In order to explore the effects of sway on keyboard input of operators in DCS and to reduce human errors, an experimental study was conducted under four sway conditions—static, low, moderate, and high. Operators' performance in numeric, alphabetic, and alphanumeric input tasks, as well as their subjective ratings, were measured. Statistical analyses were employed to examine the effects of sway and to compare performance differences across input types. The results show that sway has a significant main effect on keyboard input performance. Compared with the static, low, and moderate sway conditions, the input time and number of corrections for alphabetic and alphanumeric entries significantly increase under the high-sway condition, whereas the accuracy of numeric and alphanumeric inputs significantly decrease. Significant differences are also found among the three input types: numeric input achieves the highest accuracy and shortest completion time, while alphabetic input requires the longest time. Sway also significantly affects subjective evaluations. Under the high-sway condition, perceived input difficulty, physical discomfort, visual discomfort, and workload ratings are all significantly higher. Under low and moderate sway conditions, input performance and perceived difficulty are largely consistent with the static condition, whereas under moderate sway, the perceived difficulty and workload for alphabetic input are significantly higher than those under low and static conditions. Under high sway, overall input performance markedly declines, accompanied by pronounced increases in difficulty, discomfort, and workload.

To improve the efficiency of large-scale personnel evacuation and reduce casualties after toxic gas leakage accidents, a CFD model and wind direction statistical probability were adopted to simulate the spatiotemporal distribution of toxic gas concentrations. Combined with the facilities and personnel conditions of evacuation sites, a dynamic comprehensive risk measurement method for evacuation sites was proposed. First, based on the differences in dynamic comprehensive risks of evacuation sites, a multi-trip emergency evacuation bus scheduling model with splitable evacuation demand was constructed, with the objectives of minimizing evacuation time and total risk expectation. Then, the augmented weighted Chebyshev method was used to convert the multi-objective model into a single-objective model, and a genetic algorithm integrated with adaptive large neighborhood search was designed for solution, in which destruction and repair operators were adopted to improve the search capability of the algorithm. Finally, the rationality of the model and the effectiveness of the algorithm were verified through numerical example analysis.The results show that the emergency evacuation bus scheduling model with the objectives of minimizing evacuation time and total evacuation risk expectation can generate scheduling schemes that prioritize the evacuation of personnel from high-risk evacuation sites, thereby improving evacuation efficiency and reducing the total risk expectation. Compared with the expected total evacuation risk, evacuation time is more sensitive to changes in the latest time window of evacuation sites, and the number of required buses decreases with the extension of the evacuation time window. In comparison with the traditional genetic algorithm, the genetic algorithm integrated with adaptive large neighborhood search can reduce the objective functions by 4.46% and 2.44%, respectively.

To address the low efficiency and safety of personnel evacuation in fire scenarios amid China's accelerating urbanisation and the increasing number of densely populated venues, a single-story building fire evacuation model based on the A^* algorithm was developed. A grid-based model was introduced and combined with the PSO algorithm to comprehensively account for multiple complex factors influencing the evacuation process, including smoke concentration, high-temperature environments, hazardous gas dispersion, and personnel density. Based on the single-story building fire evacuation model, multivariate functional relationships were used to quantify the influence coefficients of various factors on personnel movement velocity, thereby achieving a precise description of the evacuation process. During numerical simulation experiments, the performance of the original and improved A^* algorithms was compared across scenarios with varying occupant numbers within the building. The results indicate that compared to the traditional A^* algorithm, the improved model reduces evacuation time by 20.9% and path length by 3.27%. It can effectively prevent evacuation paths from falling into local optima and avoid occupants entering dead ends.

To improve the prediction accuracy and reliability of slope instability in mine waste dumps, a hybrid GA-BP model was developed by integrating an improved GA with a BP neural network. The model employed GA to globally optimize the initial weights and thresholds of the BP network, and incorporated the Levenberg-Marquardt (LM) algorithm to enhance convergence speed. Ten key parameters—including bench slope angle, geotechnical internal stress, bench height, surface displacement, and pore water pressure—were selected as inputs, with the slope safety factor as the output. Training and validation was performed using 150 field case datasets. The results show that GA-BP model reduces the mean squared error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) by 46.9%, 25.4%, and 5.38%, respectively, compared to the conventional BP model. Predictions are closer to the safety threshold (F_s = 1.2), indicating enhanced sensitivity and stability. Pearson correlation analysis confirms strong relationships between surface and internal displacement (0.98) and between pore water pressure and rainfall (0.75), supporting the rationality of the input indicators. The study demonstrates that GA-BP model effectively overcomes local optima and gradient vanishing issues in BP networks, providing a reliable tool for intelligent slope stability assessment.

