In order to clarify the research context of transportation network resilience， a systematic review of domestic and international studies was provided， covering three key dimensions： network modeling， resilience assessment， and enhancement strategies. For network model construction， modeling methods for single-mode and multimodal transport networks were analyzed. Regarding the resilience assessment for transportation networks， a resilience assessment indicator system was sorted out， consisting of network topology indicators， operation indicators and attribute indicators. And four resilience assessment methods were summarized， including network indicator analysis， network performance curve evaluation， simulation modeling and data-driven techniques. Concerning resilience enhancement strategies， enhancement measures were systematically categorized into prevention， emergency response， and recovery strategies， with corresponding optimization models and algorithms summarized. Finally， the current research status of transportation network resilience was reviewed and the further research directions in this field were proposed in terms of specific disruption scenario simulation， resilience optimization strategies and multimodal transport network resilience.

The traditional zeolite with high mass transfer resistance and limited adsorption capacity， limits its application. Combining zeolite with activated carbon， diatomite， and other materials to construct composite materials with synergistic effects is the important technical ways to meet diversified needs and expand application fields. This paper conducted a critical survey on the research progress in synthesis method of zeolite-based composites， focusing on two aspects： composite zeolites and zeolite-guest material composites， and then summarized their applications in adsorption separation， catalysis， and other fields. The results showed that common synthesis strategies of composite zeolites comprise template method， sol-gel method， and zeolite silicon source method， but had disadvantages in structural instability of composites. Among these synthesis methods of zeolite-guest material composites， electrostatic self-assembly and solution blending methods can be used in combination with various types of guest materials， were promising approach to preparing zeolite- guest material composites. Meanwhile， they had disadvantages in complex processes and high solvent dependence. Despite the broad application prospects of zeolite-based composite materials in adsorption separation and industrial catalysis， research on their adsorption of gaseous organic pollutants， especially VOCs， remains limited. Universal synthesis strategies under mild conditions， such as low-temperature and low-pressure synthesis methods， as well as environmentally friendly solvent systems， should be developed in the future to promote the practical application of zeolite-based composites.

This paper focused on the effects of different solid wastes on the workability， mechanical properties， and durability of controlled low strength material（CLSM） as well as mechanisms. Various strength prediction models were systematically summarized. In addition， the strength formation mechanism and durability of cementless solid waste-based CLSM were outlined. In conclusion， future research recommendations and perspectives were given. Waste-based CLSM， offering low-carbon， highly flowable and tunable-strength advantages， is the key to greening backfill projects； establishing a waste-property/performance database， machine-learning-aided mix optimization， and systematic upgrades of durability and long-term heavy-metal safety are imperative.

Under the carbon neutrality framework， the integrated development of transportation and energy systems has emerged as a critical pathway to facilitate urban green and low-carbon transformation. Adopting a transport-energy integration perspective， this paper investigates key technologies and implementation approaches for synergistic optimization between smart grids and eg.electric bus （e-bus） infrastructure， as well as between bus energy consumption and grid power supply. The study systematically reviews major challenges and research advancements in global e-bus development， with particular emphasis on diversified charging modes. It provides an in-depth analysis of charging infrastructure configuration optimization methods considering grid constraints and their corresponding bus scheduling strategies. Building upon the development of Vehicle-to-Grid （V2G） technology， the paper examines coordinated charging-discharging scheduling strategies for e-buses and evaluates operational optimization approaches for integrated photovoltaic-energy storage-charging （PV-ES-C） bus stations. Finally， the study outlines future trends in transport-energy integration， offering both theoretical references and practical guidance for sustainable development of urban e-bus systems.

