Issue 07, 2026
Risk analysis of methanol tank area fire accidents based on Bayesian network
Chen Chao;Yuan Boyuan;Wang Haijun;Wang Bo;Zeng Tao;Wang Yujie;Jiang Liqiong;Methanol is an important chemical with both new energy potential and fundamental chemical value, primarily stored in tank farms. Methanol leaks can easily lead to fires, and during combustion, they produce minimal smoke, with gas radiation serving as the primary heat transfer mechanism. This characteristic complicates fire detection and warning, potentially resulting in domino-effect accidents. This study systematically identifies risks in the methanol tank area through fault tree analysis, utilizing vertical storage tanks and pipeline systems as analysis nodes to pinpoint key risk factors. Based on Bayesian network theory, a probability model for tank farm fire accidents was constructed to enable quantitative analysis of accident probabilities. A dynamic Bayesian network method was employed to establish a fire domino effect evolution model, revealing the chain propagation mechanism by which initial accidents trigger secondary incidents. The study identified 56 risk events within the methanol tank farm, including 1 leaf node(T), 21 intermediate nodes(M), and 34 root nodes(X). The calculated probability of a fire accident occurring in the methanol tank area is 2.34×10-6, with the probability of an ignition source being 1.75×10-2 and the probability of a methanol leak at 2.47×10-2. Through Bayesian network sensitivity analysis, the key factors contributing to fire accidents were found to be sparks from vehicle exhaust pipes, illegal smoking mars, and residual fires following hot work. The dynamic Bayesian network model illustrated the domino effect of accidents, simulating the failure process associated with tank area fires. It was observed that the failure probability of each storage tank increased continuously over 1 000 seconds, peaking at 4.22×10-2 from an initial value of 2.34×10-6. The constructed Bayesian network model effectively identifies the key contributing factors of pool fires in methanol tank farms and their potential domino effects, providing a solid foundation for developing fire prevention and control strategies in these facilities.
Application of improved BNT-FRAM model in aircraft departure process safety analysis
Yuan Leping;Wu Xiaoxue;To effectively identify and manage performance changes and propagation within the civil aviation operation system, this paper employs the functional resonance model to analyze the task flow involved in aircraft departure procedures. Utilizing the hierarchical task analysis method, we decompose the task flow based on the organization of the air transport system and the interactions among different entities. A total of 20 functional modules have been identified, including cockpit preparation(F1), ATC clearance(F2), push back request(F3), push back permit(F4), push back(F5), ground traction(F6), startup request(F7), and startup permit(F8), among others. It is important to note that this paper employs blind number theory to estimate the probability distribution of the 20 functional modules concerning changes in time and accuracy. This approach mitigates bias introduced by expert subjectivity. Additionally, this paper abstracts the ability of each functional module to return to a normal state when it confronted with internal and/or external perturbations into a resilience factor. This resilience factor is then used to calculate the variability of each functional module. Finally, the functional coupling value is calculated using the risk coupling formula, enabling the quantification of upstream and downstream functional interactions. This approach identifies key functions and paths that are susceptible to change and formulates risk control measures for managing performance variations. The results indicate that seven functional modules-takeoff request(F14), receipt of aircraft(F15), hand over from tower to approach control(F20), hand over from ramp to tower control(F13), takeoff(F18), instructing aircraft to enter the runway(F16), and takeoff permit(F17)-exhibit greater variability during the aircraft departure process.Ground taxiing(F12)-takeoff request(F14)-takeoff permit(F17)-takeoff(F18)-airbone(F19)-hand over from tower to approach control(F20) and other two key paths are identified. When the system encounters disturbance, these functional modules are more likely to resonate with connected function modules, thereby highlighting potential failure nodes and critical paths within the process is valuable for operation analysis.
