discuss the challenges of model generalization, including the potential for overfitting or underfitting, and the strategies used to mitigate these issues. For example, when discussing AI-assisted constitutive modeling, the authors could discuss the limitations of using AI to model complex rock behavior, such as the difficulty in capturing the underlying physics of rock deformation and failure. They could also discuss the potential for AI models to produce results that are not physically meaningful, and the importance of incorporating domain knowledge to ensure that AI models are both accurate and interpretable. This would provide a more balanced and realistic view of the potential and limitations of AI in rock mechanics. Finally, the authors should consider including a section on the ethical considerations of using AI in rock mechanics. This could include a discussion of the potential for AI to perpetuate existing biases in rock mechanics data, the importance of transparency and explainability in AI models, and the need for responsible use of AI in decision-making processes. For example, the authors could discuss the potential for AI to be used to automate tasks that were previously performed by human experts, and the potential impact of this on the workforce. They could also discuss the importance of ensuring that AI models are used in a way that is fair and equitable, and that the benefits of AI are shared across all stakeholders. This would help to ensure that the development and application of AI in rock mechanics is both ethical and beneficial. These suggestions are concrete, actionable, and directly address the identified weaknesses. They are also realistic in scope and can be implemented within the framework of a review paper.

❓ Questions

Several key questions arise from my analysis of this paper, focusing on areas where further clarification and exploration are needed. Firstly, regarding data quality and availability, how do the authors propose to address the challenges related to obtaining high-quality, labeled datasets for training AI models in rock mechanics? Are there any specific recommendations for creating standardized datasets that can be used to train and validate AI models across different geological settings? This is crucial because the reliability of AI models heavily depends on the quality and diversity of the training data. Secondly, concerning model generalization, what are the authors' thoughts on improving the generalization capabilities of AI models in rock mechanics? How can we ensure that models trained on specific datasets can be effectively applied to different geological settings and rock types? This is a critical issue because the variability of geological materials can significantly impact the performance of AI models. Thirdly, regarding the interpretability of AI models, the paper mentions the importance of interpretability, but what specific methods or techniques do the authors suggest for improving the interpretability of complex AI models in rock mechanics? How can we ensure that these models are not only accurate but also provide insights into the underlying physical processes? This is essential for building trust and confidence in AI applications in this field. Fourthly, regarding interdisciplinary collaboration, how can interdisciplinary collaboration be fostered to advance the field of intelligent rock mechanics? What specific roles do different disciplines, such as geology, computer science, and engineering, play in this collaboration? This is important because the integration of AI into rock mechanics requires expertise from multiple disciplines. Finally, regarding future directions, what are the most promising future directions for the application of AI in rock mechanics? Are there any specific areas or challenges that the authors believe should be prioritized in future research? This is crucial for guiding future research efforts and ensuring that AI is used effectively in this field. These questions target core methodological choices and seek clarification of critical assumptions, aiming to address key uncertainties and further the development of AI in rock mechanics.

📊 Scores

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

AI Review from ZGCA

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

📋 AI Review from ZGCA will be automatically processed

📋 Summary

This paper reviews the emergence of intelligent rock mechanics at the intersection of AI and classical rock mechanics. It surveys methods and applications from early neural networks and SVMs to modern deep learning, generative models, and especially physics-informed neural networks (PINNs) and related hybrids. Section 2 covers data-driven estimation of mechanical properties (UCS, elastic moduli, velocities) using ANNs, ANFIS, SVMs, ensembles, Bayesian models, and physics-aware approaches. Section 3 reviews image-based 3D microstructure reconstruction (from simulated annealing to VAEs/GANs) and fracture detection across 2D images and 3D point clouds, including transfer learning and hybrid deterministic–stochastic methods (DSIM). Section 4 surveys AI-assisted constitutive modeling (BP/LSTM/TCN hybrids, ensemble/probabilistic regressors) and physics-informed PDE solvers (PINN variants, PIRBN, PI-TCN, poromechanics PINNs, and MFGNNs) with integration into FEM/DEM settings. Section 6 synthesizes applications in rock mass classification, rockburst/geohazard prediction, tunneling/boring operations, slope stability, and other emerging use cases, highlighting ensembles, time-series models, and physics-aware learners. Section 7 outlines challenges (data scarcity, generalization, interpretability, computational efficiency, workflow integration, and maintenance/reproducibility) and future directions (hybrid physics-data models, digital twins, spatial reasoning, and potential LLM utilities).

