Aiming at the conflict between the modeling style and ergonomic design of sport vehicles， the general layout design of a domestic B-class SUV is taken as the main content. Firstly， the development process of automobile general layout scheme is introduced based on ergonomics to set the initial layout scheme. Secondly， the modelling and simulation software of ergonomics， RAMSIS， is used to check the proposed initial layout scheme. To solve the layout problems， the contradiction between the vehicle modeling style and the occupant arrangement in the vehicle is reconciled by adjusting several parameters and implementing partial improvement under the condition that the vehicle size definition remains unchanged. Finally， the subjective evaluation test is carried out on the automobile man-machine test platform.

Aiming to overcome the problems of high cost and long development cycle of engine calibration test， a Gaussian Process Regression model based on Particle Swarm Optimization （PSO-GPR） is proposed to deal with nonlinear and complex engine performance and emission prediction for improving test efficiency. Working with calibration tests to the ignition angle of a gasoline engine and combined with the test design of gap filling， engine operating parameters such as torque， fuel consumption， IMEPcov， HC， NO _x and CO emissions have been predicted by using the model with a small amount of test data. R²， RAAE and RMAE were introduced to evaluate the model generalization ability. On this basis， the influence of different number of training sets on the generalization ability of the model was studied， and the universality of the model has been verified with three different engines. The results show that the PSO-GPR model can predict the engine performance and emission parameters with expected accuracy which is better than traditional GPR model and Multivariate Polynomial Regression （MPR） model. The study suggests that the model is universal， which provides a reference for reducing the workload of engine calibration.

In order that the hydraulic manipulator outputs high stiffness while working，and changes stiffness to realize compliance while collision， the robot has corresponding flexibility under different working conditions and scenarios. A hydraulic servo flexible drive mechanism with both valve-controlled leakage and floating position was designed. Stiffness was reduced by internal and external unloading， so the robot could improve the safety of man-machine physical contact. Firstly， the joint structure and the principle of variable stiffness were introduced. Then， the dynamic simulation model was established by AMESim. The ranges of the valve controlled leakage and the radius of floating leakage were determined， when the following characteristics and motion characteristics of the mechanism were good. Next， the stiffness simulation model was established by MATLAB/Simulink， stiffness scatter values under different working conditions were obtained by simulation. Then the relationship between the single factor and the variable stiffness of the joint was analyzed. Finally， the multivariate nonlinear regression equation of joint stiffness， about floating position， valve control leakage and volume， was fitted based on MATLAB. And the comparison diagram of the initial stiffness scatters and the fitting four-dimensional curve was realized by the program. The results show that the stiffness of the flexible drive mechanism can be adjusted continuously in a wide range while ensuring good dynamic characteristics.

When using the traditional method to control the drive system of new energy vehicles， there is no speed deviation， resulting in the problems of slow speed， high wheel slip rate， unstable vehicle centroid displacement and high motor torque. Therefore， under the condition of spatial geometric constraints， the unified coordinated control method of new energy vehicle drive system is proposed. Firstly， the drive system model of a new energy vehicle is constrained by spatial geometry method to find and repair the fault. After the repair， the system power frequency domain of the new energy vehicle without coordinated control is analyzed， and the conclusion is obtained that the angular velocity difference of the new energy vehicle will change with the total input torque difference. Then， the engine torque of the new energy vehicle is transmitted to the transmission， the transfer function of the drive system is obtained， and the reason for the decline of the power performance of the drive system is obtained. Finally， the dynamic coordinated control method is used to regulate the speed deviation of new energy vehicles to realize the coordinated control of the drive system. The experimental results show that by the proposed control method， the speed of the new energy vehicle is relatively fast， the slip rate of the driving wheel is low， the vehicle centroid displacement is stable， and the motor torque is lower. It is proved that the proposed method can control the drive system of the new energy vehicle with good stability performance.

