As an interdisciplinary subject of object detection, action recognition and visual relationship detection, human-object interaction (HOI) detection aims to identify the interaction between humans and objects in specific application scenarios. Here, recent work in the field of image-based HOI detection is systematically summarized. Firstly, based on the theory of interaction modeling, HOI detection methods can be divided into two categories: global instance based and local instance based, and the representative methods are elaborated and analyzed in detail. Further, according to the differences in visual features, the methods based on the global instance are further subdivided into fusion of spatial information, fusion of appearance information and fusion of body posture information. Finally, the applications of zero-shot learning, weakly supervised learning and Transformer model in HOI detection are discussed. From three aspects of HOI, visual distraction and motion perspective, the challenges faced by HOI detection are listed, and it is pointed out that domain generalization, real-time detection and end-to-end network are the future development trends.

Pedestrian flow statistics have an important value of research in the fields like intelligent security. In view of the low accuracy in pedestrian flow statistics of video surveillance systems, an intelligent statistic method of video pedestrian flow is proposed with small object features considered. Key technologies are focused in this work such as the improved Faster R-CNN (Faster region-convolutional neural network) algorithm for small object detection, moving object association and matching, and intelligent statistics of bidirectional pedestrian flow. More efforts include adapting the Faster R-CNN network structure according to the small-scale characteristics of the head object, improving the feature extraction ability of the network for small objects by using shallow features of images, and realizing real-time tracking of moving objects through the tracking algorithm based on trajectory prediction. Meanwhile, an intelligent statistic algorithm for bidirectional pedestrian flow is developed to achieve accurate statistics of pedestrian flow. To prove the effectiveness of the proposed method, the experiments were conducted in scenes with different levels of density. The results show that compared with the original algorithm, the improved object detection algorithm improves the mean average precision by 7.31% and 10.71% on the Brainwash test set and Pets2009 benchmark data set, respectively. For the intelligent statistic method of video pedestrian flow, the comprehensive evaluation index F value in various scenes can reach above 90.00%, which is 1.14%−3.04% higher than the excellent methods in recent years.

In order to extend the service life of the multi-stack fuel cell system (MFCS) and ensure that the overall degradation performance of each stack during operation gradually tends to be consistent, an MFCS adaptive power allocation method is proposed for the high-power proton exchange membrane fuel cell (PEMFC) system, which can take into account the aging of the stack. During the MFCS operation, the fuel cell output power dynamically changes according to different operating conditions, which causes the aging degree of each stack to be generally inconsistent. Therefore, the voltage degradation degree (VDD) is introduced to characterize the aging degree of the fuel cell stack during operation. In addition, a semi-empirical fuel cell model is also used to simulate the effect of aging on stack performance. Finally, a hardware-in-the-loop (HIL) test platform is built on RT-LAB. The proposed method is compared with the average and chain power allocation methods. The results show that the proposed method can coordinate the output of each fuel cell to slow down the aging rate of the stack, and reduce the hydrogen consumption of MFCS by 13.59% and 8.04% respectively.

In order to reduce the large-scale variation in the output power of the single fuel cell during the operation of the multi-stack fuel cell system (MFCS), improve the average efficiency of the MFCS and ensure the long-term stable operation of each fuel cell, based on the forgetting factor recursive least squares (FFRLS), an improved chain power distribution method with online identification is proposed for the high-power proton exchange membrane fuel cell system. This method uses the real-time online identification capability of the FFRLS algorithm to estimate the maximum efficiency range (MER) of each fuel cell in operation, and uses its boundary value as the constraint reference value to update the limited range of chain power in real time. Then the output of each stack is distributed according to the load demand power change and the order of fuel cell efficiency. Finally, on the semi-physical platform RT-LAB, compared with the average and traditional chain power distribution methods, the proposed method improves the efficiency of MFCS by 0.93% and 1.95% respectively.

