Objective: Rapid and accurate assessment of dynamic threats to surrounding assets, such as transmission lines, chemical storage tanks and nuclear facilities, during the propagation of forest-grassland wildfires is critical for emergency response and disaster mitigation. Traditional fire risk assessment methods typically rely on static hazard maps or single-time fire-front predictions and therefore fail to adequately account for the continuous spatiotemporal evolution of fire behavior influenced by changing environmental factors (e.g., wind and fuel), leading to cumulative error propagation in dynamic risk prediction. Furthermore, the complex and heterogeneous failure mechanisms of different infrastructure types are seldom incorporated into a unified, dynamically updating risk framework. Therefore, this study aims to develop a novel, integrated risk assessment method capable of continuously updating infrastructure failure probabilities as wildfire intensity and extent evolve. Methods: The proposed framework synergistically integrates a data-driven wildfire spread prediction model with a dynamic Bayesian network (DBN). The fire-affected landscape and key protected assets were first discretized into interconnected spatial nodes. A DBN structure was then constructed to capture two fundamental dependencies: the temporal autocorrelation of fire conditions at each node (how the state evolves) and the spatial dependencies between adjacent nodes (how fire spreads from one location to another). The core parameters of this DBN were informed by a data-driven fire spread model. Specifically, predictions of key fire behavior metrics, namely, spread rate and fireline intensity, were generated for each node. Then, DBN conditional probabilities were updated over time using the predicted spread rate and the fireline intensity, which were further mapped to asset-specific fragility models to infer evolving failure probabilities. Results: Using the proposed framework, we built a kilometer-scale synthetic fire scenario using every single protected asset and showed that the 24-h failure probabilities of transmission lines, chemical tanks, and nuclear facilities increased by up to 22%, 68% and 20% across wind-fuel scenarios. In addition, a case study based on observations from the 2022 Oak Fire (California, USA) was conducted using multiple protected assets. The 24-h failure probabilities reached 0.24 for transmission lines, 0.14 for hazardous-chemical tanks, and 0.53 for the nuclear facility. Furthermore, we evaluated a firebreak construction strategy: as the suppression effect on the rate of spread increased, the failure probability could be reduced to as low as 0 for transmission-line nodes and to 0.12 for the nuclear facility, whereas the probability reduction for hazardous-chemical tanks was not significant due to their spatial distribution and structural differences. Conclusions: This study proposes a novel dynamic wildfire risk assessment framework that integrates data-driven spread prediction with probabilistic graphical modeling. By continuously updating a DBN using fire behavior model outputs, the proposed approach effectively mitigates cumulative prediction errors in a spatiotemporally evolving environment and provides quantitative support for rapid response, emergency resource allocation, and intervention strategy assessment.
Objective: In pulverizing systems, elevated ambient temperatures make coal dust deposited on hot surfaces prone to spontaneous ignition; once entrained by airflow into the surrounding space, such deposits may further evolve into coal dust cloud explosions, posing major risks to safe operations. The objective of this study is to reveal the critical conditions and kinetic characteristics under which self-ignition of deposited coal dust initiates explosions in such systems. Methods: In this study, a coal dust combustion—explosion experimental platform was combined with Fluent numerical simulations to conduct a systematic investigation of the self-ignition, ejection, and explosion processes of deposited coal dust under heated environments, as well as behavioral characteristics of these processes. First, a coal dust combustion—explosion experimental system capable of precise control of ambient temperature and hot-plate temperature was established; on this basis, the internal temperature distribution of coal dust layers at different hot-surface temperatures and the critical conditions for self-ignition were examined. Second, ejection tests were conducted with dust-layer center temperatures of 260-380 ℃ to analyze the formation, ignition, and explosion of deposited coal dust clouds. Finally, based on the experimental results, a Fluent-based numerical model was developed for the post-lofting processes, including moisture evaporation, devolatilization, gasphase combustion, and char combustion. The particle trajectories, velocities, and temperature evolution during lofting were analyzed, and the critical oxygen concentration and dust concentration required to induce an explosion were determined. Results: Experimental results show that the hotplate temperature required for thermal runaway decreases with increasing dust-layer thickness. For a 4 mm coal dust layer, the critical hotplate temperature for self-ignition is 255 ℃; when the thickness increases to 10 mm, this temperature drops to 225 ℃. When the internal temperature of a self-ignited dust layer lies within 275-395 ℃ (corresponding to a center temperature of 300-340 ℃), the resulting coal dust cloud can trigger an explosion. The lofting process can be divided into three stages—rapid ejection, decelerating diffusion, and free diffusion—with explosions occurring mainly in the latter two; the maximum particle velocity is approximately 60 m/s, and the peak temperature is approximately 2 400 ℃. Mechanistically, volatiles released from coal particles undergo homogeneous combustion outside the particles, whereas char undergoes heterogeneous combustion within the particle interior; these simultaneous phenomena significantly elevate the flame-core temperature. As the ambient temperature increases, both the critical coal dust mass concentration and the critical oxygen concentration for explosion decrease: when the ambient temperature rises from 120 ℃ to 200 ℃, the critical dust cloud concentration decreases from 380 to 95 g/m3, and the critical oxygen concentration decreases from 21% to 13%. Conclusions: Combustible gases (e.g., CO) generated during the self-ignition stage accumulate and are subsequently ignited by high-temperature particles, initiating gas-phase combustion. The resulting heat release then ignites suspended coal dust particles, triggering solid-phase combustion. This sequence constitutes a critical pathway by which self-ignition escalates into an explosion. These findings provide a theoretical basis for explosion prevention and control in pulverizing systems and offer practical guidance for risk mitigation measures, including hot-surface temperature control, dust-layer thickness management, ventilation and oxygen concentration limits, and operational strategies that minimize the lofting of preheated deposits.
