📋 AI Review from DeepReviewer will be automatically processed
📋 AI Review from ZGCA will be automatically processed
This paper is a broad review of intelligent rock mechanics at the intersection of AI and geomechanics. It traces the evolution from early ANN/SVM work to modern deep learning, covering: (i) data-driven estimation of rock properties (UCS, elastic moduli, velocities) using ANN/SVM/ensembles and physics-aware hybrids; (ii) image-based modeling and fracture detection, including deep generative 3D reconstructions and CNN/point cloud methods for crack segmentation and DFN extraction; (iii) AI-assisted constitutive modeling and simulation, with neural constitutive laws, hybrid FEM-DEM surrogates, and physics-informed methods (PINNs and variants) for PDEs and multiphysics; and (iv) applications in rock engineering, from rock mass classification to rockburst prediction, tunneling/boring operations, slope stability, and related emerging use cases. The review highlights ongoing challenges—data scarcity, generalization, interpretability—and advocates for physics-informed and hybrid models (notably PINNs) and for standardized datasets and reproducible workflows.
Cross‑Modal Consistency: 26/50
Textual Logical Soundness: 23/30
Visual Aesthetics & Clarity: 10/20
Overall Score: 59/100
Detailed Evaluation (≤500 words):
Visual ground truth (image‑first):
• Figure 1: Large flowchart “AI technology framework for rock mechanics and rock engineering”; many small text boxes, arrows, category bands.
• Figure 2 (rockburst case): (a) Geological cross‑section with lithology/structure legend and tunnel alignment; (b) Chainage bar showing excavated/monitored segments and advance arrow; (c) Tunnel cross‑section (≈7.2 m×6.2 m) with axes; (d) Face photo with dashed contour, “monzogranite,” “structural planes,” “rockburst area”; (e) LSTM diagram predicting “Number of events/Energy/Apparent volume”; (f) LSTM diagram predicting intensity classes (None→Extremely intense); (g) Grid of “Warning results and release time vs blasting cycle.”
Figure‑level synopsis: Field layout and lithology → monitoring scheme → LSTM architectures → warning grid; a workflow from site context to prediction and alerting.
• Figure 3 (tunneling PINN): (a) MLP with physics/data/coordinate losses; (b) TBM, Plaxis model and sampling grid with colormap; (c) 3D surface of predicted ground deformation (GD); (d) Heatmap of GD.
Figure‑level synopsis: Physics‑informed parameter inversion and deformation prediction, with sampling layout and predicted fields.
1. Cross‑Modal Consistency
• Major 1: Multi‑panel figures (Fig. 2, Fig. 3) appear without captions/numbering in the text; several panels are never referenced, blocking verification. Evidence: Unlabeled panels (a–f, g) and (a–d) shown after Sec. 5.2/5.3 with no in‑text calls.
• Major 2: Bibliometric claims lack supporting visual/table; no trend plot, keyword network, or counts provided. Evidence: “The results highlight…‘machine learning’ now surpasses ‘artificial intelligence’ in frequency” (Intro, para 2).
• Minor 1: Fig. 2 uses undefined symbols/classes (e.g., PN…PEX, GD), not introduced in prose. Evidence: Intensity outputs PN…PEX on Fig. 2(f).
• Minor 2: Fig. 1 is cited once (“as shown in Figure 1”) but its internal labels are not mapped to section terminology. Evidence: “as shown in Figure 1” (Intro, para 3).
2. Text Logic
• Major 1: Bibliometric methodology/results insufficiently specified (journals listed, but no sample size per venue, time window, or extraction protocol to support conclusions). Evidence: “Using…terms, all articles published before 2025 were systematically retrieved…” (Intro, para 2).
• Minor 1: Occasional broken words/formatting (e.g., “al ternative,” hyphenation) mildly distract clarity. Evidence: Sec. 3, “al ternative.”
• Minor 2: Some landmark claims (e.g., PINN efficiency, LSTM superiority) are asserted via citations without summarizing comparable baselines or metrics in text. Evidence: Sec. 4, “with improved accuracy‑efficiency trade‑offs.”