In the AC/DC dual-system traction power supply system, the locomotive works in different power supply modes before and after phase breaking, which results in differences in traction return current characteristics and affects the combined rail potential distribution of different sections. A model for AC/DC dual-system traction power supply system is established by using CDEGS software, and the correctness and validity of the calculation model is verified. Given the mutual interference in traction return current between the AC and DC sections, the combined rail potential limit of the AC/DC dual-system traction power supply system is proposed, and the main factors affecting the combined rail potential distribution and its sensitivity are explored. Results show that when the rail insulation joints are not installed in the power free zone, the increase in the length of the power free zone can reduce the mutual interference in the traction return current between the AC and DC sections, leading to the reduction of the combined rail potential in the power free zone. The increase in soil resistivity can aggravate the mutual interference of traction return currents between the AC and DC sections, resulting in increase in the combined rail potential in the power free zone. The greater the rail-to-ground resistance of the DC section, the higher the combined rail potential for the DC section. The insulating joints in the power free zone between the AC and DC sections can change the reflow structure and avoid the out of limits of the combined rail potential. The first dual-system line in China is taken as an example, showing that the DC component of the combined rail potential can be reduced from 103.92 V to 60.20 V when the insulation joints are installed, which can ensure personal safety.

Figuring out the operation discharge characteristics of oil-paper insulation under different hydrostatic pressures can improve the operation and maintenance strategy of oil-paper insulation equipment, and can also provide a reference for the miniaturization and weight reduction of oil-paper insulation equipment. For this reason, surface discharge tests of oil-paper insulation were carried out on the column plate model under hydrostatic pressures ranging between 0.01 and 0.60 MPa in a self-designed hermetic oil tank without local discharge. The characteristic discharge parameters were measured in the development processes of discharge over the surface of the insulated oil-paper under different hydrostatic pressures. Several partial discharge parameters from initial discharge to breakdown, white spot and bubble were recorded. According to the previous results, together with the law of gas gas dissolution and compression under different hydrostatic pressures, it is inferred that the hydrostatic pressure increase has a inhibitory effect on gas defects in oil-paper insulation. The white spot and bubble during discharge under different hydrostatic pressures verify the inferred results. Both the test results and inferred results show that the initial discharge voltage and breakdown voltage increase as the hydrostatic pressure. Under the same voltage level, the peak discharge capacity, average discharge capacity, discharge repetition rate, and discharge power decrease as the hydrostatic pressure rises.

Lack of individual heterogeneity in traditional travel behavior models causes errors in the interpretation of real choice behaviors. In order to explore the influence of individual heterogeneity on travel choice behavior, firstly, a mixed logit based choice model and a latent-class conditional logit based choice model are built. Secondly, orthogonal design method is used to generate stated preference questionnaires for an empirical survey in Chengdu regarding travel choice behaviors of time-sharing lease on new energy vehicles. Finally, the mixed logit model is calibrated by using maximum likelihood simulation and Halton sequence sampling. The latent-class condition logit model is solved by regression analysis. The results show that access time, waiting time, in-vehicle time and cost are the key factors in choosing urban traffic modes. Both two models reveal that individual heterogeneity has a significant influence on travelers’ choice behaviors. The latent-class conditional logit model has a higher goodness of fit of 0.143 and a hit ratio of 77.85%, compared to those of 0.139 and 61.28% for the mixed logit model. Besides, the latent-class conditional logit model divides travelers into three categories, and the degree of differentiation is 0.908 4. Group 1 is most sensitive to cost but insensitive to waiting time; group 2 is more sensitive to access time and waiting time than cost; group 3 has an intermediate sensitivity to time and cost.

In order to improve the driving safety and traffic efficiency of connected and automated vehicle (CAV) at non-signalized intersection, firstly, a driving safety field model of non-signalized intersection is established, the objective function considering vehicle performance and traffic risk of all vehicles at intersection is constructed, and the corresponding constraints are also proposed. The model predictive control is used to optimize driving strategy for all vehicles at the intersection. Co-simulation platform is built based on VISSIM, MATLAB and NS3, which verifies and analyzes the performance of the proposed algorithm based on vehicle collision type, driving risk improvement and traffic congestion level, respectively. The experimental results show that when the ratio of traffic flow to traffic volume is greater than 1.0, compared with the traditional actuated control system, the gain of the proposed algorithm is greater than 90%, 10%, 10% and 5% in delay time, travel time, number of conflicts and traffic capacity, respectively. When the communication delay is less than 100 ms and the data packet loss is within 35%, the vehicle traffic efficiency at the intersection can still be guaranteed.