Objective: Wildfire spread exhibits a high degree of uncertainty, largely owing to complex fire-atmosphere interactions. Heat released during combustion induces rapid temperature increases and highly variable airflow, creating a strongly nonstationary environment. However, existing studies typically rely on the Reynolds averaging and decomposition(RAAD) method to analyze fire-induced turbulence in airflow. RAAD assumes that the airflow field is stationary within a fixed averaging window, an assumption that fundamentally contradicts the transient nature of wildfire environments. Consequently, RAAD often fails to effectively separate large-scale trends and submesoscale motions from genuine turbulent fluctuations, leading to a systematic overestimation of turbulence parameters. This study aims to overcome the limitations of traditional stationarity assumptions by proposing a turbulence reconstruction method based on a nonstationary model. The primary objective is to accurately characterize the vertical distribution and heterogeneous features of turbulence over a fire zone during the passage of a heading fire, thereby providing a reliable physical basis for improving wildfire behavior prediction, smoke transport simulation, and the parameterization of coupled fire-atmosphere numerical models. Methods: High-frequency(10 Hz) three-dimensional wind speed and temperature data were obtained from a large-scale prescribed burn experiment conducted in the New Jersey Pine Barrens and archived by the USDA Forest Service Research Data Archive. The analysis focused on measurements collected from a meteorological tower located within the burn unit during the passage of a heading fire. Three sonic anemometers were installed at heights of 3, 10, and 20 m. To address the nonstationarity of the observations, empirical mode decomposition(EMD) was employed. EMD is an adaptive, data-driven signal processing technique suited for analyzing nonlinear and nonstationary processes. Unlike RAAD, which employs a fixed time-averaging window, EMD decomposes wind and temperature signals into a set of intrinsic mode functions based on local characteristic time scales. This method separated the original signal into three components: a time-varying mean wind flow, sub-mesoscale motions(low-frequency trends), and turbulent fluctuations. A frequency threshold of 0.01 Hz was applied to distinguish turbulent motion from low-frequency motions. The turbulence components were then reconstructed to calculate key parameters, including turbulence kinetic energy(TKE), friction velocity, and sensible heat flux. These results were rigorously compared with those obtained using the traditional RAAD method to quantify the biases introduced by the stationarity assumption. Results: Spectral analysis validated the proposed method, as the reconstructed turbulence power spectra at all measurement heights closely followed the theoretical -2/3 slope in the inertial subrange. The comparative analysis revealed several critical findings: 1. The RAAD method significantly overestimated turbulence intensity. Peak TKE values derived from RAAD were approximately twice those obtained using the nonstationary EMD-based method. Similarly, peak friction velocities calculated using RAAD were approximately four times higher than those derived from the proposed method. This discrepancy indicates that RAAD misinterpreted rapid nonstationary trends(e.g., the abrupt increase in wind speed induced by the advancing fire front) as turbulent fluctuations. 2. Analysis of the reconstructed data revealed distinct vertical patterns. During the passage of the heading fire, the magnitudes of TKE and turbulent momentum flux at heights of 3 and 10 m were comparable. By contrast, values at a height of 20 m were significantly higher than those in the lower layers, suggesting enhanced shear or entrainment activity near the canopy top or at the plume-atmosphere interface. 3. A crucial discrepancy was observed in the vertical distribution of sensible heat flux. The nonstationary model showed that the peak turbulent sensible heat flux was highest at 3 m and decreased with height (3 m > 10 m > 20 m), which is consistent with the physical reality that the primary heat source is located in the surface fuel bed. By contrast, the traditional RAAD method yielded physically unreasonable results, showing higher heat flux peaks at 20 m than near the surface. This error arose because RAAD failed to filter out nonturbulent, low-frequency temperature ramps caused by the approaching fire front. Conclusions: This study demonstrates that the traditional stationarity assumption is not applicable to the highly transient environment of a heading wildfire. By employing an EMD-based nonstationary model, the proposed method effectively isolated the time-varying mean flow and sub-mesoscale motions, thereby retrieving physically representative turbulence statistics. The results confirm that traditional methods systematically overestimate turbulence parameters and can distort the vertical profile of heat transport. The corrected turbulence characteristics identified in this study—specifically, the vertical attenuation of sensible heat flux and the distinct intensification of turbulent momentum flux at higher altitudes—offer critical empirical evidence for calibrating subgrid-scale turbulence parameterization schemes in coupled fire-atmosphere models (e.g., WRF-Fire). Furthermore, accurate quantification of friction velocity and TKE offers significant guidance for improving the prediction accuracy of firebrand transport (spotting) and near-field smoke dispersion.
Objective: Concave vertical structures are widely integrated into modern architectural systems, such as building facades, enclosed ventilation shafts, and recessed corridor components. Owing to their confined geometric boundaries, these structures induce fire plume behaviors that differ fundamentally from those in open spaces. Such behaviors include abnormal plume propagation, flame morphology distortion, and non-uniform temperature distribution, which may accelerate fire spread and increase evacuation and rescue risks. However, existing studies have primarily focused on fire plumes in open or simple confined spaces, with limited systematic investigation of the intrinsic evolution mechanisms of plume characteristics within concave vertical configurations. In particular, the coupling effects between structural dimensions and fire source conditions remain insufficiently understood. To address this gap, this study systematically investigates the evolution laws of key fire plume parameters (namely, average flame height and vertical temperature distribution) in concave vertical structures. Specifically, the objectives are to (a) clarify how the coupling of structural dimensions (width and depth) and fire source position regulates oxygen supply and heat feedback during plume development, (b) reveal the dominant mechanisms driving plume parameter variations in concave configurations, and (c) establish quantitative dimensionless models for characterizing average flame height and vertical temperature distribution. Ultimately, this study aims to provide a theoretical foundation for improving dynamic fire evolution models and formulating targeted firefighting strategies, thereby enhancing fire safety management for buildings incorporating concave vertical components. Methods: A 1∶10 reduced-scale experimental platform was constructed based on the Froude similarity criterion to ensure the scalability of fire dynamic behaviors. The experimental matrix included multiple concave structural configurations. The concave region width was set to 0.3—0.6 m (corresponding to 3—6 m at full scale), depth ranged from 0.3 to 0.9 m (3—9 m full scale), and the test structure height was 1.55 m (15.5 m full scale). Fire experiments were conducted under varying heat release rates (HRRs) of 23.69—65.14 kW, corresponding to 7.4—20.5 MW at full scale, to represent different fire intensities. Key parameters were measured as follows: average flame height was obtained using continuous image acquisition (50 fps) and image processing (to extract flame contours and calculate time-averaged heights); vertical temperature distribution was recorded using K-type thermocouples spaced at 0.1 m intervals, covering heights from 0.1 m to 1.2 m along thecentral axis of the concave structure. After data acquisition, dimensionless analysis was performed to normalize structural and fire parameters. A dimensionless structural factor integrating width, depth, and height was introduced to quantify geometric confinement effects. Regression analysis was then applied to establish predictive models for average flame height and vertical temperature distribution. Results: The experimental results reveal distinct variation patterns of fire plume parameters under concave vertical constraints: (1) The average flame height increases monotonically with HRR, rising from 0.31 m at 23.69 kW to 0.68 m at 65.14 kW (maximum tested condition). (2) Temperature decreases with increasing height. In the near-field region (≤0.3 m above the source), temperature drops by 40%—50% from the peak value; in the far-field region (>0.3 m), the decrease rate slows to 15%—20%. Structural width affects temperature distribution. Narrower configurations (0.3 m width) produce a more concentrated high-temperature zone confined within 0.2 m above the fire source, whereas wider structures (0.6 m width) exhibit a more dispersed thermal distribution. (3) Further analysis confirms that the coupling between structural dimensions and fire source position governs oxygen supply and heat feedback. Smaller structures restrict oxygen entrainment, leading to localized incomplete combustion, while simultaneously enhancing wall radiant feedback. These two mechanisms jointly modify plume morphology, flame height, and vertical temperature profiles. The introduction of the dimensionless structural factor enabled the development of predictive models for both average flame height and vertical temperature distribution. The models achieved coefficients of determination (R2) exceeding 0.84, demonstrating strong fitting reliability and predictive capability. Conclusions: This study systematically clarifies the evolution mechanisms of fire plumes in concave vertical structures. The main conclusions are as follows: (1) Variations in fire plume parameters are governed by the coupled effects of structural dimensions and HRR, with oxygen supply restriction and heat feedback enhancement identified as the dominant mechanisms. (2) The proposed dimensionless structural factor effectively quantifies concave geometric confinement effects, and the developed models accurately predict average flame height and vertical temperature distribution. (3) The findings enhance the accuracy of dynamic fire evolution modeling in concave architectural configurations and provide practical guidance for firefighting strategies, such as cooling structural walls in deep concave spaces to mitigate heat feedback effects. Future research should validate these conclusions via full-scale experiments and expand the analysis to different fire source types (e.g., liquid and solid fuels) to improve model generalizability.