Aiming at the problem of traffic offenders evading law enforcement by prejudging regulatory measures, this study explored the interference mechanism of compliance camouflage strategies on the reconnaissance behavior of traffic offenders, to improve the anti-reconnaissance efficiency of traffic management. Based on the dynamic game theory of incomplete information, a sequential decision-making model between managers (government/enterprise) and offenders during the reconnaissance stage was constructed. The signal distortion mechanism was used to quantify the interference effects of four camouflage strategies (honest performance, camouflage government style, camouflage enterprise style, random interference) on the Bayesian belief updating of offenders, and the optimal strategy combination was solved using the refined Bayesian Nash equilibrium. An empirical study was conducted based on the road network of Xi'an High-tech Zone. The results show that differentiated equilibrium conditions exist in four typical scenarios. The government core area strategy increases the post-reconnaissance compliance probability by 82.3%. The enterprise park strategy induces a 40.3% rise in surface reconnaissance behavior. The suburban combination strategy reduces the violation rate by 41.7%. And the transportation hub area maintains a dynamic compliance index of 0.87±0.03. The spatio-temporal evolution demonstrats that the camouflage strategy delayed the convergence period of offenders' beliefs by 42%, drove an exponential decay of the violation rate by 73.7% within 30 days, and achieved an input-output ratio of 194%. This study indicates that, within legal frameworks such as warnings and indications, dynamic camouflage strategies can reverse information disadvantage by interfering with offenders' cognitive decision-making, thereby constructing a "cognitive-spatial-economic" collaborative governance paradigm and promoting the transformation of traffic supervision toward active intervention.

To enhance the safety risk management level of UAV logistics distribution, a cascading failure evolution model for UAV logistics distribution safety risks was constructed based on complex network theory. Through simulated evolution experiments, the risk evolution characteristics in specific scenarios of UAV logistics distribution were revealed. By analyzing 63 interview transcripts, safety risk factors and their interrelationships were identified, and the credibility of these findings was verified by comparing them with relevant literature analysis results. The safety risk factors were categorized into five types: personnel, machinery, materials, methods, and environment. A directed weighted complex network with 69 nodes and 469 edges was then constructed. Using the Gephi platform, the overall topological structure characteristics of the complex network were calculated and analyzed, confirming the necessity of dynamic risk evolution analysis. The importance levels of nodes in the complex network were classified as high, medium, and general, and the development stages were classified as latent, diffusion, and occurrence, which laid the foundation for the construction of the dynamic cascading failure evolution model. Risk propagation probability and risk load redistribution rules were defined to quantitatively characterize risk evolution. A Python program was designed to conduct simulation evolution experiments. Key experimental groups were defined, and the evolution network of UAV logistics distribution safety risks was constructed. The simulated evolution results were analyzed from dimensions such as key evolution nodes and key evolution paths. The results show that in the evolution network, nodes s15, s17, and s18 in the machinery category are the most important, and the key evolution paths starting from this category of factors are the most abundant. The loss of control of nodes s54 and s57 in the environment category is the primary cause of control failure in the machinery category nodes. Therefore, strategies should be formulated to focus on controlling factors in the environment and machinery categories, such as immature external supervision and equipment signal issues, to achieve forward shifting of safety risk control.

This study analyzes the muscular fatigue characteristics of mine emergency rescue personnel during pull-down dynamometer training, aiming to provide scientific support for enhancing the safety and efficiency of mine emergency rescue training. Forty team members from the Datong National Mine Emergency Rescue Team were selected as participants. sEMG technology was used to record changes in muscle activation, muscle contribution rates, root mean square(RMS) amplitude, and MF under standardized pull-down dynamometer training conditions. The results indicate that the triceps and latissimus dorsi are the primary force-generating muscles, while the erector spinae plays a critical role in movement restoration and trunk stabilization. As the number of repetitions increases, the triceps brachii begins to appear fatigue in the second half of training. To maintain the training, the latissimus dorsi and erector spinae show enhanced compensatory activation, and the lower segment of the erector spinae exhibits a higher activation level while showing significant fatigue characteristics. This compensation pattern driven by triceps fatigue is the key inducement for increased lumbar load and elevated injury risk in rescue personnel. It is suggested that the strength endurance of triceps brachii and the stability control and anti-fatigue ability of core muscles should be strengthened in training.