To understand the evolutionary law of heterogeneous traffic flows in intelligent network, based on the improved NaSch model, the simulation experiments are conducted respectively for the early, middle and late stages of intelligent network connectivity, and the basic diagram of traffic flow is obtained via numerical simulation to analyze the intrinsic connection between the capacity and the penetration rate of connected vehicles. Through Markov chain, the orderly arrangement of connected vehicles is proved to improve the road capacity, and random simulation experiments verify the theoretical derivation. The relative entropy in terms of vehicle arrangement is introduced to quantify the order of heterogeneous traffic flow, and clarify the essence of the connected and autonomous vehicle (CAV) improving traffic conditions. The results show that: the capacity increases with the penetration of CAV; in the early stage, the increase of penetration has a little effect on the capacity improvement with the maximum of 23.5%, while in the middle and late stages, it improves the capacity by 125.0%; under certain traffic density, CAV penetration positively correlates with traffic, and the relative entropy negatively correlates with traffic; when CAVs are in a separated state, the relative entropy is low and the improvement in the randomly mixed capacity reduces with increasing CAV penetration .

In order to study the impact of optimal allocation of shared parking capacity on user travel choices, a two-level planning model is established. The upper model determines the optimal parking capacity of shared parking land based on the total revenue of the parking management platform and the minimum walking cost. The lower layer establishes a multi-user balanced allocation model of shared parking, ordinary parking and public transportation to describe the choice of shared parking users’ travel modes. The connection between the upper and lower models is realized through the shared parking selection probability. The successive average algorithm is designed to solve the lower model, and it is embedded in the differential evolution algorithm to solve the upper model. A simulation test is conducted on part of the road network in the CBD area of Jintan District, Changzhou City. The result show that under different shared parking capacity allocations, the revenue of the parking management platform increases with the increase of capacity and then decrease, while the user’s travel expenses decrease with the increase of capacity and then rising, which shows that reasonable allocation of shared parking capacity can achieve a balance between the revenue of the parking management platform and the demand for shared parking.

To reflect the complexity of real-life situation and to depict the distribution of engineering characteristic (EC) values associated with each alternative in an effective way, firstly, according to the theory of product planning house of quality (PPHoQ), interval linguistic variables are used by the customer representatives to describe their preferences over the customer requirements (CRs). The aggregation of the correlations between CRs and ECs and the autocorrelations of ECs yield the importance degrees of ECs. Secondly, the gaps between the target values and the optimal ones of ECs are analyzed to calculate the expectations of population distributions for alternative plans. Thirdly, the idea of stochastic dominance theory is introduced to construct the dominance matrix via pairwise comparisons between the alternatives. Finally, the comprehensive evaluation indexes of each alternative are determined as per the importance of ECs, the dominance matrix, and the assignment priority matrix. The proposed method is applied in the product development of jaw crushers, in which five CRs and five ECs are determined by the project team, and the alternatives are selected through the house of quality in product plan and stochastic variables. The outcome of the application validates the proposed approach.

In order to study the vibration and reliability of the transmission system during the operation of a high-speed train, the vibration acceleration data of the key components of the CRH3 high-speed train transmission system are collected for a real vehicle and the kernel density estimator (KDE) method is used in statistical processing. Through data processing, an approximate curve of the probability density function of the vibration response of each key component in all directions is obtained, and the vibration of the key components of the transmission system is evaluated using the curve. The optimal confidence interval of the vibration acceleration of each key component is calculated using MATLAB. Two states of “safe” and “failure” are defined for the transmission system and key components of the transmission system, and a Weibull model of the proportional failure rate of the key components and the Markov state transition model of the drive train are established. The current state of the drive train is the initial state. Changes in the reliability of the transmission system are analyzed using the real-time failure rate and maintenance rate. The results show that the vertical vibration is strongest for the axle box bearings, gearboxes, and motor bearings in the drive train, and the vibration acceleration is concentrated in the range of 25 times, 20 times, and 10 times the acceleration of gravity, with a probability of 99.75%, respectively. It is in the range of 20.5026 times, 17.6712 times, 11.4693 times the gravitational acceleration. The optimal confidence interval for the vibration acceleration probability of each key component to be 99.75% is calculated, which provides a reference for the optimization of the system’s vibration monitoring threshold and fault evaluation. The failure rate and maintenance rate are the key factors affecting the state probability of the transmission system. An increase in the failure rate of approximately 30% reduces the state probability of the system by approximately 10%, whereas an increase in the maintenance rate from 0.05 to 0.10 increases the reliability of the system by approximately 20%.