A finite element model of the bolt group was established， and its accuracy was verified through bolt group tightening experiments. The effects of arrangement pattern， preload， and tightening sequence on the deformation at the center of the connected parts were analyzed using the finite element method. Considering the uncertainty of bolt group connection parameters due to errors， the deformation reliability of the connected parts was calculated using the Monte Carlo simulation （MCS） method. To improve computational efficiency， a surrogate model was constructed based on the finite element model presented in this study， and specific applications for calculating deformation reliability were provided. The results indicate that the deformation of the connected parts after tightening is reduced when the thickness of the connected parts is increased and the hole spacing is decreased； the tightening sequence affects the deformation process； the surrogate model constructed in this study demonstrates accuracy； and when the uncertain parameters follow a normal distribution， the deformation of the connected parts also approximately follows a normal distribution.

Aiming at the problems such as the lack of independent learning ability of traditional robotic arms and poor adaptability in unknown environments， this paper proposes a deep reinforcement learning algorithm combining prior knowledge， and modifies the design of reward function and experience pool to solve the problems such as low data quality， slow convergence speed and unsatisfactory learning effect in the early stage of training， while enhancing its generalization. The results on Pybullet simulation platform show that compared with the original algorithm， the convergence speed of the improved model is increased by 48.5%， and the success rate is increased by 8%. Compared with other mainstream algorithms， the convergence speed and success rate are significantly improved.

This paper investigates the high-temperature fatigue performance of TA15 titanium alloy laser-deposited repairs under different stress levels at 500 ℃ after heat treatment at various temperatures， analyzing the impact of microstructural evolution on fatigue life. The results show that when heat-treated at 900 ℃and 950 ℃， the content of the primary α phase in the base material decreases， the secondary α phase coarsens， and the microstructure in the heat-affected zone transitions from a transitional structure to a basket-weave structure. When the heat treatment temperature reaches 1 000 ℃， the α phase transforms into the high-temperature β phase， and recrystallization results in the formation of extremely fine β grains. Among the three heat treatment temperatures， the specimen treated at 1 000 ℃ exhibited the longest high-temperature fatigue life. Under high stress， the fatigue life of the specimens treated at 900 ℃ and 1 000 ℃ was comparable， while the specimen treated at 950 ℃ showed shorter high-temperature fatigue life under both high and low stress conditions compared to the other two groups. Fractographic analysis indicated that in the 1 000 ℃ specimen， fatigue cracks predominantly propagated through the α phase， consuming significant energy， which contributed to its longer high-temperature fatigue life compared to the other two heat treatment conditions.

To design of the fracture-splitting load and the prefabricated stress groove parameters of the reducer housing bearing seat， a theoretical method for calculating the upper limit load of the fracture splitting based on the small-scale yield fracture condition of elastoplastic fracture mechanics was proposed. By this method， the upper limit load of the simplified sample of the reducer housing bearing seat was calculated to be 538 kN. Finite element numerical models were established for simplified sample of the reducer housing bearing seat with pre-fabricated stress groove depth range of 0.8 mm~2.4 mm. Numerical simulation analysis were conducted on the initiation crack of fracture splitting. By the critical J-integral criterion， the crack initiation loads for the 0.8 mm~2.4 mm series groove depth were calculated. The lower limit for fracture splitting the reducer bearing seat under the load of 538 kN was determined to be 1.6 mm； Simplified samples were cast. And laser prefabricated stress grooves were made and the splitting processing experiments of the reducer housing bearing seat were done. The experimental load is 533.39 kN， and the groove depth is 2.0 mm； The maximum deformation of the bearing hole diameter after fracture splitting and assembly is 0.18 mm， which is less than 0.25 mm which is the allowable deformation of reducer housing bearing seat after fracture splitting. The above results can provide references for the fracture splitting process parameters of the reducer housing bearing seat， the selection of the cylinder and system pressure of the equipment， and the laser processing parameters of the prefabricated stress groove.

A torsional theoretical model （PCA-STM） applicable to composite box girders with corrugated steel webs （CSW） is proposed based on the combined action softened truss model （CA-STM）， and an optimized algorithm is given. This model takes into account the prestressing effect and introduces a deformation coordination relationship between the CSW and the concrete slabs based on the yield state of the CSW. The accuracy of the theoretical model is verified by experiments and the results show that PCA-STM can well predict the full torsional behavior of the composite box girders with CSW， including the torque-twist curve， shear strain curves of concrete and CSW.