✅ Strengths

  • Broad and up-to-date synthesis across core subareas: property estimation (Sec. 2), imaging and fracture detection (Sec. 3), constitutive and PDE surrogates (Sec. 4), and multiple engineering applications (Sec. 6).
  • Clear thematic emphasis on physics-informed and hybrid AI for improved interpretability and scientific grounding (Abstract; Sec. 1: PINNs; Sec. 4: PINN variants; Sec. 7: trust and XAI).
  • Rich set of references, including recent work (2023–2025) on VAEs/GANs for microstructures, PI-TCN, PIRBN, multi-fidelity GNNs for FEM, and application-focused ensembles and LSTMs.
  • Practical orientation in Section 6 with concrete application areas (rock mass classification, rockburst, tunneling operations, slope stability) and discussion of data handling (e.g., k-fold CV; class imbalance mitigation such as KM-SMOTE in Sec. 6.2).
  • Challenges and future directions (Sec. 7) accurately diagnose data scarcity, interpretability, workflow integration, and model maintenance/reproducibility as central issues for adoption.

❌ Weaknesses

  • Critical omission: Section 5, "MODEL GENERALIZATION IN ROCK MECHANICS AI," is empty despite generalization being a repeatedly emphasized challenge (Abstract; Sec. 1; Sec. 7). This substantially undermines the paper’s practical guidance and core claim of enhancing trust and robustness.
  • Lack of a systematic review protocol: no inclusion/exclusion criteria, search strategy, time window, or taxonomy are provided, making coverage completeness and potential biases unclear.
  • The bibliometric overview of “17 journals” (Sec. 1) is asserted without presenting methodology or quantitative results (e.g., plots, counts, temporal trends), limiting its evidentiary value.
  • Proposed conceptual framing (e.g., a Turing-test-inspired categorization of AI in rock mechanics; Sec. 1) is mentioned but not developed into a usable taxonomy or framework.
  • Insufficient synthesis into comparative guidance: no tables summarizing tasks/datasets/metrics, no benchmarking recommendations, and limited actionable protocols for practitioners (e.g., cross-site validation, OOD detection, domain adaptation, UQ).
  • Scope at times blends rock mechanics with adjacent soils/geotechnical topics without clarifying boundaries, potentially diluting focus (e.g., resilient modulus of treated soils, landslide susceptibility).
  • Minor clarity/organization issues: very long sentences and occasional formatting breaks (e.g., split references) reduce readability.