The mandatory stability constraints of the existing trajectory planning methods lead to insufficient utilization of the vehicle's collision avoidance potential. This makes it impossible to plan an effective collision avoidance trajectory for intelligent vehicles under certain critical conditions. Aiming at this practical problem， this paper proposes the concept and definition method of emergency conditions， and based on the optimal control theory， integrates nonlinear vehicle dynamics model， stability domain information and environmental information， and considers the saturation constraints of the vehicle steering actuator to develop a method for planning the collision avoidance trajectory of intelligent vehicles under emergency conditions. The simulation results show that this method can accurately plan the safe collision avoidance trajectory when the vehicle is in a critical steady state under emergency conditions， and it has good applicability on different roads， which provides strong theoretical support for the development of collision avoidance control systems for intelligent vehicles.

To address the optimal selection problem of multi-objective parameters in casting-forging dual-control forming process， a novel method based on Back Propagation neural network optimized by Improved Particle Swarm Optimization （IPSO-BP） fusion with Finite Element Analysis （FEA） is proposed. Firstly， considering the characteristics of molding process， five processing parameters， including casting temperature of the molten metal， mold preheat temperature， mold filling velocity， casting-forging pressure and dwell time， are taken as input factors， and three parameters such as casting weight， surface defects， and tensile strength， are used as output indexes. The orthogonal latin hypercube design is adopted in order to obtain processing parameters for training samples， then the nonlinear functional relationship between the influencing factors and the optimization targets is constructed through the Back Propagation neural network （BP）. Secondly， the output error of neural network is utilized as the particle fitness， and the weight and threshold of BP are optimized by the improved particle swarm optimization algorithm， to establish the prediction model of processing parameters for multi-parameter optimization. The results from CAE finite element simulation show that this method can efficiently and accurately obtain the best combination of processing parameters in molding process， and provide guidance for adjusting and optimizing the processing parameters in casting-forging dual-control forming.

In the control of AC motors， the inverter output voltage is difficult to be accurately controlled due to the inverter disturbance， which affects the output torque and efficiency. The inverter disturbance has complex nonlinear characteristics， so it is difficult to be well compensated by simply using the inverter interlock time. Consequently， offline inverter disturbance calibration is required. In this paper， an observer-based inverter disturbance self-learning and compensation method is implemented. Inverter disturbance characteristics can be accurately obtained without offline calibration. At the same time， high-performance inverter disturbance compensation is realized. By implementing proposed compensation methods to a 1 kW prototype permanent magnet synchronous motor， the performance of the inverter disturbance self-learning and compensation are both verified.

The prominent vibration and noise problems are commonly exist in the initial design stage of gear system. To solve this problem， this paper takes the two-stage reduction gear system carried by an electric drive system as the research object. The minimum fluctuation of transmission error， minimum root mean square value of vibration acceleration in the direction of meshing line， and minimum load density of discrete points on meshing line are taken as optimization objectives. The tooth surface micro modification parameters are taken as design variable， and the algorithm NSGA-Ⅱ is used to conduct multi-objective optimization. The loaded tooth surface contact analysis（LTCA） is carried out based on the software MASTA. The results show that， compared with unmodified stage， the fluctuation of meshing transmission error and system transmission error after micro modification optimization are significantly reduced， the load distribution on the tooth surface is more uniform， the maximum contact stress and the amplitude of vibration acceleration of the bearing seat are greatly decreased， and the overall dynamic performance of gear system is improved. The research ideas and methods in this paper can provide guidance for wider optimization design of gear system.

In view of the difficulty in quantifying the parking success rate of automatic parking system in the narrow environment， a performance evaluation method for automatic parking system based on hardware in the loop simulation platform is proposed. Firstly， the boundaries of the feasible parking starting region and the needed minimum lateral space for parallel parking and perpendicular parking are established based on the combination of arc and line， and the area of the feasible parking starting region is calculated by image segmentation method. Then， the hardware in the loop simulation platform is established， and the automatic test framework is designed based on Python to test the actual parking starting region of automatic parking system. The ratio of the actual parking starting region to the feasible parking starting region is taken as the parking success rate of automatic parking system in the narrow environment， which can make up for the limitations of the existing performance evaluation methods of the automatic parking system. Finally， an automatic parking system made by an auto components company is taken as test object to validate the proposed performance evaluation method of automatic parking system， and the results show that the proposed method is feasible and efficient， and can improve the performance of test object.

