Aiming at the problem of dynamic instability of hard-brittle jointed rock surrounding in deep tunnel/roadway engineering,combining with the support concepts of"coupling rigidity with flexibility"and"ove...Aiming at the problem of dynamic instability of hard-brittle jointed rock surrounding in deep tunnel/roadway engineering,combining with the support concepts of"coupling rigidity with flexibility"and"overcoming rigidity by flexibility",the prevention and control method with"rigid-flexible coupling(R-F-C)"was put forward.Through numerical simulation calculation,the impact damage process,acoustic emission(AE)evolution characteristics,and element stress/displacement evolution characteristics of unsupported surrounding rock structure model,rigid supporting surrounding rock structure model,and"R-F-C"supporting surrounding rock structure model under horizontal bidirectional impact loading were compared and analyzed.Based on the theory of stress wave propagation,the dynamic instability catastrophe mechanism of three kinds of supporting structure models induced by horizontal bidirectional impact loading was revealed.Based on the Mohr-Coulomb strength theory,the stress discrimination methods of dynamic catastrophe of surrounding rock induced by horizontal bidirectional impact loading under three kinds of supporting structures were proposed.Combined with the above numerical simulation study,the explosion impact physical and mechanical test of"R-F-C"surrounding rock supporting plate structure was further designed and carried out.Finally,combined with the"conceptual model of ball-cliff potential energy instability",the energy driving theory and energy transformation mechanism of impact-induced rockburst under three kinds of supporting structures were discussed deeply.The research results provided a scientific basis for further promoting the effective application of"R-F-C"supporting structure in the prevention and control of dynamic instability of deep tunnel/roadway surrounding rock.展开更多
Objective:To explore the relationship between pain degree and pain catastrophe and medical coping mode in patients with chronic pain.Methods:A visual analogue score scale,medical coping style questionnaire and pain ca...Objective:To explore the relationship between pain degree and pain catastrophe and medical coping mode in patients with chronic pain.Methods:A visual analogue score scale,medical coping style questionnaire and pain catastrophe scale were used to survey 200 patients in the pain department.Results:The average scores of pain degree of patients with chronic pain were(5.97±2.29),the average score of the total score of the Pain Catastrophe Scale was(21.21±11.56),and the average scores of facing,avoidance and surrender in the Medical Response Style Questionnaire were(17.93±3.4),(16.82±2.4),and(8.87±2.83),respectively.Pain degree was positively correlated with the yield dimension in pain catastrophe and medical coping(p<0.05).The yield dimension of medical coping was positively correlated with pain catastrophe(p<0.05).Medical coping methods played a partial mediating role between pain degree and pain catastrophe,and the mediating effect accounted for 21.59%of the total effect.Conclusion:The pain level of chronic pain patients can affect the level of pain catastrophe through medical coping,and clinical medical staff should guide patients to adopt positive coping methods to promote their healthy recovery.展开更多
Following global catastrophic infrastructure loss(GCIL),traditional electricity networks would be damaged and unavailable for energy supply,necessitating alternative solutions to sustain critical services.These altern...Following global catastrophic infrastructure loss(GCIL),traditional electricity networks would be damaged and unavailable for energy supply,necessitating alternative solutions to sustain critical services.These alternative solutions would need to run without damaged infrastructure and would likely need to be located at the point of use,such as decentralized electricity generation from wood gas.This study explores the feasibility of using modified light duty vehicles to self-sustain electricity generation by producing wood chips for wood gasification.A 2004 Ford Falcon Fairmont was modified to power a woodchipper and an electrical generator.The vehicle successfully produced wood chips suitable for gasification with an energy return on investment(EROI)of 3.7 and sustained a stable output of 20 kW electrical power.Scalability analyses suggest such solutions could provide electricity to the critical water sanitation sector,equivalent to 4%of global electricity demand,if production of woodchippers was increased postcatastrophe.Future research could investigate the long-term durability of modified vehicles and alternative electricity generation,and quantify the scalability of wood gasification in GCIL scenarios.This work provides a foundation for developing resilient,decentralized energy systems to ensure the continuity of critical services during catastrophic events,leveraging existing vehicle infrastructure to enhance disaster preparedness.展开更多