❓ Questions

  • Section 5 is empty. Please provide a substantive treatment of generalization tailored to rock mechanics, including: (a) recommended validation protocols (e.g., cross-site/cross-lithology holdout, temporal splits for monitoring data), (b) strategies for domain shift (transfer learning, domain adaptation, domain generalization), (c) OOD detection/calibration, and (d) uncertainty quantification (aleatoric/epistemic) and how to report it for high-consequence decisions.
  • Can you formalize a taxonomy that organizes the surveyed methods by task (property estimation, imaging/reconstruction, fracture detection, constitutive surrogates, PDE solvers, operations) and by data modality (lab tests, well logs, micro-CT, point clouds, monitoring time series), and map typical data regimes and appropriate models to each?
  • Please detail the bibliometric analysis: data sources, search strings, time span, inclusion/exclusion criteria, and present quantitative summaries (e.g., yearly publication counts, venue distribution, keyword co-occurrence).
  • The Turing-test-inspired framework mentioned in Sec. 1 is intriguing. What are the concrete levels/criteria, and how would they guide model selection or evaluation for rock mechanics tasks?
  • For physics-informed approaches (Sec. 4), could you provide guidance on when PINNs and variants are preferable to classical solvers or to data-only surrogates, including computational cost, convergence considerations, and hybrid coupling patterns (e.g., Gauss-point RVEs, variational PINNs)?
  • Can you propose a minimal set of standardized benchmarks and datasets for each task family (e.g., UCS/Udacity-type property datasets with clear train/test splits, fracture segmentation micro-CT/field point-cloud datasets, canonical poromechanics/PDE problems), along with agreed-upon metrics?
  • Several applications report accuracy gains (e.g., Sec. 6.2, 6.3, 6.4). Could you add a consolidated table that enumerates datasets, train/test protocols, metrics, and baseline comparisons so that readers can assess comparability and reproducibility?
  • Sec. 7 references Appendix A.1/A.3 (data and LLMs). Please include those appendices or summarize them in the main text, especially examples of LLM-enabled workflows (e.g., code generation, report parsing) that are safe, verifiable, and aligned with engineering QA/QC.
  • Can you discuss data governance and sharing (IP, confidentiality, safety) for subsurface and project monitoring data, and suggest mechanisms for anonymization and federated learning to alleviate data scarcity?
  • In the applications sections, what are the recommended uncertainty reporting formats and decision thresholds suitable for engineering practice (e.g., prediction intervals for deformation, probability of rockburst class, safety-factor distributions)?

⚠️ Limitations

  • Generalization under domain shift is the principal limitation not yet treated in depth: cross-site variability, lithological diversity, stress regime changes, and evolving monitoring conditions can degrade ML performance; protocols and methods need to be articulated (currently missing in Sec. 5).
  • Data scarcity and heterogeneity: fragmented, site-specific datasets impede robust training and external validation (Sec. 1; Sec. 7).
  • Interpretability/trust: even physics-aware models require careful explanation and uncertainty quantification for high-consequence decisions (Sec. 7).
  • Computational constraints: training PINNs/3D models can be expensive; practical guidance on resource requirements and model compression is needed (Sec. 7).
  • Workflow integration and maintenance: versioning, periodic retraining, and QA/QC procedures are essential yet under-specified (Sec. 7).
  • Potential negative impacts: over-trust in models without proper uncertainty reporting could lead to unsafe design or operations; biases in site-specific datasets may yield misleading inferences when transferred elsewhere; automated image-based assessments could miss critical defect classes if OOD conditions are not detected.

🖼️ Image Evaluation

Cross‑Modal Consistency: 32/50

Textual Logical Soundness: 18/30

Visual Aesthetics & Clarity: 10/20

Overall Score: 60/100

Detailed Evaluation (≤500 words):

1. Cross‑Modal Consistency

• Visual ground truth (image‑first):

– Figure 1: Single-pane framework diagram (blue/yellow boxes, arrows). Text boxes: geology, in‑situ stress, behavior/modeling, monitoring, design; very small fonts.

– Figure 2: (a) Left schematic: “domain subdivision using subgraphs,” “message passing.” (b) Right: convergence plots (error vs NDOF), example FEA vs MFGNN fields; tiny legends/axes.

– Figure 3: (a) Geological section with labeled rockburst zones. (b) Bar/line chart of daily microseismic events and S‑value over time (multi‑color). (c) Bar/line chart of daily cumulative energy and incidence (%).

• Major 1: Unreferenced/uncaptioned Figure 3 (a–c) appears; no figure number or mention in Sec. 6.2. Evidence: Three panes labeled “(a)(b)(c)” with no accompanying caption in text.

• Major 2: Title/content mismatch for Figure 1. Caption states “Collaborative development…,” image reads “Artificial intelligence technology framework…”. Evidence: “Figure 1: Collaborative development…” vs embedded banner “Artificial intelligence technology framework…”.