Optimizing opportunistic maintenance for offshore wind turbines: addressing uncertainty in life prediction
Li Yanan;Zhang Xinsheng;Wu Chunyang;Zhang Zhenlong;Su Jia;This study introduces a multi-strategy collaborative optimization approach to enhance maintenance decision-making in offshore wind turbines amidst uncertainties in lifetime prediction. In contrast to conventional deterministic strategies, the proposed method integrates uncertainty directly into the optimization framework, ensuring a thorough balance between cost and risk considerations. Initially, a dynamic probabilistic modeling scheme is employed to represent the time-varying nature of prediction errors. By developing a residual life prediction error model grounded in time-dependent probability distributions, this approach effectively captures the evolving stochastic characteristics inherent in degradation forecasts. Second, a cumulative maintenance risk quantification model is developed within the Conditional Value-at-Risk(CVaR) framework. This model addresses the asymmetric impacts of prediction errors on maintenance outcomes and offers a comprehensive measure of risk exposure. Monte Carlo sampling is utilized to generate sets of uncertain scenarios, facilitating robust estimation of potential loss distributions. Leveraging these models, a hierarchical maintenance strategy is formulated that integrates corrective, preventive, and opportunistic maintenance into a cohesive framework. To optimize both maintenance costs and operational risks throughout the turbine's life cycle, a multi-objective maintenance optimization model is established. The Nondominated Sorting Genetic Algorithm Ⅱ(NSGA-Ⅱ) is employed to solve this model, facilitating efficient exploration of the Pareto frontier and enabling coordinated optimization of strategy selection and decision thresholds. Case study results confirm the effectiveness of the proposed approach, demonstrating that the optimized strategies significantly reduce maintenance costs while precisely quantifying and mitigating operational risks. Notably, the integration of probabilistic modeling with scenario-based risk evaluation ensures that maintenance policies remain robust against various sources of uncertainty. In conclusion, the proposed method not only enhances the economic efficiency of offshore wind turbine operations and maintenance but also establishes a rational trade-off between cost and risk. These contributions provide a systematic and practical decision-support framework that can be applied to other complex engineering systems facing similar uncertainty-driven challenges.
Research on radio fingerprint positioning method with extremely sparse reference points in GNSS-denied environments
Chen Naixin;Zhu Chunli;Chen Lei;Gao Zhen;To tackle the significant decline in positioning accuracy caused by sparse reference points in Channel State Information(CSI) fingerprinting, this paper presents a fingerprint augmentation model called L2C-GAN(Location-to-CSI Generative Adversarial Network). The proposed method reconstructs fine-grained radio maps by generating virtual fingerprints based on the learned non-linear mapping between spatial locations and Radio Frequency(RF) features. The methodology consists of three key stages. First, a Density-Based Spatial Clustering of Applications with Noise(DBSCAN) is employed to preprocess the raw CSI amplitude data. This approach adaptively identifies and retains dominant signal clusters while eliminating outliers arising from environmental noise and hardware instability, thereby ensuring the purity and spatial consistency of the training data. Second, to accommodate convolutional operations, high-dimensional CSI matrices are segmented along subcarrier indices and reshaped into single-channel 2D feature maps. Third, the L2C-GAN architecture is constructed using a conditional generation mechanism. To address the low dimensionality of spatial coordinates, a Fourier-feature-based Positional Encoding mechanism is designed to map 2D coordinates into high-dimensional embedding vectors. Both the generator and discriminator integrate Self-Attention modules to capture long-range dependencies among subcarriers, enabling the model to learn global context. Experimental validation is conducted using the OpenCSI dataset in an extremely sparse setting with only 2%(88) real reference points. The results demonstrate that L2C-GAN effectively enhances radio map reconstruction. The average positioning error decreases from 2.01 m to 1.67 m, representing a 16.92% improvement over the baseline without augmentation. Furthermore, experiments with varying parameter settings show that augmenting with 400 virtual reference points yields the optimal positioning accuracy of 1.67 m. Comparative results also indicate that the proposed method significantly outperforms state-of-the-art augmentation techniques, such as SSIM-Aug(Structural Similarity-based Augmentation), AF-DCGAN(Amplitude Feature Deep Convolutional GAN), and LESS(Adaptive Fingerprint-based Localization with Less Site Survey). Notably, the model exhibits robust performance even under limited hardware configurations, achieving a positioning error of 1.92 m with only two receiving antennas, which surpasses the unaugmented four-antenna baseline.
