In order to minimize energy consumption in rail transit systems, the coordinated power supply technology of “network-source-storage-vehicle” integrates with renewable energy power generation systems along the line. This approach establishes a new coordinated power supply technology system that enables efficient energy utilization across time and space. This paper comprehensively analyzes the fundamental composition and characteristic types of physical, informational, and social architectures within the coordinated power supply system. Building upon this analysis, it introduces a temporal and spatial matching evaluation method for "load-source" based on the core concept of asset energization from a systemic comprehensive evaluation and operational perspective. Furthermore, it elaborates on important technological systems such as multi-source integration, protection reconstruction, and elastic evaluation. Emphasizing efficient energy-saving operations, it focuses on high-efficiency and high-resilience energy self-consistency technology through coordination among network, source, storage, and vehicle components. Additionally, leveraging artificial intelligence and information technology tools is proposed to construct multi-level energy management systems aimed at achieving effective coupling of diverse energy flows while ensuring safe, stable, and cost-effective operation of the system. The paper systematically summarizes key technologies related to the “network-source-storage-vehicle” coordinated energy supply system for rail transit including architectural characteristics; evaluation; optimization; safe operation; as well as coordinated operation of the system. It also outlines technical composition systems relevant to coordinated energy supply systems providing valuable references for engineering practices.

Single-inductor dual-output (SIDO) switching converters with shared charge and discharge sequence have problems of the large inductor current ripple and cross-regulation between output branches, and the control circuit fails to work normally under the wide range of changes in circuit parameters. To solve these problems, a current-mode variable frequency control (C-VF) technique with an independent charge and discharge sequence was proposed. First, the working principle of the converter in continuous conduction mode (CCM) was specifically described, and the open-loop transfer function of the main circuit was derived. Furthermore, a closed-loop small signal model was constructed, and the closed-loop cross-regulation impedance transfer functions were derived. The cross-regulation characteristics of the converter with different output voltages and load currents were analyzed in detail. Finally, simulation and experimental verification were carried out. The results show that compared with the shared charge and discharge sequence, the C-VF CCM SIDO buck convert with independent charge and discharge sequence reduces the cross-regulation and improves the transient load response performance. When the load voltages of the two branches are different, decreasing the load of one branch can reduce the cross-regulation of this branch to another one. When the output voltages of the two branches are the same, but the loads are different, the branch with heavy loads has small cross-regulation on the branch with light load.

For pantograph-catenary contact pairs in electrified railways operating in normal and abnormal states, the friction and wear performance of pantograph strip differentiates in a wear cycle, highlighted by differences in wear rate and wear profile. When abnormal wear occurs, the wear rate of pantograph strip will have a multifold increase or even dozens of times increase, but the wear profile acts differently, revealing partial eccentric wear, wavy wear, and penetrating wear. The similarities and differences in current-carrying friction and wear platforms are summarized for pantograph-catenary systems, as well as the advantages and disadvantages of contact and non-contact detection methods. The influential factors and evolution law are analyzed in view of the structure and parameters, train operation parameters, current-carrying parameters and external environment of pantograph-catenary system. Following above work, the state of the art of pantograph-catenary wear models, including mechanism analysis model and data fitting model, are analyzed extensively, and the prospective direction and development trend are put forward, such as, the equivalent simulation of a pantograph-catenary friction pair in real service under laboratory conditions, online high-precision detection of pantograph-catenary wear performance, simulation and optimization of pantograph-catenary wear performance in complex climatic conditions and multi-physical field coupling, pantograph-catenary wear prediction using big data and intelligent algorithms, intelligent operation and maintenance strategies, and capability maintenance in the whole life cycle.

Global blackouts have indicated that vulnerable lines in the power system would bring great operation risk and threaten system security. In view of this, the identification of vulnerable lines in the power system was carried out according to the transfer characteristics of the active power flows in the case of a two-stage cascading failure. First, from the perspective of system operation, the relationship between the lines was weighted in terms of the transferred active power flows when the power system had a two-stage cascading failure, and a bidirectional weighted power flow correlation network was constructed. Then, an improved E-index-based identification method for vulnerable lines was proposed to quantify line vulnerability in cascading failures. The results show that the average value at risk and the average conditional value at risk of the top eight vulnerable lines in the identified results are 3 453.73 MW and 187.82 MW respectively for the IEEE-39 bus system, which are much higher than those obtained from other methods. Meanwhile, the residual load rate of the system is only 43.4% after the static intentional attack, and the number of power islands in the system increases rapidly.

