In order to coordinate the longitudinal and lateral driving assistance systems and vehicle stability control objectives of intelligent driving vehicle， this paper proposes a vehicle integrated control strategy by integrating different driving assistance systems and vehicle stability control in a hierarchical model predictive control architecture to achieve the optimal distribution， taking the four-wheel independent drive electric vehicle as the plant. The upper layer calculates the generalized control force to meet the vehicle motion targets according to the decision results of the driving assistance strategies. The middle layer module includes two parts： the vehicle control targets prediction model and the numerical algorithm. And the desired vehicle tire forces can be obtained through the middle layer. The lower module decides the tire slip rate and slip angle according to the expected tire forces， and the tire forces are achieved by controlling the vehicle driving motor and the braking system， and the heading angle is regulated through the front wheel steering control system. Simulation and hardware-in-the-loop （HIL） verification results show that the proposed integrated control algorithm has good computation efficiency and the longitudinal and lateral control requirements of the vehicle can be ensured at the same time. Results show that the maximum errors of path tracking and velocity control are 0.011 7 m and 0.59 km/h， respectively.

A topology optimization method of multiphase porous structures considering geometric nonlinearity effects is proposed based on SIMP interpolation model. The multi-phase porous structure is formed by controlling the local volume fraction of different material phases. To reduce the number of constraints in the optimization problem， the p-norm is used to aggregate multiple local volume constraints into a global constraint function. The additional hyperelastic technique is used to avoid mesh distortion under large geometric deformation. Finally， two optimization examples are presented to verify the effectiveness of the proposed method. The optimization results show that the proposed method can achieve reasonable distribution among different material phases， and the displacement and stress of the porous structures considering the nonlinear effect are smaller. Compared with the single material structure， the porous structures composed of multi-materials has better mechanical properties.

There is bias in estimating shape parameters（\beta >\mathrm{1}） using the weighted maximum likelihood method（WMLE） under small samples. Hence， we study the bias correction method for weighted maximum likelihood estimation of shape parameters in the small sample case （sample sizes of 3 to 15）， and two bias correction functions about the sample size are proposed， which are easy to use. By Monte Carlo simulation， the accuracy and validity of the shape parameters estimated by this bias correction method are verified under different Weibull models and different small sample sizes. The example shows that the bias correction method is feasible.

In order to improve the real-time truck load estimation of convenience and accuracy， and to help to develop real-time monitoring for truck loads in a large-scale low-grade highway networks， this paper proposed a weight estimation model（Mix-MAN） for trucks integrating the multi-head attention mechanism based on the interaction effect of dynamic and static information of trucks. Firstly， multi-head attention was introduced into the model to enhance the network's ability to extract kinematic time series features； secondly， a stacked auto-encoder was used to capture the static features of trucks； finally， a feature fusion structure was designed to extract dynamic features and static features， establish the nonlinear mapping relationship between input features and weight estimation， and then to obtain the final weight estimation result of trucks. The experimental results show that compared with the MAN model without considering static information of truck， the mean absolute value error of Mix-MAN is reduced by 6%， root mean square error is reduced by 5%， mean absolute percentage error is reduced by 0.5%. The model in this paper can provide technical support for the cargo transport supervision of highway and road maintenance.

For the three-dimensional stiffened plate， a non-nested reliability based topology optimization is proposed in order to achieve lightweight and engineering design while ensuring its safety and reliability. And the reliability concept-detailed combination optimization design is carried out for three-dimensional engineering stiffened plates. By virtue of the decoupling characteristics of the sequential optimization and reliability assessment method external optimization software with variable density method as the core was applied to carry out deterministic topology optimization to improve optimization efficiency. And considering the randomness of loads and materials， the hybrid mean value method was used to search the most probable point in the reliability evaluation stage. Furthermore， the layout of reliability conceptual design was interpreted and the size optimization was carried out to realize the engineering application. The result shows that compared with the deterministic topology optimization， the reliability topology optimization proposed in this paper can meet the requirements of structural reliability and lightweight. The combination optimization strategy of conceptual design and detailed design has realized that the engineering stiffened plate can reduce the material cost by 21.40% while simplifying the layout and ensuring the performance.