Application of YOLOv5 in intelligent recognition of subgrade diseases based on ground penetrating radar images
Zhou Suhua;Huang Chuting;Sha Linchuan;Huang Minghua;As an advanced non-destructive detection technology and one of the most effective tools available, Ground Penetrating Radar(GPR) has been widely applied in identifying subgrade diseases due to its efficiency and high resolution. However, manual interpretation of GPR images is highly challenging, time-consuming, and susceptible to subjective errors. To address these issues, this study proposes an improved YOLOv5-based method for the automated identification of GPR images depicting subgrade cracks and loosening defects. This method is specifically designed to balance real-time performance with detection accuracy. A carefully constructed dataset comprising 1 290 VOC-format annotated radar images of typical subgrade defects was utilized for model training and validation. Several key architectural enhancements were introduced: the Convolutional Block Attention Module(CBAM) was embedded before the Spatial Pyramid Pooling Fast(SPPF) layer to enhance focus on meaningful features while reducing redundant parameters. Additionally, three C3 modules in the neck of the original YOLOv5 were replaced with Context Guided Blocks(CG Blocks) to strengthen contextual feature perception and improve computational efficiency. Further refinements included a Dynamic Head mechanism to enhance scale-aware representation. Ablation studies demonstrated that the combined improvements from the attention and context-guided modules yielded optimal results. The enhanced C1-CG-YOLOv5 model achieved a mean Average Precision at IoU threshold of 0.5(mAP50) of 90.8% and a mean Average Precision at IoU thresholds of 0.5 to 0.95(mAP50-95) of 64.2%, significantly outperforming the baseline model. Moreover, the model parameters were reduced by 12.8%, indicating superior lightweight performance. Finally, we developed real-time subgrade disease detection software based on the improved model, validating its detection performance, processing efficiency, and practical utility in GPR image analysis. The proposed approach offers an efficient and reliable tool for automated GPR image interpretation, demonstrating strong potential for practical application in infrastructure maintenance and non-destructive evaluation.
A dynamic clearance-signal coordinated control method for alternating traffic on two-lane roads under traffic incidents
Xie Shikun;Yang Zhen;Zhang Zhiwei;Luan Yi;Xu Guilong;Feng Yang;This study proposes a Dynamic Clearance-Signal Coordinated Control(DCSC) method aimed at improving traffic management efficiency during single-lane closures on two-lane highways affected by traffic incidents. The DCSC approach integrates manual clearance logic with adaptive signal optimization based on an improved Webster model. It incorporates key factors such as incident-zone length, vehicle-type-dependent speeds, and acceleration differences, using the moment when the last vehicle passes the opposite end as a trigger for signal phase switching. Portable signal units and roadside sensors, including cameras and radars, automatically detect the completion of vehicle clearance and initiate phase transitions, achieving fully automated control without the need for additional fixed infrastructure. Using field data from the G318 Basu-Bomi corridor in Xizang, the proposed method was evaluated through microsimulation on the SUMO platform and compared with manual and fixed-cycle control schemes. The results demonstrate that the DCSC strategy consistently outperforms the alternatives under various traffic conditions. The average signal cycle time is 20.1% shorter than that of fixed-cycle control and 22.0% shorter than manual control. Under low-flow conditions(traffic volume <300 pcu/h), the DCSC method achieves a similar average delay to manual control but significantly outperforms the fixed-cycle scheme, reducing the maximum queue length by 8.5% and 17.6%, respectively. Under medium-to-high flow conditions(traffic volume ≥300 pcu/h), the DCSC method further improves operational efficiency, reducing the maximum queue length, average delay time, and average waiting time by 16.2%, 12.6%, and 21.5%, respectively, while increasing the average speed by up to 26.8%. Sensitivity analysis confirms the robustness of the proposed method under varying incident-zone lengths and truck proportions, maintaining stable performance even in long or truck-dominated sections. By combining the precision of dynamic clearance with the efficiency of adaptive signal timing, the DCSC method offers a practical, low-cost, and effective automated control strategy for two-lane highway work zones or emergency scenarios, significantly enhancing both traffic safety and operational efficiency.