To promote the application of biomass energy and reduce the country's dependence on fossil energy, a marine turbocharged diesel engine was transformed into a sequential turbocharged system diesel engine, and the combustion test of mixed biodiesel was carried out. When the propulsion characteristic is 25% load, biodiesel is set at 5% − 25% in total. Through experiments, the effects of different proportion of Biodiesel on the power, economy and emission of diesel were studied. The results show that at 25% load, compared with pure diesel combustion, the power performance of Sequential Turbocharged Diesel engine is reduced by about 6.5%, the fuel consumption rate is increased by about 9.2%, the NO_x emission is reduced by about 9.8%, and the soot emission is reduced by about 76.7%. When the sequential supercharged diesel engine is mixed with biodiesel for combustion, its power performance and emission are better than the original engine. Through the optimization analysis of grey decision theory, the comprehensive optimization shows that the optimal blending rate of biodiesel is 25% at 25% load. With the increase of biodiesel ratio, its power and economy decline rapidly and NO_x emission increases, so the blending rate should not be too high. The transformation into a sequential turbocharged marine diesel engine makes up for the shortcomings brought by the combustion of mixed biodiesel to a certain extent.

In view of the extensive application of curved fiber composite laminates in the lightweight design of high-speed train structures, the aeroelastic stability of elastic and viscoelastic variable-stiffness composite panels in a subsonic flow field was studied. First, classical thick theory along with a Mindlin plate was adopted for structural modeling and potential flow theory for aerodynamic modeling. An aeroelastic model of composite variable-stiffness panels with curvilinear fibers was then established adopting the principle of virtual work and the finite element method, which was solved using complex mode theory in the frequency domain. The divergence characteristics for key parameters were investigated following verification of the validity and convergence of the presented method. Numerical results show that, relative to the straight-fiber panel, the critical divergence speed can be increased by approximately 50% by varying the path orientations of the curvilinear fibers.

In order to minimize grinding amount in optimization design of heavy-haul rail grinding profile, an alignment and calculation method of the grinding amount of rail profile was established. Meanwhile, a comprehensive optimization evaluation model was designed. The model regarded wheel-rail wear index, wheel-rail contact stress and rail grinding amount as optimization sub-objectives, and the optimization results of different optimized strategies were analyzed. Firstly, an automatic alignment algorithm for rail profile was established through the theories of matrix rotation transformation, curve fitting and spline interpolation. Then the amount of rail grinding was calculated. Secondly, considering the optimization indicators such as wheel-rail wear index, contact stress and rail grinding amount, a comprehensive optimization objective function was established. The genetic algorithm was used to solve the optimized rail profile in conjunction with the vehicle track dynamics simulation model. Finally, the design results of different optimization strategies were calculated and analyzed by using the established rail profile optimization design model. The results show that, considering wheel-rail wear, wheel-rail contact stress and amount of rail grinding at the same time, the average wear index of the optimized high and low rail profile reduces by 68.9% when compared with the initial profile. The low rail contact stress obtains a decrease of 39.1%. The grinding amount gets a reduction of 21.8%. Thus the optimized effect is the best. After optimization, the high rail profile wear index and low rail contact stress decrease significantly in conditions of only considering wheel-rail wear and contact stress, but the decline rate of the grinding amount is relatively slow, reaching 11.3%. When only considering the grinding amount, the grinding amount of rail profile drops the fastest after optimization, which is 24.4%, while the wheel-rail contact stress is significantly larger.