In order to address the issue of human-machine collaboration failure in the takeover decision-making of autonomous driving and achieve precise recognition of the driver's cognitive state, this paper simulates three typical scenarios, night high-speed curves, mobile phone distracted driving, and combined scenarios of strong light and heavy rain, collecting and analyzing the dynamic interaction data of driving behavior and eye movement features construct a dynamic weight allocation(DWA) distribution feature fusion framework for multimodal perception and cognition collaboration, constructed DWA-RF model, and explore the dynamic mechanism of the driver's cognitive state and takeover decision-making behavior in complex environments. The results show that the distracted state significantly prolongs the takeover time. In scenes of strong light and heavy rain, the superimposition of distraction and environmental pressure leads to a sharp reduction in the range of scanning. Fatigue causes an increase in lane departure distance, accompanied by a reduction in pupil diameter and abnormal scanning behavior. The cognitive state classification accuracy of DWA-RF constructed in this paper has reached 93.6%, verifying the effectiveness of this model in identifying the driver's cognitive state for autonomous driving takeover decisions.

To safeguard employees' psychological health and improve the accuracy and interpretability of psychological stress evaluation methods, taking multimodal physiological time-series data as the research object, a LSTM with AM(LSTMA) method was proposed to accurately evaluate employees' psychological stress states in the paper. Firstly, using the multimodal physiological time-series data (Blood Volume Pulse (BVP), Electrocardiogram (ECG), Electrodermal Activity (EDA), Electromyogram (EMG), Respiration (RESP), Body Temperature (TEMP), and three-axis Acceleration (ACC)) from the WESAD dataset were adopted as the research carrier, the gating memory mechanism of the modal-specific LSTM modules was utilized to accurately capture cross-time-step temporal dependency features, effectively retain key physiological features strongly associated with psychological states, and filter out short-term random noise, thereby ensuring that the physiological feature data could truly characterize the dynamic evolution of employees' psychological states. Secondly, after feature fusion, the attention mechanism was introduced to adaptively assign attention weight coefficients based on the feature importance of physiological data across different modalities and time steps, enhancing key features and micro-response features sensitive to psychological stress states while suppressing the interference of redundant information. Finally, the accurate evaluation of psychological stress states was accomplished through a fully connected neural network. Experimental results show that the LSTMA method achieves an evaluation accuracy of 94.56% for the four-classification task (neutral, stress, pleasure, and meditation) of psychological stress states. After Leave-One-Out Cross-Validation (LOOCV), the accuracy is improved to 98.08%. Ablation experiments verify the synergistic enhancement effect of the modal-specific LSTM and the attention mechanism, and model interpretability analysis further confirms the scientificity and rationality of LSTMA.

To investigate the spread characteristics of forest fires under complex topography and multi-factor coupling conditions, this study develops an optimized forest fire spread model that integrates terrain-slope correction, wind-field effects, and vegetation indices. First, Gaussian filtering was applied to correct the digital elevation model (DEM) to reduce noise, and terrain slope and aspect were derived from the refined DEM. Subsequently, the enhanced vegetation index (EVI) was introduced to improve the forest fire spread prediction model, enhancing prediction accuracy in areas with dense vegetation cover. By combining the model with CA, the predicted fire spread can be visualized. Finally, the predicted fire variable values were compared with the observed data from Muli Tibetan Autonomous County to verify the scientific validity and effectiveness of the model. The results indicate that the model is highly sensitive to vegetation changes in low EVI value ranges, with an effect size of 0.870, suggesting that the introduction of EVI improves fire prediction accuracy in areas with high vegetation cover. The improved fire spread model achieved an area prediction error rate and perimeter error rate of 29.40% and 5.79%, respectively, which are lower than the pre-improvement values of 44.27% and 16.99%. The Kappa coefficient of the improved model is 0.8238, which is closer to 1 compared to the pre-improvement model.