Issue 08, 2026
A coal mine fire risk assessment model based on difference matrix and fuzzy Bayesian network
Li Yan;Wang Xue;Li Hongxia;To enhance the specificity and causal reasoning capabilities of coal mine fire risk assessment, this study constructed an integrated evaluation model that combines rough set differential matrices with fuzzy Bayesian network. Based on the causation mechanisms of coal mine fires and relevant standards, we established an indicator system encompassing four dimensions—fire prevention, emergency response, risk monitoring, and management capacity—comprising 14 secondary indicators. We employed the rough set differential matrix method to reduce indicator attributes, identifying five key indicators from the initial set: ignition source management(A3), ventilation system design and maintenance(A4), emergency plan completeness(B1), intelligent monitoring system application(C2), and personnel training and safety awareness(D2). This process, grounded in data from 45 coal mine accident cases and expert assessments, achieved objective indicator selection through the construction of a differential matrix, calculation of attribute discrimination, and determination of normalized weights. Building upon the identified key indicators, we applied fuzzy set theory to address uncertainties in expert evaluations. We converted expert linguistic assessments into triangular fuzzy numbers and derived fuzzy probabilities for each root node using weighted averaging and defuzzification. Using these as prior probabilities, we constructed a fuzzy Bayesian network structure. Through network parameter learning and conditional probability table configuration, we achieved forward probability prediction and backward causation diagnosis under conditions of uncertainty. Model application and validation results indicate that ventilation system design, emergency plan completeness, and intelligent monitoring system application are high-risk factors, each exhibiting a probability exceeding 50% of being in a “high” risk state. Reverse inference further confirms that when the system is in a high-risk state, the posterior probabilities of these nodes increase significantly, indicating they are key drivers of the risk state. Validation using the Chongqing Songzao Coal Mine “9.27” major fire accident case demonstrates that the model successfully raised the high-risk probability from a baseline state of 27.1% to 89.3% based on anomalies in key indicators. Through posterior analysis, it precisely quantified the risk contribution of each factor. This research integrates the objective attribute reduction capability of rough set differential matrices with the uncertainty reasoning capacity of fuzzy Bayesian network, effectively addressing shortcomings of traditional risk assessments, such as redundant indicators, subjective weighting, and inadequate handling of ambiguous information. The model not only significantly enhances the accuracy and interpretability of coal mine fire risk identification but also facilitates quantitative diagnosis and early warning of key causal factors. It provides a reliable decision support tool for coal mining enterprises to implement differentiated and precise fire risk management. The model and methodology developed herein offer a novel technical approach for tackling the dynamic assessment of multi-factor coupled risks within complex industrial systems.
Explosion risk assessment of powder compaction process for roman candles pyrotechnic composition based on dynamic Bayesian network
Zhang Shulin;Wang Lanning;Zhang Jingying;Li Qi;Guo Xin;Lu Yi;To scientifically evaluate the explosion risk associated with the powder compaction process of Roman Candle pyrotechnic compositions, this study integrated Fault Tree Analysis(FTA) with Dynamic Bayesian Network(DBN) methodologies. Initially, an FTA model was developed to identify and analyze the causal pathways leading to explosion accidents. This model was subsequently transformed into an equivalent Bayesian Network(BN) model. Next, eight experts were invited to assess the basic events using a five-point scale, and the reliability of their scores was verified through consistency checks utilizing a judgment matrix. Finally, the concept of time slices was introduced into the BN model to enhance its analytical capabilities. Basic events were categorized into static nodes and dynamic nodes based on the influence of time on their behavior. The prior probabilities for static nodes and the transition probabilities for dynamic nodes were calculated sequentially. A fuzzy analysis method was employed to convert expert scores into trapezoidal fuzzy numbers, allowing for the integration of expert opinions by assessing the similarity among their scores. This facilitated the computation of prior probabilities for the static nodes based on the consolidated expert assessments. Dynamic nodes were further classified into three categories—equipment failure, human error, and organizational factors—according to the types of accidents, enabling the calculation of their transition probabilities. GeNle software was utilized to simulate changes in accident probability over the next three years, using a 90-day time slice interval, under the assumption of neglecting the repair rate. The results indicated that the initial accident probability was 0.052 8, which increased to 0.268 after three years. Additionally, diagnostic reasoning was employed to calculate the posterior probabilities of basic events. By comparing the relative differences between posterior and prior probabilities, five key hazard sources were identified: low ignition energy of sulfur, improper material selection for mixing, failure of pressure relief devices, inadequate personnel training, and motor sparks resulting from line short circuits. Finally, targeted risk control measures were recommended across three dimensions: standardizing personnel operations, enhancing equipment and processes, and strengthening safety management systems. These measures aim to reduce the probability of fire and explosion incidents effectively.
