To investigate the difference in driver gazing characteristics and vehicle lateral offset under two navigation modes of full-touch screen vehicle infotainment system and smart phone, the driving data acquisition platform was constructed under two types of real-road conditions. The relevant data of 27 drivers and vehicles were collected under the two navigation modes of full-touch screen vehicle infotainment system and smart phone. Five indicators of total task duration, single time of looking away from front, number of gazing switches, total time of looking away from front and standard deviation of lane position were selected. Mean comparison, variance and significance analysis were carried out on 1 041 sets of valid data of the five indicators at the four vehicle speeds. The results show that at the four vehicle speeds, for the full-touch screen vehicle infotainment system navigation, the average values of four indicators of total task duration, number of gazing switches, total time of looking away from front and standard deviation of lane position are all higher than those of mobile navigation. Compared with mobile navigation, the average values of total task duration, number of gazing switches and total time of looking away from front are increased by respective 78.55%，77.17% and 49.20%, while the average value of single time of looking away from front is decreased by 15.98%. With the increasing of vehicle speed, the total task duration, single time of looking away from front and total time of looking away from front under both navigation modes show a certain downward trend, while the standard deviation of lane position shows an obviously upward trend. Based on the analysis results of the above indicators, compared with the mobile navigation mode, the full-touch screen vehicle infotainment system navigation mode causes greater operational risks during the navigation setting stage.

To solve the problems of poor stability and driving safety of the ego-vehicle caused by the variability of the preceding vehicle′s driving conditions, taking the hub-motor-driven electric vehicles as research object, the constant speed/follow-up cruise switching strategy was proposed with considering the inter-vehicle distance and relative speed simultaneously. The optimized safe time-headway model was designed, and the longitudinal tracking control strategy for the automotive adaptive cruise control (ACC) system was developed based on hierarchical control. The upper-level controller based on model predictive control (MPC) was established to solve the desired acceleration of vehicle. Utilizing the inverse longitudinal dynamics model, the lower-level controller was constructed to solve the output torque. The torque distribution strategy for the front and rear axles was designed, and by the electric motors, the wheel drive and braking torque were precisely controlled to achieve longitudinal tracking control for ACC. The co-simulation model was established by Matlab-Simulink and Carsim software to conduct simulation tests under three driving conditions of variable speed, variable acceleration and variable deceleration. The simulation results demonstrate that the proposed adaptive cruise control system based on hierarchical control can stably follow the preceding vehicle′s movement under different driving conditions with maintaining safe inter-vehicle distance, which exhibits excellent following stability and safety.

To solve the problem of service quality testing and evaluation for low-speed automated vehicles with low accuracy of traditional automated driving function evaluation methods, the V2I-based simulation test platform for autonomous vehicles in the loop was designed, and the FES-GALMBP evaluation algorithm was proposed. The evaluation indicators for low-speed autonomous vehicle service quality were determined by the HotSpot association rule method. The indicator weights were optimized through AHP-CRITIC subjective-objective combined weighting, and the multi-level fuzzy comprehensive evaluation model with incorporating criterion layer, standard layer and indicator layer weight calculations was constructed. The fuzzy expert system (FES) was utilized to evaluate the test sample set, and the evaluation results were used as training data for the FES-GALMBP neural network model. The Prescan/MATLAB was used to complete co-simulations and real-vehicle tests in low-speed autonomous bus scenarios. The results show that by the proposed FES-GALMBP model, the accuracy rates of 94% for operational quality and 80% for operational safety are achieved, which are significantly higher than those by the traditional BP neural network model with 59% and 53%, respectively. The AUC values for all prediction categories of the proposed model are higher than those of the traditional BP model, which illuminates that the novel model has superior classification performance.