The reinforced concrete frame with energy dissipating self-centering hinge joint （EDSC-HJ） is a new type of resilient structure. Firstly， the basic construction and the modeling method of finite element model for EDSC-HJ frame are introduced. Secondly， the fundamental principle of the endurance time method is presented. The seismic performance of the EDSC-HJ frame and the conventional reinforced concrete frame （RCF） are compared through the utilization of the seismic time-history method. The results show that the dynamic responses of EDSC-HJ frame with different joint stiffness changes significantly under earthquake action when the relative rotational stiffness ratio of nodes is within the range of ［0.2， 0.5］， and excessive node stiffness will lead to the seismic performance of EDSC-HJ frame approaching that of RCF structures. Under rare earthquakes， the maximum base shear responses and the maximum acceleration responses of the EDSC-HJ frame decrease by 68% and 76%， respectively， compared to the RCF structure. The maximum inter story displacement responses of the EDSC-HJ frame increases by 19%， which is merely 31.5% of the design limit of 1/20. Compared with the RCF， the repairability of the EDSC-HJ frame under high intensity is significantly improved.

To enhance the efficiency of public transit transfers， a multi-objective collaborative optimization study on the layout of regular bus transfer stations around subway stations was conducted.A multi-objective collaborative optimization model considering three main objectives：passenger transfer convenience，travel time cost，and bus station load balance，was constructed.A solution method based on the social spider optimization algorithm was provided，and dynamic learning rate adjustment methods and Pareto optimal solution sets were introduced to optimize the algorithm.The case analysis results show that after optimization，the average shortest path is reduced by 28.98%，the average shortest path time is reduced by 34.22%，and the average station load balance is improved by 6.5%，indicating that the model has a good optimization effect on station layout.

Based on truck and private car trajectory data， a trip group classification method using Clustering by fast search and find of density peaks（CFSFDP） was proposed， in which seven trip indicators—including trip start time， trip end time， and total operation duration—were employed as feature variables. The trucks and private cars were respectively divided into three trip groups. A BP neural network-based vehicle trip group identification model was subsequently established， with trip characteristic indicators serving as inputs and the density clustering algorithm's classification labels as outputs， thereby enabling rapid identification of different trip groups within the same vehicle type.The results indicated that the proposed CFSFDP-BP-based trip group classification and identification method demonstrated satisfactory prediction accuracy and reliability. An identification accuracy of 0.991 was achieved for truck trip groups， while 0.988 was obtained for private car trip groups， with superior performance observed for truck trip group identification compared to private cars. Three distinct truck trip groups were identified： flexible-low intensity， traditional-medium intensity， and traditional-high intensity. Meanwhile， three private car trip groups were classified as commute-low intensity， commute-medium intensity， and flexible-high intensity. These findings are expected to support traffic management authorities in formulating refined traffic management strategies and improving traffic operational efficiency.

A mechanical model for shield tunneling underpassing high-speed railway tunnels that considers ground variability is developed， which can be used to analyze the randomness of longitudinal additional responses in high-speed railway tunnels induced by shield underpassing. The model accuracy was verified based on an engineering case study. The results indicate that： the longitudinal response characteristics of existing high-speed railway tunnels caused by shield tunnel underpassing can be captured by stochastic analysis， and the envelope range of additional responses in high-speed railway tunnels can also be provided， enabling a more objective evaluation of the impact of new shield underpassing construction to be made； the relationship of "deterministic result<stochastic calculation mean<stochastic calculation 95th percentile" is satisfied by the statistical characteristics of extreme values of longitudinal bending moment and shear force in high-speed railway tunnels induced by shield underpassing， and a normal distribution consistent with the probability distribution type of foundation stiffness is basically followed； the distribution patterns of additional responses in high-speed railway tunnels can be reflected by stochastic analysis results， and the distribution range and probability characteristics of key evaluation indicators such as maximum deformation can also be provided， by which the reliability of evaluation results is improved.