Risk analysis for railway dangerous goods transportation system based on a FMEA-ETIFIAD approach
Chong Pengyun;Yao Mian;Huang Wencheng;This study introduces a risk analysis method, Failure Mode and Effect Analysis-Extended Triangular Intuitionistic Fuzzy Information Axiomatic Design(FMEA-ETIFIAD), to address the limitations of traditional risk assessment techniques in railway hazardous goods transportation systems. The method incorporates three key innovations to enhance conventional FMEA. First, Triangular Intuitive Fuzzy Numbers(TIFNs) are employed to evaluate the severity(S), occurrence probability(O), and detection possibility(D) of each risk sub-indicator, replacing the traditional discrete real-number scoring system. This fuzzy approach incorporates membership, non-membership, and hesitation values, thereby minimizing information loss and effectively managing uncertainties. Second, the method calculates the information content of S, O, and D using the Information Axiom(IA), moving away from the traditional Risk Priority Number(RPN) that relies on linear multiplication. This modification resolves the issue of multiple risk levels being assigned identical RPNs. Third, the Entropy Weight Method(EWM) is applied to objectively determine the weights of risk sub-indicators based on historical accident data from the Chinese railway hazardous goods transportation system(1985-2017), thereby eliminating biases associated with subjective expert judgments. A case study was conducted utilizing data from the Chinese railway hazardous goods transportation system. Three experts from Southwest Jiaotong University and the Chengdu Railway Bureau assessed the risk sub-indicators using TIFNs. The data were processed using MATLAB R2012b and Microsoft Excel 2013 VB to compute the information content of each risk factor. Results indicate that personnel risk(H) holds the highest total information content(1.448 342 59), followed by equipment risk(M1, 1.229 193 54), cargo risk(M2, 0.967 631 42), management risk(M3, 0.788 793 21), and environmental risk(E, 0.764 327 26). Critical sub-indicators include insufficient technical knowledge of staff, equipment failure, hazardous goods handling, and inadequate railway infrastructure. A comparison with traditional FMEA demonstrates that the FMEA-ETIFIAD method effectively distinguishes risk levels in cases exhibiting identical RPNs and provides a more reliable risk assessment that aligns closely with actual accident data.
Research on a cloud model for assessing airport runway incursion risk using integrated weighting
Zhang Yuhui;Sun Zeyuan;Liao Shujiao;In response to the current inadequacies in quantitative assessment methodologies for runway incursion risks, which hinder the accurate identification of high-risk segments in airport operational management and contribute to frequent safety incidents and operational efficiency constraints, this study proposes an integrated cloud model framework based on integrated weighting principles for refined risk identification. By systematically applying the Latent Dirichlet Allocation(LDA) topic modeling technique, thematic elements and relevant keywords were extracted from a comprehensive collection of 122 global runway incursion accident investigation reports spanning the past two decades. The derived thematic content underwent rigorous validation by a panel of aviation safety experts, ultimately establishing a well-structured runway incursion risk indicator system comprising four primary indicators—human factors, communication failures, airport environment, and management factors—along with 24 carefully defined secondary indicators. Regarding the weighting methodology, an improved Analytic Hierarchy Process(AHP) was employed to calculate subjective weights for each indicator, while the Kendall concordance coefficient was utilized to verify the reliability of these weight assignments. Simultaneously, the enhanced entropy weight method was implemented to derive objective weights based on data characteristics. Through the application of game theory principles, the final combined weights were determined to be 0.153 for subjective dimensions and 0.847 for objective dimensions, respectively. The validity of these final indicator weights was further confirmed through both additive and multiplicative combinatorial weighting approaches. Utilizing the cloud model as the theoretical foundation for risk quantification, empirical data from four representative Chinese airports were selected for comprehensive model validation: a 4E-category airport in North China, a 4F-category airport in East China, a 4C-category airport in Southwest China, and a 3C-category airport in Northwest China. The validation results demonstrated risk level assessments of medium-high(0.665), medium-low(0.553), medium-low(0.715), and medium-high(0.628), respectively, showing strong alignment with actual operational conditions. Furthermore, the model exhibited satisfactory stability under various parameter variations, providing valuable practical references for targeted improvements in specific risk factors across different airport environments, ultimately contributing to the effective reduction of runway incursion risks.