To solve the problems of long-term pedestrian trajectory prediction around autonomous vehicles with insufficient performance and poor adaptability to complex scenarios by the existing methods, the novel method was proposed. The pedestrian trajectory prediction problem was modeled, and the CrossFormer-based pedestrian trajectory prediction method was developed. The dimension-segment-wise (DSW) embedding technology was introduced to explicitly learn the correlations between adjacent time frames, and the two-stage self-attention mechanism (TSA) was combined to comprehensively capture the long-term dependencies of pedestrian trajectories. The hierarchical encoder-decoder structure was employed to adaptively capture pedestrian trajectory dependencies at different time scales for enhancing the model scalability in long-term prediction. The multi-modal information fusion, the self-attention mechanisms and the scalability optimization were innovatively integrated to achieve efficient solution for pedestrian trajectory prediction tasks. The experiments were conducted on the two datasets of ETH and Jiangsu University campus pedestrian trajectory data (JDD). The time series segmentation analysis and quantitative and qualitative evaluations were performed. The results show that by the proposed method, the values of average displacement error （ADE） and final displacement error (FDE) are respective 0.627 and 1.32 on the ETH dataset, which are significantly better than those by the traditional methods of LSTM with 0.895 and 1.74 and SR-LSTM with 0.728 and 1.66. On the JDD dataset, the values of ADE and FDE are 0.281 and 0.53, respectively, which are far superior to those of GAN with 0.562 and 1.01 and STGAT with 0.673 and 1.43. The robustness and generalization ability of the proposed method in complex scenarios are verified.

To investigate the influence of security measures on resident ride-hailing decision behavior and travel preference, the extended technology acceptance model was constructed. Based on the nationwide network questionnaire survey data, the multiple indicators and multiple causes （MIMIC）model was used to test the path of the extended technology acceptance model. The comprehensive weighted average method was used to evaluate the effectiveness of rectification measures. The results show that the security risk is the critical factor to prevent passengers from using ride-hailing. On the contrary, privacy risk has slightly negative impact on the use intention. The male, the group with high frequency of using ride-hailing and the group with old age have more positive attitude towards using ride-hailing again after safety rectification. The effectiveness analysis reveals that passengers have high recognition degree for safety response, driver reviewing and identification with scores of 6.43, 6.07 and 6.02，respectively.

To explore the phenomenon of water and gas transport and electron conduction in the cathode catalyst layer (CCL), the pore network model of CCL for the proton exchange membrane water electrolyzer was established, and the change rules of the accurate transport parameters of permeability and diffusivity and the electrochemical parameters of electronic conductivity and electrochemical activity specific surface area with the microstructure of catalyst layer were investigated. The results show that with the dcreasing of water saturation, the gas relative diffusivity and relative permeability in different directions in CCL are gradually increased, while when the water saturation is 0-0.61, the better water/gas transport characteristics are achieved. The conductivity of catalytic layer is nonlinearly increased with the increasing of platinum particle radius. When the platinum particle radius is increased from 5 nm to 20 nm, the electron conductivity is increased by 55.7%. The conductivity is decreased with the increasing of pore diameter, while the change is small. The electrochemical specific surface area of CCL is nonlinearly decreased with the increasing of platinum particle radius or pore diameter, and the decreasing is comparable.

The three-dimensional SK-type static mixer model based on the mixer without mixing elements was established to simulate the mixing behavior of gas mixtures. The operational and design parameters were investigated, and the influence patterns of mixing performance of  uniformity index and pressure drop were analyzed. The results show that when the number of mixing elements is three, the optimal balance between performance and energy consumption is achieved. The length-to-diameter ratio of the mixing elements significantly affects the mixing efficiency, while the arrangement of mixing elements has less than 5% impact on the pressure drop. For further optimization, the response surface with low interpolation error is constructed based on the Kriging surrogate model. Combined with the global optimization capability of the genetic algorithm， the optimal flow channel structure is obtained. Compared with the base model, the uniformity index at the outlet section is improved by 7.2% after optimization,and the stratification phenomenon of hydrogen and methane is completely eliminated with improved uniformity of the flow field distribution. 

To improve the work efficiency and screening yield in the circular crushing process, taking tomato straw as example, the screening performance of the spiral screening device was investigated. The numerical simulation and experimental research by EDEM software were conducted to analyze the effects of spiral speeds and charging speeds on the screening performance of spiral. The results show that with the increasing of spiral speed, the screening rate is increased with latter decreasing, and the output is gradually increased with latter unchanging. With the increasing of charging speed, the screening rate  is gradually decreased, and the output is gradually increased. Increasing the spiral speed can improve the sieving flow condition of crushed straw  particles and reduce the sieving time of unit straw  particles. The high sieving rate of 37.36% and high yield of 139.89  kg/h can be obtained when the spiral speed is 180  r/min. Increasing the charging speed can obstruct the flow of raw materials in the sieving process, which makes the sieving rate decrease. The high material concentration brings high yield, while the yield growth rate is gradually decreased. The inlet mass flow rate of 0.05 kg/s can control the screening rate more than 35% with high output, and the simulation results are verified by the experiments.