Corrosion depth prediction model for buried pipelines using intelligent optimization and feature selection
Liang Rui;Wang Nipeng;Zhou Wenhai;Li Kai;Song Yanhong;Hu Caizhi;This study presents a novel hybrid predictive framework designed to tackle the dual challenges of accuracy and interpretability in modeling pitting corrosion of buried pipelines. The methodology enhances a standard Backpropagation(BP) neural network by first employing and comparing three distinct intelligent optimization algorithms—Particle Swarm Optimization(PSO), Sparrow Search Algorithm(SSA), and Non-dominated Sorting Genetic Algorithm Ⅱ(NSGA-Ⅱ)—to globally optimize the network's initial weights and biases. A key innovation of this study is the development of a Recursive Feature Elimination(RFE) method driven by SHAP values, which systematically evaluated an initial set of 12 input features to identify a more parsimonious and robust optimal subset of 11 variables. The performance of all models was rigorously assessed on a hold-out validation set using a comprehensive suite of metrics(R2, MSE, RMSE, MAE). The results clearly demonstrate the superiority of the optimized models over the baseline network. After feature selection, all three optimized frameworks—PSO-BP, SSA-BP, and NSGA-Ⅱ-BP—achieved significant performance improvements, with their MSE values falling within a low-error range of 6.18% to 14.77% lower than their pre-RFE performance. In stark contrast, the baseline BP model saw its MSE reduced by 83.48% solely due to feature selection. Among the advanced frameworks, the NSGA-Ⅱ-BP model consistently demonstrated the best performance across all evaluation metrics, achieving an exceptionally low final Mean Squared Error(MSE) of 0.058 75, the highest R2 of 0.968 1, and the lowest RMSE of 0.242 3 and MAE of 0.182 3. In addition to its quantitative performance, the SHAP-based interpretability analysis of the top-performing NSGA-Ⅱ-BP model provided strong validation of its scientific soundness. The analysis confirmed that the model's internal logic aligns closely with electrochemical corrosion theories, accurately identifying service life, soil pH, and key corrosive ion concentrations as the most influential predictors. Additionally, the model demonstrated advanced learning capability by recognizing the information redundancy in traditional parameters such as soil resistivity. In conclusion, this research presents a predictive model that is not only highly accurate but also transparent and mechanistically coherent. The comparative analysis further validates the selection of NSGA-Ⅱ as the optimal optimization strategy for addressing this complex engineering challenge.
Construction of a tailings dam displacement early warning model based on deep learning and the cloud model
Yang Bin;Qin Shuhao;Yang Yuhao;Hu Jun;Xia Zhiguo;This paper proposes a displacement early warning method for tailings dams that integrates grey relational analysis, the Neural Basis Expansion Analysis for Time Series with Exogenous Variables(N-BEATSx) deep learning model, and a cloud model-based risk classification framework. Grey relational analysis is employed as a preliminary step to identify the dominant factors affecting dam deformation. The results indicate that time and ponding water pressure have the strongest influences, with correlation coefficients of 0.799 and 0.703, respectively. These reflect the combined effects of material creep, consolidation, and hydrostatic loading. Building on this, the N-BEATSx model is utilized to predict seven-day-ahead displacements using monitoring data from a tailings dam in Liaoning Province, with separate evaluations for horizontal displacement and settlement. The predictive performance shows that N-BEATSx significantly outperforms the baseline N-BEATS model, as the Root Mean Square Error(RMSE) and Mean Absolute Error(MAE) for horizontal displacement forecasts reach 0.423 and 0.361, respectively, while settlement forecasts achieve RMSE and MAE values of 0.398 and 0.318. Compared to Neural Basis Expansion Analysis for Time Series(N-BEATS), the N-BEATSx model improves prediction accuracy by an average of approximately 19%, demonstrating its superior ability to capture nonlinear temporal dependencies and short-term fluctuation patterns in dam deformation. To translate numerical forecasts into actionable early warning signals, a cloud model is introduced to establish quantitative thresholds and classify displacement predictions into four hierarchical risk levels. Following the 3E principle, blue, yellow, orange, and red warning grades are defined, enabling a direct mapping of predicted values to discrete safety states and transforming model outputs into operational risk indicators. This proposed method achieves both high-precision forecasting and risk quantification, providing an effective technical pathway for linking predictive modeling with engineering decision-making. The results confirm that this integrated approach not only enhances the reliability of displacement forecasting but also delivers a transparent and structured framework for risk-based early warning, offering valuable support for the safe operation and management of tailings dams.