Modelling of thermal decomposition kinetics and study of reaction mechanisms of NCM523
Wang Yan;Zhang Jingyue;Yu Tianmin;Chen Jie;Su Yang;Xu Yabei;To gain an in-depth understanding of the thermal decomposition mechanism of lithium-ion battery cathode materials, this study utilized Thermogravimetric analysis(TG) data of LiNi0.5Co0.2Mn0.3O2(NCM523) obtained at multiple heating rates and integrated these data with a Chemical Reaction Neural Network(CRNN) for kinetic modeling of the pyrolysis process. Through iterative training and validation, a 4-4 model comprising four species and four principal reactions was successfully developed. The model accurately captures the thermal mass-loss behavior of NCM523 and achieves a relative error of less than 1% across all heating rates. Additionally, the modeling results suggest that the pyrolysis follows an irreversible phase-transition pathway, progressing from a layered structure to a lithiated spinel, and ultimately to a rock-salt phase, reflecting the intrinsic structural evolution of the material. The thermal decomposition process can be divided into four distinct stages. In the first stage, the layered structure partially transforms into a lithiated spinel, with a relatively slow reaction rate due to the initial rearrangement of the lattice. In the second stage, both the lithiated spinel and the residual layered phases continue converting into a delithiated spinel, accompanied by an accelerated reaction rate as the system gains thermal activation. The third stage involves the decomposition of intermediate phases into a rock-salt structure, with the concurrent release of oxygen. During this stage, the reaction rate increases sharply in the high-temperature region before becoming diffusion-limited. In the final stage, all intermediate phases fully convert into the rock-salt structure, and the reaction rate gradually stabilizes as structural reorganization is completed. Key kinetic parameters, including activation energies and pre-exponential factors for each reaction step, were autonomously extracted by the CRNN model. These results demonstrate that the CRNN approach not only identifies complex thermal decomposition pathways but also provides physically interpretable kinetic constants without requiring prior assumptions about the reaction mechanisms. Overall, this work highlights the potential of the CRNN as an effective, data-driven tool for analyzing the thermal decomposition behavior of lithium-ion battery cathode materials.
Study on the influence of hydrogen blending on the ignition and explosion characteristics of methane-air tubes with different equivalence ratios
Li Xuejiao;Feng Dingyu;Wang Quan;Chang Weida;Zhu Wenyan;Ge Yu;Explosion experiments were conducted in a self-built stainless steel flame acceleration tube to investigate the impact of hydrogen blending ratios on the explosion behavior of methane-air mixtures. Tests were performed at hydrogen blending levels of 0, 10%, 20%, and 30%, along with equivalence ratios of 0.9, 1.0, and 1.1. Utilizing experimental equipment such as a high-speed camera, a PCB piezoelectric pressure sensor, and a photoelectric sensor, this research examined the flame development process, explosion overpressure, and average flame propagation speed within the tube. The results indicate that hydrogen enrichment enhances the flame luminosity of the methane-air mixture as observed through the observation window. Under fuel-rich conditions(Φ=1.1), hydrogen reacts preferentially with air, resulting in incomplete combustion of methane and the formation of a yellow flame. Three characteristic pressure peaks(p1, p2, p3) were identified in the pressure-time curves for all test conditions. These peaks correspond to the following events: the rupture of the vent membrane, the discharge of the burning mixture from the tube, and a secondary external explosion. The high reactivity of hydrogen enhances the peak values of p1, p2, and p3, as well as increases the flame propagation speed. Notably, the amplitude of p1 leads to two types of failures in the Polyethylene(PE) vent membrane: partial rupture and complete rupture. Complete membrane rupture allows a significant amount of unburned mixture to escape from the tube, which intensifies the external explosion while leading to a noticeable decrease in p2 and a reduction in flame propagation speed. In contrast, when the membrane is only partially ruptured, a substantial amount of unburned mixture accumulates inside the tube, resulting in a more intense reaction within the tube. The experimental data provide a foundation for the prevention and control of accidental explosions involving hydrogen-methane-air premixed gases in enclosed tubes. Additionally, this research serves as a methodological reference for studies on the explosion characteristics of other combustible gases.