During the monitoring engineering of landslides, the monitoring data of accumulated displacement are usually affected by the external factors. Therefore, the displacement curve always has step-like character, which br...During the monitoring engineering of landslides, the monitoring data of accumulated displacement are usually affected by the external factors. Therefore, the displacement curve always has step-like character, which brings some difficulties to the accurate prediction of landslides. In order to solve this problem, based on the wavelet analysis and cusp catastrophe, a new kind of analysis method is proposed in this article. First, Fourier transform method can be used to extract the frequency component of the curve of monitoring displacement. Second, the wavelet transform was adopted to inspect the breakpoints of signals, which can be used to analyze the cause of the occurrence of the step-like character in the curve of landslide monitoring. Based on the cusp catastrophe theory, a nonlinear dynamic model was established to conduct the simulation calculation of time forecasting of landslides. According to a case study of landslide, the periodical rainfall and reservoir level fluctuation are the main factors leading to the step-like changes in the curve of monitoring displacement. In addition, the results of simulation calculation are in agreement with the fact of local failure of landslides. This method can provide a new analysis way for the time prediction of landslides.展开更多
This study addresses the maneuver evasion problem for medium-to-long-range air-to-air missiles by proposing a KAN-λ-PPO-based evasion algorithm.The algorithm introduces Kolmogorov-Arnold Networks(KAN)to mitigate the ...This study addresses the maneuver evasion problem for medium-to-long-range air-to-air missiles by proposing a KAN-λ-PPO-based evasion algorithm.The algorithm introduces Kolmogorov-Arnold Networks(KAN)to mitigate the catastrophic forgetting issue of Multilayer Perceptrons(MLP)in continual learning,while incorporatingλ-return to resolve sparse reward challenges in evasion scenarios.First,we model the evasion problem withλ-return and present the KAN-λ-PPO algorithm.Subsequently,we establish game environments based on the segmented ballistic characteristics of medium and long range missiles.During training,a joint reward function is designed by combining the miss distance and positional advantages to train the agent.Experiments evaluate four dimensions:(1)Performance comparison between KAN and MLP in value function approximation;(2)Catastrophic forgetting mitigation of KAN-λ-PPO in dual-task scenarios;(3)Continual learning capabilities across multiple evasion scenarios;(4)Quantitative analysis of agent strategy evolution and positional advantages.Empirical results demonstrate that KAN improves value function approximation accuracy by an order of magnitude compared with traditional MLP architectures.In continual learning tasks,the KAN-λ-PPO scheme exhibits significant knowledge retention,achieving performance improvements of 32.7% and 8.6%over MLP baselines in Task1→2 and Task2→3 transitions,respectively.Furthermore,the learned maneuver strategies outperform High-G Barrel Rolls(HGB)and S-maneuver tactics in securing positional advantages while accomplishing evasion.展开更多
Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-h...Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-harm,long-term disability,reduced productivity,and significant societal and economic burden.Despite recent advances,detecting risk from online text remains challenging due to heterogeneous language,evolving semantics,and the sequential emergence of new datasets.Effective solutions must encode clinically meaningful cues,reason about causal relations,and adapt to new domains without forgetting prior knowledge.To address these challenges,this paper presents a Continual Neuro-Symbolic Graph Learning(CNSGL)framework that unifies symbolic reasoning,causal inference,and continual learning within a single architecture.Each post is represented as a symbolic graph linking clinically relevant tags to textual content,enriched with causal edges derived from directional Point-wise Mutual Information(PMI).A two-layer Graph Convolutional Network(GCN)encodes these graphs,and a Transformer-based attention pooler aggregates node embeddings while providing interpretable tag-level importances.Continual adaptation across datasets is achieved through the Multi-Head Freeze(MH-Freeze)strategy,which freezes a shared encoder and incrementally trains lightweight task-specific heads(small classifiers attached to the shared embedding).Experimental evaluations across six diverse mental-health datasets ranging from Reddit discourse to clinical interviews,demonstrate that MH-Freeze consistently outperforms existing continual-learning baselines in both discriminative accuracy and calibration reliability.Across six datasets,MH-Freeze achieves up to 0.925 accuracy and 0.923 F1-Score,with AUPRC≥0.934 and AUROC≥0.942,consistently surpassing all continual-learning baselines.The results confirm the framework’s ability to preserve prior knowledge,adapt to domain shifts,and maintain causal interpretability,establishing CNSGL as a promising step toward robust,explainable,and lifelong mental-health risk assessment.展开更多