In view of two-hop multi-relay transmission in energy harvesting wireless sensor networks (WSN), a wireless radio frequency power beacon (PB)-assisted energy harvesting relay model based on simultaneous wireless information and power transfer (SWIPT) was constructed. Under the condition that the relay node has the characteristics of capturing the source node, loop self interference, and PB signal energy, the outage probability and throughput of the destination node using two different receiving strategies, namely, selection combining (SC) and maximal ratio combining (MRC), were derived. Then, under multiple constraints such as ensuring communication quality of service (QoS), a relay selection algorithm was proposed to jointly optimize the time switching factor and power splitting factor with the goal of maximizing the throughput. Simulation and numerical results show that the PB transmit power, time switching factor, number of antennas, and power splitting factor significantly affect the system’s outage probability and throughput. When PB transmit power is 6 dBW, and the number of antennas is 3, compared with the random relay selection algorithm and the max-min relay selection algorithm, the system’s throughput gains under the SC strategy are 0.29 and 0.15 bit/(s·Hz), respectively, and those under the MRC strategy are 0.32 and 0.16 bit/(s·Hz), respectively.

The engineering testing of a high-speed railway signaling system (HSRSS) focuses on the complex behavior relationship and state synchronization among various equipment in the system. Since the testing modeling method for engineering testing lacks complex behavior interaction and synchronization mechanism, the engineering testing modeling method and test case generation method of HSRSS based on an extended finite state machine (EFSM) were proposed. First, the engineering testing characteristics of HSRSS were analyzed, and the testing modeling requirements for complex event interaction and state synchronization were proposed. Based on the theory of finite state machine, the state events and hierarchy were extended to meet the modeling requirements for complex behavior relationships and state synchronization in engineering testing of HSRSS. The formal definition of EFSM was given by using Z specification language. The pattern and synchronization mechanism of EFSM were defined. Then, the algorithm of transforming the EFSM into a timed automata (TA) was put forward, and the test case generation algorithm of TA was used to automatically generate test cases for engineering testing of HSRSS. Finally, with the route control in engineering testing of HSRSS as an example, an EFSM model was established, and test cases were generated. In addition, the generated test cases were evaluated through mutation analysis. The result shows that the test cases have a mutation score of 1 when testing state mutation and signal expression mutation, and they have good coverage, which can meet the requirements of engineering testing of HSRSS.

The wheel-rail creep curve influences dynamic wheel-rail interaction, which further affects rail wear. To study the effect of the measured wheel-rail creep curve on rail wear, parameters suitable for the Polach model and modified FASTSIM algorithm were obtained based on the least square method, and measured creep curves at the running speed of 40–400 km/h of the vehicle were simulated. After that, the vehicle system dynamics model was established in the SIMPACK, and measured creep curves were considered through the Polach model. Finally, the Kik-Piotrowski model and modified FASTSIM algorithm were used to calculate the non-Hertzian rolling contact, and rail wear was predicted by the USFD model. The discrepancies of rail wear under ideal and measured creep curves were compared. The research shows that the rail wear depth under the ideal creep curve is more obvious than that under the measured creep curve. As more vehicles pass the rail, the rail wear distribution range under ideal conditions is larger, and the distribution ranges of inner and outer rail are respectively 1.5 and 1.3 times those under the measured creep curve; the friction coefficient and wear rate significantly influence the magnitude and distribution range of rail wear, so it is necessary to consider the measured wheel-rail creep curve in vehicle dynamics simulation and rail wear calculation. A pre-processing program is developed to determine parameters of the measured creep curve, which can serve for vehicle dynamics simulation and rail wear calculation and effectively guide maintenance work such as rail grinding.

