To solve the problem of tire blowouts caused by tire leakage and excessive radial force during the operation of rubber-tired metro vehicle, the dynamic model of rubber-tired metro vehicle was established, and the dynamic performance analysis of vehicles under the condition of running wheel failure was carried out. By Sobol method, the sensitivity analysis was conducted on the main factors affecting the operational stability of rubber-tired metro vehicle. Based on the non dominated sorting genetic algorithm (NSGA-II), the optimization study was conducted on the suspension parameters which have significant impact on the dynamic performance of rubber-tired metro vehicle. The results show that under the condition of running wheel failure, the operational stability and curve possibility of rubber-tired metro vehicle can still be met. By coordinating and optimizing parameters, the radial force on the running wheel of rubber-tired metro vehicle can be reduced, and the power performance of the rubber-tired metro vehicle is improved.

To meet the dual requirements of off-road vehicle power responsiveness and fuel economy, the energy management strategy based on driving style recognition under off-road conditions was proposed. The four off-road conditions were constructed by nuclear principal component analysis and clustering algorithm, and the condition identification model was constructed by random forest algorithm. The condition data was obtained by sliding window, and the energy management strategy based on model predictive control theory was established according to the online condition identification results. The driving style identification factor was introduced to optimize the objective function. The simulation results on the AVL Cruise platform show that compared to the optimized rule-based strategy , the model predictive control energy management strategy can reduce the 0-100 km/h acceleration time by 10.10% and improve the longitudinal speed following capability by 15.50%. The real vehicle verification results show that the 0-80 km/h acceleration time is 6.39 s with the 0-50 m acceleration time of 4.91 s, and the power abilities are respectively improved by 24.12% and 7.49% on average compared with those of the rule-based strategy.

The subway station is an important part of the urban rail transit network， and accidents at the station can seriously affect the normal operation of subway network.To identify the key stations, the Passenger-flow-K-shell (PK) evaluation indicator with considering passenger flow attributes was proposed based on the k-core idea. The comprehensive importance evaluation method for stations was constructed with integrating degree, intermediate number and eigenvector centrality by multi-criteria decision analysis. The simulation test of damage resistance for network was designed to verify the importance of key nodes to the network performances. Taking Beijing rail transit as example, the experiment was conducted to rank the comprehensive importance of the Beijing subway station led by Ping′anli Station. The research results show that the damage resistance performance of Beijing rail transit is greatly affected by key nodes. After the key nodes are removed, the network performance is seriously lost. Subway departments should focus on the protection of stations with high comprehensive importance and improve the network survivability.

To solve the problem of poor adaptability of teaching spraying robots for multiple varieties and small batches of workpieces, the robot spraying method for complex workpieces based on model parameterized reconstruction by 3D point cloud was investigated. The line laser vision sensor was used to scan the workpiece and obtain 3D point cloud, and the background and hooks were removed from the point cloud to segment the point cloud of workpiece. The size of workpiece was obtained by projecting and slicing in the main vector direction of the point cloud of workpiece, and the parametric modeling of workpiece was conducted. Based on the type of feature surface, the concave surface and the surface with complex connected areas were decomposed into closed and convex surfaces. The scanning line method was used to generate the spraying path and carry out robot trajectory planning, and the robot spraying system was built. The multiple test results show that the system can reconstruct 3D models and plan robot trajectories based on the point cloud of workpiece. The robot can move along the planned trajectory and is not sensitive to the requirements of suspension posture, which meets the actual needs of spraying.

The electrically driven unfolding structure with efficient thermally/electrically triggered shape memory effect was prepared by reserving  conductive path in the polyurethane acrylic printing matrix and spraying graphene conductive solution. The electrically responsive shape memory behaviors of the structure were tested at room temperature and underwater environments. The results show that the unfolded structure can quickly respond and recover under low voltage conditions in different environments. The prepared electrically driven structure has excellent shape recovery function in underwater environment, and the shape changes of different surface structures under electrical stimulation have significant deceleration effect on the underwater vehicle. This study provides new ideas for the application of 4D printing in underwater intelligent actuators and electronic devices.

To fully and reasonably utilize the contextual features of actions, category-related features and predicted future features for rapid action detection in online action detection, the online action detection method based on linear attention and category association feature learning was proposed. The Transformer architecture was improved by employing lightweight linear self-attention based on the Hadamard product to reduce computational cost for video contextual feature learning. The action features from training samples were clustered to associate the video sequence context with action category features for achieving the effective learning of category-associated feature representations. By integrating contextual features, category-associated features and predicted future features, the action discrimination at corresponding moments was enhanced. The performance experiments were conducted on typical datasets to realize the hyperparameter selection analysis, and the working accuracy and operational efficiency of different methods were compared. The ablation experiments and visualization analysis were provided. The results show that on the Thumos14 (TSN-Anet), Thumos14 (TSN-Kinetics) and HDD datasets, the mAP values of the proposed method are respectively improved by 0.2, 0.5 and 0.2 percentage point compared with the Colar method, which indicates that the new method outperforms the currently advanced Colar method.