AZ91 and Az91-xGd（x=0.5，1.0，1.5）magnesium alloys were prepared by extrusion method. The microstructures， hydrogen charging and hydrogen embrittlement behavior of AZ91 and Az91-xGd alloys were studied by scanning electron microscopy， electrochemical oxidation and slow strain rate tensile methods. The results show that with the addition of Gd， the size and quantity of β-Mg₁₇Al₁₂ in AZ91 decrease， and the distribution is uniform. Under the same hydrogen-charged conditions， with the increase of rare earth Gd content， the hydrogen concentration in extruded magnesium alloy AZ91 gradually decreases， which solves the problem of partial polymerization of hydrogen in the alloy to a certain extent， greatly reduces the hydrogen embrittlement susceptibility， and effectively improves the hydrogen embrittlement resistance.

In order to improve the accuracy of road target detection in the field of road traffic safety， an innovative infrared and visible fusion and detection network is established by borrowing the idea of multi-level feature image fusion for fusion in image fusion technology and the idea of Ghost bottleneck module building in GPNet to reduce the complexity of the algorithm. The network is divided into three parts： selective image fusion module， lightweight target detection module and fusion quality and detection accuracy discriminative network. Three sets of experiments were conducted as data sets under urban working conditions with an average vehicle speed of 30-40 km/h in daytime， nighttime and special weather （rain， fog， etc.）. Experimental results： the highest average gradient lift of 5.648 81， cross-entropy of 0.936 68， edge strength of 56.945 7， information entropy of 0.925 208 781， mutual information of 1.000 548 571， peak signal-to-noise ratio 3.053 893 252， Qab0.342 882 208， Qcb0.208 983 81 and mean square error reduction of 0.08. The AP， mAP and Recall of the output of the lightweight target detection network are all at the optimal level， which verifies the advantages of the innovative application of infrared and visible light technologies for road obstacle detection.

To clarify the effects of high-density interchange environments on the mental load of drivers， a field driving test with 38 drivers was conducted in high-density interchange scenarios of “Shanghai Chongqing Expressway” （No.G50） in Chongqing， China. And the ECG data were collected under natural driving conditions using PhysioLAB physiological instrument. Subsequently， a quantitative model of mental workload was developed using factor analysis based on heart rate and heart rate variability indexes to explain drivers' mental workload differences under different interchange spacing. The results show that： The reduction of interchange spacing increases drivers' mental workload， and the average mental workload in small spacing interchanges increases by 15.15% compared with normal-spacing interchanges. Between the different driving styles of drivers and other genders， there are differences in the mental workload. And the reduction in interchange spacing significantly impacts anxious drivers and female drivers. The mental workload decreases after the increased familiarity with the driving environment. Vehicle weaving can lead to a significant increase in the mental workload of the driver. The average mental workload of drivers during the merging of small spacing interchanges increasesby 13.89% compared to the diverging process.

In order to explore the influence of temporal instability on crash injury severities， a random parameter approach with heterogeneity in means and variances was constructed to analyze the heterogeneity of elderly driver injury severities. The 2015 to 2017 traffic crash data in a state of the United States was used for the analysis. 19 influencing factors were selected based on the characteristics of elderly， vehicle， road， road environment， and built environment. The average marginal effect was used to reflect the effect of each factor on the crash injury severities. The log likelihood ratio was used to test the global and local temporal instability of the influencing factors of crash injury severities. The results show that there is significant temporal instability in the influencing factors of elderly driver injury severities. The influence of three factors on the injury severities， namely， escape after collision， rural area and dry road surface environment， changes positively and negatively with years. In the built environment characteristics， the influence of crash occurring in commercial district on the serious injury severities changes positively and negatively with years. There is temporal instability in the heterogeneity of influencing factors of elderly driver injury severities. The random parameters are related to years and the influencing factors of their mean and variance are significantly different in years.