Study on the mechanism of CO2 influence on H2 explosion characteristics
Hu Shoutao;Gong Hao;Nie Baisheng;Zhang Zhiwei;Hong Zijin;To explore the mechanism of CO2's influence on the explosion characteristics of H2, experiments on H2/CO2/air mixture explosions were conducted, and the impact patterns of varying volume fractions of CO2 on H2 explosions were obtained. Key elementary reactions and radicals during the explosion process were extracted using Chemkin-Pro, and subsequently, key elementary reactions were simulated using Materials Studio software. The experimental results indicate that, with the increase in the volume fraction of CO2, both the maximum explosion pressure and the rate of pressure rise exhibit a decreasing trend. Conversely, as CO2 is incorporated into the H2 mixture, an increase in the concentration of CO2 correlates with a prolonged duration to achieve the peak explosion pressure. Additionally, this leads to a reduction in the peak explosion pressure, thereby exerting a more pronounced inert effect on H2 detonations. When the volume fraction of H2 reaches 25%, the impact of CO2 volume fraction on the maximum explosion pressure gradually becomes apparent, and its effect on the maximum pressure rise rate is at a stage of controllable gain. When the H2 volume fraction reaches 29.5%, further increasing the CO2 volume fraction particularly enhances the detonation inerting effect. Simulation results indicate that CO_2+·H = CO+·OH is a critical branching reaction in the impact of hydrogen explosions. CO_2+·H readily accepts electrons, while CO+·OH more readily donates electrons. Compared to CO+·OH = ·H+CO2, the reaction CO_2+·H = CO+·OH involves a lower intermolecular energy barrier, making the reaction more likely to proceed. As the volume fraction of CO2 increases, the chemical inerting effect strengthens, while the contribution of chemical inerting weakens. Under conditions where the CO2 and H2 volume fractions are 5% and 29.5%, respectively, the chemical inerting contribution reaches its maximum, accounting for 21.81% of the total inerting effect. The chemical inerting effect of CO2 primarily manifests in its consumption of ·H free radicals to form intermediates, thereby reducing the reaction rates of key chain reactions such as ·H+O2 =·O+·OH and ·H+H_2O = ·OH+H2. This reduction tends to saturate as CO2 concentration increases.
Mechanistic effects of carboxyl and amino groups on toluene pyrolysis over activated carbon
Zhao Jiangping;Liang Zhenghao;This study aims to clarify the thermal decomposition mechanisms of toluene adsorbed on activated carbon functionalized with carboxyl(—COOH) or amino(—NH2) groups through reactive molecular dynamics simulations. Functionalized activated carbon fragments were constructed in Materials Studio, geometrically optimized using the conjugate gradient method, and annealed under an NVE ensemble to relieve local stress. Amorphous cells with a density of 0.6 g/cm3 were generated to replicate porous structures, and toluene was subsequently packed into the activated carbon to achieve a final density of 0.8 g/cm3 using the COMPASS Ⅲ force field. The simulations were conducted in LAMMPS under periodic boundary conditions within a temperature range of 2 000-3 000 K. Each system was first equilibrated at 300 K under the NVT ensemble for 10 ps, followed by 200 ps of high-temperature decomposition at each specified temperature, using a time step of 0.2 fs. Structural characterization—including density, radial distribution function, surface area, pore volume, and porosity—demonstrated good agreement with available experimental data. The results revealed distinct decomposition behaviors between the two functionalized systems. In the —COOH system, initial exothermic decarboxylation is followed by endothermic reactions, resulting in a rapid increase in enthalpy and a heightened risk of thermal runaway. Conversely, the —NH2 system maintains relatively stable enthalpy through deamination and hydrogen transfer, leading to milder decomposition. Radical analysis indicates that O· radicals generated by carboxyl decarboxylation in the —COOH system preferentially participate in oxidative consumption reactions, thereby suppressing the reverse recombination of decomposition fragments. In contrast, the —NH2 system exhibits high concentrations of H· radicals, inducing a “hydrogen pool effect” that triggers reverse reactions and fluctuations in decomposition behavior. Product distribution further distinguishes the two systems: the —COOH system primarily produces H_2O, CO2, CO, and char, with suppressed tar formation. In contrast, the —NH2 system generates substantial amounts of H2, NH3, and HCN at low temperatures, while also promoting char decomposition at higher temperatures. Intermediate analysis reveals that C—H—O species dominate in the —COOH system, stabilizing the reaction pathways. In the —NH2 system, however, C—H—N intermediates accumulate, which enhances the formation of toxic gas precursors. These contrasting radical-mediated mechanisms, influenced by surface functional groups, govern decomposition kinetics, product variation, and thermal stability. The findings underscore the significant role of functionalized activated carbon in regulating Volatile Organic Compound(VOC) decomposition pathways. Furthermore, this research provides theoretical guidance for the design of efficient, low-risk materials for VOC treatment and for the safe management of thermal decomposition processes.