The existing ball screw degradation assessment method usually assumes that sufficient labeled data sets are available. However, it is difficult to obtain massive labeled data sets under practical projects due to excess failure cost and difficulty of obtaining labels. To solve the above problems, an intelligent state evaluation method based on multi-scale adversarial domain adversarial learning is proposed. Combining an attention convolution neural network module and a domain adversarial learning module, a deep learning model is established by using sensor signals collected under different working conditions, so as to learn domain invariant features adaptively and realize efficient knowledge reuse and feature migration. The experimental data sets are constructed by using the ball screw degradation signals collected under multiple working conditions to verify the effectiveness of the method. The results show that the proposed method achieves a recognition accuracy higher than 89.02% in six sub-tasks of degradation state identification of ball screw under cross-working conditions with missing labels. The proposed method can fully migrate key features with labeled data and achieve the degradation state identification of target operating conditions under missing label samples.

In order to study the longitudinal vibration characteristics of the stepped lifting pipe with a complex structure under the action of the ocean current and mining vessel heave motion, the longitudinal vibration performance of a 5000 m long lifting pipe was analyzed using the vibration theory of continuous elastic rod. Firstly, a mathematical model of the longitudinal vibration of the lifting pipe was established according to the D’Alembert principle, and the natural frequency equation of the pipeline was derived by the method of separating variables. Then, the mass normalization of the vibration mode was carried out. Finally, a finite element model of the lifting pipe was established using ABAQUS software to study the longitudinal dynamic response of the pipeline. The results show that the first-order longitudinal resonance frequency of the lifting pipe is in the frequency band where the wave energy is concentrated in the mining area. With an increase in the buffer mass, the natural frequency of the lifting pipe decreases gradually, and the influence of the buffer mass on the high-order natural frequencies becomes more obvious. With the wave frequency increasing, the longitudinal amplitude, axial force and axial stress first increase and then decrease, and reach their peaks at the first-order natural frequency, which occurs at 5000, 0 and 1000 m along the lifting pipe length, respectively. As the heave amplitude of the mining vessel increases, the dynamic response of the lifting pipe increases gradually; after the heave amplitude is greater than 1.5 m, the growth rate of the dynamic response of the lifting pipe slows down. When the first-order longitudinal resonance occurs in the lifting pipe, the vibration displacement and axial force first increase and then make constant-amplitude steady-state oscillation. With the increase of seawater depth, the vibration amplitude along the length of the pipe increases gradually, the vibration equilibrium position moves down, and the vibration response time is delayed; meanwhile, the axial force and axial stress decrease gradually, and the axial stress between each pipe string segment increases sharply.

As there is angle deviation between the car body coordinate system and smarphone coordinate system, a systematic correction method for the smartphone coordinate alignment is proposed to make smartphone sensor data truly reflect the vibration acceleration of the vehicle body. This method corrects the smartphone vertical acceleration according to gravity direction, and the smartphone horizontal acceleration by means of the orthogonality of the lateral and longitudinal acceleration of the vehicle body. The maximum likelihood principle is used in the estimation of angular deviation to ensure the reliability of smartphone angle correction. Field test results indicate that the detection data of two smartphones obtained a vertical angle deviation of 0.008° and 0.007° relative to gravity direction, the horizontal angle between both smartphones is 29.75°, and the deviation from the test placement angle 30.00° is 0.25°. The amplitude and main frequency of the vehicle body acceleration respectively detected by the smartphone and the high-precision sensor are consistent in time domain and frequency domain.

An uneven humidity field forms inside double-block ballastless tracks long located in the water environment, and humidity will affect the mechanical properties of track structures. In order to study the macroscopic mechanical properties of concrete under different humidity conditions inside the double-block ballastless tracks in the water environment, combined with the humidity distribution of the double-block ballastless track in the water environment, the molecular dynamics of the nano-component (C-S-H) of the concrete matrix is established. Multi-scale calculation of concrete matrix and two-level homogenization analysis are conducted. The results show that the surface humidity of double-block ballastless tracks in the water environment show distinct gradient differentiation, and the maximum humidity difference inside the tracks can reach up to 38.41%; the elastic modulus and Poisson’s ratio of concrete increase with saturation; when concrete saturation is increased from 0 to 100%, the increases of the elastic modulus of the bearing layer, the track bed slab and the sleeper concrete attain 35.0%, 19.5% and 16.2%, respectively.