In order to study the characteristics of lane changing conflicts involving trucks on the expressway， reduce the risk， and explore the significant factors that cause traffic conflicts among lane changing behaviors of trucks. Firstly， this paper considered the traffic conflict measurement indicators of time and space dimensions， proposed judgment indexes， and used the extracted traffic conflict indexes as a category of samples in the conflict prediction model. Secondly， the XGBoost algorithm was adopted to build the prediction model， and its performance was compared with the Random Forest （RF） and Gradient Boosting Decision Tree （GBDT） algorithms. Finally， the SHAP algorithm was used to study the impact of traffic conflicts on truck lane changes， including the lane changing characteristics of trucks， the relative motion state between different vehicles， and traffic flow. The results show that the XGBoost model has a better prediction effect， with an accuracy and F1-score of 83.87% and 84.85%， respectively. The lane changing safety of trucks is greatly affected by the motion status of the preceding vehicle in the original and target lanes， and the occurrence of collision is negatively correlated with the vehicle length and traffic volume of the road section per minute. Under the interaction of multiple factors， when the traffic volume of a section is more than 70 pcu per minute， the probability of collision when trucks change lanes increases with the increase of the proportion of trucks in the traffic flow. When the vehicle length is less than 8 m， the possibility of truck collision is positively correlated with the speed along the lane line.

In light of the poor real-time performance， low accuracy， weak generalization， insufficient training data and susceptible to the external changes in compaction quality assessment of earth-rock dams. This paper proposes a compaction quality assessment model based on the binary muti-population genetic algorithm fused back propagation neural network. This model addresses the problem of small sample learning in the intelligent compaction assessment of rockfill materials on-site by adjusting the transfer activation function， improving the migration updating mechanism， and constructing a training loss function considering the epistemic uncertainty and aleatoric uncertainty. The results show that the assessment performance of the proposed model is better than 9 comparison models， and the combinatorial uncertainty can improve the generalization and data error tolerance of the model， which features generalizability and application value in other engineering scenarios.

In order to demonstrate the durability of asphalt mixture designed by vertical vibration method （VTM） under the environment of light and repeated load， this paper firstly compares the mechanical properties of AC-16 mixture formed by VTM method， Marshall method and rotary compaction method （GTM）. The anti-aging performance and fatigue resistance characteristics of VTM and Marshall design methods were deeply analyzed. The fatigue equation was established by Weibull distribution， and the fatigue life was compared. The results show that VTM forming specimens have the characteristics of high compactness， low porosity and ore gap ratio. The mechanical properties of GTM and VTM asphalt mixture are at least 35% higher than those of Marshall. GTM method， VTM method， Marshall method and field correlation were 94%， 92% and 72%， respectively. After aging， the low temperature performance of asphalt mixture designed by the two methods is similar. After short-term aging， the performance of VTM design asphalt mixture recovery asphalt changes more than Marshall's； After long-term aging VTM design asphalt mixture recovery asphalt performance changes less than Marshall； Weibull model presented good linear correlation. The fatigue resistance of asphalt mixture based on VTM design is better than Marshall method under different failure probabilities.

The standard subgrade uniformity evaluation method， which relies on data variability， struggles to quantify the spatial distribution properties of the compaction data and is unable to identify regional compaction surface non-uniformity. To solve this issue， test fields constructed under the Songtong Expressway's subgrade were used to obtain spatial data of compaction. Based on geostatistical approaches， a half-variance function model with intelligent compaction parameters were established， to reveal the probability distribution and spatial autocorrelation of the subgrade， and then evaluate the regional uniformity of the subgrade. The results reveal that， the vibration modulus followed the normal distribution； The spatial autocorrelation distance was approximately 5~8 meters and the maximum autocorrelation coefficient was about 0.6~0.8； The nugget and geographic heterogeneity were positively correlated， while the range and patchiness were negatively correlated. The nearest neighbor index revealed the level of weak point concentration， and if it is larger than 1， it signifies that the weak points are separated and the spatial uniformity is good.

A profiled steel sheeting-concrete composite bridge deck for medium- and small-span steel plate girder bridges was proposed. Two continuous deck specimens were fabricated and tested to study the failure mode and load capacities subjected to symmetrical load at two mid-span. The mid-span deflection， the strain of concrete and reinforcement， and the distribution of cracks and crack widths at the hogging moment region were obtained and discussed. The two slab specimens are identical except for the position of perfobond connectors： 9.25 m long and 1.5 m wide， the thickness of the profiled steel sheeting is 3 mm， and the depth of concrete is 24 cm. The test results showed that the failure mode of the specimens was flexure； the load capacities of two specimens were 1 128.5 kN and 1 095.7 kN， respectively； the composite slab with perfobond connectors arranged in the crest had greater bending stiffness； the distribution range of concrete cracks due to the hogging moment is about twice the width of up flange of the girder， which should be paid attention to in design. The load capacities of the continuous composite bridge deck calculated by the plastic theory fit well with the experimental results.