During the outbreak of an earthquake, the pier columns in reinforced concrete (RC) structures are usually subjected to horizontal forces and vertical forces under reciprocating loads. When the transverse constraints in the structure are insufficient, the longitudinal reinforcement may have obvious transverse deformation. The bearing capacity and ductility of the structure are thus significantly reduced, and it is difficult to repair. Therefore, it is of great significance for structural design and construction to study the buckling mechanism of longitudinal reinforcements under compression. Firstly, the research on longitudinal reinforcement buckling based on the direct compression test of a single reinforcement was summarized, and the analysis shows that the main influencing factors of reinforcement buckling are the slenderness ratio and yield strength. Secondly, the research progress of reinforcement buckling based on the direct compression test of RC short columns was summarized, and the possible factors influencing longitudinal reinforcement buckling under complex interaction were described. Moreover, combined with the development status of civil engineering, the research progress of longitudinal reinforcement buckling in pier columns with new materials and structures was discussed. Finally, it is concluded that the current research on longitudinal reinforcement buckling lacks a theoretical method for determining the direction and extent of reinforcement buckling. Further investigations are needed to address the overall buckling issues of longitudinal reinforcements and the longitudinal reinforcement buckling concerns in novel structures. Subsequent studies should comprehensively consider the effect of factors such as corrosion, loading history, cross-sectional geometry, and interactions among materials on longitudinal reinforcement buckling.

Four-point bending tests were conducted on six ultra-high performance concrete T-shaped (UHPC-T) section beams to establish a reinforcement stress calculation method for reinforced UHPC beams and study the variation law of reinforcement stress. Based on the mechanism of force balance and deformation coordination between reinforcement and UHPC, a reinforcement stress calculation formula was derived using the differential equations of equilibrium, deformation, and bond-slip established by micro elements, which could comprehensively reflect the influence of bond-slip between reinforcement and UHPC interfaces and the contribution of steel fibers to tensile strength. By simplifying the calculation of the strain non-uniformity coefficient and the reinforcement stress in cracked sections, a simplified formula for reinforcement stress suitable for engineering applications was proposed. The results show that the increase in reinforcement stress under unit load decreases with the increase in reinforcement ratio, but it is not related to the change in steel fiber volume fraction. Compared with ordinary concrete beams, the reinforcement stress in UHPC beams is relatively small in the cracked section, but the uneven distribution of reinforcement stress between adjacent cracks is intensified. The calculation value of the suggested formula for reinforcement stress is in good agreement with the experimental values in this article and existing literature. The average ratio of the calculated value of the simplified formula for reinforcement stress to the experimental value is 1.03, and the coefficient of variation is 0.06, indicating that this simplified formula can be used for calculating reinforcement stress in UHPC beams.

Influenced by rainfall, strong earthquakes, and intensive engineering activities, the slope at the tunnel entrance may experience collapses, landslides, debris flows, and other geological disasters, especially under the complex conditions of violent earthquakes, high geostress, abrupt slopes, and extremely cold environments. To explore the stability of tunnel entrance slopes under different working conditions in complex disaster-prone environments, the Jiarishan tunnel in southwest China was taken as an example, and a “space-air-ground” integrated investigation technology combining unmanned aerial vehicle (UAV) photography, surface trenching, laboratory experiments, field tests, and cave surveys was employed to obtain accurate engineering geological information. Typical failure characteristics of the studied slope were systematically revealed, and the two failure causes and evolution models of the slope were discussed. In addition, the slope stability under different working conditions was qualitatively analyzed using the limit equilibrium method and three-dimensional numerical simulation. The results indicate that the stability factor of the slope at the Jiarishan tunnel entrance is always greater than 1.15 under natural, rainstorm, and earthquake working conditions. Overall, the slope is stable except for the local deformation and instability of the shallow horizon of the posterior margin. This research can provide theoretical guidance and technical support for similar tunnels in terms of siting, construction, and operation safety.

To elucidate the applicability and feasibility of the calculation methods for impact force and impact depth of falling rocks based on Hertz contact theory, as well as the determination method for the reinforcement coefficient of buffering soil layers, full-scale model tests, inversion analysis, and mathematical statistics methods were employed. The falling rock impact tests were carried out, involving cubic rock and conical rock with a spherical top, each with a volume of approximately 1 m³ and a weight of about 2 t. They were dropped from heights ranging from 1 to 10 m onto buffering soil layers with thicknesses of 0.5–2 m. Then, the reinforcement coefficient of buffering sand layers was determined, and theoretical and test results for impact force and impact depth of falling rocks were comparatively analyzed. The research conclusions are as follows: Based on inversion analysis of the test results, it is recommended that the reinforcement coefficient of buffering sand layers within the 99.7% confidence interval should range from 0.25 GN/m^5/2 to 10.00 GN/m^5/2. The theoretically calculated average impact force of cubic rock is 140% larger than the test value, while that of conical rock with a spherical top is 21% larger than the test value. The theoretically calculated average impact depth of cubic rock is 112% larger than the test value, while that of conical rock with a spherical top is 5% larger than the test value. Within the 99.7% confidence interval of the reinforcement coefficient, the range of calculated impact force and impact depth of rock can encompass 100% of the test results. Under the same conditions, the test value of the impact depth of conical rock with a spherical top is greater than that of cubic rock. The impact depth increases with the increase in the thickness of the buffering layer, while the variability in the impact force of falling rocks shows no significant correlation with the rock shape and buffering layer thickness.