Modern intelligent systems,such as autonomous vehicles and face recognition,must continuously adapt to new scenarios while preserving their ability to handle previously encountered situations.However,when neural netwo...Modern intelligent systems,such as autonomous vehicles and face recognition,must continuously adapt to new scenarios while preserving their ability to handle previously encountered situations.However,when neural networks learn new classes sequentially,they suffer from catastrophic forgetting—the tendency to lose knowledge of earlier classes.This challenge,which lies at the core of class-incremental learning,severely limits the deployment of continual learning systems in real-world applications with streaming data.Existing approaches,including rehearsalbased methods and knowledge distillation techniques,have attempted to address this issue but often struggle to effectively preserve decision boundaries and discriminative features under limited memory constraints.To overcome these limitations,we propose a support vector-guided framework for class-incremental learning.The framework integrates an enhanced feature extractor with a Support Vector Machine classifier,which generates boundary-critical support vectors to guide both replay and distillation.Building on this architecture,we design a joint feature retention strategy that combines boundary proximity with feature diversity,and a Support Vector Distillation Loss that enforces dual alignment in decision and semantic spaces.In addition,triple attention modules are incorporated into the feature extractor to enhance representation power.Extensive experiments on CIFAR-100 and Tiny-ImageNet demonstrate effective improvements.On CIFAR-100 and Tiny-ImageNet with 5 tasks,our method achieves 71.68%and 58.61%average accuracy,outperforming strong baselines by 3.34%and 2.05%.These advantages are consistently observed across different task splits,highlighting the robustness and generalization of the proposed approach.Beyond benchmark evaluations,the framework also shows potential in few-shot and resource-constrained applications such as edge computing and mobile robotics.展开更多
This study examines how native pore structures and loading conditions influencethe fracture size distribution and the predictability of catastrophic failure in rocks.Four lithologies with distinct pore characteristics...This study examines how native pore structures and loading conditions influencethe fracture size distribution and the predictability of catastrophic failure in rocks.Four lithologies with distinct pore characteristics,i.e.granite,limestone,red sandstone,and marble,were tested under uniaxial compression and Brazilian splitting.Nuclear magnetic resonance(NMR)was used to characterize pore structures,while acoustic emission(AE)monitoring captured the temporal evolution of microcracking.The relationships among pore properties,AE b-values,and failure predictability were systematically evaluated.Results show that the overall b-value is primarily controlled by native pore size rather than loading condition.Rocks with larger pores display higher b-value and greater temporal variability,whereas those with smaller pores exhibit lower and more stable b-value.To assess failure predictability,the AE count rate was incorporated into an inverse power law model.The model demonstrates higher predictive accuracy for high-porosity rocks.The average predicted failure time(t_(p))decreases monotonically with porosity:under uniaxial compression,t_(p)for granite,marble,limestone,and sandstone are 2.32,1.82,1.42,and 0.03,respectively;under Brazilian splitting,3.54,3.30,0.10,and 0.03.Among the four rock types,sandstone with the highest porosity exhibits the smallest discrepancy between predicted and actual failure time,whereas granite with the lowest porosity shows the largest.As porosity decreases,prediction accuracy progressively declines for limestone and marble.Overall,the findings indicate that native pore heterogeneity governs both fracture scaling behavior and failure predictability,and that these effects are largely independent of the loading conditions examined in this study.展开更多
2022年6月北江流域遭遇百年一遇特大洪水,对下游江心洲的形态与植被覆盖产生了显著影响。研究基于2022年多时相哨兵2号遥感影像及实测河道地形数据,结合归一化差分水体指数(Normalized Difference Water Index,NDWI)和归一化差分植被指...2022年6月北江流域遭遇百年一遇特大洪水,对下游江心洲的形态与植被覆盖产生了显著影响。研究基于2022年多时相哨兵2号遥感影像及实测河道地形数据,结合归一化差分水体指数(Normalized Difference Water Index,NDWI)和归一化差分植被指数(Normalized Difference Vegetation Index,NDVI),分析了北江下游芦苞浅段5个江心洲的平面形态、断面地形、NDVI均值及植被覆盖的年内演变特征。结果表明:洪水期间各江心洲面积与周长均显著减小,其中未建土堤的中间洲、芦苞洲和太监洲面积减少幅度更大,分别达55.4%、69.0%和66.4%。河道断面地形显示洪水后主河槽以轻微淤积为主,江心洲高程未发生显著变化。建堤江心洲(邓塘洲、四姓洲)堤内植被受洪水影响较小,NDVI均值在洪水后仍保持较高水平,而江心洲堤外及未建堤江心洲植被面积普遍减少32.5%~74.4%。研究分析了特大洪水对北江下游芦苞浅段江心洲形态及植被覆盖的差异化影响,为区域河道治理与生态修复提供了科学依据。展开更多