Research on the tunnel wind field characteristics induced by mobile fans and optimization of fan combination
Xiao Feng;Li Yuanzhe;Zhang Jiaqing;Tao Haowen;Yang Yao;The mobile fan, as a portable smoke exhaust device, serves as an essential backup for smoke control when fixed smoke exhaust systems fail. This paper analyzes the wind field characteristics, wind speed, and wind pressure generated by various combinations of mobile fans(differing in the number of fans, spacing, and lifting height) within a tunnel through numerical simulation methods. The aim is to explore the smoke exhaust capabilities of the combined mobile smoke exhaust system in tunnel environments. The research findings indicate that the lift pressure produced by the combined fans in the tunnel results from the interference and mutual interaction of each fan's jet development. Notably, the actual increase in lift pressure does not follow a simple linear relationship with the number of fans. As the number of fans increases, the boost coefficient of the combined fan also rises, demonstrating that more fan energy is converted into the kinetic energy of the airflow within the tunnel. Conversely, increasing the spacing between the mobile fans reduces the interaction between multiple jets generated by the fans, which subsequently decreases the actual lift pressure and boost coefficient. When forming the fan matrix combination, appropriately elevating the height of the fan matrix enhances the uniformity of the induced airflow over shorter distances, thereby improving the stable wind speed in the pressure ventilation section. Under the conditions investigated in this study, the boost coefficient provided by the combined mobile fan ranged from 0.20 to 0.36. Based on these conclusions, recommendations for optimizing the configuration of the combined mobile smoke exhaust system are proposed. This study provides a foundation for the practical application of mobile fans in tunnels and offers valuable insights for the implementation of mobile smoke gas control equipment.
Enhancing flame and smoke detection using YOLOv13 with multi-scale convolutional attention
Zhou Xihua;Wu Pengfei;Traditional flame detection methods often struggle with low feature utilization efficiency, which hampers their effectiveness in meeting the demands of modern fire safety monitoring. The You Only Look Once(YOLO) framework is an efficient deep learning-based object detection architecture that enables rapid and accurate localization and identification of flame and smoke objects within images. However, the application of the state-of-the-art YOLOv13 model for flame and smoke detection remains relatively underexplored. Consequently, addressing challenges such as significant scale variations among flame and smoke targets, as well as enhancing the practical deployment potential of detection models, continues to be a key research objective. In this study, we first perform a comparative analysis between YOLOv13 and other classic models from the YOLO series to evaluate its performance advantages specifically for flame and smoke detection. To effectively manage scale variations and enhance the model's applicability, we propose a Multi-Scale Convolutional Attention(MSCA) module. This module is integrated into YOLOv13, resulting in an improved architecture termed YOLOv13s-MSCA. Experimental results demonstrate that YOLOv13 offers a more favorable balance of detection accuracy, inference speed, and practical utility compared to other versions of YOLO. Notably, the proposed YOLOv13s-MSCA model demonstrates impressive performance on the FASD dataset: flame detection accuracy under high-precision localization(mAP75) increases by 3.5 percentage points, while overall high-precision localization/detection accuracy(mAP75) improves by 2.4 percentage points. Additionally, the overall detection accuracy(mAP50-95) rises by 1.4 percentage points. On the enhanced D-Fire dataset, smoke recall improves by 1.9 percentage points. The model also showcases an improved capability for detecting small-scale flame and smoke targets with high precision. While challenges remain in accurately identifying heavily occluded objects against complex backgrounds, the proposed model exhibits strong robustness and generalization abilities, ensuring reliable detection performance across diverse and intricate scenarios. This model holds significant potential for application in real-world fire warning systems.