In order to study the mechanical deformation characteristics of the new horseshoe prefabricated initial support structure and compare its difference with the spray anchor initial support structure, the prototype loading tests of two types of structures were carried out. The structural design, lining prefabrication, test loading, and result analysis were introduced systematically, and the test results were deeply analyzed. The results show that the ultimate bearing capacity of the prefabricated initial support structure is 2.80 times the design load, and that of the spray anchor initial support structure is 1.32 times the design load. The ultimate bearing capacity of the former is about 2.10 times that of the latter. The prefabricated initial support structure is affected by the positive bending moment and negative axial force at the vault and the arch waist, as well as the negative bending moment and positive axial force at the arch shoulder and the arch bottom. The spray anchor initial support structure is similar. The maximum moment and the maximum axial force of the prefabricated initial support structure are about 1.39 and 1.45 times that of the spray anchor initial support structure near the destruction phase, respectively. The concave and convex deformation trend of the prefabricated initial support structure is basically the same as that of the spray anchor initial support structure. The right arch shoulder is convex, and the right arch waist is concave, with steel and concrete spalling when the structure is damaged, and the ultimate deformation capacity of the prefabricated initial support structure is about 1.20 times that of the spray anchor initial support structure.

The hard and brittle surrounding rock in tunnels is prone to rockburst under high ground stress conditions. Currently, the support structure design for rockburst tunnels in China mainly adopts the engineering analogy method. In order to quantify the impact load of the rockburst on the support structure, the energy conversion relationship during the rockburst process was analyzed from an energy perspective. Then, the kinetic energy theorem and energy conservation principle were used to calculate the impact load. Combined with the loose pressure of the tunnel, a load calculation method for rockburst in tunnels was proposed. At the same time, the influence of different factors such as tunnel diameter on the range of rockburst action was explored. Finally, the rationality was verified based on the rockburst section of a high ground stress tunnel. The research results indicate that the dynamic load factor in the formula for rockburst impact load is positively correlated with the structural stiffness used for tunnel support. Under the same span and ground stress conditions, the rockburst depth of a circular tunnel is smaller than that of a horseshoe tunnel, and the lateral range of the rockburst is larger than that of a horseshoe tunnel. Under the same tunnel shape and ground stress conditions, as the tunnel span increases, the depth and lateral range of the rockburst also increase. Under the same tunnel shape and span conditions, a higher ground stress value indicates a greater depth and lateral range of the rockburst. The rockburst load of Grade Ⅱ surrounding rock in a single-track tunnel ranges from 12.02 kPa to 337.75 kPa, and that of Grade Ⅲ surrounding rock in a single-track tunnel ranges from 25.36 kPa to 352.12 kPa. Moreover, the rockburst load of Grade Ⅱ surrounding rock in a double-track tunnel ranges from 8.54 kPa to 288.55 kPa, and that of Grade Ⅲ surrounding rock in a double-track tunnel ranges from 33.11 kPa to 300.83 kPa.

In order to systematically investigate the response law of the internal relative humidity of concrete at an early age to different tensile stress levels, a test method of the internal relative humidity of the concrete under constant axial tension was developed in this paper, and the response law of the relative humidity under different tensile stresses was studied experimentally. According to the experimental results and theoretical analysis, a linear model of the relative humidity and tensile stress of early-age concrete under one-side drying conditions was presented. The results show that the tensile stress causes the instantaneous decrease in the internal relative humidity of the concrete. When the tensile stress increases from 0.8 MPa to 3.2 MPa, the relative humidity change at the depth of 50, 75, and 100 mm of the concrete increases from 0.5%, 0.4%, and 0.3% to 0.8%, 0.7%, and 0.6%, respectively. At the same time, with the increase in tensile stress, the decrease in the relative humidity gradually increases. Under the same tensile stress, the response of relative humidity of the concrete, close to the exposed surface, to tensile stress which is is more obvious. The relative humidity gradually recovers during the tensile stress loading, and the time is about 2.5 h. A similar phenomenon also occurs during the compressive stress loading, and the time is about 20.0 h. Therefore, the relative humidity recovery time is shorter during the tensile stress loading.