Based on the concept of energy intensive and replaceable design， a segmental assembled pier with external replaceable corrugated steel plate and embedded steel pipe tenon and mortise is proposed. To verify the replaceability of prefabricated bridge piers， specimens with corrugated steel plates and perforated corrugated steel plates were designed and manufactured， respectively. The quasi-static cyclic loading test was carried out on one specimen. The fatigue loading test was conducted on the other one. After the test， the damaged corrugated steel plate was replaced by a new one， and then a quasi-static reciprocating loading test was conducted again. The seismic performance and replaceability of the pier were evaluated by comparison of the damage process， energy dissipation capacity， bearing capacity. The results show that external perforated corrugated steel plates have a better matching design bearing capacity with embedded parts compared with external corrugated steel plates. Although the strength and stiffness of the steel plate were reduced， the energy consumption capacity is better. Since the energy consumption of the bottom pier is mainly concentrated on the external corrugated steel plate， there is no crushing failure and damage of the bottom concrete section before and after the test. The external corrugated steel plate can achieve replaceable performance under the design target drift ratio.

The current researches on pavement mechanics usually focus on the derivation of analytical solutions and the exploration of the factors affecting them， there are few studies on the difference of analytical solutions. In order to solve this problem， transfer matrix method， stiffness matrix method， wave transfer method and reflection and transmission matrix method were applied in the deduction process of the analytical solutions considering the transverse isotropy of materials and the contact state between layers. On this basis， the calculation efficiency and the range of adaptability of the above four analytical solutions are studied by combining four kinds of pavement structures. The results show that： ①for the matrix scale， reflection and transmission matrix method， transfer matrix method， stiffness matrix method and wave transfer method increase in turn； ② for the solution speed， the transfer matrix method takes much less time than the other three methods， while the reflection and transmission matrix method consumes the most time； ③in the numerical stability， the transfer matrix method is very easy to overflow data when calculating the dynamic response of finite thickness， while the other analytical solution methods have good numerical stability； ④in practical application， the transfer matrix method is particularly suitable for the surface of semi space problems， and the stiffness matrix method can only calculate the dynamic response of the surface of each structural layer， however， the wave transfer method and the reflection and transmission matrix method can solve the dynamic response of any position in the pavement.

To address the problem of insufficient security of public data caused by its characteristics of wide sources， large information volume， and high sensitivity， as well as the vulnerability of one-way authentication networks to various attacks， an optimized public data security algorithm applying a lightweight mutual authentication protocol was proposed. A lightweight mutual authentication protocol was obtained by combining Physical Unclonable Functions （PUF） with synchronized random numbers in this algorithm， which reduces the computational and communication overheads during the optimization process. The credibility of user readers during protocol access was authenticated through trusted network connections. Synchronized random numbers were deployed at both ends of user readers and public data electronic tags， and PUF was integrated into the public data electronic tags to enhance the resistance of keys in massive public data electronic tags against various attacks. Thus， mutual identity authentication between user readers and public data electronic tags was completed， and secure optimized protection of sensitive data was achieved. The research results show that the algorithm can effectively resist 10 types of attacks， including de synchronization attacks， cloning attacks， and man in the middle attacks， during the storage and transmission of public data， and has low latency， which can ensure the security of public data.

The TUST 50^th percentile Chinese male pedestrian injury bionic model was constructed based on CT medical imaging data of a volunteer with Chinese 50^th percentile male anthropometric characteristics， and its effectiveness was verified by the reconstruction of cadaveric blunt impact tests. In accordance with the pedestrian model certification requirements outlined in the Euro NCAP protocols， a walking?posture model is further constructed， and simulation analyses of pedestrian?vehicle collisions at various collision angles were performed. Results indicate that in lateral collisions between pedestrians and vehicles， the head impact time was delayed by approximately 10 ms compared to frontal and rear impacts， and the overall injury severity was relatively mild. The highest risk of head trauma was exhibited in rear impacts， whereas severe injuries to the lower extremities were more likely to be caused by frontal impacts. These findings demonstrate that significant influence of collision angle on pedestrian injury was exerted. Therefore， collision angle should be carefully considered in the optimal design and assessment of pedestrian protection systems. The TUST injury bionic model developed in this study provides crucial technical support for research on pedestrian?vehicle collision injury mechanisms， optimization of intelligent safety systems for vehicles， and the construction of virtual evaluation frameworks.