• Minor 1: Figure 2 appears to have two sub‑panes but the caption and text do not identify (a)/(b). Evidence: “Figure 2: Machine learning methods for solving FEM problems(Black & Najafi, 2022)”.

• Minor 2: Figure 1 is first cited in Intro but placed far from first mention. Evidence: Figure is separated by multiple paragraphs/sections.

2. Text Logic

• Major 1: Section 5 (“MODEL GENERALIZATION IN ROCK MECHANICS AI”) is empty, breaking the promised narrative. Evidence: Section heading present without content between Sec. 4 and Sec. 6.

• Minor 1: Broken line/citation in Intro around “Li et al., [newline] 2025”. Evidence: “(Li et al., [line break] 2025)”.

• Minor 2: Occasional typographic inconsistencies in captions (missing space before parenthesis). Evidence: “problems(Black & Najafi, 2022)”.

3. Figure Quality

• Major 1: Illegible at print size. Critical labels in Figure 1 (module names, arrows) cannot be read. Evidence: Framework boxes rendered with tiny fonts at ≈270×510 px.

• Major 2: Figure 3 (a–c) axes, legends, and values are too small to read, blocking comprehension. Evidence: Low‑resolution panels with dense annotations.

• Minor 1: Figure 2’s axes/legends are barely legible; sub‑pane labels absent. Evidence: Small tick labels in convergence plot and field maps.

Key strengths:

  • Comprehensive, current survey with strong coverage across properties, imaging, constitutive AI, and applications.
  • Good triangulation of methods (classical ML, DL, PINNs, GNNs) with domain problems.
  • Citations are broad and up‑to‑date.

Key weaknesses:

  • Missing Section 5 content.
  • Figure problems: illegibility, mislabeled/mismatched titles, and an uncaptioned 3‑pane figure.
  • Limited figure‑based evidence supporting key claims; captions need richer, stand‑alone explanations.

Actionable recommendations:

  • Fill Section 5 with definitions, benchmarks, and UQ/generalization strategies.
  • Re‑export all figures at publication resolution; ensure color‑blind‑safe palettes and readable fonts.
  • Align captions/titles with image content; number and reference Figure 3 explicitly; label sub‑panes (a/b/c) in captions and text.
  • For each quantitative visual, add clear axes, units, legends, and a one‑sentence takeaway to pass the “figure‑alone” test.

📊 Scores

Originality:2
Quality:2
Clarity:2
Significance:2
Soundness:2
Presentation:2
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 provides a comprehensive review of the integration of artificial intelligence (AI) and machine learning (ML) techniques into the field of rock mechanics. It chronologically outlines the evolution of AI methodologies, starting from early methods like backpropagation and support vector machines to more recent advancements such as deep learning frameworks, including convolutional neural networks and transformer architectures. The authors highlight the growing integration of physics-informed neural networks (PINNs) and graph-based learning, which aim to bridge data-driven inference with physical interpretability. A significant portion of the paper is dedicated to showcasing the diverse applications of these AI techniques in rock engineering, including data-driven estimation of rock properties, image-based modeling and fracture detection, AI-assisted constitutive modeling, and simulation. The paper also touches upon the challenges and future directions in the field, emphasizing the need for standardized datasets, interdisciplinary collaboration, and the development of transparent and reproducible AI workflows. While the paper provides a valuable overview of the current state of AI in rock mechanics, it falls short in critically evaluating the limitations of these methods and their practical implementation in real-world scenarios. The paper's focus on listing applications without a deeper analysis of their impact and the challenges they address makes it read more like a catalog than a critical review. Additionally, the paper lacks a thorough discussion on the practical aspects of implementing AI in rock engineering, such as the specific steps involved in data collection, model training, and validation in real-world projects. The absence of a detailed comparison with traditional methods and the limited discussion on the limitations of AI in this field further weaken the paper's overall contribution. Despite these limitations, the paper serves as a useful starting point for researchers interested in the intersection of AI and rock mechanics, providing a broad overview of the field and highlighting the potential of AI in advancing rock engineering.