In order to study the effect of fly ash and silica fume contents on the properties of alkali-activated slag-based concrete (AASC), the changes in setting time, cubic compressive strength, cubic splitting tensile strength, flexural strength, and elastic modulus of AASC were investigated by conducting tests on setting time and basic mechanical properties. Based on the test results, a regression analysis method was used to establish the conversion relationship equation of cubic splitting tensile strength, flexural strength, and elastic modulus with cubic compressive strength, and the effect of fly ash and silica fume on the properties of AASC was revealed according to the microstructure and phase composition. The results show that the fly ash and silica fume can prolong the setting time of AASC; the mechanical property indicators of AASC tend to strengthen and then weaken with the increase in the contents of fly ash and silica fume, and the optimal contents of fly ash and silica fume are 20% and 10%, respectively. The proposed empirical formulae for cubic splitting tensile strength, flexural strength, and elastic modulus of AASC have a high fitting precision. The appropriate contents of fly ash (silica fume≤20%) and silica fume (silica fume≤10%) can promote the hydration reaction of AASC and the denser microstructures.

Foamed concrete is susceptible to environmental factors such as temperature during the period from pouring to curing and molding, and adverse environmental factors can lead to performance degradation and deterioration of foamed concrete after curing and molding. In order to study the performance and microporous structure of foamed concrete affected by temperature (−15–70 ℃) during the pouring period, a single factor indoor test program was designed by using dry density, compressive strength, and water absorption to evaluate the physical property of foamed concrete, the construction quality, and the operational durability of foamed concrete, respectively. In addition, the evolution of the microporous structure of foamed concrete was analyzed by using an image data acquisition system and Image J software. The results show that the dry density of foamed concrete generally decreases in a stepwise manner as the temperature rises. The compressive strength first decreases, then increases, and finally decreases. The water absorption rate takes 0 ℃ as the cut-off point, and it tends to increase and then decrease when the temperature decreases or increases. The equivalent pore size shows a trend of increasing, decreasing, and increasing. The pore roundness value first increases and then decreases. The fractal dimension of pore distribution first decreases and then becomes larger. The results of macroscopic performance and microporous structure show that the recommended temperature range for the construction of foamed concrete during the pouring period is −5–40 ℃.

In order to study the influences of the number of fiber reinforced polymer (FRP) layers, the type of FRP, and the volume of steel fiber on the axial compression performance of ultra-high performance concrete (UHPC) circular stub columns, 21 FRP-confined UHPC circular stub columns were tested under axial compression. The typical failure characteristics and stress mechanism of the specimens were analyzed. In addition, the influence of various parameters on the ultimate strength and ultimate strain of the specimens was studied. The experimental results show that the ultimate strength of UHPC circular stub columns can be improved by increasing the number of FRP layers. The ultimate strength of C12, C22, and C32 is 17.8%, 25.4%, and 23.4% higher than that of C11, C21, and C31, respectively. With the increase in the volume of steel fiber, the ultimate strength and ultimate strain, and the ductility of UHPC circular stub columns are improved. The ultimate strength and ultimate strain of C31 are increased by 2.9% and 15.1%, respectively, compared with C21, as well as 4.7% and 50%, respectively, compared with C11. Under the same FRP layers and volume of steel fiber, the improvement of the ultimate strength of confined UHPC circular stub columns by carbon fiber reinforced polymer (CFRP) is significantly better than that by glass fiber reinforced polymer (GFRP). The ultimate strength of C11, C12, and C13 is 9.7%, 7.8%, and 7.2% higher than that of G11, G12, and G13, respectively. In view of the constraint of the steel fiber, calculation models of compressive strength and ultimate strain of FRP-confined UHPC circular stub columns are proposed. Furthermore, the constitutive model of FRP-confined UHPC is given.