To solve the problems of pavement type recognition models with large volume and low accuracy, the improved ShuffleNet V2 network pavement type recognition model was proposed. The efficient channel attention (ECA) module was added to the ShuffleNet V2 network structure to achieve cross channel information interaction by attention mechanism. The size of convolution kernel was adjusted according to the number of input channels. The ReLU function was replaced by the LeakyRelu function to avoid the invalidation of activation function. In order to improve the feature extraction ability and generalization ability of the model, the module composed of inflated convolution was introduced to obtain the wider range of image information with the image resolution unchanged. According to the classification characteristics of pavement types, the number of each module stacked and the overall architecture of the network were adjusted to reduce the model′s computational and parametric quantities. The improved algorithm was verified on the road surface classification dataset (RSCD). The results show that the enhanced ShuffleNet V2 model achieves parameter quantity of 4.67×106, representing reduction of 1.4×105 compared to the original model. The accuracy reaches 95.53% with improvement of 0.71 percentage point over the pre-optimization level. The inference time is reduced by 31%, and the accuracy of road surface type recognition and response speed are improved.

To meet the requirements of high torque, wide speed range and high efficiency of drive motors in the field of electric vehicles, the weak magnetic current minimum speed control strategy based on cross axis magnetic field regulation was proposed with considering the characteristics of FLCPM motors with high speed, low magnetic flux and heavy load high magnetic flux. By utilizing the cross axis magnetic field to reconstruct the distribution of id and iq currents in weak magnetic control, the leakage flux of the permanent magnet was limited and indirectly controlled to achieve air gap flux control. In order to reduce the impact of leakage flux changes on current distribution calculation and demagnetization current calculation, the minimum is current weak magnetic constraint condition and online adjustment judgment were introduced, and the id and iq distribution quantities were optimized to improve the efficiency of weak magnetic operation. The iq current weak field control and minimum weak field current control experiments were conducted on the FLCPM motor, and the current magnitude, speed regulation characteristics and weak field efficiency characteristics were analyzed. The results show that the traditional id current weak field on FLCPM motors has excessive weak field, while the minimum weak field current control has smaller is current, wider speed range and higher weak field efficiency. The current is reduced by up to 2.83%, and the weak field efficiency can reach up to 91.0%.

To solve the problem of AC components of the three-phase inductance asymmetry of linear motors caused by the unique longitudinal end effect in the position and speed estimation errors of sensorless control, the improved phase-locked loop (PLL) based on second-order notch filter was proposed for linear flux-switched permanent magnet (LFSPM) motor. The mathematical model of the LFSPM motor with considering inductance asymmetry was established, and the sliding mode observer was designed based on the stationary coordinate system. According to the analysis results, due to the inductance asymmetry, the extended back electromotive force could contain positive and negative sequence components, resulting in AC components in the PLL position error that could not be suppressed by conventional PI controllers. The principle of the improved PLL suppressing the AC components of position error was theoretically explained. The simulation model of the LFSPM motor sensorless control system was built using MATLAB/Simulink, and the sensorless control effects of the conventional PLL and the improved PLL were discussed. The experimental platform was constructed to test the proposed control strategy. The results show that compared with the conventional sensorless control methods, the improved PLL scheme based on the second-order notch filter can effectively improve the sensorless control accuracy of the LFSPM motor with reducing the speed error by 35.71%, the position error by 44.46% and the thrust ripple by 12%, respectively.

To solve the performance degradation of model predictive current control caused by parameter mismatch in five-phase permanent magnet synchronous motor (PMSM) under open-circuit fault, the novel model-free predictive fault-tolerant control strategy was proposed. The five-phase PMSM model was built in the rotary synchronous frame under open-circuit fault condition, and the total disturbance caused by parameter uncertainties and unmodeled dynamics was analyzed. Based on the ultra-local model, the extended state observer was designed to estimate the disturbance in the fundamental and the third harmonic subspaces. The disturbance was employed as compensation of the voltage references which were obtained on the basis of deadbeat control. The proposed MPFTC model for five-phase PMSM was built in MATLAB/Simulink, and the performance comparison simulation analysis with the traditional MPFTC was conducted. The five-phase PMSM experimental platform was established, and the relevant experiments were carried out by the traditional MPFTC strategy and the proposed MF-PFTC strategy, respectively. The results show that by the proposed method, excellent steady-state, dynamic and robust performance are achieved under both healthy and open-circuit fault conditions, and the proposed strategy is verified.

To solve the problem of Kalman filter algorithm with insufficient accuracy in estimating the state of charge of lithium-ion batteries, the improved particle swarm optimization (IPSO) based dual Kalman filter method was proposed. On the basis of particle swarm optimization algorithm, the spider movement strategy based black widow optimization algorithm (BWOA) was introduced in the particle speed update method. The improved particle swarm optimization algorithm was used to optimize the noise Covariance matrix of the dual Kalman filter. Based on the experimental data, the parameter identification and state of charge（SOC）estimation were achieved by the second-order resistor-capacitor circuit（RC）equivalent circuit model. The parameter identification and SOC estimation by Kalman filter algorithm were compared with those by particle swarm optimization Kalman filter algorithm. The results show that by the improved method, the accuracies of parameter identification and SOC estimation can be significantly improved with better anti-interference ability, and the accuracy improvement range of parameter identification estimation is 7.9%-38.5%, while the accuracy improvement range of state of charge estimation is 41.0%-51.4%.

