The behaviors of adjacent lane target vehicles cutting into the own lane were investigated during the commercial autonomous vehicles cruising or following in the current lane. By the lateral safety distance threshold method, the lateral entry behavior of the target vehicle was predicted and identified. Four longitudinal distance models of active fuzzy alternative safety measurement, responsibility sensitive safety model, headway and collision time were investigated and simulated. The results of simulation testing and open source data environment testing show that the active fuzzy replacement safety measurement model can fully utilize the vehicle motion performance limitations and the road environment information transmitted by the perception module, which ensures driving safety with relatively small impact on traffic flow. The algorithm with entry recognition can adopt gentle braking before the vehicle enters the self driving lane for avoiding the dangerous behavior of emergency braking in the original algorithm and improving the driving safety and riding comfort of autonomous driving.

To solve the problem of huge cost caused by the collision of agents with obstacles when the environmental information was unknown, the robust reinforcement path planning algorithm based on Bayesian optimization and quadrature was proposed. The grid map of the environment was established for setting environmental rewards. Bayesian optimization was used to establish Gaussian Process surrogate model for historical action sets, and the mean and variance of rewards were estimated by Gaussian Process. The upper confident bound (UCB) method was selected to balance exploration and exploitation, and the action sequence was selected to avoid over-exploration and over-exploitation. Bayesian quadrature method was used to actively learn the environment, and the uncertainty of environmental information was reduced by minimizing the expected posterior variance of Q value for avoiding collision and improving robustness. The Q table was updated iteratively, and the Q learning method was used to plan the path. The simulation experiment was carried out to compare the proposed algorithm with the classical Q-learning path planning algorithm. The results show that compared to the classical algorithm, the proposed algorithm has higher learning efficiency for the environment, smaller collision probability, faster convergence speed of the optimal path steps and more effective path planning.

To solve the problems of poor generalization and defective fault dataset in the existing fault diagnosis research of vehicle steering system, the fault diagnosis method of intelligent steering system based on the random forest (RF) algorithm was proposed. The physical model of the intelligent steering system was built based on the joint simulation of Simscape and CarSim. The several critical faults of steering system were simulated, and the fault datasets under various driving conditions were collected. The fault diagnosis model based on RF algorithm was established. The input data was classified to realize the fault diagnosis of intelligent steering system, and the proposed method was compared with several typical algorithms. The steering system bench was built to collect the imbalance sample data set for further validation. The experimental results show that using the RF algorithm for fault diagnosis of the simulated fault dataset, the fault diagnosis accuracy rate is approximate 87.43%. In the experimental verification, the fault diagnosis accuracy rate is 99.93% with fast diagnosis speed and good generalization.

The fourth-order fuel cell air system model was established in Simulink, which included voltage model, air compressor model, cathode model and supply manifold model. PID, fuzzy PID and sliding mode controllers were designed to control the voltage of the air compressor, so that the system could track the optimal oxygen excess ratio in time when the load current of the fuel cell was changed. The simulation results show that when the load current changes abruptly, for the reasonable output air compressor voltage ，the sliding mode control shortens the oxygen excess ratio adjustment time by about 1.5 s compared with PID and fuzzy PID control, and the optimal oxygen excess ratio can be reached faster with  better control effect, while the net output power of the system can be kept above 17 kW.

Based on the structural parameters of compact spiral wound heat exchangers with large mass flux, the physical and numerical models of shell side two-phase flow were developed to predict the variation of shell side Reynolds number with vapour quality and mass flux. The effects of vapour quality and mass flux on the pressure distribution and velocity flow line in the shell side of the heat exchanger were investigated for three different working conditions and operating conditions in the shell side of the segmented heat exchanger design. The results show that when the vapour quality is increased from 0.2 to 0.9, the liquid Reynolds numbers are respectively decreased by 87.18%, 82.41% and 90.39% in the three working sections. The pressure in the shell side neighboring layer winding channel is characterized by staggered distribution of high and low pressure locally.

The electrohydrodynamic atomization of ethanol in nanoscale was simulated by molecular dynamics (MD) method. The simulation system was established with 1 235 ethanol molecules, and ethanol molecules were placed in the gold atom-constructed nozzle for molecular dynamics computation. The effects of electric field strength and liquid flow rate on the electrohydrodynamic jet atomization process were discussed. The jet morphologies and molecular interaction mechanism of ethanol under different operating parameters were obtained. The results show that when the liquid flow rate is constant, with the increasing of electric field strength, the Taylor cone formed by ethanol atomization is gradually elongated to form a jet, and the jet length is increased. When the electric field strength is constant, with the increasing of liquid flow rate, the length of ethanol jet is increased, and the jet transforms from full shape to sharp shape, while the jet finally breaks.