The purpose of this study was to test and evaluate the stiffness of the damping pad of a floating slab track and to provide accurate calculation parameters for the dynamic simulation analysis of the floating track. In this study, the load application range of the damping pad test samples was calculated in a finite element simulation, and the static stiffness and dynamic stiffness (5.0、10.0、20.0、30.0 Hz) of the damping pad were tested and evaluated using a mechanical testing machine equipped with a temperature control box and combined with time-temperature superposition. On the basis of the actual mechanical characteristics of the anti-vibration damping pads, the effects of using traditional 4.0 Hz parameters and actual frequency-dependent parameters on the simulated natural frequency and admittance characteristics of the floating slab track were compared and analyzed. The results show that the static stiffness of damping pads should be tested in three different load ranges according to the deformation, static analysis and analysis of the bending deformation of the base plate. The dynamic stiffness of damping pads should be tested under three different preloading conditions according to the tuning frequency, safety and insertion loss analysis of the floating slab track. In the case of no vehicle load (under a vehicle load), the natural frequency of the floating slab obtained using the 4.0 Hz parameters of the polyurethane damping pad is 27.0 Hz (15.5 Hz) , whereas the true natural frequency after considering the frequency-dependent stiffness of the damping pad is 31.5 Hz (18.3 Hz) . The natural frequency of the floating slab track would be underestimated and the vibration isolation frequency band and vibration isolation effect would be overestimated if parameters obtained at 4.0 Hz are used to analyze the vibration transfer characteristics of the floating slab track. The admittance results obtained with the parameters at the first-order frequency of the floating slab track are basically consistent with those obtained using the actual frequency-dependent characteristics.

In order to study the axial compressive behavior of rectangular concrete-filled stainless steel tubular (CFSST) short columns, axial compression tests were conducted on seven groups of rectangular CFSST short columns with different cross-sectional sizes. The failure modes, load-displacement curves, load-circumferential strain curves, load-longitudinal strain curves, and load-cross-sectional aspect ratio curves of different specimens under axial compression were obtained. The influence of aspect ratio of the rectangular cross-section on the bearing capacity of the specimens was analyzed. The results show that the typical failure mode of rectangular CFSST short columns under axial compression was local outward buckling. Under the same aspect ratio, when the wall thickness of the specimens increases from 4 mm to 6 mm, the bearing capacity of the specimens increases by 25%–57%. With the fixed wall thickness, when the aspect ratio of specimens increases from 1 to 2, the bearing capacity decreases by 22%–30%. Comparison of the test results with the calculated results by relevant codes and standards indicates that the bearing capacity of CFSST stub columns is more than 14% higher than that of the concrete-filled conventional carbon steel tubular stub columns with the same section. Furthermore, a formula for calculating the axial compression bearing capacity is obtained by numerical fitting of the data obtained in this study and those from literature, which can well predict the bearing capacity of rectangular CFSST short columns.

To accurately predict the full-range mechanical behavior of prestressed concrete composite box girders with corrugated steel webs (PCCBGCSWs) subjected to pure torsion, a theoretical model called “the improved softened membrane model for torsion (ISMMT)” was proposed based on the softened membrane theory. First, the equilibrium equations, compatibility equations, constitutive laws of materials and general solution algorithm of the ISMMT were briefly introduced. On this basis, a simplified solution algorithm was additionally presented for the stage when both steel bars, the prestressing steel and the corrugated steel webs (CSWs) are in the elastic state, which contains only one iteration loop. To validate the feasibility and accuracy of the ISMMT, a PCCBGCSW specimen was tested under pure torsion. The results, including the overall torque-twist curve, shear strains in the CSWs and concrete flanges, and strains in the prestressing steels and steel bars, were obtained from the test. Then, the experimental results were compared with the theoretical results calculated by the ISMMT. The comparison of solving efficiency between the general computer program and the simplified one was also conducted. Results show that the ISMMT not only can provide accurate prediction for the full-range torque-twist curve of PCCBGCSWs under pure torsion, but also can make precise predictions for the entire stress evolution in components of the PCCBGCSWs during loading, including the concrete flanges, CSWs, prestressing steels, and steel bars. When using the ISMMT to predict the full-range behavior of the PCCBGCSW specimen under pure torsion in this study, the number of iterations could be as high as 7.9 × 10⁶ if the general computer program was used. However, if the simplified computer program was employed, the number of iterations could be significantly reduced (5 times at least and 193 times at most), which is a great improvement in solving efficiency in comparison with the general computer program. Therefore, the ISMMT provides an effective way to analyse the full torsional behaviour of PCCBGCSWs under pure torsion.