To explore how sustained loading and drying-wetting cycles of chloride salt affect the flexural behavior of reinforced concrete (RC) beams, the corrosion tests and flexural capacity tests of 19 RC beams were carried out at first. The effects of various sustained loading grades and drying-wetting cycle periods were studied on the crack distribution, the mass corrosion rate of longitudinal reinforcements, and the flexural behavior of corroded members. The relationship was summarized between the maximum and average mass corrosion rate of longitudinal reinforcements and the flexural bearing capacity reduction ratio. The study results show that the corrosion degree of longitudinal reinforcements was higher in the pure bending segment than in other areas, and distributed non-uniformly along the circumference of the longitudinal reinforcements. There was no noticeable correlation between the position of the extreme mass corrosion rate of longitudinal reinforcements and the occurrence position of initial transverse cracks. The maximum mass corrosion rate of longitudinal reinforcements increased with the loading grade and drying-wetting cycle period, and was more notably affected by the drying-wetting cycle period. The peak load of the beam decreased after corrosion. When the mass corrosion rate of the longitudinal reinforcements was low (average value below 3% or maximum value above 6%), the correlation between the mass corrosion rate and the reduction ratio of bearing capacity was low, and the maximum mass corrosion rate has less correlation with the reduction ratio of bearing capacity than the average mass corrosion rate. When the mass corrosion rate of longitudinal reinforcements increased (average value above 3% or maximum value above 6%), both the maximum and the average mass corrosion rates of longitudinal reinforcements had an increased correlation with the reduction ratio of bearing capacity. When the average mass corrosion rate of longitudinal reinforcements is the same, the reduction of bearing capacity of specimens with natural corrosion is higher than that of specimens with external current corrosion.

Vehicle trajectory data provides abundant spatial-temporal traffic flow information, which can be used for traffic research. Traditional vehicle trajectory models mostly focus on the artificial driving environment and fail to consider the impact of mixed traffic flows composed of regular vehicles (RVs), connected vehicles (CVs), and connected automated vehicles (CAVs). To solve this problem, a full sample vehicle trajectory reconstruction model of signalized intersections in connected automated environments was proposed. Firstly, the composition of vehicles at signalized intersections of urban roads and the passage of queues in connected automated environments were analyzed. Secondly, a model for estimating the number of trajectories of mixed traffic flows on urban roads was constructed, and the concept of virtual vehicles was further proposed to estimate the traffic status of different vehicles according to the queuing of front and rear vehicles. Finally, a numerical simulation test was designed to analyze the influence of traffic flow density and penetration rate of CAVs and CVs on the model, and the model was verified by NGSIM data. The results show that the error of the number and position of the model decreases with the increase in traffic flow density and the penetration rate of CAVs and CVs. For example, when the traffic flow density increases from 20 veh/km to 50 veh/km, both the error of the number and position of the model shows a decreasing trend, and the maximum error is no more than 6.88% and 8.02 m. Compared with that of CVs, the penetration rate of CAVs has a greater impact on the model results.

As a key factor affecting location quality of cellular phone data, location frequency has an important influence on the extraction accuracy of travel mode. In order to quantify the change rule between the location frequency and accuracy of travel mode extraction, a travel mode extraction model based on random forest is proposed. Second, with the help of communication operators, through a field data collection, individual cellular phone data and corresponding real travel information were simultaneously acquired. The dataset is used to verify the travel mode extraction model. Finally, a series of cellular phone datasets with different location frequencies are built through data sampling. With this series of datasets, the extraction accuracy of traffic modes under different location frequencies is evaluated. The evaluation results show that the overall extraction accuracy for walking, non-motorized vehicles, cars, and buses is 79.2%, and the sensitivity of each travel mode to location frequency is different. The sensitivity of non-motorized vehicles and buses is higher, and the sensitivity of walking and cars is relatively low. As the location frequency is decreased from 48 seconds per data to 241 seconds per data, the overall accuracy of non-motorize vehicles and buses is decreased by 19.2% and 21.5%, respectively, while that of walking and car is decreased by 12.8% and 11.5%, respectively. Owning to the requirements of extraction accuracy and computing efficiency, 60 seconds per data is recommended as the optimal threshold for user screening and data sampling.