Objective: Steeply inclined cable tunnels are critical components of pumped storage power stations, and any fire incident within these components poses a severe threat to the operational safety of power stations. Unlike conventional horizontal tunnels, steeply inclined tunnels are characterized by unique characteristics, namely high drop, extended length, and large slopes. These unique geometric features create distinct fire dynamics, which render standard horizontal-tunnel assessment methods inadequate for evaluating specific risk profiles. Fire incidents in steeply inclined cable tunnels may be triggered by various factors, such as electrical short circuits, which can cause large-scale losses. Therefore, it is imperative to develop a specialized fire risk assessment framework for steeply inclined cable tunnels. Methods: This study developed a risk-matrix-based fire-risk assessment method for steeply inclined cable tunnels. First, 19 basic events that may induce fires within steeply inclined cable tunnels were identified via literature review, historical case analysis, and field investigations. Based on these identified events, a fault tree was established to analyze fire probabilities under varying slopes, evaluate the significance of each basic event, and determine key basic events requiring prioritized prevention and control. Second, severity rating standards for fire consequences were formulated based on three technical criteria: ignition temperature of the cable, bearing capacity of the tunnel structure, and fire resistance performance of the fire compartment. Using computational fluid dynamics (CFD), the ceiling temperature and fire duration in the tunnel under fire scenarios were simulated, after which the severity ratings for fire consequences were determined. Finally, a comprehensive fire risk matrix was established, followed by the determination of the overall fire risk level for steeply inclined cable tunnels based on the integrated fire occurrence probability and fire consequence severity level. Results: Using the steeply inclined cable tunnel of the Jixi pumped storage power station in Anhui Province, China, as a case study, the fire risk levels were evaluated for four slopes (0°, 15°, 30°, and 45°). The results were as follows: 1) the fire risk ratings for slopes 0°, 15°, 30°, and 45° were found to be 3A, 3A, 2A, and 2A, respectively, all of which fall within the low risk range. 2) As the slope increased, the fire occurrence probability increased from 6.50×10-5 to 8.88×10-5. Meanwhile, as the slope increased, the ceiling temperature and fire duration within the tunnel decreased from 633℃ to 311℃ and from 0.65 h to 0.53 h, respectively, decreasing the fire consequence severity. 3) The importance indices of the fault tree revealed that X13 (failure of the high-voltage grounding system), X11 (decrease in the dielectric strength), and X8 (aging of the insulating layer) represent the key basic events that exert the strongest impact on the fire-occurrence probability. Thus, the monitoring and control of the events must be prioritized. Conclusions: By analyzing the data obtained from fault tree analysis and CFD simulations, this study quantified fire risk levels for steeply inclined cable tunnels across varying slopes. Furthermore, key basic fire-triggering events were identified. Accordingly, targeted prevention and control measures are proposed. Compared with conventional horizontal cable tunnel assessment methods, this framework accounts for the influence of slope changes on the fire occurrence probability and fire-consequence severity, providing results that are more suitable for steeply inclined cable tunnels of pumped storage power stations.
Objective: The simultaneous occurrence of fire-related smoke and combustible gas leakage, particularly methane, poses a significant and complex threat in various environments, including residential kitchens, commercial catering establishments, and industrial workshops. Conventional detection systems are typically designed to monitor a single hazard type, requiring multiple sensors for comprehensive risk coverage. This research aimed to develop a novel, highly integrated, and cost-effective composite detector that can simultaneously monitor smoke and methane. Methods: The foundation of our proposed method is a composite detection model utilizing the optical properties of a single 1 653.7 nm laser. This wavelength was specifically chosen because it aligns with a characteristic absorption line for methane and is capable of inducing Mie scattering from smoke particles. The main methodological challenge was to decouple the composite signals, as methane absorption and smoke extinction attenuate the transmitted laser intensity. To achieve this, a sophisticated signal processing strategy was developed. We employed tunable diode laser absorption spectroscopy in a time-division multiplexing scheme using a single triangular wave scan. The rising edge of the scan incorporated high-frequency harmonic modulation, enabling wavelength-modulation spectroscopy (WMS) to extract the second-harmonic signal. By contrast, the falling edge of the scan remained unmodulated, allowing for direct absorption spectroscopy (DAS) to provide robust measurements for higher methane concentrations. The decoupling of the smoke signal was achieved by analyzing the baseline of the scattered light signal, which corresponds to wavelengths where methane absorption is negligible, making its intensity directly proportional to smoke particle concentration. In addition, to mitigate the effects of smoke on methane measurements, all absorption signals were normalized during processing. To further enhance the detector's performance, a novel optical structure was designed to address the inherent weaknesses of the 1 653 nm laser, specifically, the relatively weak scattering signal from smoke particles and the need for a long optical path for sensitive methane detection. We implemented a multimirror, "cross-beam" optical path within a compact chamber, which effectively folds the laser beam, causing it to traverse the detection volume multiple times before reaching the photodetector. This innovative configuration resulted in a sixfold extension of the effective absorption path length for methane, increasing it from the standard 10 cm to 60 cm. By optimizing the placement of the mirrors and the angle of the scattering detector, we configured the system to collect scattered light from high-intensity beam-crossing regions, leading to a fourfold enhancement of the collected smoke scattering signal. Results: In single-analyte tests, the detector exhibited high sensitivity, achieving a lower detection limit of 0.008% volume fraction for methane. The hybrid WMS-DAS approach proved effective, with WMS yielding more accurate results for concentrations below 0.25%. A linear regression analysis of the measurement data indicated excellent linearity (R2=0.9833). For smoke detection, the device achieved a detection limit of 0.05 dB·m-1. The critical validation was the composite detection experiment, in which methane detection was performed under various stable background smoke densities (0, 0.1, 0.5, and 1.0 dB·m-1). The results conclusively demonstrated the effectiveness of the signal decoupling methodology, showing that the presence of smoke, even at high densities, did not significantly affect the accuracy of methane concentration measurements. However, at the highest smoke densities (0.5 and 1.0 dB·m-1), the WMS signal exhibited increased fluctuations, indicating that extreme smoke levels can degrade the signal-to-noise ratio, setting a performance boundary for the current system. Conclusions: This study proposed and validated a novel method for the synchronous detection of smoke and methane using a single 1 653 nm laser source. A functional prototype of the composite detector demonstrated the feasibility and effectiveness of the approach. Major innovations include a composite detection model that enables signal decoupling through a time-division WMS/DAS scheme and a performance-enhancing cross-beam optical path that extends the absorption length sixfold and enhances the scattering signal fourfold. Most importantly, it demonstrated robust performance in composite scenarios with minimal cross-interference. By overcoming the limitations of conventional systems, this single-source composite detection method offers a promising, low-cost, and highly integrated solution, providing a new technical pathway for advanced early warning systems in complex, multihazard environments.