Numerical and experimental methods were used to study the NACA 0015 airfoil structure ultrasonic deicing system. For the ice-coated double-layer plate model， the effects of the thickness and Young's modulus of ice layer on the interface shear stress concentration coefficient （ISCC） and deicing frequency were explored by the global matrix method （GMM）. At the same time， specific piezoelectric actuators were designed based on the theoretical research results and the accuracy of the finite element simulation was verified by the impedance curve. Finally， according to the geometric characteristics of the airfoil， the location of piezoelectric actuators and deicing efficiency were studied. Related deicing experiments were conducted to verify the feasibility of the ultrasonic deicing system. The experimental results showed that the deicing power of this airfoil structure is only around 16% of the traditional electro-thermal deicing.

To explore the traffic safety characteristics of the multi-vehicle interactive lane-changing at the merging area， the study of real-time multi-vehicle interactive risk assessment was carried out. Considering the characteristics of vehicle dynamics and multi-vehicle interaction， a Bayesian hierarchical model of real-time risk assessment was constructed based on the physical state layer， multi-vehicle interaction layer and risk probability layer of vehicles. The model parameters were calibrated by MCMC Gibbs sampling method. The model validity was tested by posterior prediction of p value and variation of parameter quantile. And the predictive performance of the model was evaluated by simulation analysis. The method of k-means clustering was adopted to classify the risk level of multi-vehicle interaction. The results show that the statistical BGR of model parameters are all less than 0.1. And the p value of Durbin-Watson's posterior， normal hypothesis， test distribution symmetry and kurtosis are all close to 0.5. So the risk assessment model is convergent and has good fitting. The AUC results show that the model is more accurate in identifying low risk， higher risk and highest risk， which has good performance. The research can evaluate the vehicle operation risk in a certain period of time and provide a reference for driving decision-making.

In order to accurately detect the impact resistance of superstructure of high?pier bridge and reduce the response time of structural impact resistance detection， an impact resistance detection method of superstructure of high?pier bridge under strong earthquake was proposed. According to the peak ground acceleration and spectral acceleration， the seismic intensity factor parameters were selected， the IDA curve of IM parameter component structure of high?pier bridge was constructed， and the structural dynamic characteristics were analyzed. The characteristic data was quantified horizontally， and the curvature corresponding to different anti?attack states was used. The impact resistance of the superstructure of high? pier bridge under strong earthquake in damage state was defined， and the impact resistance of the superstructure of high?pier bridge was detected according to the dynamic structure characteristics of seismic factors. The simulation results show that the response time of the proposed method is less than 0.5 s， and the maximum experimental error is 0.0615%. This method can quickly complete the structural impact resistance detection with high discrimination ability.

The typical unit of PHC rigid pile-raft composite foundation in high-speed railway is selected. Considering the undrained objective characteristics of rigid pile， the consolidation equations of soil in reinforcement area and underlying layer are established respectively， and the influence of foundation design parameters on consolidation characteristics is revealed in depth combining with examples. The results show that the consolidation of rigid pile-raft composite foundation has a time-varying effect. The filling period is mainly affected by the consolidation rate of the reinforcement area， and the resting period is mainly affected by the consolidation rate of the underlying layer. Design parameters such as replacement rate， relative reinforcement depth and pile-soil compression modulus ratio all have an impact on the consolidation rate of composite foundations， and the critical ranges are 0.03-0.06， 0.5-0.8 and 100-150， respectively. The rationality of the algorithm in this paper has been verified， and the research results can provide a theoretical basis for the design of rigid pile-raft composite foundations for high-speed railways in soft soil areas.