The concrete ring beam at the entrance of subway station is generally connected with the tunnel segment by bolts. The bolts are often embedded in the ring beam of subway station, and there is bond-slip deformation between the bolts and the concrete wrapping them. This may have an impact on the opening width of the ring joint and the damage development of the ring beam structure. Based on the existing constitutive model of bond-slip between concrete and steel, a high-precision finite element program function was utilized to investigate the concrete nonlinear behavior and cracking mechanism, and a model describing the behavior of bond-slip mechanism was built up to measure the influence of the slip displacement on the width of ring gap. For three different types of connection bolts with long anchorage and short anchorage, simulations were performed to analyze the bond-slip behaviors of these bolts in the ring beam concrete and the width increasing of the gap between ring beam and segment. Meanwhile, through quantitative analysis of distributions of bond stress and bolt stress along the bolt length, the mechanism for the contribution from the displacement and deformation of connecting blots to the width increasing of the gap was revealed. Results show that the bolt connection stiffness derived from the bond-slip models is between those derived from the completely fixed model and the spring model, and the influence of the bond-slip deformation on the width of the annular gap between the shield segment and the station ring beam cannot be ignored. In the case of long anchorages, the contribution of bond-slip to the gap width can reach up to about 30% when the bolt begins to yield. After the bolt yields, the slip proportion will decrease to less than 8% with the expansion of the circumferential seam. Affected by this, the tensile stiffness of the bolt considering bond slip is at least about 1/3 of the fully fixed model

Conveyers on the bridge deck change the aerodynamic shape of the side girder. In order to explore the vortex-induced vibration performance and countermeasures of the side girder with a conveyer, a 1.00∶20.00 rigid segment model test of free suspension is carried out in wind tunnel. Firstly, the vortex-induced vibration performance of the side girder beam section with a conveyer is studied, and tests are conducted as to how it is affected by structural damping ratio. Secondly, the cases of whether a conveyer is equipped are compared. Finally, aerodynamic measures such as air nozzles, stabilizing plates at beam bottom, and horizontal baffles are used to optimize the vortex-induced vibration performance of the main girder section. The results show that the vortex-induced vibration performance of the side girder with a conveyer is poor at the specified 0° and ± 3° wind attack angles, and the maximum exceeds the specification limit value by 286%. The deck conveyer reduces the vortex-induced vibration stability of the main girder, and the peak value of the vortex-induced vibration response increases by 44%. The installation of stabilizing plates at beam bottom is beneficial to improve the vortex-induced vibration performance of the main beam, and the effect of stabilizing plates with the same height as the bottom of the beam becomes better with the increase in the number of stabilizing plates. The vortex-induced vibration suppression effect of the main beam is 93% when installing three stabilizing plates with a depth of 1.5 m. The 2.0 m high middle stabilizing plate extending 0.5 m from beam bottom can completely suppress the vortex-induced vibration. The nozzle has a little influence on the vortex-induced vibration performance of the main beam, but it has an optimal angle value in a certain range. When a horizontal baffle is separately arranged at the beam bottom, the peak value of vortex-induced vibration response is reduced by 17%. A combined measure of a nozzle, nozzle horizontal splitter plate, and horizontal baffle of 1 m width is adopted to optimize the main beam section, and the peak value of the vortex-induced vibration response of the main beam is reduced by 92%, which is far lower than the specification limit.