Objective: In automated aircraft assembly, multifunctional end effectors and one-shot drilling, reaming, and countersinking processes are commonly used for riveting holes in stacked structures. The pressure foot integrated into the end effector monitors and compensates for workpiece deformation effects on countersinking accuracy. However, due to structural constraints of the pressing unit and application-specific limitations, the pressure foot cannot adaptively conform to the workpiece surface contour, leading to systematic compensation errors. To understand the mechanisms behind countersinking errors and compensation errors during automated drilling of stacked structures, this study designs a frame-type stacked specimen that accurately reflects the deformation characteristics of aircraft structures during countersinking. Methods: Finite element simulations and machining experiments were conducted to quantitatively analyze how typical aircraft structural parameters (e.g., geometric features) and hole location parameters influence deformation, countersinking errors, and depth compensation errors. Results: Countersinking errors manifest as depth, normal vector, and apex angle errors. In the frame-type stacked setup, the lower layer influences the upper layer's deformation through its support effect, while the ribs increase frame's stiffness and reduce localized deformation of the upper layer. Even when normal vector detection and adjustment are performed before drilling to ensure the pressure foot and tool are perpendicular to the stack, the axial drilling force causes stack deformation. Since the pressure foot is rigidly attached to the pressing unit and cannot adapt to workpiece deformation, a relative deviation occurs between the tool's feed direction and the actual normal vector at the drilling point. Although the displacement sensor in the pressing unit can monitor the pressure foot's displacement and compensate for stack deformation online, the pressing position of the pressure foot does not align with the center of the countersinking hole. This mismatch leads to different local deformations at these two points, which is the fundamental cause of the systematic error in the pressure foot-based compensation method. The size of this systematic error increases as the structural rigidity at the drilling location decreases. Under a constant axial drilling force, the deformation behavior depends on the structural geometric parameters of the frame-type stack and the hole position. Specifically, when the hole is farther from the intermediate rib, the lower layer mainly undergoes bending deformation of the frame's top surface, and the deformation characteristics to the right of the vertical rib significantly impact countersinking errors. The frame height and structural layout to the left of the vertical rib primarily influence the overall torsional deformation of the stack; increasing the frame height improves stiffness and reduce torsion under axial load. While structural parameters have limited effect on normal vector and apex angle errors, they greatly influence depth and compensation errors. Notably, the compensation error remains stable within a certain range of structural parameters. Conclusions: Finite element simulations enable a quantitative analysis of the systematic deformation compensation error via the pressure foot and help identify structural parameter thresholds where the compensation error stays stable. This regular pattern makes it possible to eliminate systematic errors using a feedforward strategy. In engineering practice, zone-specific compensation values for countersinking depth can be defined in advanced, ensuring accurate and efficient depth control in frame-type stacked structures.
Objective: With the increasing complexity of operating conditions in humanoid robots, the risk of collision and impact failure in harmonic reducers has considerably increased. A critical issue arising from severe working environments is the generation of excessive impact forces inside the reducer. Such excessive forces not only accelerate gear wear and degrade transmission precision but also trigger a series of related faults, which in turn directly undermine the service life and operational reliability of the robot. Methods: To address this issue, this study proposes a novel calculation method for contact impact force to evaluate the impact resistance of harmonic gear drives. First, a common-tangent double-circular-arc flexspline tooth profile is constructed. On this basis, the circular spline tooth profile is established using the improved kinematic method. Thereafter, the initial assembled meshing model is built in accordance with the kinematic principles of the harmonic gear drive mechanism. Subsequently, based on this model, the meshing common normal line equation and backlash equation are established. All actual meshing impact points of the harmonic gear drive under load conditions are calculated, and the relative contact velocity under backlash conditions is derived. Based on the law of conservation of energy, the proposed method integrates the Hertzian contact theory with the Hunt-Crossley hysteretic damping model to establish an accurate contact-impact force prediction framework. Transient dynamic analysis is conducted to validate the proposed theoretical model. A comprehensive parametric study was subsequently conducted to quantify the effects of critical flexspline geometric parameters (convex arc radius, convex tooth center offset, convex tooth center displacement, and root wall thickness) with respect to the impact characteristics. Results: By calculating the parameters for the harmonic gear drive system, the following results are derived: (1) when the harmonic gear drive mechanism is subjected to a load torque of 92.95 Nm, with increasing contact time of the harmonic gear drive, the contact force of each tooth initially increases and then decreases, which is consistent with the meshing-in impact process between the flexspline and circular spline. (2) As the contact time increases continuously, the contact force of each tooth initially increases and then decreases. The maximum contact deformation occurs at Tooth No. 11, with a value of 2.03×10-4 mm, which is consistent with the amped collision model proposed by Hunt and Crossley. (3) In a dynamic calculation using the theoretically constructed model, as the impact velocity increases, the meshing force of the loaded tooth profile pairs gradually increases. The error for Tooth No. 8 is 4.65%, while that for Tooth No. 11 is 11.95%. Finite element analysis verifies the accuracy of the dynamic meshing force theory and calculation method. (4) The results of the parametric study reveal that the convex arc radius exerts the most pronounced effect on the peak impact force, whereas the root wall thickness has the least influence. Conclusions: This study not only provides a reliable analytical framework for predicting the evolution of the meshing impact in harmonic drives but also offers practical design guidelines for improving the load-carrying capacity and impact resilience of the drives in humanoid robotic joints and other precision transmission systems.
Objective: With advancements in three-dimensional integration technology, wafer stacking has become a critical process for enhancing semiconductor device performance in the post-Moore era. The reliability of interfacial electrical interconnections depends on the bonding overlay accuracy, which is now primarily limited by residuals at the 50 nm level. This study addresses the challenge of bonding residuals in high-precision wafer bonding, which arise from the coupled effects of wafer elastic deformation, clamping constraints, and bond wave propagation. Existing models often lack comprehensive multiphysics coupling or fail to establish a link between specific process parameters and residual formation, limiting their use in process optimization. Therefore, developing a high-fidelity coupled model is essential for understanding the residual generation mechanism and devising effective suppression strategies. Methods: A multiphysics coupling analysis model was developed that comprehensively considers wafer anisotropy, clamping boundary effects, and bond wave propagation behavior. The framework integrates anisotropic thin-plate elasticity (incorporating crystal orientation transformation tensors), gas film dynamics (governed by a modified Reynolds equation with bonding stress), and contact mechanics (solved via the augmented Lagrangian method). Bond wave propagation is governed by an energy criterion at the wavefront, balancing effective bonding energy against strain energy and the work performed by gas film and mechanical contact pressures. A finite element model for the 300 mm wafer bonding process was developed, achieving submicron accuracy. Key numerical strategies included a staggered iterative scheme for updating the wavefront, bonding force, gas pressure, and structural deformation; adaptive time stepping based on residual variations; and stabilization damping to suppress rigid body motion. Model validity was confirmed through comparison between simulation predictions and experimental pattern wafer geometry measurements. Results: The simulation accurately captured the nonuniform bond wave propagation induced by wafer anisotropy. The stress and residual distributions exhibited a distinct fourfold symmetry consistent with the crystallographic orientation. Residuals were primarily concentrated near the wafer edge, with additional significant residuals observed at the center-consistent with previous reports. Experimental validation showed strong agreement between simulated and measured residual distribution patterns. Systematic parameter studies revealed that using a flat bond head reduced the 2-norm of the residual vector by 41% compared with a spherical head (0.63 μm vs. 1.07 μm). Employing a lower-stiffness material (polyethylene) for the bond head reduced the residual 2-norm by 23% compared with PEEK plastic. Moreover, an increase in the initial wafer gap correlated with a higher residual 3σ value. Conclusions: This study establishes a robust multiphysics coupling and process co-optimization framework for high-precision wafer bonding. The proposed model effectively captures the combined effects of wafer anisotropy, gas film dynamics, and contact mechanics on residual formation, enabling high-fidelity simulation of the bonding process and quantitative analysis of key process parameters. The findings demonstrate that optimizing the bond head design-with a flat surface and low-stiffness material-and minimizing the initial wafer gap can significantly suppress bonding residuals. This work provides a theoretical basis and practical design guidelines for optimizing wafer bonding processes to achieve superior overlay accuracy.