To solve the problem of low spectrum recognition rate under low signal-to-noise ratios (SNRs) conditions in cognitive radio spectrum sensing, the blind spectrum sensing algorithm based on Pietra-Ricci Index (PRI) and Support Vector Machine (SVM) was proposed. The PRI sensing decision metric was constructed by sampling the covariance matrix. The SVM was trained by the calibrated feature samples to obtain the optimal classification model for spectrum occupancy states. The PRI was utilized as feature quantity to effectively characterize the variation characteristics of the received signal. By introducing kernel function, the signal feature space was mapped to the high-dimensional space, which was expected to facilitate sample discrimination. The spectrum sensing classifier combining PRI and SVM was constructed. Using PRI as decision metric, the algorithm flow and complexity analysis were provided, and the algorithm was simulated and analyzed. The results show that the new algorithm can accurately classify the user signals and noise under low SNRs conditions, and it achieves lower computational complexity compared to similar algorithms. Compared to the existing algorithms, for the false alarm probability of 0.1, the detection probability reaches 89.4% by the proposed algorithm, which is increased by 20.0% than that by Cholesky decomposition-based method with only 69.4%. The proposed algorithm can significantly enhance the accuracy of primary user signal identification in cognitive radio systems.

To solve the problem of low computational efficiency caused by the poor selection and acceptance rate of the initial points when the classical Markov chain Monte Carlo (MCMC) algorithm was used to solve the information of river water pollution sources of discharge, discharge time and discharge location, the two-dimensional diffusion model of pollutants was constructed through COMSOL simulation software. The effects of two above aspects on the traceability results of point source water pollution by different algorithms were compared and analyzed, and the two-stage multi-chain Metropolis-Hastings algorithm based on Equidistant random sampling (ERS-TSMH) was proposed. The simulation results show that the traditional MH algorithm and TSMH algorithm are easy to fall into the local optimum or non-convergence during solving, and the acceptance rate of the former is around 20%, while that of the latter reaches nearly 50%. The multi-chain ERS-MH algorithm improves the accuracy of inversion, while it is converged and inefficient after about 10 000 iterations. The multi-chain ERS-TSMH algorithm can guarantee the traceability accuracy, while it is converged after about 5 000 iterations with significantly improved efficiency and high stability and reliability.

To optimize the anti-interference performance of Beidou satellite navigation receiver in complex electromagnetic environment, the Beidou navigation signal interference suppression algorithm based on the versoria function was proposed. The traditional minimum mean square adaptive algorithm was replaced by the improved versoria function to change the step size iteration formula and improve the convergence speed of algorithm with maintaining small steady-state error and low computational complexity. The error function was used to suppress the interference of the digital signal processing module of Beidou receiver, and the carrier tracking error and code tracking error after the suppression were simulated. The influence of the proposed algorithm on the tracking error accuracy of Beidou receiver was analyzed, and the pseudorange measurement error of receiver was calculated according to the interference ratio before and after interference suppression. The results show that the proposed algorithm can effectively reduce the impact of interference on the receiver performance and improve the tracking ability and positioning ability of receiver, and the positioning accuracy is nearly 3 times higher than that before the suppression with better stability.

To explore the repairing effect of microbial mineralization technology on load-induced cracks in concrete beams, the diffusion method was proposed to repair the vertical flexural cracks of cracked beams under loading and unloading conditions by microbial mineralization. The crack repairing effect was evaluated by chloride attack test. The results show that during the repairing process of microbial mineralization, the calcium carbonate generated in the vertical flexural crack is gradually accumulated from the bottom to top due to gravity until the crack repair is completed. The chloride ion mass fraction at the repaired crack section is significantly lower than that at the same depth of unrepaired crack section. Under the sustain-loading condition, the long-term performance improvement coefficient of chloride attack resistance of cracked beams with average crack width less than 0.120 mm after repairing is more than 74.95%. Within the same range of crack width, the long-term performance improvement coefficient after crack repairing under the unloading condition is about 15% higher than that under the loading condition.