Hydrogen sulfide distribution in coal mining face and return airway under airflow field effects
Liu Ajian;Liu Yang;Liu Hao;Su Jiahao;As coal mining progresses to greater depths, the issue of abnormal hydrogen sulfide(H_2S) emissions has intensified, posing significant threats to the safety of underground personnel and the stability of production due to H_2S's high toxicity, flammability, and corrosiveness. Existing studies have predominantly focused on static analyses at the coal seam scale and simplified diffusion models, resulting in considerable discrepancies between predicted H_2S diffusion patterns and actual mining conditions. The transport mechanisms of H_2S under dynamic mining disturbances—particularly those induced by shearer operations—remain insufficiently investigated. To address this gap, this study employs a coupled multi-physics numerical simulation approach, integrating fluid flow, gas transport, and mechanical disturbance effects, to systematically analyze H_2S dispersion behavior within a fully mechanized mining face and its connected return airway. The geometric model was constructed based on the 1802 working face of Xinzhuang Coal Mine in Gansu Province, with meshing refined in regions of high gradient flow. Simulations utilized the realizable k-ε turbulence model and species transport equations, with H_2S release sources calibrated according to empirical data. Boundary conditions incorporated ventilation velocity, wall roughness, and dynamic updates to simulate shearer movement. The results indicate that the airflow field in the return airway can be categorized into four distinct zones: high-speed flow, low-speed flow, vortex, and boundary layer regions. A stable vortex zone forms 7-15 m from the inner side of the return airway, identified as a high-risk area for H_2S accumulation, where concentrations reach 3-5 times those found in the mainstream zone. Within the working face, the shearer's obstruction creates higher pressure upstream and lower pressure downstream. Due to the contraction of the flow passage cross-section downstream of the shearer, a localized acceleration zone forms, increasing the maximum airflow velocity by approximately 40% compared to the intake airflow. This acceleration intensifies turbulent mixing, which subsequently affects H_2S dilution and transport. Based on the flow field structure and gas concentration distribution, a three-zone classification for H_2S risk is proposed, along with a differentiated prevention and control strategy tailored to the aerodynamic features of each zone. The methodology and findings provide a valuable reference for H_2S hazard management in deep, high-sulfur coal mines with similar geological and mining conditions.
Investigation of moisture content effects on spontaneous combustion characteristics and thermophysical parameters of non-caking coal
Ma Li;Guo Yangyang;Liu Hongliang;Wei Gaoming;Feng Jinbo;The risk of spontaneous combustion in residual coal within goafs varies significantly under water-immersed conditions. To investigate the effects of moisture content on the spontaneous combustion behavior and exothermic characteristics of residual coal, programmed temperature-rise experiments, thermophysical property tests, and infrared spectroscopy were conducted. This study examined the spontaneous combustion characteristics, thermophysical parameters, and evolution of functional groups in non-caking coal with five different moisture contents(mass fractions of 3.10%, 5.48%, 6.93%, 12.20%, and 17.10%). The experimental results indicate that, as moisture content increases, the production of CO, oxygen consumption rate, and heat release intensity of the coal samples initially increase and then decrease. The coal sample with a moisture content of 6.93% exhibits the highest coal-oxygen reaction rate. Additionally, increasing moisture content leads to an increase in specific heat capacity, while decreasing thermal diffusivity and thermal conductivity, thereby reducing the efficiency of heat diffusion and conduction within the coal matrix. Regarding the functional groups in the coal samples, the mass fraction of —OH increases with rising moisture content, while those of —CH2 and C=O initially increase and then decrease. The mass fraction of C=C remains relatively stable. Moisture content influences the spontaneous combustion process of coal through three key mechanisms:(1) modifying the quantities of essential functional groups and active sites, thereby regulating the intensity of chain reactions;(2) affecting thermophysical parameters and altering the efficiency of heat diffusion, storage, and transfer within the coal matrix;(3) reshaping the pore structure of the coal, thereby adjusting the contact area between coal and oxygen. This study provides valuable insights for predicting the spontaneous combustion of residual coal in water-immersed goaf environments.