Objective: The sliding guide used in the telescopic mechanisms of morphing aircraft typically exhibits large clearance and low stiffness, resulting in complex dynamic behavior that significantly affects mechanism performance. Most existing studies on sliding guide dynamics focus on machine tool guides, which differ substantially in structural and operating conditions. Therefore, a specialized dynamic model that accounts for guide flexibility and clearance effects is needed to accurately characterize the dynamic performance of telescopic sliding guides. Methods: A dynamic model of a sliding guide with clearance was developed. The guide rail was represented by finite element beam elements based on the Euler–Bernoulli beam theory to capture its elastic deformation. A multipoint contact detection method was introduced to avoid missed detections of contact states between the slider and the deformed guide. Contact may occur in either line or area contact modes. To compute the contact force in area contact mode, a variable stiffness contact force model was proposed; its damping coefficient was chosen manually. Because different contact mode calculations could produce large discrepancies in contact force and cause numerical instability, a modified variable stiffness contact force model was introduced for area contact, in which the damping coefficient was corrected using the material restitution coefficient. For line contact, the Flores contact force model was adopted. Friction forces were calculated using the Ambrósio modified Coulomb friction model. The dynamic equations of the slider and the guide were formulated using the Lagrange multiplier method with Baumgarte stabilization. A unified numerical solution strategy based on MATLAB's ode15s solver was implemented to simulate the dynamic response. Results: Numerical simulations revealed the influence of key parameters on dynamic behavior. When guide elasticity was included, peak contact forces decreased, but the lateral displacement of the slider increased, accompanied by sustained oscillation owing to cantilevered guide vibration. Larger clearance sizes yielded higher peak contact forces and larger amplitude oscillations in guide tip deflection, while the time-averaged friction force decreased. In a dual-stage, dual-guide configuration, the system exhibited more frequent collisions and chaotic lateral motion, with notable jamming caused by asynchronous deformation of the two guides. Reducing the interguide spacing mitigated this jamming effect. Experimental validation using a prototype with adjustable clearance showed that the equivalent friction coefficient decreased with increasing clearance under different actuation speeds and modes, consistent with simulation trends. This effect was more pronounced during deployment than during retraction and at lower speeds. The effect diminished at larger clearances, exhibiting nonlinear saturation. The deviation between simulated and experimental friction coefficients was within 30.00%, confirming the validity of the proposed dynamic model. Conclusions: This paper presents a comprehensive dynamic modeling framework for sliding guides in telescopic mechanisms that incorporates guide elasticity and clearance effects. The proposed contact detection method and modified contact force model increase modeling accuracy and numerical stability. The simulation and experimental results demonstrate that guide elasticity, clearance size, and guide configuration considerably affect dynamic behavior. These findings provide a theoretical foundation for the design and control of sliding guides in deployable aerospace mechanisms.
Objective: Clarifying the causal relationships among states, actions, and rewards in reinforcement learning (RL) for robotic control is crucial for enhancing policy interpretability and for ensuring safe and reliable decision-making. Many RL algorithms still rely on traditional neural network structures and are therefore treated as black boxes that cannot reveal the causal relationships between policy and observation space. Moreover, in high-dimensional and dynamically evolving state-action spaces, conventional attention mechanisms are not effective enough to capture the long-term causal dependencies between state variables and actions. This limitation restricts the explainability of autonomous control systems and poses safety risks when deployed in complex real-world environments. Methods: Therefore, this paper proposed a robotic motion skill interpretation framework based on a graph neural network-neural causal model (GNN-NCM). By replacing attention-based components with GNNs, the model inferred and captured causal influences in sequential decision-making. First, this paper applied conditional independence testing to discover the underlying causal graph and to identify how different state and action variables influenced one another over time. Using the learned causal structure, a GNN was trained to jointly represent nodes (states and actions) and edges (causal dependencies) and to perform both qualitative and quantitative causal inference. The GNN framework integrated structural causal discovery with neural message passing, enabling efficient learning of high-dimensional relationships while preserving interpretability. this paper implemented and validated the algorithm in two representative robotic control environments, LunarLander and Hopper-V4, which differ in control complexity and state dimensionality. this paper used multiple analytical tools, including state decomposition, action separation, and heatmap-based visualization, to assess causal strength and directionality of state-action-reward relationships. This work captured causal weights during decision-making and improved the precision of causal weight prediction, thereby revealing deeper information encoded in the causal model. Results: Experimental results demonstrated that the proposed GNN-NCM method substantially improved causal inference accuracy, interpretability, and prediction performance relative to conventional attention-based and causal explanation baselines. (1) In the LunarLander environment, the causal prediction error of the GNN inference network decreased by an average of 62%, demonstrating a superior ability to capture stable causal dependencies in continuous control tasks. (2) The model successfully identified state factors that made little contribution to the overall reward while still guiding specific reward components (for example, fuel consumption and landing smoothness). (3) Heatmap visualizations revealed distinct causal interaction patterns among state dimensions, showing, for example, how particular joint angles or velocities causally contributed to reward fluctuations over time. Quantitative evaluation of causal strengths enabled precise attribution of performance outcomes to particular control variables, improving both the interpretability and trustworthiness of learned policies. Conclusions: The proposed GNN-NCM framework offers a novel, interpretable approach to causal modeling in high-dimensional RL for robot control. By integrating causal structure discovery with neural graph inference, the method narrows the gap between black-box deep RL models and transparent, causality-aware policy representations. It enhances the interpretability, safety, and reliability of decision-making in autonomous robotic systems and demonstrates clear advantages in modeling accuracy and computational efficiency. The results demonstrate that graph-based causal reasoning offers a promising direction for future research in areas such as interpretable RL, interpretable robot control, and safe AI decision-making systems. Further extensions could apply this approach to multi-agent environments and real-world robotic applications, thereby driving the development of reliable and causally based intelligent control frameworks.
Objective: After an earthquake, helicopters and unmanned aerial vehicles (UAVs) can be effective means of emergency rescue response when roads are destroyed, power outages occur, and communications are disrupted in hilly and mountainous areas. Owing to the different functionalities and uses of UAVs and helicopters, which operate in different airspaces, existing research tends to schedule and optimize these two types of rescue equipment separately. If they do not account for the cooperation between the two heterogeneous aircraft, the overall rescue efficiency is reduced. Methods: This study aims to optimize helicopter and UAV teams for search and rescue operations in a post-earthquake environment. This study proposes an innovative parallel rescue framework for the simultaneous coordination of heterogeneous aircraft, rather than classical sequential methods. This study focuses on a performance-based hybrid task allocation model that systematically leverages the specificities of different aircraft types while simultaneously considering the satisfaction of the rescue as a key optimization. This optimization goal balances operational benefits with timely mission accomplishment. It is measured by the total distance flown and the overall satisfaction with task completion. The mathematical model dynamically adapts the priority of affected areas. It also includes several operational constraints, such as the number of aircraft, the frequency of service, payload capacity, endurance limits, and time windows for effective rescue response. To overcome this complex, multiobjective optimization issue, this study designed an improved evolutionary algorithm called greedy-enhanced non-dominated sorting genetic algorithm Ⅱ(GE-NSGA-Ⅱ), which was developed based on an improved mutation strategy adapted to population distribution characteristics. Results: The algorithm created a mechanism for adjusting the greedy-adaptive intensity and ensured robust randomness during the initial search phase. Over time, as the process evolved, the local search intensity improved. Even with adjustments, the method remained stronger and more robust than others due to improved global search capabilities. Moreover, the derived solutions remained disparate owing to the intensity provided, thus improving the overall process. The experimental results confirmed the effectiveness and superiority of the model using data from the 2008 Wenchuan earthquake. The ordinary NSGA-Ⅱ and improved GE-NSGA-Ⅱ algorithms were compared for optimal scheduling in three different post-earthquake rescue scenarios. In scenario 1, the total flight distance decreased by 23.04%, whereas satisfaction increased by 5.98%, showing the most significant optimization effect. In all three scenarios, with the increased scale of rescue operations and resource allocation complexity, the total flight distance increased by 210.00%, whereas satisfaction decreased from 0.855 1 to 0.611 5. Sensitivity analysis of parameters was based on population size, crossover probability, and mutation probability. Conclusions: The findings of this study show that the proposed framework ensures that critically injured individuals receive priority search and rescue coverage in disaster scenarios. Moreover, the framework can dynamically adapt to continuously evolving operational requirements. The flexibility of a cooperative system is characterized by the aircraft's ability to allocate tasks according to its performance profile. For example, UAVs can be used effectively for clustered assessment missions in enemy zones, whereas helicopters can perform long-range heavy-lift operations. The results of this in-depth comparison show that the proposed algorithm is better than the traditional optimization algorithms at achieving the quality of the generated task allocation schemes and promoting maximum efficiency in resource utilization and timely rescue. This study provides a scientific decision-support framework for rescue commanders to coordinate the dispatch of heterogeneous aerial assets. The operational efficiency is greatly improved during the crucial golden rescue time. This study has immediate applications in earthquake or disaster response. In addition, it can be useful for more general coordination problems involving complex multiagents arising in other emergency situations that require dynamic allocation of heterogeneous resources, and responses must occur under severe constraints and in a time-critical environment.