Nine groups of notched PPFRC beam with different proportions are tested for three-point bending test to study the control variable as polypropylene fiber dosage and length/diameter ratio. Based on double-K fracture model and four-stage fracture model， the influence of polypropylene fiber on different fracture toughness is discussed respectively. The bilinear softening curve of PPFRC is concluded， The influence coefficient is introduced to solve the problem of modifying the empirical calculation formula. The XFEM（Extended finite element method）of ABAQUS was used to simulate the fracture process， and the feasibility of the model was verified by parameter analysis. The results show that the four-stage fracture model is more suitable for judging the failure of PPFRC structure. The macro-crack initiation toughness in the model is positively correlated with the fiber dosage and negatively correlated with the length/diameter ratio. The structural failure toughness is positively correlated with the fiber dosage and length/diameter ratio. The macro crack initiation toughness and structural failure toughness of 12 kg/m³ and 20 mm polypropylene fiber concrete are 9.80 MPa·m^1/2 and 55.32 MPa·m^1/2. The β_p of macro fiber introduced in bilinear softening constitutive equation is 0.849； The value of β_p of micro fiber is in the range of 0.409-0.552 and increases with the increase of aspect ratio.

In order to investigate the dynamic properties of SRM under cyclic loading， the law of permanent deformation of soil-rock mixture （SRM） with two different particle size types， the evolution of hysteresis circle， dynamic elastic modulus and damping ratio development law of SRM under different reinforcement methods are investigated by large dynamic triaxial test. The experimental results show that the axial strain （ε）-vibration number （N） curve of the specimen under different dynamic stress amplitude is consistent with the shakedown theory and more in line with the hyperbolic model development law. The reinforcement effect significantly improves the dynamic stability of the SRM specimen； when the dynamic stress amplitude is larger， the occlusal friction effect of the SRM specimen mixed with large gravel（GL） is more significant than that of the SRM specimen mixed with small gravel（GS）， and the accumulated plastic deformation is smaller. As an inhomogeneous engineering material， the hysteresis loop shows a "sawtooth" shape at low cycle times during cyclic loading. Before and after 30 loading cycles， the growth rate of the dynamic elastic modulus of the geogrid-reinforced specimen increases， and the damping ratio begins to decrease from the unstable fluctuation. Comparing different reinforcement test data， the results show that the geocell reinforcement enhances the resilient and stiffness of SRM better than the biaxial geogrid reinforcement.

The macroscopic mechanical properties， autogenous shrinking properties and microstructural properties of engineered cementitious composites （ECC） prepared from granitic porphyry powder （GP）， limestone powder （LP） and quartz powder （QP） with the mass of 20%， 40%， 60%， 80% and 100% instead of river sand were studied. The mechanical property test results show that： The peak stress and ultimate strain of each tensile specimen first decreased and then increased with the increase of stone powder substitution rate， while the uniaxial compressive strength and flexural strength increased with the increase of stone powder substitution rate， when the stone powder substitution rate was 100%， the mechanical properties of each specimen were optimal， and the peak tensile stress reached 4.4 MPa and above， and the ultimate strain exceeded 4.2%， its uniaxial compressive strength is more than 50 MPa， and its flexural strength is more than 18 MPa. Compared with the reference group， the mechanical properties of the specimens were greatly improved. Autogenous shrinkage test results show that： the autogenous shrinkage of each specimen becomes larger with the increase of stone powder substitution rate， when completely substituted， the autogenous shrinkage of doped GP， LP， QP increased 117.3%， 127.3%， 119.5% respectively compared with the benchmark specimens， for this reason， the stability of the matrix is adversely affected by the replacement of river sand by stone powder. The microscopic test results show that adding stone powder can be used as the nucleation matrix of hydration products and promote the reaction， and the limestone powder has the best promotion effect. When the content of each stone powder is large， the dispersion of polyvinyl alcohol（PVA） fiber in the matrix is improved， and the matrix and fiber can jointly bear the load and coordinate the deformation， making full use of the strength of the matrix and fiber. The logistic model was used to fit the autogenous shrinkage test results， and the obtained ECC self-shrinkage prediction model predicted better， and the validation found that this model has certain applicability.