基金Project(2023AH051167)supported by the Natural Science Research Project of Anhui Educational Committee,ChinaProject(AHBP2024B-04)supported by the Foundation of Anhui Engineering Research Center of New Explosive Materials and Blasting Technology,China+1 种基金Project(GXZDSYS2023103)supported by the Open Fund for Anhui Key Laboratory of Mining Construction Engineering,ChinaProjects(52274071,52404155)supported by the National Natural Science Foundation of China。
文摘Aiming at the problem of dynamic instability of hard-brittle jointed rock surrounding in deep tunnel/roadway engineering,combining with the support concepts of"coupling rigidity with flexibility"and"overcoming rigidity by flexibility",the prevention and control method with"rigid-flexible coupling(R-F-C)"was put forward.Through numerical simulation calculation,the impact damage process,acoustic emission(AE)evolution characteristics,and element stress/displacement evolution characteristics of unsupported surrounding rock structure model,rigid supporting surrounding rock structure model,and"R-F-C"supporting surrounding rock structure model under horizontal bidirectional impact loading were compared and analyzed.Based on the theory of stress wave propagation,the dynamic instability catastrophe mechanism of three kinds of supporting structure models induced by horizontal bidirectional impact loading was revealed.Based on the Mohr-Coulomb strength theory,the stress discrimination methods of dynamic catastrophe of surrounding rock induced by horizontal bidirectional impact loading under three kinds of supporting structures were proposed.Combined with the above numerical simulation study,the explosion impact physical and mechanical test of"R-F-C"surrounding rock supporting plate structure was further designed and carried out.Finally,combined with the"conceptual model of ball-cliff potential energy instability",the energy driving theory and energy transformation mechanism of impact-induced rockburst under three kinds of supporting structures were discussed deeply.The research results provided a scientific basis for further promoting the effective application of"R-F-C"supporting structure in the prevention and control of dynamic instability of deep tunnel/roadway surrounding rock.
文摘Objective:To explore the relationship between pain degree and pain catastrophe and medical coping mode in patients with chronic pain.Methods:A visual analogue score scale,medical coping style questionnaire and pain catastrophe scale were used to survey 200 patients in the pain department.Results:The average scores of pain degree of patients with chronic pain were(5.97±2.29),the average score of the total score of the Pain Catastrophe Scale was(21.21±11.56),and the average scores of facing,avoidance and surrender in the Medical Response Style Questionnaire were(17.93±3.4),(16.82±2.4),and(8.87±2.83),respectively.Pain degree was positively correlated with the yield dimension in pain catastrophe and medical coping(p<0.05).The yield dimension of medical coping was positively correlated with pain catastrophe(p<0.05).Medical coping methods played a partial mediating role between pain degree and pain catastrophe,and the mediating effect accounted for 21.59%of the total effect.Conclusion:The pain level of chronic pain patients can affect the level of pain catastrophe through medical coping,and clinical medical staff should guide patients to adopt positive coping methods to promote their healthy recovery.
基金This work was funded in part by the Alliance to Feed the Earth in Disasters(ALLFED).
文摘Following global catastrophic infrastructure loss(GCIL),traditional electricity networks would be damaged and unavailable for energy supply,necessitating alternative solutions to sustain critical services.These alternative solutions would need to run without damaged infrastructure and would likely need to be located at the point of use,such as decentralized electricity generation from wood gas.This study explores the feasibility of using modified light duty vehicles to self-sustain electricity generation by producing wood chips for wood gasification.A 2004 Ford Falcon Fairmont was modified to power a woodchipper and an electrical generator.The vehicle successfully produced wood chips suitable for gasification with an energy return on investment(EROI)of 3.7 and sustained a stable output of 20 kW electrical power.Scalability analyses suggest such solutions could provide electricity to the critical water sanitation sector,equivalent to 4%of global electricity demand,if production of woodchippers was increased postcatastrophe.Future research could investigate the long-term durability of modified vehicles and alternative electricity generation,and quantify the scalability of wood gasification in GCIL scenarios.This work provides a foundation for developing resilient,decentralized energy systems to ensure the continuity of critical services during catastrophic events,leveraging existing vehicle infrastructure to enhance disaster preparedness.