Objective: With the rapid development of unmanned technologies, unmanned monitoring ships have become indispensable platforms for marine fishery monitoring and illegal fishing detection, offering advantages such as high flexibility, extended endurance, and remote operability. These platforms significantly expand monitoring scope, improve detection precision, and reduce labor costs. However, the evaluation of intelligence for water-based unmanned platforms remains in its initial stage. Existing evaluation frameworks are predominantly designed for land-based autonomous vehicles and fail to address the unique characteristics of unmanned monitoring ships, including complex marine environments, dynamic non-cooperative targets, and adversarial interactions. Current methods focus primarily on static functional descriptions or virtual environment testing, lacking the capability to reflect real-world dynamic interactions and environmental uncertainties. To address these critical limitations, this study aims to establish a comprehensive, scientific, and targeted intelligence evaluation methodology specifically tailored for unmanned monitoring ships operating in non-cooperative scenarios. Methods: This paper proposes a multi-layered evaluation system comprising three core components. First, the UP-RAGAs (Unmanned Platform-based Retrieval-Augmented Generation Assessment) framework is constructed by reconstructing the traditional RAGAs framework into three stages—task input, platform response, and action output—thereby adapting it to real-scenario operational logic and supporting both algorithm-level software testing and system-level integrated evaluations. Second, a function mapping grading method is developed, classifying intelligence into 0—9 grades across three dimensions: autonomy (independent decision-making capability), cooperativity (multi-sensor data fusion and interoperability), and study ability (adaptive improvement through environmental interaction). Complementing this qualitative approach, an ontology capability evaluation method is proposed, involving 12 key capability indicators distributed across four perception stages: detection (optical imaging, multi-channel fusion, target detection), recognition (target identification, key component detection, behavioral intention analysis), positioning (distance measurement, coordinate transformation, track generation), and tracking (target acquisition, stable tracking, trajectory prediction). Third, to address non-cooperative scenarios characterized by dynamic target behaviors and environmental uncertainties, an innovative OODA+E (Observe-Orient-Decide-Act plus Effect) evaluation method is developed by integrating OODA loop theory with environmental and target state factors. Furthermore, a Bayesian-based multi-factor linkage model is established to dynamically quantify the influence of target movement levels, relative distances, viewing angles, and weather conditions on perception performance, enabling comprehensive assessment under dynamic adversarial conditions. Results: Validation through a typical operational scenario demonstrates the effectiveness and feasibility of the proposed methodologies. The ontology capability evaluation successfully quantifies intelligence using normalized data across the four perception stages, revealing specific performance characteristics. The OODA+E method effectively incorporates dynamic factors, revealing that high target mobility significantly degrades tracking stability, while relative distance inversely affects detection accuracy and positioning precision. Additionally, weather condition variations impact optical imaging performance following expected probabilistic distributions. These quantitative results confirm that the proposed framework can accurately assess intelligence levels under complex, non-cooperative conditions where traditional static evaluation methods would prove inadequate, providing granular insights into factor-specific influences on system intelligence. Conclusions: The proposed evaluation system successfully integrates static grading, quantitative capability assessment, and dynamic multi-factor analysis, effectively addressing the limitations of existing methods that are confined to cooperative scenarios and static testing environments. By realizing multi-dimensional and multi-level intelligence evaluation spanning algorithm performance to system-wide operational effectiveness, this research provides reliable technical support for the standardized application, performance optimization, and mission planning of unmanned monitoring ships in marine-related fields. The innovative incorporation of Bayesian-based multi-factor linkage analysis significantly enhances the scientific rigor and practical applicability of intelligence assessment, offering a systematic and comprehensive solution for advancing the deployment of intelligent unmanned platforms in complex maritime operations characterized by uncertainty and adversarial dynamics.
Objective: The increasing complexity and dynamic nature of modern mission environments pose significant challenges to the coordinated control of multirotor unmanned aerial vehicle (UAV) swarms. Variations in environmental conditions-such as weather, terrain, and electromagnetic interference-combined with shifting mission requirements demand adaptive and intelligent control strategies to ensure efficient and robust swarm performance. Traditional methods often fall short in addressing real-time target allocation under resource constraints and maintaining stable formation control in the presence of dynamic disturbances. This study proposes a novel intelligent cooperative control method that integrates an optimized target allocation strategy with an elastic formation model to enhance the adaptability, precision, and robustness of multirotor UAV swarm operations. Methods: The proposed method comprises two main components: target allocation optimization and cooperative formation control. First, a multiobjective target allocation function is formulated, considering path cost and average mission time, and the allocation problem is solved using a hybrid Hungarian-genetic algorithm, which combines the efficiency of the Hungarian method with the global search capability of genetic algorithms. This approach accommodates various UAV-to-target ratio scenarios and incorporates constraints such as maximum range, mission time limits, task priorities, and dynamic threat avoidance. Second, a dual-layer cooperative control architecture is designed. An elastic formation model is developed by integrating formation principles with a virtual elastic structure to maintain swarm cohesion. A sliding mode controller based on an improved exponential reaching law is designed for formation control, while a desired position controller translates formation commands into individual UAV actions. The stability of the controller is verified using Lyapunov theory. Simulations are conducted in a combined MATLAB/Simulink and Gazebo environment using quadrotor models based on industrial UAV specifications under various conditions, including static obstacles and dynamic threat zones. Results: Experimental results demonstrate that the proposed method achieves high-quality target allocation under dynamic conditions. Energy efficiency analysis shows that the average remaining battery power remains stable even after multiple task allocations, indicating effective load balancing. Compared to traditional methods, the proposed method reduces maximum power consumption and average flight distance, thereby improving overall energy efficiency. In terms of cooperative control performance, the proposed method achieves a 20.35% improvement in control accuracy and a 15.42% reduction in mission completion time relative to benchmark methods. Trajectory comparisons show that UAV swarms using the proposed method successfully avoid hazardous zones and reach targets more consistently. Robustness tests under simulated disturbances-such as localized turbulence and communication link interruptions-reveal that the proposed method maintains higher mission completion rates, greater average remaining power, and fewer UAV failures than existing approaches. Conclusions: This study presents an intelligent cooperative control method for multirotor UAV swarms that integrates optimized target allocation with an elastic formation control framework. The hybrid Hungarian-genetic algorithm enables efficient and adaptive task assignment, while the dual-layer controller enhances formation stability and responsiveness under dynamic conditions. Experimental validation confirms that the proposed method significantly improves control precision, mission efficiency, and system robustness, even in the presence of environmental disturbances and communication disruptions. The proposed method offers a viable solution for enhancing the autonomous cooperative capabilities of UAV swarms in complex operational scenarios.