To analyze the significant influencing factors of the entrainment coefficient of the ejector, a two-dimensional numerical model of the compressible flow of the ejector with air as the working medium was established, and its calculation accuracy was validated by the experimental data. Meanwhile, the calculation matrix was designed by utilizing the D-optimal experimental design method. Based on the least-squares method, the response surface prediction model of the entrainment coefficient with a second-order form was constructed, and the significant parameters of the ejection coefficient and their interaction were simulated based on the constructed model. The research results show that the coincidence between the predicted and calculated values of the entrainment coefficients proves the accuracy of the response surface prediction model; the interaction between the length of the diffuser section, the mixing section length, the diameter of the mixing section, and the distance from the nozzle outlet to the inlet of the mixing section, the interaction between the mixing section diameter and mixing section length, and the interaction between the mixing section length and the diffusion angle of the diffuser section are the key factors affecting the entrainment coefficient because their P-values are less than 0.001. Meanwhile, the interaction of the diffusion angle of the diffuser section, the distance from the nozzle outlet to the inlet of the mixing section and the length of the diffuser section, and the interaction between the length of the mixing section and the distance from the nozzle outlet to the inlet of the mixing section have an important influence on the entrainment coefficient. Therefore, they are the main parameters affecting the entrainment coefficient. In addition, it is found that among the significant interaction of parameters, the significant factors affecting the entrainment coefficient will change with the value range of the two parameters with significant interaction.

A dense crowd counting network based on multi-scale perception was proposed to solve the problems of diverse target scales and large-scale changes of crowds in dense crowd scenes. Firstly, since the small-scale targets account for a relatively large proportion of the images, a dilated convolution module was introduced based on the visual geometry group 2016 (VGG-16) network to mine the detailed information in the images. Then, by utilizing the multi-scale information of the target, a novel context-aware module was designed to extract the contrast features between different scales. Finally, In view of the continuous change of target scales, the multi-scale feature aggregation module was designed to improve the sampling range of dense scales, enhance the interaction of multi-scale information, and thus improve the model performance. The experimental results show that mean absolute errors (MAEs) of the proposed method are 62.5, 6.9, and 156.5, and the root mean square errors (RMSEs) are 95.7, 11.0, and 223.3 on ShangHai Tech (Part_A/Part_B) and UCF_CC_50 datasets, respectively. Compared with the optimal method of comparison model, the MAE and RMSE are reduced by 1.1% and 4.3% on the UCF_QNRF dataset and by 8.7% and 13.9% on the NWPU dataset.

The existing deep learning methods have low precision and poor generalization ability in calculating dense correspondence between non-rigid point cloud models. To address these issues, a novel method for calculating unsupervised three-dimensional (3D) point cloud model correspondence based on a feature sequence attention mechanism was proposed. Firstly, the feature extraction module was used to extract the features of the input point cloud model pair. Secondly, the transformer module learned context information by capturing self-attention and cross-attention and generated a soft mapping matrix through the correspondence prediction module. Finally, the reconstruction module reconstructed the point cloud model based on the obtained soft mapping matrix and used the unsupervised loss function to complete training. The experimental results on FAUST, SHREC’19, and SMAL datasets show that the average correspondence errors of this algorithm are 5.1, 5.8, and 5.4, respectively, which are lower than those of the classical algorithms including 3D-CODED, Elementary Structures, and CorrNet3D. The correspondence between non-rigid 3D point cloud models calculated by the proposed algorithm has higher accuracy and stronger generalization ability.

To address challenges of unclear correlation, intricate knowledge retrieval, and difficult knowledge application across diverse domains of high-speed trains, the organizational structure involving multi-source heterogeneous knowledge pertaining to high-speed trains was first analyzed, and a knowledge graph pattern layer and knowledge graph of the high-speed train domain was developed based on the product structure tree and stage domain of high-speed trains. Subsequently, the bidirectional encoder transformer-bidirectional long short-term memory network-conditional random field (BERT-BILSTM-CRF) model was employed for entity recognition, so as to establish the mapping of stage domain ontology. Then, the entity attributes of high-speed trains were categorized into structured and unstructured attributes. The Levenshtein distance and the continuous bag of words-bidirectional long short-term memory network (CBOW-BILSTM) model were utilized to calculate the similarity of corresponding attributes, resulting in aligned entity pairs. Ultimately, the knowledge fusion graph of high-speed train domain fusion was constructed by using the coding structure tree of high-speed train products for mapping and fusion. The proposed method was applied to high-speed train bogies for verification. The results reveal that in terms of named entity recognition, the entity recognition accuracy of the BERT-BILSTM-CRF model reaches 91%. In terms of entity alignment, the F1 values (the harmonic mean of accuracy and recall) of entity similarity calculated by the Levenshtein distance and the CBOW-BILSTM model are 82% and 83%, respectively.