Crumb rubber modified asphalt（CRMA） has technical advantages in terms of cost saving， performance enhancement， etc. However， CRMA produces a large amount of harmful gases during pavement construction， and it is urgent to quantitatively analyze the emissions of CRMA. In this paper， a method for sampling CRMA emissions was developed， and gas chromatograph-mass spectrometer （GC-MS） with internal standard method was used to quantitatively analyze the chemical compositions of CRMA emissions. The main conclusions are as follows. The standard recovery rate is good， and there is no adsorption breakthrough during the process of emissions collection. The standard curves of different concentration range have good linear relationships. The addition of crumb rubber significantly increases the emission of benzene series， benzothiazole， trimethyl dihydroquinoline and other CRMA characteristic emissions. Compared with neat asphalt， the emission of n-alkanes from CRMA decreases. The emission of PAHs from CRMA is similar to that from neat asphalt. Naphthalene is the main component of the PAHs released from CRMA and neat asphalt.

The concrete-filled steel tubular （CFST） arch bridges with circular tube ribs usually adopt the sparsely arranged transverse braces， which have a large shear deformation in the out-of-plane instability， it is not suitable to use the existing calculation method based on the assumption that shear deformation is not considered. Taking the lattice columns with sparsely battened plates as the research object， the calculation formula of elastic stability bearing capacity of lattice columns considering the shear deformation is derived. Based on the calculation formula of elastic stability bearing capacity of lattice column and by employing the equivalent beam-column method， the calculation formula of out-of-plane elastic stability bearing capacity considering shear deformation for CFST arch bridge with circular tube ribs is proposed. The parameter analysis on the standard CFST arch bridges with circular tube ribs are carried out by finite element （FE） method， the results are compared with the calculation results according to formulas considering and without considering shear deformations， and formulas recommended by the existing codes and references. The results of research show that in the scope of structure dimensions commonly used in engineering， the results obtained by recommended formulas in existing codes and references are one order of magnitude smaller than those obtained by finite element method， they are not suitable for the calculation of out-of-plane elastic stability bearing capacity of CFST arch bridge with circular tube ribs. The error between FE analysis and proposed calculation formula of out-of-plane elastic stability bearing capacity considering shear deformation for CFST arch bridge is within 15%， which has a large precision improvement and good applicability.

In this paper， changes of skid resistance， water seepage and noise reduction performance of drainage asphalt pavement in Guangdong Province in recent years were compared and analyzed. The reduction ranges of road traffic accidents before and after the construction of drainage asphalt pavement were studied and the water seepage coefficients in different positions of drainage asphalt pavement lanes were tested. A new test method to test and evaluate the lateral drainage capacity of drainage asphalt pavement was put forward. The results show that the initial anti-sliding performance of the small particle drainage asphalt pavement is relatively high， but the anti-sliding performance begins to decline two years after construction. The anti-sliding performance of large particle drainage asphalt pavement continues to show an upward trend after two years of construction. The attenuation of water permeability coefficient of small particle size drainage asphalt pavement is faster than that of the large particle size drainage asphalt pavement. The water seepage performance of the vehicle wheel track belt is greater than that of the middle of the wheel track belt and the road shoulder. The thickness of the drainage asphalt pavement with large particle size is twice of that with small particle size， but its lateral drainage capacity is 2.39 times of that with small particle size. After the paving of the drainage asphalt pavement， the number of accidents in sunny days， rainy days and in all of the days in the statistical sections were reduced by 64.3%， 57.1% and 59.5% respectively. Drainage asphalt pavement plays a positive role in improving the traffic safety of drainage difficult points in expressways. This paper can provide reference for engineering decision in the construction and maintenance of drainage asphalt pavements.

Aiming at the problem of blurred background information and reduced clarity in rain images due to rain streaks and rain veiling effect， an image de-rain method based on deep frequency feature attention mechanism is proposed. Firstly， a parameter estimation network based on the frequency de-rain model is constructed， and a frequency feature decomposition module based on spatial scale transformation and octave convolution is designed to distinguish the frequency features of parameters， the spatial redundancy of low-frequency features is reduced， and the operation efficiency is improved. Secondly， in the rain streaks detection module， the multi-frequency channel attention mechanism is used to map the weight information of the rain streaks layers to enhance the weight feature diversity and improve the rain streaks detection performance. Finally， a de-rain repairing network based on focal frequency loss is designed to further repair the estimated image of the de-rain model， the image repair quality by narrowing the frequency gap is improved， and the anti-interference ability of the network is enhanced. The experimental results show that the proposed method can effectively remove the rain streaks in the real rain image scene， suppress the rain veiling effect， keep the background information of the image intact， and the image has high definition.