This paper proposes a mixed precision integer quantization method for the layer of a trained dialogue summarization model， which uses a more accurate method based on augmented Hessian matrix to evaluate layer sensitivity， considering inter-layer correlation while effectively preserving intra-layer information. Due to the large number of outliers in the conversation summarization model activation， quantifying them directly will result in significant decrease in model accuracy. In this paper， we propose a method of smoothing first and then mixed precision quantization. Activation outliers are first translated to weights using smoothing factors， and then weights and activations are quantified with mixed precision based on sensitivity assessment scores. This method can quantize the model from 16 bit or 32 bit floating-point to 4-16 bit mixed precision， which reduces the storage requirement of the model and accelerates the inference speed. On the benchmark dataset SAMSum， a significant performance improvement is achieved compared with the classical baseline system， which is almost the same as the performance of the non-quantized model.

To solve the problem of storing and transmitting star maps by space detection cameras， a method of compression of star maps of region of interest based on real-time star location is proposed in this paper. First， the top hat algorithm of the ring structure element is used to separate the star object and the background in the star map. Then， through the improved star location algorithm， the star target is detected and the coordinates are output while scanning the star map. Then， the star image is obtained by opening the window in real time according to the star coordinates of the original star map. The star image is lossless compressed by JPEG-LS image compression algorithm， and the original star map background is compressed by near lossless compression with adjustable parameters. FPGA simulation results show that the processing time of the proposed method increases by less than 6% on the basis of scanning a star map with less than 20 star points. When the background compression parameter is 2， the compression ratio of the proposed method is more than 2 times higher than that of JPEG-LS lossless compression algorithm， and the PSNR is greater than 40 dB and SSIM is greater than 0.95. This method has obvious advantages in compression performance and has significant engineering application value.

In order to solve the problem that model-based diagnosis （MBD） methods are time-consuming when observations are uncertain due to sensor failures or other factors. This paper mainly focuses on software fault detection and proposes the Bucket Sort for Diagnoses （BSD） method to solve the problem of MBD with uncertain observation.The BSD method reduces the program spectrum using maximum satisfiability， and then uses bucket sorting to determine the priority of diagnosis. Diagnosis based on feasible observations with fewer faulty sensors have higher priority. Within every bucket， the Spectrum-Based Fault Localization （SBFL） method is used to calculate the suspicion of the components， and the algorithm finally return an ordered diagnosis set. For 350 sets of test instances， the experimental results show that the number of faults identified by the BSD algorithm under the TOP-1 metric is increased by 3.91 times and 7.56 times compared with MstLikeDiag and Incremental-O2D， 6.15 times and 4.79 times under the TOP-5 metric， and 6.43 times and 4.13 times under the TOP-10 metric. Meanwhile， the average efficiency of the BSD method is 92.1 times and 106.5 times than that of the MstLikeDiag and Incremental-O2D methods for 350 test instances.

Facing complex video scenes with large motion amplitude， real-time video super-resolution algorithms are difficult to reconstruct texture details and occluded regions， based on generative adversarial network， a real-time video super-resolution method based on efficient gating and target region attention is proposed. The method firstly uses an efficient gating reconstruction network as a generative network to maintain high efficiency while adaptively selecting complex region information through a simplified gating mechanism to enhance the reconstruction results. Further， the method proposes a target region attention discriminative network to provide multiscale and spatio-temporal information feedback for the generative network， and acquires multiscale information of the complex video through the multiscale mechanism and ReLU linear attention； through the significant spatio-temporal discriminative module， it restricts the discriminative network to focus on the complex regions， and better acquires the spatio-temporal information of the complex regions. The experimental results show that the proposed method exhibits significant superiority over other SOTA algorithms， the model efficiency achieves an inference delay of 13.36 ms and a real-time score of 65.806， which indicates the efficient real-time performance of the model.