✅ Strengths

One of the paper's notable strengths is its comprehensive and chronological overview of the evolution of AI methodologies in rock mechanics. The authors effectively trace the development of AI techniques from early methods like backpropagation and support vector machines to more recent advancements such as deep learning frameworks, including convolutional neural networks and transformer architectures. This historical perspective is valuable for understanding the progression of AI applications in the field. The paper also provides a detailed account of the diverse applications of AI in rock engineering, covering areas such as data-driven estimation of rock properties, image-based modeling and fracture detection, AI-assisted constitutive modeling, and simulation. The authors present a wide range of studies, which demonstrates the breadth of AI's impact on rock mechanics. The inclusion of recent advancements like physics-informed neural networks (PINNs) and graph-based learning is particularly commendable, as it highlights the growing trend towards integrating physical interpretability with data-driven approaches. The paper's emphasis on the potential of AI to transform rock mechanics from an empirical discipline to a data-driven, intelligence-enabled science is a strong and forward-looking statement. The authors also acknowledge the importance of interdisciplinary collaboration and the need for standardized datasets, which are crucial for the advancement of the field. Overall, the paper's strengths lie in its broad coverage of AI applications in rock mechanics and its optimistic outlook on the future of the field.

❌ Weaknesses

Despite its comprehensive overview, the paper has several significant weaknesses that undermine its overall contribution. Firstly, the paper's structure and writing style are more akin to a catalog of applications and methods rather than a critical review. The 'INTRODUCTION' section, for instance, reads more like a chronological listing of AI advancements and their applications in rock mechanics, lacking a clear articulation of the paper's scope, objectives, and key contributions. This makes it difficult for readers to grasp the core message and the paper's unique value. The 'REVIEW' section follows a similar pattern, systematically listing various AI methods and their applications without providing a critical analysis of the limitations or challenges associated with each method. For example, while the paper mentions the use of convolutional neural networks (CNNs) for microseismic event localization, it does not delve into the specific challenges of applying CNNs to this task, such as the need for large, labeled datasets and the difficulty in interpreting the results in a physically meaningful way. This lack of critical evaluation is a recurring issue throughout the paper, making it less insightful and more of a descriptive overview. Secondly, the paper's claim of being a 'comprehensive review' is overstated. The paper does not provide a detailed discussion of the practical aspects of implementing AI in rock engineering, such as the specific steps involved in data collection, model training, and validation in real-world projects. The 'CHALLENGES AND FUTURE DISCUSSIONS' section mentions data limitations, model interpretability, computational efficiency, integration with domain workflows, and maintenance of AI systems, but it does not offer concrete solutions or best practices for addressing these challenges. This omission is particularly problematic, as practical implementation is a crucial aspect of the field's development. Thirdly, the paper's discussion of the limitations of AI in rock mechanics is insufficient. While the authors acknowledge issues like data scarcity and model interpretability, they do not provide a balanced perspective on the potential drawbacks and challenges of using AI in this field. For instance, the paper does not address the 'black box' nature of many AI models, the risk of overfitting, or the ethical considerations associated with AI-driven decision-making in rock engineering. The absence of a dedicated section or a more thorough discussion on these limitations weakens the paper's credibility and makes it less useful for researchers and practitioners. Lastly, the paper's contribution is limited by its lack of novel insights or a unique perspective. The paper primarily summarizes existing research without offering new methodologies, frameworks, or critical evaluations that advance the field. The 'CONCLUSION' section reiterates the potential of intelligent rock mechanics but does not provide a compelling argument for the paper's originality or its impact on the broader scientific community. These weaknesses collectively suggest that the paper, while informative, falls short of being a comprehensive and critical review of AI in rock mechanics. The lack of critical analysis, practical guidance, and a balanced discussion of limitations are particularly concerning and need to be addressed to enhance the paper's value.