In order to avoid the drawbacks caused by network node attacks， a complex network node attack detection algorithm based on incremental learning is proposed. Firstly，this method utilizes threshold self-learning to denoise the signals contained in the detected complex network；Secondiy， uses a combination of support vector machine and network node attack feature extraction principles to achieve accurate extraction of complex network node attack features； Finally， the extracted node attack features are input into an improved hybrid neural network and incremental learning is carried out to achieve the classification of node attack features and achieve precise detection of complex network node attacks. Through experimental verification， the proposed method can accurately detect node attacks in various complex networks smoothly and efficiently.

Mass spectrometry is commonly used for disease prevention and diagnosis， but the large number of mass spectrometry data features and the wide variation of features among different diseases make the task of multi-disease diagnosis complex and difficult. To solve the above problems， this paper proposes the generative adversarial autoencoder integrated voting algorithm msDAGVote based on mass spectrometry data. The msDAGVote feature extraction framework uses a dual autoencoder-based generative adversarial network， and after the network has been trained by mass spectrometry data， the generator sub-network is used for feature construction. Evaluated using mass spectrometry datasets of 10 different disease types， the experimental data show that msDAGVote extracts better features than comparative method， significantly reduces the number of features required for classification while providing excellent diagnostic power for disease classification， with classification AUC over 0.98 on six datasets and 0.87 on the remaining challenging datasets.

The papers proposed an improved ant colony optimization algorithm（SACO） based on the adaptive Lévy diversity mechanism to enhance convergence accuracy and the ability to avoid local optimum. The new algorithm was applied to welded beam engineering optimization problem. SACO combines the mechanism to enhance the population diversity， making the algorithm avoid local optimum. The paper designed a series of experiments to test the performance of SACO. Experimental results show that the algorithm shows better convergence， accuracy， and the ability to avoid local optimization in function experiments. Meanwhile， the proposed algorithm is applied to the welded beam design problem， obtaining significantly better results than other comparison algorithms. SACO shows competitiveness in function optimization and welded beam design optimization， which can be used as an effective tool for solving real-life engineering problems.

Aiming at the problems of large data demand， audio and video data alignment， and noise robustness in audio visual speech recognition technology， this paper analyzes in depth the features and advantages of the four types of core models， namely， connectionist temporal classification， long short term memory， Transformer， and Conformer， summarizes the applicable scenarios of each model， and puts forward the ideas and methods to optimize the performance of the models. Then the model performance is quantitatively analyzed based on mainstream datasets and commonly used evaluation criteria. The results show that CTC has large performance fluctuations under noisy conditions， LSTM can effectively capture long temporal dependencies， and Transformer and Conformer can significantly reduce the recognition error rate in cross-modal tasks. Finally， future research directions are envisioned at the levels of self-supervised training and noise robustness.

With the continuous development of network communication technology， intrusion attacks have become increasingly covert and difficult to detect， posing a serious threat to data security. To this end， a hybrid encryption method for biological information databases based on AES and Hash algorithms is designed. AES algorithm is used to map the Information mapping of biological data to the state matrix to obtain the key. After multiple rounds of byte replacement， row displacement， column confusion and round key addition， encryption operations are completed to obtain ciphertext. Use the MD5 function in the Hash algorithm and binary method to authenticate ciphertext data information. When an error message is found， trigger the reader to modify it to ensure data integrity. When a visitor sends an access request to the database， their identity information is reviewed through a reader. If the review is successful， they can access the database， and vice versa. Using AES algorithm to decrypt ciphertext， check if there is secondary identity information in the access permission identity of this file. If there is， read this data， otherwise， prohibit reading， and achieve mixed encryption of the two. The experimental results show that the proposed method can ensure the integrity of biological data information， and the time for data encryption， data decryption， and identity authentication is extremely short.