Multiple-input-multiple-output non-orthogonal multiple access（MIMO-NOMA） technique involves the NOMA idea into MIMO systems， which can effectively improve the system energy efficiency and throughput. Nevertheless， complex spatial structures and rapidly varying spatial channels of the MIMO-NOMA system decrease the system's throughput and block its application. To address this issue， this paper proposes a deep-learning-based framework for MIMO-NOMA systems that employs an efficient Convolutional Deep Neural Network（CCDNN） tailored to MIMO-NOMA communications. The method comprises multiple convolutional and hidden layers and is able to solve the power-allocation problem through a dedicated algorithm， thereby improving both the transmission rate and energy efficiency of the MIMO-NOMA system. Numerical simulations are conducted for validation， and the results demonstrate that the proposed CCDNN framework outperforms conventional approaches， offering a viable solution for enhancing MIMO-NOMA power-allocation performance and laying a theoretical foundation for future research in this area.

Integrated sensing and communication has a broad application prospect， and integrated waveform design is one of its key technologies. In this paper， a new Integrated sensing and communication waveform design method based on Costas encoding is proposed for the current problem，aimed to promote the communication rate and spectral efficiency of the waveform. The waveform embeds information within the sub-pulse of Costas waveform through PSK（Phase Shift Keying） modulation， and at the same time， combines the multicarrier characteristics of Costas waveform， utilizing the waiting time of radar reception to transmit information. Simulation results show that the proposed waveform has a “thumbtack-shaped” fuzzy function， slightly improved delay resolution， the same frequency shift resolution and bit error rate as the comparison waveform， while the communication rate and spectral efficiency can be several times of the comparison waveform. The proposed waveform solves the problems of low communication rate and spectral efficiency of the current Costas-DPSK（Differential Phase Shift Keying） waveform， and it can be applied to more passive integrated sensing and communication application scenarios.

Aiming at the problems of communication delay between controller and actuator in path tracking， strong nonlinearity of the vehicle system， and output constraints of the controller， this paper proposes a nonlinear path tracking controller considering communication delay to improve the performance of vehicle path tracking by taking an electric vehicle as the research object. Firstly， a two-degree-of-freedom vehicle model is established， a tire model is established by using the "magic formula"， and a nonlinear path tracking model with communication delay is established； Secondly， according to the output constraints of the controller， a nonlinear robust controller is designed by using the backstepping adjustment function to meet the system performance indexes and robustness requirements； Finally， the path tracking performance is verified by simulation through MATLAB/Simulink under sinusoidal and double-shifted line conditions， respectively. The results show that： The root mean square （RMS） value of the lateral error of the path tracking of the controller proposed in this paper is reduced by 0.061 9 and 0.010 1 for sinusoidal condition and 0.097 2 and 0.034 7 for double-shift condition compared to the PID and MPC methods， respectively. The robustness of the controller in this paper is better than that of the PID and MPC methods when disturbances are introduced.

To address the issues of the high computational load for fingerprint matching and the dependence of positioning accuracy on the fixed K valuein the weighted K-nearest neighbor （WKNN） algorithm for bluetooth pedestrian localization， a modified WKNN （MWKNN） algorithm is proposed for the initial localization of the target. To further enhance positioning accuracy， a particle filtering algorithm is introduced， using the initial localization resultsfrom the MWKNN algorithm as measurement values to update the particles，and ultimately achieves precise target localization. Experimental results indicate that the proposed MWKNN-PF algorithm improves positioning accuracy by 41.8% and reduces computation time by 59.6%. This algorithm provides more accurate and real-time indoor pedestrian localization for related applications.