To enhance the assembly efficiency and connection performance of beam-column assembly joints in frame structures, the novel plug-in external sleeve beam-column assembly joint was proposed. The three-dimensional solid model of joint was established by the finite element software of ABAQUS. Based on the nonlinear finite element analysis method, the joint model was subjected to low-cycle repeated loading simulations. The seismic performance indicators of the joints of failure modes, hysteresis curves, skeleton curves, energy dissipation curves and equivalent viscous damping coefficients were analyzed. The results show that the hysteresis curve of the joint is full and plump with no significant pinching effect, and the overall stiffness degradation trend is stable. The joint exhibits excellent energy dissipation capacity and ductility.

To solve the problems of poor adaptability and weak recognition ability of traditional damage identification methods for arch structures, the damage identification method of arch structure based on DenseNet121 and transfer learning was proposed. The acceleration response of the arch structure was converted into time-frequency diagram by continuous wavelet transform to divide the training set, validation set and test set samples. Based on the transfer learning strategy, the DenseNet121 model was used to learn the features of training set, and the performance was optimized through model optimization for damage identification. The test set was sent into the trained optimal model to realize damage identification. The proposed method was verified by the test of concrete filled steel tube arch structure. The results show that the proposed method has good stability during the training process and can accurately identify the damage degrees of arch structure with identification accuracy of 93.33%. Compared with ResNet50, VGG16 and AlexNet, the identification accuracy of the proposed method is improved by 2.94%, 5.53% and 10.53%, respectively. The proposed method has better comprehensive recognition effect according to five evaluation indexes of accuracy rate, precision rate, loss value, recall rate and F1.

The novel baffle with corrugated holes was proposed, and the three-dimensional liquid sloshing problem in cylindrical containers equipped with the novel baffle was investigated based on the fluid subdomain method. The fluid domain was divided into several non-axisymmetric three-dimensional subdomains, and the formal solution in each subdomain was derived by generalized Fourier transforms. Substituting the formal solutions into the free surface conditions, the rigid boundary conditions of baffle and tank wall and the continuity conditions among subdomains, the characteristic equation relating undetermined coefficients and sloshing frequencies was obtained by the weighted integral method for eliminating spatial coordinates. Based on the modal superposition method, the dynamic response equation of fluid under horizontal excitation was established. The results show that by the proposed baffle, the vortex dissipation at baffle edges is effectively enhanced, and the sloshing response amplitude is reduced. Compared with the ADINA numerical model, the fluid subdomain method is verified with high efficiency and accuracy. Additionally, the sloshing natural frequencies can be adjusted by changing the baffle parameters for influencing the steady-state response, and under equal-area conditions, the novel baffle exhibits better sloshing control performance than the conventional ring baffle.

To reveal the synergistic mechanism of nanoparticles, the graphene/copper nanocomposites with copper mass fractions of 1%, 5%, 10% and 20% were prepared by molecular level mixing method and were characterized by scanning electron microscopy, X-ray diffraction and Raman spectroscopy, which showed that the copper particles were uniformly distributed on graphene lamellae. The graphene/copper nanocomposites and the graphene and copper particles mechanically mixed in the same proportion as the composites were blended into biodegradable W40 base oil, and the two nanofluids were prepared by mechanical stirring, ultrasonic dispersion and chemical dispersion. The graphene/copper hybrid nanofluids were evaluated according to the indexes of dispersion stability, viscosity and wetting angle. The results show that the dispersion stability of graphene/copper nanofluids by molecular level mixing method is better than that by mechanical mixing method. The contact angles of both hybrid nanofluids with AISI 304 stainless steel are smaller than that of base oil, while the viscosity values are higher than that of base oil. The contact angle is decreased with the increasing of mass fraction of nanoparticles. The viscosity is increased with the increasing of mass fraction and decreased with the decreasing of temperature.