In view of the deficiencies of poor watermark robustness and low data rate in PE files， this paper encrypts the sub-watermark information based on chaotic perturbation， which improves the robustness of the watermark； disperses and encodes the watermark information into the whole code segment based on chaotic hashing， which improves the hiddenness of the watermark； and applies the dynamic graph structure of MPPCT_DF， which increases the amount of the node encoding of the graph， and at the same time， increases a static data field in each node， which enlarges the information storage capacity. Based on this， a chaotic loading PE file dynamic graph watermarking （CBDW_PE） technology scheme is proposed， and gives the main component modules， key algorithms， shows the system implementation and conceptual data model， and finally compares and analyzes the algorithms. After comparison， the technology has better effect on shearing， tampering and reverse attack， improves the security and concealment of watermark， effectively improves the problem of low data rate of PE files， and facilitates the loading of meaningful watermark.

Aiming at the problem of low accuracy of bearing fault diagnosis based on convolution neural network in noisy environment， a multi-scale convolution neural network anti-noise model （MACNN） with channel attention was proposed. When extracting features of different scales， the model adaptively selected channels containing fault features by using channel attention to improve the anti-noise ability of the model and suppressed the influence of noise； In addition， the one-dimensional convolution of adaptive size was used to adjust the channel weights of features of different scales， and the features of different scales were adaptively fused； Finally， feature classification is carried out through full connection layer. Experimental results on two bearing datasets show that MACNN has better fault diagnosis ability than other methods under noise interference with different signal-to-noise ratios.

Convolutional neural network applied to image compressive sensing reconstruction at low sampling rate， which contains less information， and it is difficult for the reconstruction network to pay attention to the context information of the image. To overcome this problem， a compressive sensing image reconstruction based on deep unfolded self-attention network is proposed. The network combines sampling matrix and self-attention mechanism for image depth reconstruction， and fully utilizes the information of measurement values through multi-stage reconstruction module to enhance the quality of image reconstruction. The experimental results show that the network proposed can make full use of the sampling information of the image， outperform the existing state-of-the-art methods on different datasets， and the visual effect of the reconstructed image is better.

A unified deceleration is usually adopted among all train units when an emergency occurs during the operation of a virtually-coupled train. Affected by the actual operation conditions， safety risk is likely to be increased under the control of uneven spacing between train units. To solve this problem， this paper designs a multi-vehicle cooperation-based brake control algorithm. Firstly， the emergency braking scenario of a virtually coupled train is analyzed， and total risk coefficient is defined to evaluate the safety risk. Then， with the goal of minimizing the total risk coefficient， a specific algorithm is designed to calculate the train braking rate of each unit according to the actual working conditions of the virtually-coupled train. Finally， simulation results show that the algorithm has lower total risk coefficient and fewer collision counts， which proves the correctness and effectiveness of the proposed multi-vehicle cooperation-based brake control algorithm.

In order to solve the annoying boundary effects in correlation filter （CF） trackers induced by cyclic shift when sampling training patches， and improve the tracking performance， an adaptive content aware spatially regularized correlation filter （ACSRCF） is proposed. Firstly， real negative samples are generated from the background area around the target object， so as to alleviate the filter degradation induced by the fake negative samples generated from the circularly shifted object patches. Secondly， the locality sensitive histogram （LSH） based foreground feature is extracted and incorporated with the spatial regularization weight which is updated adaptively according to the varied object-oriented appearances. Thirdly， the CF model is optimized using the alternative direction method of multipliers （ADMM） in which the model is decomposed into two sub-problems and the LSH-based features are used in iteration for obtaining the optimal solutions. Finally， the proposed method is evaluated on 5 public benchmark datasets. The experimental results show that the accuracy and success rate scores of our method on OTB50 dataset are 90.3% and 66.1%， respectively， exceeding the other CF trackers. The data on the OTB100 dataset are 92.2% and 69.2%， and the accuracy ranks first among all the trackers， while the success rate is ahead of other CF trackers.

In actual working environments， exoskeleton devices for speech systems have poor voice command recognition performance due to the influence of environmental noise. This paper presents speech characteristics based on the Gammatone Frequency Cepstrum Coefficient using discrete orthogonal Stockwell transform. Time domain information of speech signal energy and zero crossing rate is characterized by discrete path transformation and composed into hybrid features. Redundancy， irrelevance， and information complementarity between the features are considered under low signal-to-noise ratios. The improved correlation fast filtering feature selection algorithm is used to obtain the optimal feature subset for the voice system of exoskeleton device control commands. Experimental results show that the optimized hybrid features are more robust under low signal-to-noise ratios， and the recognition rate of traditional Mel cepstral coefficients improves by about 20% under pink noise with zero signal-to-noise ratios.