 Based on the peridynamics theory and the updated Lagrangian particle hydrodynamics model, the nonlocal coupled method was proposed for the water entry problem of wedge-shaped bodies. The nonlocal differential operators were used to solve the Navier-Stokes equations of Newtonian fluids. Through the simulations of different working conditions, the effects of the water entry height and angle of wedge-shaped bodies with the same mass on the motion of wedge-shaped bodies and the evolution process of typical flow fields near the wedge-shaped bodies were quantitatively analyzed. The relationship between the immersion length of wedge-shaped body and the water entry depth under different water entry angles was also investigated. The simulation results were compared with the existing experimental results and analytical results to verify the proposed method. The results show that nonlocal fully coupled method has two major advantages of maintaining the consistency with partial differential equations and its spatial gradient interpolation with second-order accuracy. The proposed method is verified through the simulation of the water entry problem of wedge-shaped bodies. When the entry angles are 15°, 20° and 25°, the immersion length of the wedge-shaped body is linearly increased with the water entry depth. The trend is consistent with previous research findings, which further validates the accuracy of the nonlocal fully coupled method.

To investigate the influence of floor composite effect on the seismic performance of prefabricated steel reinforced concrete（SRC）column-steel beam joints, the finite-element model of the bolted-welded hybrid connection at edge joint was developed in ABAQUS and validated based on the quasi-static cyclic test data. Based on the validated model, the hysteretic response of the prefabricated steel-concrete composite edge joint with an added floor slab was analyzed with considering floor width, floor thickness, floor reinforcement ratio and column axial compression ratio. The results show that the floor structural parameters and the axial compression ratio markedly affect the stress magnitude and distribution within the joint core and the floor. However, the overall seismic performance of the joint is slightly influenced by the variations in floor structural parameters. The load-carrying capacity and the energy dissipation are significantly increased with the increasing of axial compression ratio and accompanied by heightened risk of brittle failure. The best comprehensive seismic performance of bearing capacity and ductile energy dissipation is achieved when the axial compression ratio is 0.3.

To enhance the identification efficiency, Fourier transform infrared spectroscopy (FTIR) combined with machine learning was employed to analyze the brands and grades of asphalt under various aging states. Attenuated total reflection-Fourier transform infrared spectroscopy (ATR-FTIR) was utilized to extract features from 72 asphalt spectral samples under multiple aging conditions. Feature extraction and spectral data simplification were performed by successive projections algorithm (SPA), principal component analysis (PCA) and feature peak absorbance analysis. The processed spectral data were served as input to establish classification models by support vector machine (SVM) and linear discriminant analysis (LDA). Cross-validation and confusion matrices were applied to evaluate the model performance. The results show that the fused model integrating spectral feature selection of SPA, PCA and feature peak absorbance analysis with LDA achieves 100% classification accuracy in identifying asphalt aging states with cross-validation accuracy exceeding 98%. The combination of spectral feature extraction and LDA significantly improves the generalization capability and classification precision, which can effectively distinguish asphalt brands and service conditions under diverse aging states.

To investigate the parameter sensitivity of deformation of adjacent shallow-foundation buildings induced by deep excavation, the orthogonal experimental design was adopted. The key parameters were selected with soil cohesion, elastic modulus, internal friction angle, building foundation depth, structural load, foundation stiffness and building-to-excavation distance. The maximum settlement and horizontal displacement of the adjacent building foundation were used as primary evaluation criteria. The numerical simulations were performed using Midas GTS software. The model was verified by the comparison with field measurement data from a case study, and the range analysis was applied to the results to evaluate parameter sensitivity. The results show that under the constant excavation method, the soil elastic modulus is the key influencing parameter on the maximum foundation settlement and horizontal displacement. The building-to-excavation distance, structural load and foundation depth are important parameters. Foundation stiffness, soil internal friction angle and soil cohesion are general influencing parameters.

 To improve the utilization rate of locally sourced vesicular basalt, the vesicular basalt rocks and aggregates from Tianchang County of Anhui Province were taken as research subject. The mineral compositions and chemical constituents of the rocks were analyzed, and the physical properties and mechanical indicators of the vesicular basalt rocks and aggregates were tested to identify the key indicators. The results show that the rocks with high strength are mostly bluish in color, while those with low strength tend to be gray or grayish-red. The crushing characteristics of basalt aggregates are comparable to those of limestone with relatively low crushing values, while the crushed aggregates account for significant proportion. There is nonlinear negative correlation between the bulk specific gravity and water absorption of the aggregates. In general construction, the saturated compressive strength of igneous rock is typically required to be no less than 100 MPa with the minimum value of 80 MPa. For vesicular basalt, the saturated water absorption must be controlled below 1.07% and 1.71%, respectively. The smaller the particle size of the basalt aggregates is, the higher the content of weak particles is. To improve the utilization rate of locally sourced materials, the use of aggregates with particle size smaller than 9.50 mm can be reduced.