To realize the high-precision motion trajectory tracking control of pneumatic actuators, the proportional directional valve-controlled pneumatic cylinder system was taken as research object, and the robust controller with online adaptive parameters was designed based on the  control strategy. The proportional directional valve flow model, the cylinder kinematic model, the cylinder friction model and the cylinder thermodynamic model of system were established. The robust controller was designed based on the backstepping method to suppress the effects of parameter estimation errors, uncertainty nonlinearities and external disturbances of system, and the online parameter adaptive rate was designed for the controller to effectively estimate the parameter uncertainty term in the system model in real time. The real-time control system for adaptive robust tracking control of the pneumatic motion trajectory was built by the XPC-Target module in MATLAB/Simulink. The experimental results show that by the designed controller， good trajectory tracking control performance is achieved. The maximum tracking error is 0.80 mm for tracking 0.4 Hz sinusoidal trajectory, which is about 2.67% of the amplitude.

To address the inefficiency of existing Fine-Tuning methods in handling minor visual variations, the transfer reinforcement learning model based on semantic selection was proposed. Inspired by the inattentional blindness that irrelevant background stimuli were overlooked during attentional shifts, the model was designed to decouple visual task transfer from policy control tasks through unsupervised semantic segmentation and semantic weight selection for preserving only policy-relevant visual features. The Flappy Bird game variant was utilized as experimental environment, and the transfer tests were conducted under varying visual interference conditions. The multiple soft attention mechanisms were compared in terms of transfer performance. The results show that the proposed method outperforms attention-based transfer approaches in efficiency, interpretability and adaptability to complex backgrounds. The significantly enhanced robustness is demonstrated, particularly in environments with substantial visual interference.

In the traditional knowledge distillation methods, the semantic segmentation tasks were not optimized, and the spatial-level feature alignment was primarily focused on with the problems of limitations in teacher-student architecture constraints and redundant knowledge transfer. To address these issues, the channel-wise features from backbone, feature enhancement layers and prediction outputs were utilized with incorporating self-attention mechanisms, softmax normalization and correlation matrices. The adaptive channel-wise feature distillation framework was developed with incorporating three key components of adaptive feature distillation for backbone layers, channel correlation distillation for fusion layers and channel significance distillation for label layers. To verify the framework, the systematic experiments were conducted on the PSPNet semantic segmentation network. The results show that by the proposed method, the segmentation accuracy is improved by 5.79% with maintaining the inference efficiency of PSPNet-ResNet18 student model. The framework can support the heterogeneous backbone architectures between teacher and student models.

To solve the problem that adversarial samples easily caused misjudgments in deep learning-based intrusion detection systems (IDSs) for industrial wireless sensor networks (IWSN) and led to decline in detection accuracy, the IWSN intrusion detection model was constructed based on generative adversarial networks (GAN) to resist adversarial attacks and improve detection accuracy. Three adversarial algorithms of fast gradient sign method (FGSM), basic iterative method (BIM) and projected gradient descent (PGD) algorithm were employed to efficiently generate adversarial samples. The GAN was utilized to conduct integrated training on three types of adversarial samples for producing novel adversarial samples. To address the prevalent mode collapse issue in GAN, the Wasserstein distance and gradient penalty constraints were incorporated into the model design. The multi-layer perceptron (MLP) was integrated into the system to perform attack detection. The proposed detection model was verified through the experiments on the real-world industrial natural gas pipeline dataset and by the TensorFlow 2 framework and Pycharm software. The experimental results show that the detection rate of the proposed model for adversarial samples outperforms those of FGSM, BIM, PGD and ensemble adversarial training-based defense methods, which can effectively defend against adversarial attacks.

The stability of nonlinear stochastic systems with Lévy noise and Markov switching was investigated. The state feedback controller was designed to ensure the mean-square exponential stability of the closed-loop system. The continuous-time state feedback controller was discretized to accommodate the requirements of practical control systems. Through derivation, it was determined that under the discretized controller, the second moment of the difference between the system state and the state under the continuous controller was bounded. The simulation results show that the discretized controller can still maintain the system stability, and the stability of the closed-loop system under the action of discrete controller is verified.