Significance: Driving fatigue is a major contributor to traffic accidents and is characterized by its prevalence, concealment, and associated risks. It impairs drivers' physiological functions and depletes their psychological cognitive resources. Therefore, a comprehensive understanding of its underlying mechanisms, coupled with the development of accurate monitoring, early warning, and intervention technologies, is crucial to enhancing road safety. However, existing reviews often lack a systematic integration of research on fatigue mechanisms and advancements in associated monitoring and intervention technologies. To address this gap, this paper systematically reviews the causal factors and mechanistic hypotheses underlying fatigue and evaluates the current status of monitoring, warning, and intervention technologies. Current research indicates that multisource information fusion, which integrates physiological signals, visual imagery, and driving behavior data, significantly enhances the robustness and accuracy of fatigue detection compared to single-modality approaches. Meanwhile, early-warning and intervention strategies are evolving from passive monitoring toward active prediction, incorporating temporal analysis and contextual awareness. These strategies include the use of in-vehicle visual, auditory, and tactile stimuli; the development of autonomous vehicle takeover systems; and the optimization of road infrastructure design. This review synthesizes recent progress and emerging trends in the research on driving fatigue mechanisms and associated monitoring-warning technologies, and proposes integrated strategies to support effective fatigue management. Progress: In physiological monitoring, electroencephalography remains the gold standard, with deep learning models such as convolutional neural network (CNN)-attention methods achieving accuracy rates of up to 97.8%. Electrocardiography and heart rate variability measurements are also widely used, but their applicability is limited by the intrusiveness of the sensors. Therefore, current research focuses on noninvasive alternatives such as photoplethysmography and miniaturized devices, although challenges such as motion artifacts and environmental interference persist. Visual monitoring leverages computer vision and deep learning, including CNNs and long short-term memory models, to analyze features such as eye closure percentage, yawn frequency, and head pose, achieving accuracy rates of up to 92.7%. However, its performance is susceptible to lighting conditions, occlusions, and individual differences, necessitating improvements in real-world robustness. Driving behavior monitoring uses data on vehicle dynamics such as steering wheel variability and lane deviation, and machine learning models trained on these data achieve accuracies of up to 91.2%. While this method is privacy-preserving and readily deployable, it suffers from detection latency and is influenced by road conditions and driving habits, which limit its effectiveness for early warning. Multisource data fusion overcomes these limitations by integrating physiological, visual, and behavioral data using architectures such as multicolumn CNNs, achieving superior accuracy. However, high computational demands, data heterogeneity, and model interpretability issues pose challenges for real-time deployment. In terms of warnings and interventions, conventional systems rely on real-time detection to trigger alerts. Current research is shifting focus toward proactive prediction using bio-mathematical models and recurrent neural networks, achieving accuracies of up to 88.2% in forecasting fatigue several minutes in advance. Large language models enable intelligent, adaptive dialogue for graded intervention, supporting integrated "monitoring-assessment-response" frameworks. Autonomous driving technologies, particularly conditional automation, can reduce fatigue by allowing drivers to perform non-driving tasks and by providing emergency vehicle takeover capabilities. Road design and managerial measures complement these technological solutions within a holistic "human-vehicle-road-environment- management" framework. Conclusions and Prospects: This review outlines the mechanisms underlying fatigue, the associated monitoring technologies, and intervention strategies to address fatigue, emphasizing the key role of multisource data fusion in improving fatigue detection accuracy. The shift from passive warning to proactive intervention, supported by artificial intelligence and autonomous systems, represents a critical technological pathway. However, challenges remain in areas such as the precision of real-time predictions, the comfort of wearable devices, and the computational efficiency of multisource data fusion models. Future research should prioritize dynamic mechanism modeling, cross-scenario adaptive algorithms, and human-machine collaborative intervention to develop more reliable and scalable fatigue mitigation solutions.
Objective: The effectiveness of traditional automatic emergency braking (AEB) systems in mitigating traffic accident severity has been validated; however, their performance remains significantly inadequate under rainy conditions, at high speeds, at intersections, and in complex traffic scenarios. To address these limitations, this paper proposes an enhanced AEB system designed for multiscenario operation. By incorporating a high-precision map (prior information) and a systematic risk modeling mechanism, the system aims to improve the recognition of collisions and the rationality of braking activation, thereby enhancing adaptability in specific traffic environments. Methods: First, system-theoretic process analysis was performed to analyze the information flow across the perception, decision, control, execution, and environment layers of the AEB system. This analysis identified safety-critical control behaviors within the control loop and combined them with typical failure modes to construct a structured set of unsafe control behaviors, providing traceable targets for root cause analysis and strategy refinement. Building on this framework, the root causes of unsafe control behaviors were categorized into two coupled mechanisms: input bias on the perception side and temporal-logical defects on the decision side. The former included positioning and environmental information errors arising from sensor hardware limitations, environmental interference, and information fusion defects. The latter involved decision instability caused by inadequate risk assessment and inappropriate strategies. To quantify the relative influence of multidimensional risk factors and guide parameter optimization, an analytic hierarchy process-based risk weighting model was developed. This model assigned weights to factors such as vehicle motion state, road geometric constraints, and environmental interference, thereby forming a quantitative risk weighting system that linked scenario characteristics to triggering behaviors. Building on this foundation, an enhanced collision time metric, T1, that integrated high-precision maps was developed. Using AHP-weighted scenario coefficients, T1 is dynamically adjusted, enabling a more rational determination of AEB triggering timing based on roadway geometry, traffic semantics, and environmental conditions. Finally, real-vehicle tests were conducted at the Dongfeng Intelligent Connected Vehicle Demonstration Zone using a BYD Han EV platform for validation. Results: Real-world test results demonstrated that the proposed AEB system significantly outperforms traditional AEB systems in representative scenarios. Relative to conventional AEB strategies, the proposed system achieved a 27.9% reduction in average collision speed at high speeds and a 75.0% increase in the collision avoidance rate. Under rainy conditions, the collision speed decreased by 48.7%, and the avoidance rate improved by 79.9%. In pedestrian-related intersection tests, the conventional and proposed systems brought the vehicle to a complete stop before a collision; however, the latter system achieved a stopping distance closer to the ideal safety margin range of 1.0—1.5 m, indicating reduced overconservative intervention and a lower false-trigger rate. In the combined high-speed and rainy scenario, the collision speed was reduced by 31.2%, and the collision avoidance rate increased by 44.4%. The T1 metric integrated with high-precision maps enabled earlier intervention at high speeds and delayed triggering at intersections, enhancing decision consistency and braking activation rationality without compromising deceleration capability. Conclusions: The proposed model provides an interpretable, practical, and robust approach for improving the adaptability and reliability of AEB systems in complex traffic environments. By leveraging high-precision maps to achieve scenario-adaptive risk perception and trigger optimization, the proposed model effectively addresses the limitations of traditional approaches and offers important methodological support for designing next-generation safety-critical braking systems in intelligent vehicles. Future work will further consider multiparticipant interactions, refined environmental modeling, and variations in vehicle load to extend the model's applicability.