There is a large amount of duplicate， conflicting， and redundant data in sensors， and traditional data fusion methods can only fuse partial source data， resulting in low credibility of target monitoring data. Therefore， a multi-source heterogeneous sensor data fusion model based on fuzzy theory is proposed. The D-S evidence theory is used to design data fusion rules， and the probability distribution function and distance matrix are used to reduce the calculation difficulty of multi-source data fusion rules. The binary function is used to convert each data source， and the support function is used to calculate the support value of each data source. The weight value of OWA operator is obtained with the help of fuzzy theory quantitative operator. According to these two values， conflicting data is removed， duplicate data is fused， and data source fusion is completed. The experimental results show that the proposed method can effectively reduce the fusion error of various source data， improve the reliability of monitoring data， and ensure the shortest fusion time cost.

The dynamic programming in the global energy management strategy of fuel cell/lithium battery/supercapacitor hybrid trams is effective and can improve the fuel economy， but it requires a priori operating conditions， is computationally intensive， time-consuming， and cannot be controlled online， and an improvement strategy is proposed to address the above problems. Firstly， the Markov power state transfer matrix is established based on the historical tram operation data to predict the tram operation conditions； secondly， the fuel cell power data in the tram is updated by using the sliding window with a large number of continuous and infinitely fast observation data to speed up the iteration of the dynamic planning algorithm， and the minimum hydrogen consumption is used as the objective function to obtain the optimal control by solving the forward search in the inverse direction. Finally， the proposed strategy is compared with the state machine strategy and the traditional dynamic planning strategy. The results show that the proposed strategy significantly reduces the operation time compared with the traditional dynamic planning， effectively reduces the number and amplitude of fuel cell high-current discharges， and improves the durability of the fuel cell； the reasonable power distribution of the main and auxiliary power sources during the real-time operation of the tram improves the fuel economy and average efficiency of the hybrid system.

In order to meet the energy absorption requirement of the landing buffer， this paper starts with the buffer filling material of the buffer， draws lessons from the Kelvin structure and spiral structure， and uses the engineering bionic principle to design and establish two kinds of structure models： footballene bionic structure and multi-helix bionic structure. Considering the requirement of energy absorption and reuse of the buffer structure of the lander， the biomimetic structure samples were prepared by NiTi alloy with shape memory effect and were made by additive manufacturing technology. The mechanical properties， energy absorption ability and recoverability of the samples were analyzed and verified by simulation and experiments. The results show that the accuracy of the numerical simulation is verified by comparing the force-displacement curves of the simulation test and the isometric static pressure test， in which the multi-helix bionic structure has better mechanical properties， and its maximum energy absorption is 3 096.23 J； the recovery rates of the two structures are as high as 98.02% and 97.12%， respectively， and the recovery rate of footballene bionic structure is slightly higher. In this study， the bionic buffer structure of the leg lander is prepared by the method of adding materials， which provides a reference for the bionic design of the buffer structure of the lander.

Aiming at a series of problems restricting the mechanization of maize harvesting in hilly and mountainous areas， such as small plot， difficult turning and many sloping land， a cross-roll ear picking and cutting platform was designed to solve the mechanization of maize harvesting in hilly and mountainous areas. The length of the whole machine is 0.7 m shorter than that of the conventional corn harvesting and cutting table， which effectively reduces the turning radius of the whole machine and improves the adaptability of the harvester to complex terrain. The Box-Behnken test method was used to study the influence of the angle of transverse roll picking， the forward speed of the machine and the rotation speed of the roll on the ear loss rate in the process of ear picking. The regression equation between the test factors and the index was established， and the corresponding response surface was generated. The results showed that the configuration angle of ear picking roll， the forward speed of the machine and the rotation speed of ear picking roll had significant effects on the ear loss rate in the harvest process. Finally， the optimal combination was determined as the configuration angle of ear picking roller 26°， the forward speed of the machine was 2 km/h， and the rotation speed of ear picking roller was 1 050 r/min. Under these conditions， the average ear loss rate was 1.227%， which was lower than the requirements of the national technical standard of corn harvester.