The China University of Geosciences (simplified Chinese: 中国地质大学(武汉); traditional Chinese: 中國地質大學(武漢); pinyin: Zhōngguó Dìzhì Dàxué (Wǔhàn); abbreviated 地大 or CUG) is a key national university directly under the administration of the Education Ministry of the People's Republic of China. It is located in Wuhan, the capital of Central China's Hubei Province. It is a Chinese Ministry of Education Double First Class Discipline University, with Double First Class status in certain disciplines.It is regarded as one of the top geosciences university in China and exerts considerable influence on the Chinese mining and oil industry. Its notable alumni include Wen Jiabao, the Premier of China's State Council between 2003 and 2013, who attended the China University of Geosciences when it was known as the Beijing Institute of Geology (BIG). The motto "Being austere and simple, keeping on practice and acting for truth" is from him.
Reconstructing causal dynamic networks from multivariate time series is a foundational problem in complex systems science. Yet, the key scientific issue is not simply causality detection, but causal interpretability. Interpretable causality is the foundation for testable mechanistic hypotheses, transferable forecasting, and principled decision-making for intervention and control. In real-world complex systems, causal inference is often compromised by noise, missing data, high dimensionality, nonlinearity, time delays, heterogeneity, and partial observability. Classic approaches to interpretable causality yield explicit, inspectable quantities such as causal graphs, coefficients, and governing equations. However, this methodological shift has heightened expectations: AI-extended approaches are increasingly required to recover explicit causal mechanisms rather than opaque predictive dependencies, thereby preserving interpretability. This review summarizes four classic methods and their AI-extended counterparts based on time series data, including Granger frameworks, information-theoretic measures, nonlinear state–space/manifold reconstruction, and mechanistic differential-equation learning. Next, we elucidate their motivations, core principles, the origins of interpretability, and the assumptions required for meaningful conclusions. Finally, we highlight representative applications across climate studies, neuroscience, epidemiology, finance, social science, ecology and molecular biology, followed by a discussion of comparative analysis, open challenges, and future research directions.
As an ultrawide-bandgap semiconductor for extreme-environment applications, ,B-Ga2O3 exhibits remarkable radiation resistance. However, its behavior under extreme electron irradiation exceeding conventional energy regimes remains largely unexplored. Here, we combine 10 MeV electron irradiation (fluences: 1 & times; 1017 and 5 & times; 1017 cm-2), first principles calculation, and machine-learning molecular dynamics (ML-MD) simulations parameterized to match these conditions to systematically reveal the anisotropic structural limits and phase evolution mechanisms of ,B-Ga2O3. Electron irradiation induces pronounced anisotropic lattice degradation, electronic and optical responses across different crystal planes (100), (010), (001), and (2 & strns; 01). The (100) plane undergoes severe near-surface degradation via the collapse of GaO4 tetrahedra, evidenced by the disappearance of the B(2) g , A(8g) and A(9) g Raman modes. For the (010) orientation, the pronounced channeling effect reduces scattering cross-sections and favors dispersed point defects, broadening the bulk X-ray diffraction (XRD) rocking curve full width at half maximum (FWHM) from similar to 42 to over 155 arcsec. Across the four planes, irradiation induces a similar to 690 nm infrared emission while differentially suppressing their intrinsic blue-violet luminescence. The (001) plane demonstrates the highest overall tolerance. ML-MD simulations reveal that electron irradiation-generated point defects reorganize into a low-fraction, dispersed gamma-phase inclusion that remains structurally coherent with the defective ,B-phase matrix. These insights into the orientation-dependent structural, electronic, optical properties, and phase evolution establish a strategic framework for designing radiation-resistant ,B-Ga2O3 devices with sustained performance under extreme conditions. (c) 2026 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.
Given the rapid advances in artificial intelligence (AI), research and deployment of intelligent emergency evacuation navigation are urgently needed. This study focuses on resident evacuation after toxic hydrogen sulfide (H₂S) release in mountainous gas fields and develops an intelligent emergency evacuation navigation framework integrating multiobjective dynamic route optimization, road failure resilience assessment, and uncertainty-aware variable message sign (VMS) deployment screening. The framework uses real-time toxic gas and traffic state information to update route impedance and recompute evacuation routes by jointly considering road availability, clearance time, health-risk consequences, and congestion. The composite benefit-per-unit index (CBUI) of candidate VMS node combinations is predicted using a LightGBM-based surrogate model to support limited VMS deployment, and quantile prediction with conformal calibration addresses prediction uncertainty. Jointly considering toxic gas risks, traffic congestion, and clearance time provides the most balanced evacuation performance. VMS-based dynamic guidance helps maintain evacuation function and reduce evacuation function recovery delay under connected road failure scenarios. The CBUI indicates that a small number of properly configured VMS units can yield substantial evacuation benefits, whereas further deployment produces diminishing marginal returns. When road failures occur and some VMS units become unavailable, the framework supports more conservative VMS deployment decisions under limited resources.
Deep saline aquifers represents one of the most geologically promising solutions for large-scale carbon dioxide (CO2) sequestration, boasting a global storage potential of up to 10,000 Gt of CO2, which far exceeds other subsurface options. This comprehensive review systematically examines recent advancements in aquifers-based CO2 sequestration through experimental studies, simulations, pore-scale analyses, and field applications. Major trapping mechanisms, including structural, residual, solubility, and mineral trapping, are critically evaluated together with key factors affecting storage efficiency, injectivity, and long-term containment security. This study highlights the potential of CO2 foams for enhancing storage efficiency through improved mobility control and sweep efficiency within saline aquifers as overlooked by previous reviews. Additionally, it provides a detailed pore-scale analysis of CO2-brine-rock interactions, providing new insights into multiphase flow dynamics and trapping mechanisms at the microscopic level, which are crucial for accurate reservoir-scale predictions. Moreover, it explore the emerging role of microbial processes in CO2 sequestration, including their impact on mineralization and the security of long-term storage through biogeochemical interactions. In addition, the mechanical integrity of saline aquifers during CO2 injections was discussed in details as not discussed in previous reviews. A dedicated discussion of salt precipitation is presented, including deposition stages, influencing factors such as brine chemistry and injection conditions, and mitigation strategies for maintaining injectivity. The review also identifies key challenges and research gaps for futures researches related to long-term storage prediction, coupled process modeling, and economic feasibility. By integrating insights from pore-scale mechanisms to field-scale applications, this work provides a multidisciplinary framework for advancing aquifers-based CO2 sequestration as a safe and scalable climate mitigation technology.
Distributed parallel machine scheduling has attracted much attention in recent years. Mainstream methods are mainly search-based algorithms. However, they lack a self-organizing strategy to synergize multiple local search operators. This work proposes a cooperative memetic algorithm with deep reinforcement learning, named RLMA. RLMA uses a multi-rule initialization method, a cooperative global search strategy, and several local search operators. A fully connected deep Q-network (DQN) is used to learn the relationship between an encoded schedule and a suitable local search operator. During the search process, the DQN selects a promising operator for each solution. An elite archive is also used to store high-quality nondominated solutions and preserve useful search experience. RLMA is compared with five recent algorithms on 20 test instances. The results show that the DRL module improves operator selection in the memetic algorithm. RLMA also obtains better average results on the tested instances. This study provides a simple way to combine DRL and evolutionary search for shop scheduling problems.