Experimental study of cylindrical samples under uniaxial cyclic compression was conducted to explore the effect of loading frequency and stress level on fatigue performance of ultra high toughness cementitious composite （UHTCC） under traffic loading. The results show that accumulation rate of stress-cycle curves could be disassembled into three stage with increase in number of cycles. Meanwhile， nonlinearity of strain increases gradually. The increase of loading frequency and stress level would reduce the fatigue life of specimens. The constitutive model for UHTCC was proposed based on the basis of uniaxial compression cyclic tests. Moreover， the stress path predicted by the model is in good agreement with the experimental results， which verifies the correctness and applicability of this model. Fatigue damage propagation is characterized by dissipated strain energy （DSE） based on energy-based mechanistic approach （EBM）. The initial damage （about 2%） is obtained and the development law of fatigue damage factor is revealed. Additionly， the fatigue cracking is fast at high stress ratio.

In order to overcome the shortcomings of traditional reinforced concrete buckling restrained brace， such as heavy weight， high cost and slow construction speed， it was necessary to find a kind of all steel buckling restrained brace which was light， high strength， low cost， high energy consumption， easy to install and replace. Inspired by the research of other scholars on the weakening of the core elements of buckling restrained braces， combined with the fractal Mandelbrot function， a new type of fractal buckling restrained braces prototype was conceived by using MATLAB. With the help of ABAQUS software， the advantages of fractal buckling restrained braces over traditional buckling braces， fractal buckling braces and traditional buckling restrained braces were compared and analyzed. The fractal dimension， length width ratio， width thickness ratio， interval thickness ratio and the thickness of externally restrained steel plate of partial interface and full interface restrained fractal buckling restrained brace were compared and analyzed， and the optimal energy dissipation specimen was selected. Meanwhile some real data were collected to verify it. It may provide a theoretical basis and a new design idea for the research and application of subsequent engineering reinforcement. The results showed that the stress of the fractal buckling braces was more dispersed than that of the traditional buckling braces， and the stress and displacement at the end joints were smaller. In the late stage， the weakened parts gradually yielded under the action of external forces， and the energy dissipation was better. Compared with traditional buckling braces （TR-0）， fractal buckling braces had better energy dissipation， which could increase by 5.11% ~ 8.97%. Compared with the partial fractal buckling braces （FD-1） with all the out-of-interface constraints （FD-1）， the force of the core element was more dispersed， the hysteresis curve was fuller， and the energy dissipation effect was better.

Aiming at the problems of low matching precision and complex calculation of traditional 3D scene surface reconstruction feature points， a new algorithm based on virtual reality technology is proposed. The image texture information is processed by wavelet decomposition， the image is encoded quantitatively and the distance between the visual points and images is taken as the measurement value. The appropriate critical value is selected to complete image compression and the redundant data of 3D scene image is eliminated. The heuristic information in the image is calculated， the pheromone matrix is updated， the critical value of the pheromone matrix is used to determine whether the pixel point is the edge point of the image， and the edge information of the 3D scene image is extracted. The minimum recognition distance of 3D scene is calculated by virtual reality equipment， and the deviation between theoretical image projection value and actual projection value is calculated. The pixel value of image surface reconstruction is calibrated by deviation value， and the surface reconstruction target of high precision 3D scene image is completed. The simulation results show that the proposed method can capture the key features of 3D scene image effectively， and the resolution of reconstructed image is improved significantly.

In the unsupervised domain adaptive transfer learning process， domain-independent features lead to the degradation of model segmentation performance， but there is no effective feature selection method for transfer learning segmentation model at present. To solve this problem， a general feature selection module for transfer learning was proposed based on optimal transport， which can be applied to various unsupervised domain adaptive image segmentation models. In this module， the optimal sample subsets of two domains are selected by weighted optimal transport of segmentation accuracy， and then the features of sample subsets are subjected to entropy regularized optimal transport， so as to obtain a descending list of similarity between two domains to remove domain-independent features. The universal feature selection module is applied to three unsupervised domain adaptive models to solve the problem of Covid-19 image segmentation， which improves the model performance to a certain extent.