Some image fusion methods do not fully consider the illumination conditions in the image environment, resulting in insufficient brightness of infrared targets and overall low brightness of the image in the fused image, thereby affecting the clarity of texture details. To address these issues, an infrared and visible image fusion algorithm based on attention mechanism and illumination-aware network was proposed. Firstly, before training the fusion network, the illumination-aware network was used to calculate the probability that the current scene was daytime or nighttime and apply it to the loss function of the fusion network, so as to guide the training of the fusion network. Then, in the feature extraction part of the network, spatial attention mechanism and depthwise separable convolution were used to extract features from the source image. After obtaining spatial salient information, it was input into a convolutional neural network (CNN) to extract deep features. Finally, the deep feature information was concatenated for image reconstruction to obtain the final fused image. The experimental results show that the method proposed in this paper improves mutual information (MI), visual fidelity (VIF), average gradient (AG), fusion quality (Qabf), and spatial frequency (SF) by an average of 39.33%, 11.29%, 26.27%, 47.11%, and 39.01%, respectively. At the same time, it can effectively preserve the brightness of infrared targets in the fused images, including rich texture detail information.

Aiming at the poor interpretability of modulation recognition methods based on time-frequency deep learning, an interpretable framework of a modulation recognition network is proposed, utilizing time-frequency gradient-weighted class activation mapping (Grad-CAM). Through the key features of the hidden layer in the Grad-CAM visual deep model, the significance of the deep features extracted from the network hidden layer are illustrated in terms of correct and error recognition, revealing the decline of network performance in the environment of low signal-to-noise ratio (SNR). The contribution values of different convolution cores at each network layer are quantified and sorted to determine the network redundancy. The simulation results verify the interpretable framework of the time-frequency deep learning network for modulation recognition. The interpretable analysis results reflect that there is a large amount of noise present in the feature extraction region of the network in a low signal-to-noise ratio environment, and the tested modulation recognition network exhibits a high degree of redundancy.

To address the challenges of identifying and annotating built-up areas in synthetic aperture radar (SAR) images, a novel semi-supervised method for extracting built-up areas that combined improved pseudo-labeling techniques with an edge enhancement strategy was proposed. Initially, SAR images from the same location but at different time were introduced as a natural data augmentation method, and the pseudo-labels were determined by voting based on the prediction results of multi-temporal images. Subsequently, an edge-enhancement auxiliary module was designed, which corrected the body features of the built-up areas through feature map warping and improved edge features with skip connections. Separate supervision for the body and edge features was performed. Moreover, a dataset for extracting built-up areas in multi-temporal SAR images, which included two types of sensors and two urban areas, was constructed. This dataset contains 1,000 annotated images and 800 groups of unlabeled temporal images. Experimental validations based on this dataset have demonstrated that on the constructed test set, the baseline method trained with full data achieves an intersection over union (IoU) of 63.43%, while the proposed method reaches an IoU of 63.46% and 68.24% when using 10% and full data, respectively. Remarkably, using only 10% of the annotated data, the proposed method can achieve the precision that the baseline method has obtained with full annotated data.

Accurate prediction of microscopic traffic parameters in atypical complex scenes is a prerequisite to ensure stable operation of the intelligent vehicle infrastructure cooperative systems (IVICS). To solve the problem of vehicle speed distribution disorder and difficulty in prediction caused by bottleneck phenomenon during peak hours in the merging area under IVICS conditions, First, using the UAV video, the full-sample high-precision vehicle trajectory data of the intertwined area during peak hours are extracted from a wide-area view. Then, as bidirectional long short-term memory (Bi-LSTM) networks cost long time and affect the prediction performance of the model when training parameters are manually set, a BHO-Bi-LSTM (bayesian hyperparameter optimization bidirectional long short-term memory) integrated vehicle speed prediction model based on Bayesian hyperparameters optimization is proposed. Finally, the classical multiple linear regression model and Bi-LSTM model of vehicle speed prediction are constructed for comparison. The results show that the BHO-Bi-LSTM model outperforms other models, with a goodness-of-fit and rank correlation of 91.05% and 94.87%, respectively, and error mean, error standard deviation, mean square error, root mean square error, and normalized root mean square error of 0.0561, 0.4556, 0.2106, 0.4589, and 0.0785, respectively, which can overcome the disadvantage in prediction of complicated traffic speeds during peak hours.