To suppress the torque ripple caused by cogging torque in axial field flux-switching permanent magnet (AFFSPM) motor, the adaptive extended Kalman filter (EKF) based on cogging torque suppression method was proposed. Based on the analysis of cogging torque and the mathematical and loss models of AFFSPM motor, the cogging torque induced torque ripple was treated as extended state variable. Together with the feedback current from the current loop, the system extended state-space model was constructed. The forgetting factor was introduced into the state estimation process to enhance the accuracy and speed of observation. Comparative experiments with the conventional harmonic current injection method were conducted under the steady and transient conditions. The results show that by the proposed method, the torque ripple is reduced by 43.5%, and the electrical loss is reduced by 14.8% at low speeds, which can effectively suppress cogging torque induced ripple and improve system efficiency.

To efficiently improve the performance of double stator vernier permanent magnet（DS-VPM） motors with considering parameter fluctuation, the multi-level robust optimization design approach was proposed. By the proposed multi-level robust optimization approach, the relevance between motor design parameters and motor performance was investigated based on flux modulation theory, and the key harmonics affecting the electromagnetic performance of motor were identified. The sensitivity analysis method was used to reduce the dimension of motor design parameters. To comprehensively consider the influence of motor design parameter fluctuations on motor performance, the multi-level robust optimization design approach was adopted to realize the optimal robust motor design with high output torque, low torque ripple and low iron loss under different operation modes. The results show that the proposed multi-level robust optimization approach is verified by the performance comparison and experiments.

Considering the consolidation rate of shaft foundation accelerated by the vacuum combined surcharge preloading and the influences of multi-stage surcharge preloading, time- and depth-dependent well resistance, radial and vertical combined seepage flow, soil nonlinear characteristics and soil gravity stress distribution on the consolidation rate of the shaft foundation, based on the vacuum combined surcharge preloading model, the nonlinear consolidation model of shaft foundation under vacuum combined multi-stage surcharge was established. Through formula derivation, the corresponding control equation system was established, and the corresponding numerical solution was obtained by the finite difference method. By comparing with the existing analytical solutions and engineering measured values, the numerical solutions were verified, and the in-depth analysis of the consolidation characteristics of the shaft foundation under various conditions was conducted. The results show that the consolidation rate of the shaft foundation is decreased with the increasing of variable well resistance parameter and increased with the increasing of final drainage capacity parameter. When the ratio of the compression index to the permeability index of soil is constant, the rate of consolidation in the early stage of the shaft foundation is decreased with the increasing of the ratio of two-stage loads, and the rates of consolidation in the middle and later stages are increased with the increasing of vacuum negative pressure. When the ratio of shaft excavation depth to the radius of shaft influence zone is greater than 20, the influence of vertical seepage on the consolidation rate can be ignored.

To investigate the micro-scale mechanisms of the mass transfer process in the folding sieve tray, the mass transfer model based on the volume-based overall liquid-phase mass transfer coefficient and the second-order irreversible chemical reaction kinetics equation were coupled. The computational fluid dynamics (CFD) methodology was employed to simulate CO2 absorption by monoethanolamine (MEA) solution on the folding sieve trays. The results show that compared with no chemical reaction, the CO2 absorption rate with chemical reactions is significantly improved. The average volumetric time enhancement factors are 32.9, 31.4 and 29.2 for apparent gas velocities of 0.56, 0.64 and 0.72 m/s, respectively, which are in the same range as those reported for similar flow rates. The simplification with considering overall mass transfer coefficient has good accuracy. Among the four affecting factors of apparent gas velocity, liquid load, mole fraction of inlet gas CO2 and mole fraction of inlet liquid MEA, the change of mole fraction of inlet liquid MEA exhibits the strongest influence on the removal rate of CO2 in the gas phase.

To solve the problem of balancing the co-processing capacity of cement kilns and controlling carbon emissions, the cement kiln co-processing compatibility matching model was established mainly based on carbon constraints. For the optimization goals of minimizing carbon emissions, maximizing waste disposal capacity, minimizing comprehensive cost and maximizing economic benefits, with the quality and safety of cement clinker as constraints, the matching method based on multi-objective mayfly optimization algorithm (MO-MOA) was proposed to optimize the matching process. The test results show that compared with the manual matching method by relying on expert experience, the proposed MO-MOA matching method reduces the labor costs, shortens the matching cycle and improves the accuracy and reliability of optimization scheme, which can avoid the problem of poor matching scheme effect caused by insufficient experience in manual operation. The comparison results with other multi-objective optimization algorithms can confirm the superior performance of MO-MOA in solving the compatibility optimization problem, and it performs better in terms of the solution  set convergence and the comprehensive quality.