Because the existing methods fail to extract the neighborhood features of the fencing continuous action image， the image segmentation results are not ideal， the anti-noise performance is poor， and the segmentation operation time increases. A fencing continuous action image segmentation method based on spatial neighborhood information is proposed. The feature points of the body contour were extracted automatically， the adaptive neighborhood geometric feature covariance matrix about the feature points as feature point descriptors was set， the similarity of each descriptor was measured， and the neighborhood features of the continuous fencing action were extracted. Then， through the neighborhood feature， the spatial constraint term function was added to the objective function， the spatial neighborhood membership function was set using the prior probability of the pixel， and the kernel function was added at the same time to optimize the image characteristics of the fencing continuous action. A new weighted membership function was obtained through membership function modification， which strengthens the proportion of neighborhood information clustering and completes image segmentation. Finally， an experimental test was carried out. After testing， it is proved that the proposed method can obtain the ideal segmentation effect， and at the same time can improve the anti-noise performance and reduce the segmentation operation time.

To study the internal relationship between features， a feature engineering algorithm based on sparse autoencoder （BioSAE） was proposed to encode given datasets， and it was assumed that the features encoded by sparse autoencoder might become better disease biomarkers. A comprehensive evaluation and experiment were carried out using 3494 methylation samples from 6 cancer types from TCGA. First， the encoded features were obtained through the sparse autoencoder， and then these features were analyzed and compared with the original methylation features. The experimental results show that in most modeling experiments conducted in this study， the BioSAE-encoded features are better than the original methylation features. Applying this algorithm to the datasets in the other research areas， such as image data， has also achieved a similar improvement.

Aiming at the problem that data integrity and privacy protection are difficult to take into account at the same time in wireless sensor networks， a data fusion privacy protection method PPMLEC with low energy consumption and integrity verification is proposed. PPMLEC expresses the ID number of a node by the combination of cluster number and node number， transmits the ID number implicitly， calculates the disturbance value by hash function， disturbs the data collected by the node， and improves the security of the data， and verifies the integrity at the base station by constructing two fusion trees. Experimental results show that， on the premise of not increasing network energy consumption， PPMLEC reduces the traffic by about 4.7% compared with the existing integrity protection scheme based on distributed authentication and by about 9.5% compared with DCSA scheme.

Aiming at the problems of large amount of data of national drug inspection business and low efficiency of storage and sharing of big data of drug inspection， a three-chain fusion block chain topology was proposed. The blockchain structure consists of a primary blockchain and two secondary blockchains. The main blockchain consists of a large number of fog nodes， and it is an editable lightweight blockchain. Each fog node attach two auxiliary chains externally. One of the secondary blockchains is an editable field fast detection blockchain. Another secondary blockchain is an editable field investigation blockchain， and it is a scalable twin chain blockchain. Bloom Filter is introduced into each block header extension region in each blockchain to improve the efficiency of blockchain data query. The introduction of fog-chain-cloud storage mode solves the problem of low efficiency of shared data storage， and it can reduce the pressure of data storage and high-frequency access on blockchain. Finally， an efficient drug inspection big data platform is built. Simulation experiments show that the three-chain fusion blockchain topology and various introduced blockchain measures can adapt to the business structure of drug inspection data sharing， and improve the efficiency of data sharing storage and access.

In view of the large workload of off-line bench calibration of the optimal operating line （OOL） of hybrid electric vehicle engines， and the difficulty of obtaining the optimal speed under actual road conditions， an online optimization method for the optimal operating line of hybrid electric vehicle engines was proposed. First， a hybrid electric vehicle simulation model was established， and the accuracy of the model was verified through simulations under winter urban condition. Secondly， a target gradient estimation method based on recursive least squares （RLS） with forgetting factor was proposed，the accurate calculation of the specific fuel consumption target gradient in the process of optimizing the optimal operating line was realized using the real?time data. Then， an online optimization method for the optimal operating point of the engine based on gradient descent was proposed to realize the real-time calculation of the optimal engine speed. The comparative simulation of traditional calibration methods verifies the superiority of proposed method in real-time and control effect.

