To quantify the rescue effect of medical supplies delivered to demand points at different times, the concept of rescue utility is introduced, and a rescue utility quantification function was constructed based on the time difference of medical supplies arriving at demand points. On this basis, a path planning model for multi-modal distribution of medical supplies in flood disasters was established with the objective of maximizing total rescue utility and minimizing total distribution time. According to the characteristics of the model, a HNSGA-II was designed for solution. Experiments were conducted using multiple types of examples. The results demonstrate that HNSGA-II achieves a 62% and 29% improvement in total rescue utility compared to the traditional Non-dominated Sorting Genetic Algorithm-II (NSGA-II) and the Multi-Objective Artificial Bee Colony Algorithm (MOABCA), respectively. Additionally, the average satisfaction level of material delivery time is enhanced by 13% and 6.1%, respectively. These findings indicate that HNSGA-II significantly improves emergency rescue outcomes, exhibits superior multi-objective optimization capability, and ensures that disaster victims receive timely and effective treatment under emergency conditions.

In order to improve the efficiency of facepiece-wearing detection for tunnel operation workers, a feature fusion-based facepiece detection model was proposed. First, high quality query images were selected, and an image gallery was established. An image retrieval method was adopted to obtain samples and measure the similarity between query and gallery images, thereby iteratively expanding the dataset scale. Then, Histogram of Oriented Gradients (HOG) and Fisher features were extracted from the images. The Ant Lion Optimizer (ALO) was introduced to compute the optimal weight combination for the two types of features, which were subsequently fused. Finally, based on the fused features, a SVM was utilized to train a facepiece detection model, and experimental evaluations were conducted on the self-constructed dataset. The results indicate that the proposed model effectively accomplishes the task of facepiece-wearing detection in tunnel operation scenarios. Feature fusion enhances the image description and improves the detection accuracy of the model. Compared to using only HOG features or Fisher features, the accuracy is increased by 6% and 14%, respectively. The model meets the accuracy requirements for facepiece-wearing detection of workers in tunnel construction environments.

To reduce the occupational health risks of communication base station maintenance personnel and improve their physical and mental health, a moderated mediation model was built in this study. It was based on COR to explore the psychosocial factors affecting these personnel's physical and mental health. First, the action path of prosocial work orientation on occupational health risks was analyzed based on the COR, and four theoretical hypotheses were proposed. Second, a hypothetical model encompassing prosocial work orientation, meaningful work, neuroticism, and occupational health risks was established. Finally, SPSS and PROCESS were employed to analyze 3 559 valid sample data from a large communication company to verify whether the hypotheses held true, and corresponding management implications were put forward based on the data. The results show that prosocial work orientation significantly negatively predicts occupational health risks. Meaningful work plays a partial mediating role in the relationship between prosocial work orientation and occupational health risks. Neuroticism not only positively moderates the negative relationship between meaningful work and occupational health risks but also further positively moderates the strength of the aforementioned mediating path—specifically, the mediating effect is stronger among individuals with high neuroticism.

To achieve accurate identification and evaluation of the service state of heavy-haul railway lines, a heavy-haul railway line comprehensive inspection vehicle was designed and developed, integrating six major inspection systems: a track condition inspection system, a rail profile measurement system, a track geometry measurement system, a rail flaw detection system, a wheel-rail force measurement system, and a vibration measurement system. A positioning and synchronization system is employed to enable synchronized acquisition and correlation analysis of multi-source line-condition data, and to provide graded alarms for suspected defects. The results show that, since its commissioning in July 2023, the heavy-haul railway line comprehensive inspection vehicle, operating at speeds up to 80 km/h, has continuously collected multi-source data including track geometry parameters, surface inspection images, internal B-scan images, wheel-rail forces, and vibration signals, and has established a database of typical defects. By combining the positioning and synchronization system with a multi-source spatiotemporal mapping model, the system realizes automatic alignment and correlation analysis of inspection data, enabling accurate tracing of defect causes and reducing both false positives and missed detections. Furthermore, the proposed graded alarm mechanism for suspected defects supports differentiated responses such as speed restriction, reporting, and record-keeping according to exceedance levels, thereby ensuring train operation safety while improving both heavy-haul railway maintenance efficiency and train operation efficiency.

In order to improve the identification accuracy of kick risk during drilling, a multi-category kick risk intelligent identification model was proposed by integrating feature engineering and machine learning techniques. Firstly, a wavelet transform was employed to achieve noise suppression based on field-measured kick data. Secondly, the dynamic variation trends of key parameters were extracted using smooth spline functions, and the abnormal fluctuation behaviors of kick-related characteristic parameters were analyzed. Based on this, a three-level risk classification criterion (low, medium, and high) was established, and kick risks were labeled according to the variation features of drilling data. Then, the sparrow search algorithm (SSA) was introduced to optimize the extreme learning machine (ELM), and a multi-classification kick risk intelligent identification model based on IELM was constructed. Finally, the performance of the model was validated through training, tuning, and testing on the constructed risk dataset. The results show that the IELM model outperforms the original ELM and back-propagation (BP) neural network model in terms of classification accuracy and discrimination stability, and is capable of identifying different levels of kick risks more accurately and efficiently.