3. Figure Quality
• Major 1: Fig. 1 text and many labels in Fig. 2(a,b,c) are illegible at print size; legends/axes unreadable, hindering comprehension. Evidence: Tiny fonts in Fig. 1 boxes and Fig. 2(a–c) chainage/elevation.
• Minor 1: Color palettes in deformation maps (Fig. 3c–d) lack unit/scale description in the manuscript text. Evidence: Colorbar shown, units unspecified.
Key strengths:
• Broad, up‑to‑date survey spanning properties, imaging, constitutive ML, and applications.
• Good coverage of physics‑informed and hybrid trends with diverse citations.
Key weaknesses:
• Cross‑modal gaps: uncaptioned/under‑referenced figures; illegible panels.
• Bibliometric claims lack presented evidence; several assertions not tied to quantitative visuals.
• Figures fail the “figure‑alone” test; add captions, define symbols/units, and enlarge fonts.
Recommendations:
• Add full captions, numbering, and explicit in‑text references for Fig. 2–3; define all symbols/units.
• Provide bibliometric plots/tables (yearly counts, keyword co‑occurrence) supporting Intro claims.
• For case figures, report datasets, metrics, and baselines directly on the panels or captions.
📋 AI Review from SafeReviewer will be automatically processed
This paper presents a comprehensive review of the application of artificial intelligence (AI) and machine learning (ML) techniques in the field of rock mechanics. The authors meticulously trace the evolution of AI methodologies, from early methods like backpropagation and support vector machines to modern deep learning frameworks, including convolutional and transformer architectures. The core contribution of this work lies in its synthesis of the diverse applications of these AI techniques across various aspects of rock mechanics, ranging from microstructure reconstruction and mechanical parameter estimation to constitutive modeling and real-time hazard prediction. The paper highlights the increasing integration of physics-informed neural networks (PINNs) and graph-based learning, which aim to bridge the gap between data-driven inference and physical interpretability. A significant emphasis is placed on the potential of large language models (LLMs) to automate code generation and enhance decision support in geotechnical analysis. The authors acknowledge the challenges that persist, such as ensuring data quality, achieving model generalization, and enhancing interpretability. They advocate for standardized datasets, interdisciplinary collaboration, and transparent, reproducible AI workflows to address these issues. The paper concludes with a forward-looking perspective, envisioning the development of next-generation intelligent frameworks that couple physical knowledge, spatial reasoning, and adaptive learning, ultimately propelling rock mechanics towards fully autonomous, intelligent systems. The paper's strength lies in its broad coverage of the field and its ability to connect various AI techniques to specific problems in rock mechanics. However, the paper's depth of analysis and the novelty of its insights are areas that could be improved. The paper's focus on applications, while valuable, sometimes overshadows the fundamental research challenges in adapting AI to the unique complexities of rock mechanics. The paper's lack of a dedicated limitations section and its somewhat superficial engagement with the theoretical underpinnings of the discussed AI methods also represent areas for improvement. Despite these limitations, the paper provides a useful overview of the current state of AI in rock mechanics and identifies key areas for future research.
This paper demonstrates several notable strengths. Firstly, the authors have successfully provided a comprehensive overview of the application of AI and ML in rock mechanics. The paper effectively traces the historical development of AI methodologies and their integration into the field, showcasing a wide range of techniques from basic machine learning to advanced deep learning architectures. This broad coverage is a significant strength, as it provides a valuable resource for researchers and practitioners interested in the intersection of AI and rock mechanics. The paper's ability to connect specific AI techniques to concrete problems in rock mechanics is another key strength. For instance, the discussion of CNNs for microseismic event localization and LSTMs for modeling rheological behavior provides clear examples of how these techniques are being applied in practice. Furthermore, the paper highlights the growing importance of physics-informed neural networks (PINNs) and graph-based learning, which are crucial for bridging the gap between data-driven models and physical understanding. The emphasis on the potential of large language models (LLMs) for automating code generation and decision support also points towards promising future directions. The paper's forward-looking perspective, envisioning the development of intelligent frameworks that couple physical knowledge, spatial reasoning, and adaptive learning, is a valuable contribution. Finally, the paper's identification of key challenges, such as data quality, model generalization, and interpretability, is important for guiding future research in this area. The paper's ability to synthesize a large body of work and present it in a clear and organized manner is a testament to the authors' efforts. The inclusion of a detailed list of references also enhances the paper's credibility and provides a valuable starting point for researchers interested in delving deeper into specific topics. The paper's focus on the practical applications of AI in rock mechanics is also a strength, as it highlights the potential of these techniques to solve real-world problems in geotechnical engineering.