At present， the research of cooperative adaptive cruise control algorithm mainly focuses on the longitudinal direction of single lane， and little consideration is given to the lateral control. However， the scenes like turning and lane changing are essential in actual driving process. Therefore， the dynamic model of a single vehicle in a platoon is decoupled in lateral and longitudinal directions based on Frenet framework in this paper. Aiming at the longitudinal control problem of the vehicle， the tracking performance of the controlled vehicle to its adjacent one and the leader is guaranteed by satisfying the exponential convergence condition， and the tracking weight is balanced by the sigmoidal function. As for the vehicle lateral control problem， the Lyapunov method is used to design the control algorithm. Finally， the effectiveness of the proposed control algorithm is verified by simulation experiments.

In order to solve the problem of real-time independent attitude planning of hexapod robot in omnidirectional marching， the coupling reason between the robot's motion planning and attitude planning was analyzed. A model based on the idea of body equivalent linkage was proposed， and the omnidirectional triangular gait and attitude control were realized based on the model. The experimental results of the prototype show that the control algorithm based on this model can be solved in real time in STM32F407 microcontroller， and the adjustment time of closed-loop attitude control is about 80% shorter than that of the traditional dual-PI control method.

To solve the problems of low straw burying rate and obvious straw returning and standing when plow body of the straw burying part of rapeseed direct seeding was designed with traditional turning way in rice oil rotation area of middle and lower reaches of Yangtze River， the technology scheme of side overturning and burying based on first lifting and then buckling of soil ridge was proposed， and a kind of buckling plough matched with rape direct seeder was designed， which can realize the function of side overturning and burying of soil ridge when the previous stubble was high stubble rice straw. The main structure and working process of the buckle plough were analyzed， and the mechanical model of the buckle plough stage was constructed. The buckle plough's buckle mode was defined， and the fact that the buckle plough can better realize the buckle function when the width depth ratio was 1.5~1.96 was determined. Based on the buckle way， the main structural parameters of plow surface were determined， and the operation resistance was analyzed. The results of field performance of buckle plough showed that the optimal parameters were： plowshare width of 150 mm， working depth of 180 mm， the resistance of 5.54 kN， and the rate of 92.25%； the field comparative test of button plow and turn plow showed that the rate of bucking was 34.54% higher than that of turning plow， and the performance was better than turning plow. The field verification test was carried out on the rape direct seeder with plough and rotary combination under the optimal parameters of bucking plough. The results showed that the average straw coverage rate was 91.41%， the evenness of compartment surface was 16.22~21.05 mm， and the broken soil rate was higher than 90.89%， which met the requirements of rape seed bed tillage. This study can provide a reference for the design of straw burying device and plow surface.

Due to the scattered spatial distribution， irregular shape and close to the spectral characteristics of surrounding vegetation， it is a very challenging work to extract tea plantations from satellite images. In order to solve this problem， a technical route of tea plantations extraction based on Random Forest feature selection method and Landsat-8 OLI image was proposed. In this study， Anji County in Zhejiang Province was selected as the research area. The spring， autumn and winter landsat-8 OLI images were used as the main data source. The importance evaluation， ranking and feature selection of initial features were carried out using Random Forest. The single season initial feature set， single season optimal feature set and multi-seasonal optimal feature set were designed， and nine groups of tea garden extraction experiments were carried out. The results show that multi-seasonal optimal feature set， which combines multi-seasonal information and feature selection advantages， has the best performance （PA = 87.5%； OA = 92.4%； Kappa = 0.897）. The technical route is reliable and practical for extracting tea plantations with scattered spatial distribution and irregular shape， and achieves relatively high accuracy while feature dimensionality reduces.