In order to promote the optimization and upgrading of metal dust explosion-proof technology and establishment of a safety protection and control system, this study systematically investigated the reaction kinetics of aluminum powder with excess water under varying particle sizes and stacking masses, based on the self-developed visualization experimental platform for the reaction between stacked metal dust and water. Through quantitative characterization of the effects of particle size and stacking mass on key reaction parameters-including the maximum temperature, total gas production, maximum hydrogen concentration, maximum pressure, and total reaction time-the kinetic mechanism of the aluminum-water reaction was revealed. The results demonstrate that at a fixed particle size, all characteristic parameters exhibit positive correlations with increasing stacking mass. At a fixed stacking mass, the maximum temperature follows a V-shaped trend (decreasing then increasing) with increasing particle size, while gas production, hydrogen concentration, and pressure show inverted U-shaped trends (increasing then decreasing). Total reaction time increases monotonically with increasing particle sizes. The primary reaction products are Al(OH)₃ and H₂, and the reaction process comprises three distinct stages: slow hydrogen evolution, violent reaction, and attenuation and termination.

To optimize the layout of the outbound transportation system and enhance system resilience, a "Sensitivity-Response" dual-dimensional evaluation index system was constructed. The comprehensive methods including set pair analysis, spatial Markov chain, and obstacle degree model were used to analyze the dynamic characteristics and driving mechanisms of vulnerability in 225 inland cities in China. The findings reveal that vulnerability demonstrates significant multi-polarization and spatial club convergence. The overall vulnerability decreases annually, evolving from bi-polarization to multi-polarization, with a reduction of absolute difference. Vulnerability exhibits significant spatial auto-correlation and spillover effects, forming a spatial pattern where "high-vulnerability is clustered along the borders, while low-vulnerability diffuses from urban agglomerations". A low-vulnerability neighborhood environment increases the probability of a city's vulnerability transferring downward by 64.5%. The obstructive effects of both freight turnover volume and total import-export trade value weaken annually. In contrast, the hindering effect of shortest inland-port road travel time shifts from a relatively stable to a consistent increase. Furthermore, city hub level and connectivity impose significant constraints on vulnerability of non-hub cities.

To develop reasonable emergency material allocation plans for flood disasters, prospect theory was used to characterize the risk perception levels of decision-makers. The reference points for the risk perception of "disaster relief center-disaster site" two-tier decision-makers were defined. A simulation model was constructed using system dynamics to examine the impact of decision-makers' risk perception on the allocation of emergency supplies for flood disasters. The model's validity was verified with case study of 2023 Zhuozhou flood. The study revealed that satisfaction rate at disaster sites and degree of supply-demand imbalance were influenced by risk perception levels of two-tier decision-makers, with a curve representing these effects. The findings show that the risk perception levels of decision-makers at both disaster sites and relief centers have a marginal increasing effect on the disaster site satisfaction rate and supply-demand imbalance. Additionally, under different levels of material availability, the difference in satisfaction rates at disaster sites becomes smaller as decision-makers' risk perception becomes more pessimistic.

Existing methods for predicting temperature from subway carriage and tunnel fires mainly rely on physical models and empirical methods that are valid only under narrowly-defined environmental conditions. To solve this issue, this study adopts an artificial-intelligence-based approach. A GA-BPNN network model is constructed by optimizing BPNN using a GA. The GA is employed to global optimize BPNN's weights and thresholds, after which the model is trained to predict the temperature distribution of both the subway carriage and the tunnel, thereby achieving intelligent inversion of the fire temperature field. The results show that, for subway carriage temperature prediction, GA-BPNN model yields a mean absolute error (MAE) of 8.17, a root mean square error (RMSE) of 9.76, and a coefficient of determination (R²) of 0.99. For tunnel temperature prediction, MAE is 3.95, RMSE is 5.63, and R² reaches 0.98. By comparing the results with those of the traditional BPNN, it is found that the GA-BPNN model outperforms the conventional BPNN in both prediction accuracy and generalization capability.

To mitigate dust dispersion during electric shovel excavation and loading operations, an externally mounted high-pressure air curtain closed-loop jet dust removal system was designed. This system isolated dust generated during dumping operations at the shovel front from the surrounding environment, creating a low-disturbance zone. When only the airflow system was activated, dust was drawn into the negative-pressure dust collection port through a predetermined closed-loop path via the combined action of jet outlets and the collection port. Subsequent activation of the spray system intensified droplet fragmentation under high-velocity jet impact, enabling thorough mixing of mist particles with dust-laden airflow before settlement. Numerical simulations were conducted to analyze the shovel's external flow field characteristics and droplet distribution during system operation. Four jet deflection angles were evaluated to determine optimal flow field distribution. Simulation results show that the optimal closed-loop entrainment is achieved when the jet velocity is 30 m/s and the deflection angle is 30°. Spray activation enhances droplet fragmentation through high-speed airflow impact, significantly improving dust wetting coverage in front of the shovel body. Airflow field simulations were confirmed through smoke visualization tests showing high similarity to observed flow patterns. Dust concentration measurements after spray activation demonstrated the best dust control effect around the shovel when the jet deflection is 30°, verifying the simulation accuracy of the external flow field and establishing optimal operational parameters.