💡 Suggestions

To enhance the paper's contribution and address its limitations, several concrete and actionable improvements are recommended. Firstly, the paper should adopt a more critical and analytical approach, moving beyond a simple cataloging of methods and applications. Each section should not only describe the AI techniques used but also provide a detailed analysis of their limitations, challenges, and the specific contexts in which they are most effective. For example, when discussing the use of convolutional neural networks (CNNs) for microseismic event localization, the paper should delve into the specific challenges of applying CNNs to this task, such as the need for large, labeled datasets, the difficulty in interpreting the results in a physically meaningful way, and the potential for overfitting. Similarly, the discussion of physics-informed neural networks (PINNs) should include an analysis of the challenges in formulating the governing equations and the computational cost associated with training these models. This would provide a more nuanced and insightful perspective on the current state of AI in rock mechanics. Secondly, the paper should include a dedicated section on the practical aspects of implementing AI in rock engineering. This section should provide concrete guidance on data collection, model training, and validation in real-world projects. The authors should discuss the specific steps involved in preparing data for AI models, the computational resources required for training, and the methods for validating the models in practical scenarios. For instance, the paper could explore the use of transfer learning to address data scarcity, discuss the importance of feature engineering for improving model interpretability, and provide case studies that demonstrate the successful implementation of AI in rock engineering projects. This would make the paper more useful for practitioners and researchers looking to apply AI in their work. Thirdly, the paper should provide a more balanced discussion of the limitations of AI in rock mechanics. While the paper acknowledges issues like data scarcity and model interpretability, it should also address the 'black box' nature of many AI models, the risk of overfitting, and the ethical considerations associated with AI-driven decision-making in rock engineering. The authors should discuss the potential for AI models to perpetuate biases present in the training data and the need for transparency and accountability in AI-driven rock engineering solutions. This would provide a more comprehensive and realistic view of the current state of AI in rock mechanics and help to guide future research in this area. Finally, the paper should clearly articulate its unique contribution and avoid making overly broad claims. Instead of aiming to be a comprehensive review of all AI applications in rock mechanics, the paper could focus on a specific theme or challenge within the field, such as the use of AI for real-time monitoring and prediction, or the integration of AI with physics-based models. This would allow the paper to provide a more in-depth analysis of a specific area and make a more significant contribution to the field. The paper should also include a forward-looking perspective, discussing the potential impact of AI on the future of rock mechanics and identifying key areas for future research. This would help to position the paper as a valuable contribution to the field and provide a roadmap for future research in this area.

❓ Questions

1. What specific criteria were used to select the studies and applications included in the review? How did the authors ensure that the selected studies represent the most significant and impactful work in the field of AI in rock mechanics?

2. Could the authors provide a more detailed discussion on the practical challenges of implementing AI models in real-world rock engineering projects? For example, what are the specific steps involved in data collection, model training, and validation, and how can these steps be optimized for practical use?

3. How do the AI models discussed in the paper handle uncertainty and variability in geological data? What methods are used to quantify and mitigate these issues, and what are the limitations of these methods?

4. What are the ethical considerations associated with the use of AI in rock engineering, particularly in high-consequence decision-making scenarios? How can these ethical concerns be addressed to ensure responsible and transparent AI-driven solutions?

5. Could the authors elaborate on the integration of AI models with existing rock mechanics software and workflows? What are the current limitations and challenges in this integration, and what steps can be taken to improve it?

6. How do the AI models discussed in the paper compare to traditional methods in terms of accuracy, efficiency, and interpretability? What are the specific scenarios where AI models outperform traditional methods, and where do they fall short?

7. What are the key areas for future research in the field of AI in rock mechanics? How can the limitations of current AI models be addressed, and what new methodologies or frameworks are needed to advance the field?

📊 Scores

Soundness:2.5
Presentation:2.75
Contribution:2.0
Rating: 3.5

Keywords

Click the button to extract keywords

Insights

Click the button to extract insights
Version 1 ⚠️ Not latest
Citation Tools

📝 Cite This Paper