Flat box girder has been used in most long-span bridge because of its excellent flutter performance. To facilitate bridge designers to quickly evaluate the flutter performance of flat box girders in the preliminary design stage of long span bridges, a deep neural network model based on ensemble learning was proposed for quickly predicting flutter derivatives of flat box girders. Firstly, the flutter derivatives of 15 typical flat box girders were obtained by forced vibration wind tunnel tests, and the accuracy of flutter derivatives was verified by combining the free vibration wind tunnel test and two-dimensional flutter analysis. Then, a flutter derivative dataset with the size of 525 was constructed based on wind tunnel testing data. The proposed ensemble deep neural network model was trained and tested based on the dataset. The results show that the proposed ensemble deep neural network model can accurately and quickly predict the 8 flutter derivatives at different reduced wind speeds by relying only on the box geometry properties of the flat box girder, and only using 60% of the training dataset for training can obtain acceptable prediction results with enough precision. Compared with the traditional polynomial regression model and the single artificial neural network model, the ensemble deep neural network model proposed in this paper has higher prediction accuracy and can be directly applied to the geometry selection and flutter prediction procedure in the preliminary design stage of bridges.

To make the stress obtained by hydraulic fracturing in a high geo-temperature environment more truly predict rockbursts, it is proposed that the thermal stress of rock masses generated by geothermal gradients should be considered during the rockburst prediction. Combined with the theory of elasticity, the theoretical stress solution of hydraulic fracturing in a high geo-temperature environment was obtained firstly. Then the rockburst prediction of a circular tunnel was conducted based on this theoretical stress solution. Finally, according to the thermal stress formula of rock masses, the rockburst prediction for the Sangzhuling tunnel was applied. The results show that the instantaneous shut-in pressure and the horizontal principal stress are increased by about one time the thermal stress of rock masses, and the reopening pressure is increased by about two times the thermal stress of rock masses. In addition, vertical stresses do not change. If the stress measured by hydraulic fracturing in a high geo-temperature environment is directly used for rockburst prediction, the estimated grade of rockbursts is high when horizontal in-situses are greater than vertical in-situ stresses, and the predicted position of rockbursts is inconsistent with reality when horizontal in-situ stresses are between gravitational stresses and vertical in-situ stresses. In addition, the estimated grade of rockbursts is low when horizontal in-situ stresses are less than gravitational stresses. The thermal stress of rock masses for the Sangzhuling tunnel is about 61% of the gravitational stress of rock masses. The rockburst prediction can lead to serious errors if the thermal stress is not considered.

In order to investigate the characteristics of slope with bedrock induced by rainfall, the failure process and mechanism of slope with bedrock under rainstorm were studied systematically by laboratory model test. The variation characteristics of soil moisture content and pore water pressure in time and space before and after rainfall were discussed to reveal the mechanism of slope instability induced by rainfall. Besides, the variation law of slope bearing capacity after rain was studied by loading at the top of slope. The results show that with the development of rainfall, the liquefaction flow of soil appears first at the foot of the slope, and then the local shedding of the soil occurs there. With the continuous rainfall, the scope of soil shedding damage gradually increases, which leads to the increase of the free surface of the upper soil. After the soil is damaged, it is saturated and softened by the rain and slides downward, and the rear soil is further eroded, resulting in the slope failure of a certain depth and width. The increase of soil moisture content and pore water pressure in the slope are the main factors leading to the instability and failure of the slope. After the rainfall stops, the ultimate load that the slope can bear increases at first and then decreases, and finally tends to be stable, while the failure of the slope with bedrock shows either a global slip mode or a local slip mode under the top static load.

Weak layers have a significant effect on the stability of slopes. The stability of slopes is usually calculated by the limit equilibrium method in current designs, in which multiple equilibrium equations need to be established and solved. Compared with the limit equilibrium method, the limit analysis method is more rigorous, and requires only one energy equation to be solved. In order to analyze the stability of slopes with weak interlayers, a stability calculation model was established by the limit analysis method, and then the accuracy of the solution was verified by the limit equilibrium method. Finally, the effects of load, weak layer shape, and weak layer strength on stability were analyzed. The results show that the slope safety factor decreases with an increase in the load intensity. When the acceleration amplification factor increases from 1.0 to 1.6 the safety factor decreases from 1.20 to 0.89. With a higher frequency of the external load, the slope is easier to be damaged in advance. Besides, the shape of the weak layer has a significant effect on the slope safety factor, especially when it is close to the top and surface of the slope. The safety factor decreases approximately linearly with the decrease of the friction angle and cohesion of weak layers. When the cohesion strength decreases from 9 kPa to 5 kPa, the safety factor decreases by about 30%.