基金supported by the National Natural Science Foundation of China (Nos. 40202028, 50609026)Postdoctors Foundation of China (No. 20060400256)Excellent Young Teacher Science and Technology Program of Faculty of Engi-neering, China University of Geosciences
文摘During the monitoring engineering of landslides, the monitoring data of accumulated displacement are usually affected by the external factors. Therefore, the displacement curve always has step-like character, which brings some difficulties to the accurate prediction of landslides. In order to solve this problem, based on the wavelet analysis and cusp catastrophe, a new kind of analysis method is proposed in this article. First, Fourier transform method can be used to extract the frequency component of the curve of monitoring displacement. Second, the wavelet transform was adopted to inspect the breakpoints of signals, which can be used to analyze the cause of the occurrence of the step-like character in the curve of landslide monitoring. Based on the cusp catastrophe theory, a nonlinear dynamic model was established to conduct the simulation calculation of time forecasting of landslides. According to a case study of landslide, the periodical rainfall and reservoir level fluctuation are the main factors leading to the step-like changes in the curve of monitoring displacement. In addition, the results of simulation calculation are in agreement with the fact of local failure of landslides. This method can provide a new analysis way for the time prediction of landslides.
文摘This study addresses the maneuver evasion problem for medium-to-long-range air-to-air missiles by proposing a KAN-λ-PPO-based evasion algorithm.The algorithm introduces Kolmogorov-Arnold Networks(KAN)to mitigate the catastrophic forgetting issue of Multilayer Perceptrons(MLP)in continual learning,while incorporatingλ-return to resolve sparse reward challenges in evasion scenarios.First,we model the evasion problem withλ-return and present the KAN-λ-PPO algorithm.Subsequently,we establish game environments based on the segmented ballistic characteristics of medium and long range missiles.During training,a joint reward function is designed by combining the miss distance and positional advantages to train the agent.Experiments evaluate four dimensions:(1)Performance comparison between KAN and MLP in value function approximation;(2)Catastrophic forgetting mitigation of KAN-λ-PPO in dual-task scenarios;(3)Continual learning capabilities across multiple evasion scenarios;(4)Quantitative analysis of agent strategy evolution and positional advantages.Empirical results demonstrate that KAN improves value function approximation accuracy by an order of magnitude compared with traditional MLP architectures.In continual learning tasks,the KAN-λ-PPO scheme exhibits significant knowledge retention,achieving performance improvements of 32.7% and 8.6%over MLP baselines in Task1→2 and Task2→3 transitions,respectively.Furthermore,the learned maneuver strategies outperform High-G Barrel Rolls(HGB)and S-maneuver tactics in securing positional advantages while accomplishing evasion.
基金supported by the National Research Foundation of Korea(NRF)grant funded by the Korea government(MSIT)(RS-2025-00518960)in part by the National Research Foundation of Korea(NRF)grant funded by the Korea government(MSIT)(RS-2025-00563192).
文摘Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-harm,long-term disability,reduced productivity,and significant societal and economic burden.Despite recent advances,detecting risk from online text remains challenging due to heterogeneous language,evolving semantics,and the sequential emergence of new datasets.Effective solutions must encode clinically meaningful cues,reason about causal relations,and adapt to new domains without forgetting prior knowledge.To address these challenges,this paper presents a Continual Neuro-Symbolic Graph Learning(CNSGL)framework that unifies symbolic reasoning,causal inference,and continual learning within a single architecture.Each post is represented as a symbolic graph linking clinically relevant tags to textual content,enriched with causal edges derived from directional Point-wise Mutual Information(PMI).A two-layer Graph Convolutional Network(GCN)encodes these graphs,and a Transformer-based attention pooler aggregates node embeddings while providing interpretable tag-level importances.Continual adaptation across datasets is achieved through the Multi-Head Freeze(MH-Freeze)strategy,which freezes a shared encoder and incrementally trains lightweight task-specific heads(small classifiers attached to the shared embedding).Experimental evaluations across six diverse mental-health datasets ranging from Reddit discourse to clinical interviews,demonstrate that MH-Freeze consistently outperforms existing continual-learning baselines in both discriminative accuracy and calibration reliability.Across six datasets,MH-Freeze achieves up to 0.925 accuracy and 0.923 F1-Score,with AUPRC≥0.934 and AUROC≥0.942,consistently surpassing all continual-learning baselines.The results confirm the framework’s ability to preserve prior knowledge,adapt to domain shifts,and maintain causal interpretability,establishing CNSGL as a promising step toward robust,explainable,and lifelong mental-health risk assessment.