Objective: Detecting apparent defects is fundamental for assessing the health of structures and provides guidance for preventive maintenance toward mitigating engineering hazards. Deep learning-driven computer vision methods have recently gained prominence in intelligent defect identification. However, the missed detection of small-scale targets and the unbalanced accuracy across defect categories are limitations of existing algorithms, owing to substantial variations in spatial scales. To overcome these limitations, this study introduces a detection approach that effectively captures deep network feature representations, mitigates the loss of semantic information for small targets, and maintains balanced recognition performance across multiple defect types. Methods: A lining apparent defect detection network (LADDNet) that integrates an information-sharing backbone with an adaptive and hierarchically organized feature fusion mechanism is proposed. First, a three-branch collaborative feature extraction architecture is constructed by incorporating lightweight ShuffleNet and GhostNet modules into a CSPNet framework. This design facilitates complementary feature learning across branches, thereby enhancing the stability of gradient propagation. Thereafter, an attention-integrated multi-receptive field adaptive fusion (AMFAF) module is developed. This module employs parallel convolutions with diverse receptive fields to extract multi-level spatial information and combines them via attention-based weighting, allowing the network to automatically emphasize discriminative features associated with cracks, seepage regions, and spalling contours. An attention-based intra-scale feature interaction (AIFI) module is also introduced to enhance semantic consistency within individual feature scales by promoting effective cross-channel communication and suppressing redundant responses. Finally, an adaptively spatial feature fusion detection head (ASFF-Head) is incorporated to refine multi-scale feature aggregation, improve the localization precision, and reduce missed detections of small or low-contrast targets. By integrating these modules into a unified framework, the proposed network supports end-to-end training and inference. Results: LADDNet exhibits significant advantages in the overall detection performance and inference efficiency. On the validation set, the model achieves F1, mAP@0.5, and mAP@0.5: 0.95 scores of 0.831, 0.848, and 0.595, respectively. On the test set, the model attains an F1 score of 0.794, mAP@0.5 of 0.830, and mAP@0.5: 0.95 of 0.579. Compared with a range of representative detection models, LADDNet achieves consistent improvements in both the F1 score and mAP@0.5. In terms of inference efficiency, LADDNet achieves real-time performance with a per-image latency of 9.2 ms, only 14.07×106 parameters, and 21.4×109 FLOPs, delivering substantially faster inference than RT-DETR. Furthermore, when detecting defects in images containing handwritten markings or interference from auxiliary tunnel facilities, LADDNet continues to demonstrate strong robustness. For the identification of mesh cracks, water seepage, and spalling, the model demonstrates high confidence, low missed-detection rates, and precise localization. Conclusions: The proposed LADDNet model affords markedly enhanced intelligent detection of diverse tunnel-lining defects by integrating information sharing, multi-receptive-field feature extraction, adaptive fusion strategies, and intra-scale interaction mechanisms. It delivers notable gains in accuracy, robustness, and computational efficiency, effectively overcoming long-standing challenges in multi-scale and multi-type defect recognition. These advances position LADDNet as a reliable visual perception module for automated tunnel inspection, structural condition evaluation, and long-term operational monitoring. Overall, the approach shows strong promise for real-world engineering applications and broad deployment in next-generation intelligent maintenance systems. Its versatility further underscores its value for future infrastructure management initiatives.
Objective: With the continuous development of power grids, overhead transmission lines inevitably pass through regions with complex terrain and climatic conditions. In extremely cold environments, ice accumulation occurs on these lines. Under wind loads, the accumulated ice layers vibrate and detach, potentially causing accidents such as broken lines and tower collapses. The DC deicers serve as the primary equipment for power networks to defend against freezing disasters. They utilize Joule heating to raise line temperatures and melt ice and snow. However, existing DC deicers suffer from large size, heavy weight, and difficult transportation. As a core component of DC deicers, the rectifier transformer accounts for more than 70% of the total weight. To improve the mobility of these devices, it is necessary to optimize the structure of the rectifier transformer and reduce its weight. Methods: Based on the transformer theory, a three-phase three-dimensional (3D) symmetric core and its lamination method for rectifier transformers are proposed. The electromagnetic field equations and the phasor relations of main and mutual fluxes were derived according to Ampere's law. A finite element model of a 4 MV·A rectifier transformer was established in Ansys Electronics Desktop. The time-varying law of the magnetic field and the characteristics of the loss density distribution were analyzed. The limitations of existing numerical calculation methods in dealing with the microscopic behavior of magnetic domains and the additional loss in the vertical symmetry plane were analyzed. To verify the analysis results, a 4 MV·A prototype was fabricated and tested under thermal cycling and no-load conditions. Finally, an additional loss shape function expression with the average flux density gradient as the independent variable was proposed. The undetermined coefficients and loss correction formula were obtained by collecting no-load test data of samples with different capacities. Results: The mutual fluxes of the three-phase 3D symmetric core were separated by the vertical symmetry plane of each phase. The core flux density was
Objective: Data sharing and exchange play a critical role in promoting the intelligent and digital transformation of government services. However, existing government data sharing and exchange systems typically adopt cascaded architectures, resulting in long supply-demand paths at the architectural level. In addition, limited capabilities in data quality inspection and rapid construction of underlying routing mechanisms lead to low real-time performance, difficulty in ensuring data quality, and inadequate support for scenarios involving large volumes of frequently used data. Methods: To address these challenges, a systematic research approach is adopted. First, guided by the principles of distribution, high reliability, and flexible configuration, a distributed architecture for government data sharing and exchange is proposed. A distributed data exchange network composed of peer nodes is constructed, in which node relationships are equal, thereby shortening data forwarding paths and improving exchange efficiency. By decoupling the control layer from the transport layer, the control layer is dedicated to routing management and node status monitoring, while the transport layer focuses on efficient and reliable data transmission, clarifying the functional structure and operational mechanism of the architecture. Second, a data quality inspection algorithm based on large-model intelligent agents is introduced. Using a unified inspection strategy, the algorithm evaluates data quality across four dimensions—semantic consistency, format standardization, logical consistency, and data integrity—ensuring high-quality data provision. Third, a geographic-aware routing algorithm is proposed by integrating administrative geographic information into distributed Hash tables. A hybrid routing strategy is designed, combining cross-layer routing based on a multiway tree structure with intralayer routing based on a binary tree structure, thereby reducing routing hops during data addressing. Finally, a series of experimental validation processes is employed to verify the effectiveness of key algorithms and the overall architecture. Results: Compared with the benchmark method, the proposed geographic-aware routing algorithm reduced the average hop count by 76.82%. The intelligent-agent-based data quality inspection algorithm achieved an average precision of 93.06%, an average recall of 93.50%, and an average F1-score of 0.72. Based on the distributed government data sharing and exchange architecture, three typical business scenarios—real-time transactions, unstructured transactions, and batch transactions—were evaluated. Deployment in Heilongjiang Province enabled on-site performance testing under real-world conditions. The results showed that the average response time for real-time transaction scenarios was 722 ms, with a median of 594 ms. In unstructured transaction scenarios, large-file upload speeds reached 220.0-235.0 MB/s, while batch transaction scenarios achieved an average throughput of 1.5 MB/s and an average write speed of 1 961 records/s. Compared with the theoretical performance peak of traditional cascaded systems, the performance was improved by 50% and 91%, respectively. Conclusions: The proposed distributed government data sharing and exchange system significantly enhances real-time performance and ensures data quality in government data sharing and exchange. It provides a new technical pathway for intelligent and digital government transformation and offers valuable insights into the large-scale application of "artificial intelligence + data elements" in the public sector.