Despite its strengths, this paper exhibits several weaknesses that warrant careful consideration. A primary concern is the paper's lack of depth in its analysis of the fundamental research challenges associated with applying AI to rock mechanics. While the paper provides a broad overview of various AI techniques and their applications, it does not delve deeply into the specific limitations of these techniques when applied to the unique complexities of geological materials. For instance, the paper mentions the challenges of anisotropy, discontinuities, and multiphysics coupling, but it does not thoroughly explore how these factors impact the performance of different AI models. The paper's focus on applications sometimes overshadows the need for a more critical examination of the underlying research challenges. This is evident in the paper's structure, which is organized around applications rather than a deep analysis of the limitations of AI in this context. The paper's lack of a dedicated limitations section further exacerbates this issue. The paper does not explicitly discuss the limitations of the review itself, such as potential biases in the selection of papers or the scope of the review. This omission weakens the paper's overall credibility and transparency. The paper's engagement with the theoretical underpinnings of the discussed AI methods is also somewhat superficial. While the paper mentions various AI techniques, it does not provide a detailed explanation of their theoretical foundations or the specific adaptations required for rock mechanics. For example, the paper mentions CNNs, LSTMs, and GANs, but it does not delve into the specific modifications or considerations needed to apply these methods effectively to the unique challenges of rock mechanics data. This lack of theoretical depth limits the paper's ability to provide a comprehensive understanding of the subject matter. The paper's insights, while valuable, are not particularly novel. The paper primarily synthesizes existing research rather than presenting groundbreaking new findings or methodologies. While the paper's synthesis is useful, it does not push the boundaries of knowledge in this field. The paper's discussion of the limitations of AI in rock mechanics is also somewhat limited. While the paper mentions data quality, model generalization, and interpretability as challenges, it does not fully explore the complexities of these issues in the context of rock mechanics. For example, the paper does not discuss the challenges of obtaining high-quality, representative data for training AI models in rock mechanics, or the difficulties of interpreting the results of complex AI models in a physically meaningful way. The paper's lack of a detailed discussion of the validation of AI models is also a weakness. While the paper mentions the use of validation datasets, it does not provide a thorough discussion of the specific metrics used for validation or the challenges of validating AI models in the context of rock mechanics. The paper's discussion of the integration of AI into engineering workflows is also somewhat limited. While the paper mentions the potential of AI to transform rock mechanics, it does not provide a detailed discussion of the practical challenges of integrating AI models into existing engineering practices. Finally, the paper's discussion of the ethical implications of using AI in rock mechanics is absent. The paper does not address the potential ethical concerns associated with the use of AI in this field, such as the potential for bias in AI models or the impact of AI on employment in the geotechnical engineering sector. These weaknesses, while not invalidating the paper's contributions, highlight areas where the paper could be improved to provide a more comprehensive and insightful analysis of the role of AI in rock mechanics. My confidence in these identified weaknesses is high, as they are supported by direct examination of the paper's content and structure.