基金supported by the Gansu Provincial Natural Science Foundation(grant number 25JRRA074)the Gansu Provincial Key R&D Science and Technology Program(grant number 24YFGA060)the National Natural Science Foundation of China(grant number 62161019).
文摘Modern intelligent systems,such as autonomous vehicles and face recognition,must continuously adapt to new scenarios while preserving their ability to handle previously encountered situations.However,when neural networks learn new classes sequentially,they suffer from catastrophic forgetting—the tendency to lose knowledge of earlier classes.This challenge,which lies at the core of class-incremental learning,severely limits the deployment of continual learning systems in real-world applications with streaming data.Existing approaches,including rehearsalbased methods and knowledge distillation techniques,have attempted to address this issue but often struggle to effectively preserve decision boundaries and discriminative features under limited memory constraints.To overcome these limitations,we propose a support vector-guided framework for class-incremental learning.The framework integrates an enhanced feature extractor with a Support Vector Machine classifier,which generates boundary-critical support vectors to guide both replay and distillation.Building on this architecture,we design a joint feature retention strategy that combines boundary proximity with feature diversity,and a Support Vector Distillation Loss that enforces dual alignment in decision and semantic spaces.In addition,triple attention modules are incorporated into the feature extractor to enhance representation power.Extensive experiments on CIFAR-100 and Tiny-ImageNet demonstrate effective improvements.On CIFAR-100 and Tiny-ImageNet with 5 tasks,our method achieves 71.68%and 58.61%average accuracy,outperforming strong baselines by 3.34%and 2.05%.These advantages are consistently observed across different task splits,highlighting the robustness and generalization of the proposed approach.Beyond benchmark evaluations,the framework also shows potential in few-shot and resource-constrained applications such as edge computing and mobile robotics.
基金supported by the National Natural Science Foundation of China(Grant No.42172316)the Major National Science and Technology Project for Deep Earth(Grant No.2024ZD100380X)the Natural Science Foundation of Hunan Province of China(2025JJ20030).
文摘This study examines how native pore structures and loading conditions influencethe fracture size distribution and the predictability of catastrophic failure in rocks.Four lithologies with distinct pore characteristics,i.e.granite,limestone,red sandstone,and marble,were tested under uniaxial compression and Brazilian splitting.Nuclear magnetic resonance(NMR)was used to characterize pore structures,while acoustic emission(AE)monitoring captured the temporal evolution of microcracking.The relationships among pore properties,AE b-values,and failure predictability were systematically evaluated.Results show that the overall b-value is primarily controlled by native pore size rather than loading condition.Rocks with larger pores display higher b-value and greater temporal variability,whereas those with smaller pores exhibit lower and more stable b-value.To assess failure predictability,the AE count rate was incorporated into an inverse power law model.The model demonstrates higher predictive accuracy for high-porosity rocks.The average predicted failure time(t_(p))decreases monotonically with porosity:under uniaxial compression,t_(p)for granite,marble,limestone,and sandstone are 2.32,1.82,1.42,and 0.03,respectively;under Brazilian splitting,3.54,3.30,0.10,and 0.03.Among the four rock types,sandstone with the highest porosity exhibits the smallest discrepancy between predicted and actual failure time,whereas granite with the lowest porosity shows the largest.As porosity decreases,prediction accuracy progressively declines for limestone and marble.Overall,the findings indicate that native pore heterogeneity governs both fracture scaling behavior and failure predictability,and that these effects are largely independent of the loading conditions examined in this study.