To address the identified weaknesses, I propose several concrete and actionable suggestions. Firstly, the paper should include a more in-depth analysis of the fundamental research challenges associated with applying AI to rock mechanics. This could involve a more detailed discussion of the limitations of specific AI techniques when applied to the unique complexities of geological materials, such as anisotropy, discontinuities, and multiphysics coupling. The paper should also explore the specific adaptations or modifications required to apply these techniques effectively to rock mechanics data. This could involve a more detailed discussion of the theoretical underpinnings of the discussed AI methods and the specific considerations needed for their application in this context. Secondly, the paper should include a dedicated limitations section that explicitly discusses the limitations of the review itself. This could involve a discussion of potential biases in the selection of papers, the scope of the review, and the limitations of the authors' own perspectives. This would enhance the paper's credibility and transparency. Thirdly, the paper should provide a more detailed discussion of the validation of AI models. This could involve a more thorough discussion of the specific metrics used for validation, the challenges of validating AI models in the context of rock mechanics, and the importance of using appropriate validation datasets. The paper should also discuss the need for uncertainty quantification in AI model predictions and the importance of considering the limitations of AI models when making engineering decisions. Fourthly, the paper should include a more detailed discussion of the integration of AI into engineering workflows. This could involve a discussion of the practical challenges of integrating AI models into existing engineering practices, the need for user-friendly tools and interfaces, and the importance of training engineers to use AI models effectively. The paper should also discuss the potential impact of AI on the roles and responsibilities of engineers in the geotechnical field. Fifthly, the paper should include a discussion of the ethical implications of using AI in rock mechanics. This could involve a discussion of the potential for bias in AI models, the impact of AI on employment in the geotechnical engineering sector, and the need for responsible development and deployment of AI in this field. Sixthly, the paper should consider reorganizing its structure to focus more on the fundamental research challenges and less on the applications of AI in rock mechanics. This could involve a more thematic organization, where the paper discusses the challenges of data scarcity, model interpretability, and the integration of physical principles into AI models, rather than organizing the paper around specific applications. Finally, the paper should strive to provide more novel insights and contributions. This could involve identifying gaps in the current literature, proposing new methodologies, or presenting groundbreaking findings. These suggestions, if implemented, would significantly enhance the paper's depth, rigor, and overall contribution to the field of rock mechanics and AI. These changes are within the scope of a review paper and would provide a more comprehensive and insightful analysis of the role of AI in rock mechanics.
Several key uncertainties and methodological choices in this paper warrant further clarification. Firstly, given the paper's focus on the application of AI in rock mechanics, I am curious about the specific adaptations or modifications that are typically required to apply standard AI techniques to the unique challenges of rock mechanics data. For instance, how are CNNs adapted to handle the anisotropy and discontinuities inherent in rock materials? What specific considerations are needed when applying LSTMs to model the time-dependent behavior of rocks? These questions are crucial for understanding the practical implications of using AI in this field. Secondly, the paper mentions the use of physics-informed neural networks (PINNs) and graph-based learning. I would like to understand in more detail how these techniques are being used to bridge the gap between data-driven inference and physical interpretability. What are the specific advantages and limitations of these techniques in the context of rock mechanics? How do they compare to more traditional AI techniques? These questions are important for assessing the potential of these emerging techniques. Thirdly, the paper highlights the challenges of data quality, model generalization, and interpretability. I would like to understand in more detail how these challenges are being addressed in the current literature. What specific strategies are being used to improve data quality? How are researchers working to improve the generalization capabilities of AI models? What techniques are being used to enhance the interpretability of complex AI models? These questions are crucial for understanding the current state of AI in rock mechanics and identifying areas for future research. Fourthly, the paper mentions the potential of large language models (LLMs) for automating code generation and decision support. I would like to understand in more detail how these models are being used in the context of rock mechanics. What are the specific tasks that LLMs are being used for? What are the limitations of these models in this context? These questions are important for understanding the potential of LLMs in this field. Finally, given the paper's broad scope, I am curious about the specific criteria that were used to select the papers that were included in the review. What were the inclusion and exclusion criteria? How was the literature search conducted? These questions are important for assessing the comprehensiveness and representativeness of the review. Addressing these questions would provide a more complete understanding of the paper's methodology and the current state of AI in rock mechanics.